DROUGHT CHARACTERIZATION IN THE NORTHEAST BRAZIL: A MULTISCALE WATERSHED ANALYSIS AND REMOTE SENSING MONITORING PEDRO RODRIGUES MUTTI NATAL / RENNES December 2020 DROUGHT CHARACTERIZATION IN THE NORTHEAST BRAZIL: A MULTISCALE WATERSHED ANALYSIS AND REMOTE SENSING MONITORING PEDRO RODRIGUES MUTTI Doctoral thesis submitted to the Programa de Pós- graduação em Ciências Climáticas (PPGCC) of the Universidade Federal do Rio Grande do Norte (UFRN) and the École Doctorale Sociétés, Temps, Territoires (ED-STT) of the Université Rennes 2 (UR2) as part of the requirements to obtain the degree of PhD in Climate Sciences and Geography. Supervisors: Prof. Dr. Bergson Guedes Bezerra and Prof. Dr. Vincent Dubreuil JURY MEMBERS Prof. Dr. Sylvain Bigot (Université Grenoble Alpes) Prof. Dr. Josiclêda Domiciano Galvíncio (UFPE) Prof. Dr. Lara de Melo Barbosa Andrade (UFRN) Dr. Damien Arvor (LETG – Université Rennes 2) NATAL / RENNES December 2020 Universidade Federal do Rio Grande do Norte - UFRN Sistema de Bibliotecas - SISBI Catalogação de Publicação na Fonte. UFRN - Biblioteca Setorial Prof. Ronaldo Xavier de Arruda - CCET Mutti, Pedro Rodrigues. Drought characterization in the Northeast Brazil: A multiscale watershed analysis and remote sensing monitoring / Pedro Rodrigues Mutti. - 2020. 161f.: il. Tese (Doutorado) - Universidade Federal do Rio Grande do Norte, Centro de Ciências Exatas e da Terra, Programa de Pós- graduação em Ciências Climáticas. Natal, 2020. Orientador: Bergson Guedes Bezerra. Coorientador: Vincent Dubreuil. 1. Climatologia - Tese. 2. Semiarid - Tese. 3. Water balance - Tese. 4. Climate extremes - Tese. 5. Data series gap-filling - Tese. 6. Desertification - Tese. 7. NDVI - Tese. I. Bezerra, Bergson Guedes. II. Dubreuil, Vincent. III. Título. RN/UF/CCET CDU 551.58 Elaborado por Joseneide Ferreira Dantas - CRB-15/324 “Eu, em sonho um beija-flor Rasgando tardes vou buscar Em outro céu Noite longe que ficou em mim Noite longe que ficou em mim Quero lembrar Era um domingo de lua Quando deixei Jatobá Era quem sabe a esperança Indo à outro lugar Barcarola do São Francisco Velejo agora no mar Sem leme, mapa ou tesouro De prata ou luar” (Geraldo Azevedo) ACKNOWLEDGEMENTS The ending of this chapter of my academic journey inevitably leads to a series of reflections regarding what it means to develop and write a thesis. Moreover, to develop and write a thesis under a co-tutoring convention between two institutions separated by the Atlantic Ocean. As I see it, the lessons learned go well-beyond the scientific, climatological and geographical domains. Writing a thesis in co-tutoring requires trust, inventiveness, courage, teamwork and passion. Trust was taught by both my advisors, Prof. Bergson and Prof. Vincent. When I first met Bergson at the beginning of my Master’s in 2016 and Vincent in 2017 shortly before starting my doctoral studies, both listened to what I had to say and took a leap of faith. I am thankful to you for that. You also taught me trust in the course of our countless exchanges by pointing towards the correct direction when I seemed lost and unmotivated, by humbly sharing your experiences, and for accepting the challenges that came up due to my restless spirit. On that matter, Bergson and Vincent also taught me to be inventive. Scientific research requires great planning, but also requires the knowledge on how to act when things do not go as planned. Without your guidance, none of this would have been possible. I must say I deeply admire you both as researchers, as teachers, as colleagues and as human beings, and your passion about science inspires me daily. Developing a thesis in two institutions simultaneously also requires courage. Courage to brave the seas and face all the challenges associated with living in a country with a different culture, language and climate, far away from our roots and beloved ones. Thus, I am thankful to everyone that helped make this process more pleasant: all the new friends in France, and all the old friends in Europe and in Brazil. But even so, carrying out this undertaking was only possible with a team of collaborators that helped me one way or another to see this through. I am thankful to my fellow Phd students at the LETG-COSTEL Rennes, especially Gwenaël Morin for all the support, kindness and patience; and compatriots Geisa and Ayobami. I am also thankful to all other researchers and professors at the LETG and the Department of Geography of Université Rennes 2, particularly Valérie Bonnardot and Samuel Corgne; and to the administrative staff of the LETG and Rennes 2. Furthermore, I would like to thank the staff, professors and researchers at the LETG-Nantes, particularly Brazilian friend Beatriz Funatsu, with whom I hope to consolidate future collaborations. My teamwork acknowledgements extend to all colleagues and staff at the UFRN, the PPGCC and GEOMA: Thiago Valentim, Keila Mendes, Lizandro de Abreu, and many others. A special thanks to Prof. Cláudio Moisés and Prof. Lara Andrade, for the invaluable contributions and support throughout the development of the thesis and my academic formation; and to Prof. Paulo Sérgio Lucio for helping establish contact with Prof. Vincent at the beginning of it all. I would also like to thank the members of the qualification jury: Prof. Cristiano Prestrelo and Damien Arvor for the discussions, comments, and propositions that certainly inspired me to look at things differently; and the members of the thesis defense jury: Prof. Sylvain Bigot and Prof. Josiclêda Galvíncio for the invaluable contributions. I am thankful to the institutions that financed my doctoral studies: the Coordination for the Improvement of Higher Education Personnel – CAPES for the PRINT and DS scholarships; and the French Ministry for Europe and Foreign Affairs for the Eiffel Excellence scholarship. I am thankful to all national and international agencies responsible for measuring and disclosing the data used in the thesis: ANA, INMET, EMPARN, AESA, FUNCEME, CRU, NOAA and NASA. However, all this network of collaborators that directly or indirectly influenced the development of the thesis would not be sufficient if it were not for passion. Passion for the duty, for science, for discovery, for learning, for teaching, for connecting. Passion channeled through me by my parents and my family in Rio de Janeiro, Natal and João Pessoa. I am thankful to you all. And most importantly, passion channeled through me by my lifelong partner Amanda, my fiercest supporter, my morale booster, my fortitude and my inspiration. Thank you for your patience, for your trust and for your unrelentless faith in me. ABSTRACT Drought is a recurrent phenomenon in the Northeast Brazil (NEB) region, especially in its semiarid inlands, which are characterized by a remarkable climate variability, the expansion of desertification areas and persistent water-use conflicts. Furthermore, future climate change projections indicate that drought events will become more frequent and intense, exerting more pressure over the already vulnerable region. Although several drought studies have been carried out at the NEB, some important methodological aspects inherent to data quality and control, specificities of the used techniques, and spatial scale still need to be further discussed. Therefore, the objective of this thesis is to characterize different aspects of drought in the NEB considering different spatial scales, meteorological data characteristics, and remote sensing monitoring alternatives. This characterization was carried out in the São Francisco watershed (SFW), which presents a remarkable climate diversity, in the Piranhas-Açu watershed (PAW), which is mostly located in the semiarid NEB, and in desertification hotspots. In the first study, we thoroughly validated Climate Research Unit Time Series (CRU TS) gridded precipitation and potential evapotranspiration data over the SFW. CRU TS data presents overall good correlation with observed data. Then, we compared the applicability of the evaporation deficit and the Standardized Precipitation-Evapotranspiration Index as drought indices in the SFW. Results show that periods of water shortage are becoming more frequent and more intense in the coastal and middle zones of the basin, indicating an expansion of aridity. In the second study, we used gap-filled observed rainfall time series in order to propose a comprehensive climatological analysis in the PAW. A rainfall anomaly index was used to identify drought events, which are mostly associated with El Niño events and the anomalous warming of the Tropical North Atlantic Ocean. Finally, in the third study, different stochastic models were tested in order to forecast remotely sensed Normalized Difference Vegetation Index data (MOD13A2 product) over six desertification hotspots in the NEB. Results show that the tested models satisfactorily forecast short-term dry and degraded vegetation states. The results of this thesis contribute to the current general knowledge associated with drought assessment over semiarid regions, and the specific results of each study can be further explored by management agencies or local entities in the development of specific strategies to face and adapt to drought in the NEB. Keywords: Semiarid, water balance, climate extremes, data series gap-filling, desertification, NDVI, SPEI SUMMARY ACKNOWLEDGEMENTS ..................................................................................................... 3 ABSTRACT .............................................................................................................................. 5 SUMMARY ............................................................................................................................... 6 LIST OF ABBREVIATIONS AND ACRONYMS ................................................................ 9 LIST OF FIGURES ................................................................................................................ 14 LIST OF TABLES .................................................................................................................. 18 GENERAL INTRODUCTION ............................................................................................. 20 Objectives of the thesis .................................................................................................... 24 Thesis structure ............................................................................................................... 25 Context of the development of the thesis ....................................................................... 27 CHAPTER 1: DROUGHT AND THE NORTHEAST BRAZIL ................................... 29 1.1. Drought: definitions, concepts and approaches .................................................. 29 1.2. Main climate and vegetation features of the Northeast Brazil .......................... 38 CHAPTER 2: METEOROLOGICAL DROUGHT IN THE SÃO FRANCISCO WATERSHED ................................................................................................................. 45 Chapter 2A: Assessment of gridded CRU TS data for long-term climatic water balance monitoring over the São Francisco watershed, Brazil ................................................. 49 2.1. Introduction ............................................................................................................ 49 2.2. Material and methods ............................................................................................ 52 2.2.1. CRU TS v4.02 data ........................................................................................... 53 2.2.2. Point-based measurement data ......................................................................... 54 2.2.3. Thornthwaite’s potential evapotranspiration ................................................... 54 2.2.4. Observed data quality control .......................................................................... 55 2.2.5. Statistical assessment ........................................................................................ 57 2.3. Results ..................................................................................................................... 59 2.3.1. Overall spatial and temporal performance....................................................... 59 2.3.2. Seasonal performance ....................................................................................... 64 2.3.3. Trends and change-point comparisons ............................................................. 66 2.4. Discussion ................................................................................................................ 70 2.5. Conclusion .............................................................................................................. 73 2.6. Supplementary materials - Summary of the observed data characteristics and quality control ..................................................................................................................... 74 2.7. Appendix A ............................................................................................................. 83 Chapter 2B: Drought characterization in the São Francisco watershed using the climatic water balance: methodological aspects and spatiotemporal dynamics ...................... 87 2.8. Introduction ............................................................................................................ 87 2.9. Material and methods ............................................................................................ 91 2.9.1. Data................................................................................................................... 91 2.9.2. Thornthwaite and Mather climatological water balance ................................. 91 2.9.3. Standardized Precipitation-Evapotranspiration Index (SPEI) ......................... 93 2.9.4. Characterizing meteorological drought ........................................................... 95 2.10. Results and discussion ........................................................................................... 98 2.10.1. Comparison between the DE and the SPEI ...................................................... 98 2.10.2. Meteorological drought incidence patterns .................................................... 105 2.11. Conclusion ............................................................................................................ 121 CHAPTER 3: CLIMATOLOGICAL DROUGHT ASSESSMENT OVER SMALL- SCALE SEMIARID WATERSHEDS: THE CASE OF THE PIRANHAS-AÇU BASIN 123 A detailed framework for the characterization of rainfall climatology in semiarid watersheds ...................................................................................................................... 126 CHAPTER 4: REMOTE SENSING MONITORING OF VEGETATION OVER DESERTIFICATION HOTSPOTS: ALTERNATIVE APPROACHES ................. 127 NDVI time series stochastic models for the forecast of vegetation dynamics over desertification hotspots ................................................................................................. 130 FINAL CONSIDERATIONS .............................................................................................. 131 REFERENCES ..................................................................................................................... 136 9 LIST OF ABBREVIATIONS AND ACRONYMS AAO Antarctic Oscillation AAO_P Pure Antarctic Oscillation scenario ACF Autocorrelation functions AET Actual evapotranspiration AIC Akaike Information Criterion AMM Atlantic Meridional Mode AMM_P Pure Atlantic Meridional Mode scenario ANA National Waters Agency AR Auto-Regressive models ARIMA Auto-Regressive Integrated Moving-Averages models ARIMAX Auto-Regressive Integrated Moving-Averages with an Explanatory Variable models ARMA Auto-Regressive Moving-Averages models AWC Available water capacity BEST Bivariate ENSO Timeseries BJT Box-Jenkins-Tiao model CAB Cabrobró desertification hotspot CAPES Coordination for the Improvement of Higher Education Personnel CHIRPS Climate Hazards Group InfraRed Precipitation with Station data CRU TS Climate Research Unit Time Series 10 CRU Climate Research Unit DE Deficit of evaporation EN_INT Intensified El Niño scenario EN_P Pure El Niño scenario ENSO El Niño Southern Oscillation ESRL Earth System Research Laboratory EWD Easterly wave disturbances EXC Excess water/water surplus FS Frontal systems GHCN Global Historical Climatology Network GIL Gilbués desertification hotspot GPCC Global Precipitation Climatology Centre HW Holt-Winters model IDW Inverse Distance Weighting INH Inhamúns desertification hotspot INMET National Institute of Meteorology IRA Irauçuba desertification hotspot ITCZ Intertropical Convergence Zone JAG Jaguaribe desertification hotspot LMSF Lower-middle São Francisco LN_INT Intensified La Niña scenario 11 LN_P Pure La Niña scenario LSF Lower São Francisco MA Moving-Averages models MAD Median absolute deviation MAPE Mean absolute percentage error MD Mahalanobis distance MLAD Multiple Regression Least Absolute Deviations MODIS Moderate Resolution Imaging Spectroradiometer mRAI Modified Rainfall Anomaly Index MSF Middle São Francisco NDVI Normalized Difference Vegetation Index NEB Northeast Brazil NOAA National Oceanic and Atmospheric Administration NOR Normal scenario P Precipitation PACF Partial autocorrelation functions PAW Piranhas-Açu watershed PB Paraíba state PBIAS Percent bias PCA Principal component analysis PDSI Palmer Drought Severity Index 12 PET Potential evapotranspiration pMD Proportional Mahalanobis distance RAI Rainfall Anomaly Index RMSE Root Mean Square Error RN Rio Grande do Norte State SACZ South Atlantic Convergence Zone SARIMA Seasonal Auto-Regressive Integrated Moving Averages model SARIMAX Seasonal Auto-Regressive Integrated Moving Averages with an Explanatory Variable model SBE Single Best Estimator SER Seridó desertification hotspot SFW São Francisco watershed SNHT Standard Normal Homogeneity Test SPEI Standardized Precipitation-Evapotranspiration Index SPI Standardized Precipitation Index SST Sea surface temperature TeD Testing data TrD Training data TM-CWB Thornthwaite and Mather climatological water balance TRMM Tropical Rainfall Measuring Mission USF Upper São Francisco UTCV Upper tropospheric cyclonic vortices 13 WMO World Meteorological Organization 14 LIST OF FIGURES GENERAL INTRODUCTION Figure 1 – Organizational chart of the thesis. ....................................................................................... 27 CHAPTER 1: DROUGHT AND THE NORTHEAST BRAZIL Figure 1.1 – Conceptual model for the triggering and development of different drought types. Adapted from Wilhite (2000). ..................................................................................................................... 31 Figure 1.2 – Hypothetical description of the differentiation between natural drought events, and human-induced or -modified drought events. Simulated values (dashed line) are obtained by neglecting potential anthropic influences and comparing with observed values (solid line). It is important to note that anthropic effects can be either negative or positive. Adapted from Van Loon et al. (2016b). ...................................................................................................................... 32 Figure 1.3 – Graphical representation of the theory of runs applied to the identification of droughts by using a given index. X0 is the inferior threshold, T0 is the date of the onset of the drought event, Tf is the date of the end of the drought event, D is the duration of the event, s0 is the intensity of the event at its onset, and sn is its intensity at its end. ........................................................................ 34 Figure 1.4 – Example of drought monitoring proposed by the Monitor de Secas for the month of November 2019. Adapted from FUNCEME (2019). ................................................................... 36 Figure 1.5 – Scheme representing the main atmospheric systems causing rainfall in the Northeast Brazil: the Intertropical Convergence Zone (ITCZ), easterly wave disturbances (EWD), the South Atlantic Convergence Zone (SACZ) and frontal systems (FS); and its climate types according to Köppen’s classification. Adapted from Cavalcanti (2012) and Alvares et al. (2014). ...................................................................................................................................................... 39 Figure 1.6 – Effects of the El Niño Southern Oscillation phases on the Walker circulation and rainfall over the Northeast Brazil (NOAA, 2020). .................................................................................... 41 Figure 1.7 – Effects of the sea surface temperature (SST) gradient of the Tropical Atlantic Ocean on the positioning of the Intertropical Convergence Zone (ITCZ). Adapted from da Franca and Mendonça (2016). ......................................................................................................................... 42 Figure 1.8 – Effects of the positive and negative phases of the Antarctic Oscillation (AAO) on the positioning of the South Atlantic Convergence Zone (SACZ) over Brazil. ................................. 43 CHAPTER 2: METEOROLOGICAL DROUGHT IN THE SÃO FRANCISCO WATERSHED Figure 2.1 – Climate types in the São Francisco watershed and mean climatological behavior for each subregion: Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São 15 Francisco – LMSF; and Lower São Francisco – LSF. Adapted from Alvares et al. (2014) and Maneta et al. (2009). ..................................................................................................................... 46 Figure 2.2 – Location of the measuring stations used in the study and the CRU TS dataset grid (0.5° × 0.5°. USF: Upper São Francisco; MSF: Middle São Francisco; LMSF: Lower-middle São Francisco; LSF: Lower São Francisco. ......................................................................................... 54 Figure 2.3 – Example of the visual analysis of time series with the aid of the standard normal homogeneity test (SNHT) statistic (Ti) to find potential heterogeneities in data.......................... 57 Figure 2.4 – Point-based correlation between observational precipitation data and CRU TS gridded data considering the 75 years period (1942–2016), and three 25 years periods (1942–1966; 1967– 1991; 1992–2016). Overall correlation (r) and total number of stations (n) in each period are shown at the top of each map. USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. ............................................................................................................. 60 Figure 2.5 – Scatter plot of CRU TS rainfall data versus observed data in the: USF—upper; MSF— middle; LMSF—lower-middle; LSF—lower São Francisco watershed. n indicates sample size and r indicates the correlation coefficient. The blue line indicates the 1:1 rapport and the red line indicates the linear relationship between data. ............................................................................. 61 Figure 2.6 – Point-based correlation between observational Thornthwaite’s potential evapotranspiration data and CRU TS gridded data considering the 75 years period (1942–2016), and three 25 years periods (1942–1966; 1967–1991; 1992–2016). Overall correlation (r) and total number of stations (n) in each period are shown at the top of each map. USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. ......................... 63 Figure 2.7 – Scatter plot of CRU TS Thornthwaite’s potential evapotranspiration calculated using temperature data versus observed data in the: USF—upper; MSF—middle; LMSF—lower- middle; LSF—lower São Francisco watershed. n indicates sample size and r indicates the correlation coefficient. The blue line indicates the 1:1 rapport and the red line indicates the linear relationship between data. ............................................................................................................ 64 Figure 2.8 – Monthly root mean square error (RMSE) and mean percent bias (PBIAS) of CRU TS rainfall and Thornthwaite’s potential evapotranspiration data in relation to observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. ... 65 Figure 2.9 – Smoothed mean rainfall and potential evapotranspiration derived from point-based observations and the CRU TS dataset in the: USF—upper; MSF—middle; LMSF—lower- middle; LSF—lower São Francisco watershed. The linear trendline and the detected change- points are also indicated. .............................................................................................................. 67 Figure 2.10 – Spatial distribution of the slope of the smoothed monthly rainfall and Thornthwaite’s potential evapotranspiration (PET) time series derived from the CRU TS dataset (only the grids with significant trends are plotted) and from selected stations. USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. .................................................. 70 16 Figure 2.11 – Time series of the 12-month SPEI and the 12-month accumulated difference between excess water (EXC) and deficit of evaporation (DE) in: (a) Upper São Francisco – USF; and (b) Lower-middle São Francisco – LMSF. ........................................................................................ 99 Figure 2.12 – Identification of the driest years in the: (a) Upper São Francisco – USF; and the (b) Lower-middle São Francisco – LMSF, considering 12-month SPEI in December of each year and the yearly accumulated difference between excess water (EXC) and deficit of evaporation (DE). The scale of the graphics is not comparable. .............................................................................. 100 Figure 2.13 – Difference between monthly excess water (EXC) and deficit of evaporation (DE) (colored bars) and: (a) 12-month SPEI (black line); (b) 1-month SPEI (black line) in the Upper São Francisco – USF region during the period from Jan/2005 to Jan/2011. .............................. 101 Figure 2.14 – Difference between monthly excess water (EXC) and deficit of evaporation (DE) (colored bars) and: (a) 12-month SPEI (black line); (b) 1-month SPEI (black line) in the Lower- middle São Francisco – LMSF region during the period from Jan/2011 to Dec/2016. .............. 102 Figure 2.15 – Relationship between 1-month SPEI and the difference between precipitation (P) and potential evapotranspiration (PET) in December for the Upper São Francisco – USF and September for the Lower-middle São Francisco – LMSF. ......................................................... 103 Figure 2.16 – Spatial pattern of the evolution of drought in the period from July 2011 until June 2012 according to the 1-month Standardized Precipitation-Evapotranspiration Index (SPEI) and the monthly deficit of evaporation (DE). ......................................................................................... 105 Figure 2.17 – Interannual variability of the sequential water balance in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF, in the period from 1942 to 2016. .................................................................... 106 Figure 2.18 – Monthly frequency of the sequential water balance in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF, in the periods from: 1942 to 1966, 1967 to 1991 and 1992 to 2016. ................................. 108 Figure 2.19 – Spatial distribution of the mean deficit of evaporation (DE) over the São Francisco watershed in the months from January to June in the period from 1942 to 2016. The smaller maps show the proportion of the mean DE in each 25-year period: 1942 to 1966, 1967 to 1991 and 1992 to 2016. .............................................................................................................................. 110 Figure 2.20 – Spatial distribution of the mean deficit of evaporation (DE) over the São Francisco watershed in the months from July to December in the period from 1942 to 2016. The smaller maps show the proportion of the mean DE in each 25-year period: 1942 to 1966, 1967 to 1991 and 1992 to 2016. ....................................................................................................................... 111 Figure 2.21 – Mean and lower quartile (Q1) of the number of months with water surplus in the São Francisco watershed in the period from 1942 to 2016. The variation in the number of months in relation to the mean is shown in the smaller maps: 1 – from 1942 to 1966, 2 – from 1967 to 1991 and 3 – from 1992 to 2016. ........................................................................................................ 112 17 Figure 2.22 – Mean and upper quartile (Q3) of the number of months with water deficit in the São Francisco watershed in the period from 1942 to 2016. The variation in the number of months in relation to the mean is shown in the smaller maps: 1 – from 1942 to 1966, 2 – from 1967 to 1991 and 3 – from 1992 to 2016. ........................................................................................................ 114 Figure 2.23 – Annual behavior of the number of months in which P > PET and the number of months with deficit of evaporation (DE) higher than 40 mm in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF. Significant (α = 5%) linear trends (black line) are marked with a star (*). ................................ 115 Figure 2.24 – 12-month moving average behavior of the area with water deficit or surplus in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF. Significant (α = 5%) linear trends (black line) are marked with a star (*). ........................................................................................................................................ 116 Figure 2.25 – Annual anomaly of the accumulated deficit of evaporation (DE) during the wet season in each teleconnection scenario in relation to the normal scenario (NOR): for EN_P, AMM_P and EN_INT: February to May; for LN_P, AAO_P and LN_INT: November to March. The anomalies in the NOR scenario in relation to the mean of the entire 1942-2016 period is also shown for comparison. ............................................................................................................... 118 CHAPTER 3: CLIMATOLOGICAL DROUGHT ASSESSMENT OVER SMALL-SCALE SEMIARID WATERSHEDS: THE CASE OF THE PIRANHAS-AÇU BASIN Figure 3.1 – Main geographical and climate features of the Piranhas-Açu watershed. Adapted from Mutti (2018) and Mutti et al. (2019)........................................................................................... 124 CHAPTER 4: REMOTE SENSING MONITORING OF VEGETATION OVER DESERTIFICATION HOTSPOTS: ALTERNATIVE APPROACHES Figure 4.1 – Land cover classification and location of Brazilian desertification hotspots. ................. 128 Figure 4.2 – Magnitude of trends in NDVI (2000-2018) over six desertification hotspots in Brazil: Irauçuba – IRA; Jaguaribe – JAG; Cabrobró – CAB; Seridó – SER; Inhamúns – INH; and Gilbués – GIL. Figure extracted from Mutti and Bezerra (2018). .............................................. 129 18 LIST OF TABLES CHAPTER 1: DROUGHT AND THE NORTHEAST BRAZIL Table 1.1 – Estimative of expenses and economic losses associated with drought events in the Northeast Brazil. Elaborated with data from Marengo, Torres and Alves (2017). ....................... 37 CHAPTER 2: METEOROLOGICAL DROUGHT IN THE SÃO FRANCISCO WATERSHED Table 2.1 – Rainfall and potential evapotranspiration (PET) climatological characteristics and number of measuring stations selected in the subregions of the São Francisco Watershed. USF: Upper São Francisco; MSF: Middle São Francisco; LMSF: Lower-middle São Francisco; LSF: Lower São Francisco................................................................................................................................ 58 Table 2.2 – Summary of the trends and change-point detection analysis for observed and CRU TS rainfall and Thornthwaite’s potential evapotranspiration data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. Statistically significant trend slopes and change-points (α = 1%) are highlighted in italic. ........................................................ 68 Table 2.3 – Classification of months according to the water balance and the intensity of the deficit of evaporation (DE). Adated from Mounier (1977). ......................................................................... 93 Table 2.4 – Classification of anomalies identified by the Standardized Precipitation- Evapotranspiration Index. Adapted from Loukas and Vasiliades (2010). .................................... 95 Table 2.5 – Summary of the general expected effects of the different large-scale circulation mechanisms on drought over the São Francisco watershed (SFW). USF: Upper São Francisco, MSF: Middle São Francisco, LMSF: Lower-middle São Francisco, and LSF: Lower São Francisco....................................................................................................................................... 97 Table 2.6 – Characterization of the different scenarios related to the acting of large-scale circulation patterns and the respective representative years selected in the 1942-2016 period (+ positive, - negative, or n neutral). .................................................................................................................. 98 Table 2.7 – Frequency of occurrence of Standardized Precipitation-Evapotranspiration Index (SPEI) and deficit of evaporation (DE) classes for the months of December in the Upper São Francisco – USF and September in the Lower-middle São Francisco – LMSF. ........................................... 104 Table 2.A1 – Monthly mean (µ) ± standard deviation (σ) of CRU TS rainfall estimations and observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. The reliability of CRU TS estimates in each month is also shown based on whether monthly root mean square error is less than 50% of the mean observed value (where ✓ indicates reliable results and  indicates nonreliable results). ..................................................................... 83 19 Table 2.A2– Monthly mean (µ) ± standard deviation (σ) of CRU TS Thornthwaite’s potential evapotranspiration obtained through temperature estimations and observed data in the: USF— upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. The reliability of CRU TS estimates in each month is also shown based on whether monthly root mean square error is less than 50% of the mean observed value (where ✓ indicates reliable results and  indicates nonreliable results)................................................................................... 84 Table 2.S1 – Summary of the main characteristics of the measuring stations selected in the study. ANA: National Water Agency; INMET: National Institute of Meteorology; GHCN: Global Historical Climatology Network; USF: upper São Francisco; MSF: middle São Francisco; LMSF: lower-middle São Francisco; LSF: lower São Francisco. n refers to the original sample size. The last two columns refer to the number of removed observations in each respective step. ...................................................................................................................................................... 75 Table 2.S2 – Summary of observational rainfall data availability after each step of the quality control procedure at the seasonal scale. .................................................................................................... 81 Table 2.S3 – Summary of observational temperature (Thornthwaite’s potential evapotranspiration) data availability after each step of the quality control procedure at the seasonal scale. ............... 81 20 GENERAL INTRODUCTION Drought is a recurrent and inevitable climatic phenomenon characterized by the persistence of a water deficit over time. This phenomenon is considered a natural hazard with the potential to affect water availability and, consequently, the ecosystem services of regions under the influence of almost all climate types (ESFAHANIAN et al., 2017; LI et al., 2017; MISHRA; SINGH, 2010; WILHITE, 2000). Depending on its duration and intensity, drought may result in the collapse of the capacity of a natural environment to cope with human and ecosystem needs for water. The effects of this collapse in agriculture, vegetation and water supply are normally associated with important socioeconomic impacts (HAYES et al., 2011). Indeed, it is the natural disaster which affects most people worldwide (CUNHA et al., 2015; HEWITT, 1997). On the other hand, the magnitude of the observed socioeconomic impacts is directly related to the characteristics of the affected population and region. During persistent drought events, for example, it is common for more vulnerable regions, such as the Northeast Brazil (NEB), which is mostly under the influence of a semiarid climate, to declare state of emergency or public calamity since municipalities are unable to cope with the inflicted damage (GUTIÉRREZ et al., 2014; MARENGO; TORRES; ALVES, 2017). In this context, understanding these impacts is a complex undertaking, since they are a function of not only the severity of the water deficit, but also of the local geographical susceptibility to its effects (DUBREUIL, 1996; ZARGAR et al., 2011). Due to this complexity, this phenomenon is usually studied through four basic approaches: meteorological drought, agricultural drought, hydrological drought and socioeconomic drought (WILHITE; GLANTZ, 1985). In this conceptual model, the initial water deficit (meteorological drought) is triggered by the natural variability of the climate system. The persistence of this deficit leads to the reduction in soil moisture and water stress in plants (agricultural drought); as well as to the reduction of streamflow and surface water availability (hydrological drought). Finally, these effects observed in the natural environment will lead to a response by the social and economic system (socioeconomic drought), according to its characteristics. However, it is important to highlight that these drought typologies do not occur at the same time scale, since they usually depend on the duration and intensity of the precursor event and develop as chained events (BARKER et al., 2016; WANG et al., 2016; WILHITE; GLANTZ, 1985). 21 Currently the characterization and monitoring of the different drought typologies are carried out mainly by the use of drought indices. These indices are usually based on specific variables associated with the drought type being assessed (NIEMEYER, 2008; ZARGAR et al. 2011). For example, remote sensing data is usually used to characterize agricultural drought, since they allow the mapping of soil moisture conditions through the evaluation of surface and land cover parameters. There are also aggregated indices, which combine different information to provide a global response to the impacts of drought. Using indices is advantageous because they allow: almost real-time detection and monitoring of drought; definition of the onset and the end of a drought event; the establishment of drought severity levels; their regional representation; and the identification of their relation to different spatial and temporal scales (MISHRA; SINGH, 2010; ZARGAR et al., 2011). Although individual studies on each drought type are of the uttermost importance, they present some limitations. On the one hand, using specific indices allows to focus the analysis on drought effects over a specific domain. On the other hand, they do not allow the detailed understanding of the connections and links between different drought types (MELO et al., 2016). Oppositely, aggregated indices allow a global evaluation of drought impacts, but are usually insufficient to describe specific impacts and their interrelations (ESFAHANIAN et al., 2017). A comprehensive approach to drought should encompass specific indices and methods to characterize its causes (meteorological drought), its impacts by typology: in the vegetation and in the soil (agricultural drought), in water availability (hydrological drought); and its global impacts (socioeconomic drought). This comprehensive approach, however, requires a robust coordination and expertise in different domains, with the manipulation of a variety of datasets. Overall, integrated or individual drought studies are particularly important over vulnerable regions marked by the recurrence of conflicts for water use and the degradation of natural environments due to human activities (VAN LOON et al., 2016a), where water deficits are frequent and the conservation of natural resources is crucial (SAADI et al., 2018). This is the case of the NEB, which occupies approximately 18% of the Brazilian territory and is mostly under the influence of a semiarid climate (CUNHA et al., 2015; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). The NEB is characterized by a remarkable interannual climate variability, which modulates the occurrence of extreme rainfall (DA SILVA et al., 2018; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2014) and therefore makes persistent droughts a recurrent event in the region, historically affecting millions of people and drastically impacting regional and national economy (MARENGO; TORRES; ALVES, 2017). 22 In the semiarid portion of the NEB, the Caatinga (“white forest” in the tupi-guarani native language) is the main native vegetation type, which is composed by xerophyte deciduous and semi-deciduous species largely influenced by the persistence of dry conditions. This ecosystem is considered a hotspot of global biodiversity despite being systematically under anthropic pressure due to the expansion of agricultural areas and the lack of preservation public policies (BEUCHLE et al., 2015; KOCH; ALMEIDA-CORTEZ; KLEINSCHMIT, 2017). Furthermore, recent studies have shown that the productivity of the Caatinga biome is largely driven by the occurrence of rainfall, and that it might play an important role in regional and global carbon balances (CAMPOS et al., 2019; MARQUES et al., 2020; MENDES et al., 2020). Thus, the combination of anthropic influence and the incidence of drought events in the region might aggravate soil degradation, loss of vegetation and ecosystem productivity, and the expansion of areas under risk of desertification (CUNHA et al., 2015; MARIANO et al., 2018). These effects, in turn, feedback the aridification process, influencing the local hydrological cycle and impacting the associated ecosystem services (ADAMS, 2007; D’ODORICO et al., 2013). The interactions between climate aspects, vegetation and water resources in the NEB might become even more complex if the projections for future scenarios in the region are taken into account. For example, an increase in the demand for irrigation water is expected in several semiarid areas due to the transposition of the São Francisco river, the main water body transporting water from subhumid zones of the country towards its semiarid zones (BEZERRA et al., 2019; STOLF et al., 2012). In addition, recent studies indicate that the region will be affected by a decrease in rainfall rates and an increase in mean temperature over the next decades (FRANCHITO; FERNANDEZ; PAREJA, 2014; MARENGO; BERNASCONI, 2015; VIEIRA et al., 2015), which can already be observed as an increase in aridity (DUBREUIL et al., 2019). The expansion of desertification areas in the NEB have also been associated with a potential reduction in precipitation over the region (DE SOUZA; OYAMA, 2011). Furthermore, projections by the Intergovernmental Panel on Climate Change (IPCC, 2007) indicate an increase in the frequency and intensity of drought extreme events due to global climate change. For these reasons, there is a vast literature regarding drought assessment studies in the NEB, particularly in its semiarid region. Studies at the regional scale have been developed on the teleconnections between droughts and large-scale circulation patterns (ANDREOLI; KAYANO, 2006; HASTENRATH, 2006; LIU; NEGRÓN JUAREZ, 2001; MOURA; 23 SHUKLA, 1981); drought frequency, severity and duration (BRITO et al., 2018); drought monitoring through remote sensing (ANDERSON et al., 2016; BARBOSA; KUMAR, 2016; CUNHA et al., 2015; 2018; ERASMI et al., 2014; FERREIRA et al., 2018; MARIANO et al., 2018); characterization of specific drought events (CUNHA et al., 2019; DE MEDEIROS; DE OLIVEIRA; TORRES, 2020; DE MEDEIROS et al., 2020; MARENGO et al., 2017; MARTINS et al., 2018; RAO; HADA; HERDIES, 1995); trends in extreme events indices (DA SILVA et al., 2018; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017); and desertification simulation and monitoring (BARBOSA; HUETE; BAETHGEN, 2006; DE SOUZA; OYAMA, 2011; FERREIRA et al., 2017; OYAMA; NOBRE, 2004; TOMASELLA et al., 2018; VIEIRA et al., 2015). However, there are still some important issues to be discussed and investigated regarding the development of drought studies in Brazil. For instance, one recurrent problem is the lack of consistent, gap-free, long-term data series of meteorological variables in the NEB, which is crucial to develop reliable long-term drought monitoring studies (PAREDES-TREJO; BARBOSA; KUMAR, 2017; RODRIGUES et al., 2020; XAVIER; KING; SCANLON, 2016). Although alternative interpolated or satellite-derived gridded datasets are used in many cases to overcome that issue, they usually do not overgo careful specific validation before use. Also, most drought indices used to describe meteorological drought (such as the popular Standardized Precipitation Index – SPI) have well-known issues regarding their applicability in semiarid regions (STAGGE et al., 2015), such as the inefficiency in handling skewed distribution of data due to the inflation of zero precipitation values during the dry season. Nonetheless, they are being used in studies in the NEB with little discussion on alternatives to tackle this issue. Furthermore, despite the undeniable relevance of regional scale studies, more studies at the watershed scale should be encouraged since they are considered by the National Waters Agency (ANA – Agência Nacional de Águas) as the primary unit for hydrological planning and water resources management in Brazil (BRASIL, 1997). Thus, understanding the spatial and temporal characterization of droughts at the watershed scale allows the design of management strategies that consider the specific aspects of drought occurrence in each management unit. In addition, studies at the watershed scale may corroborate and provide further details on the results usually discussed when analyzing drought at the regional scale. Finally, drought monitoring by remote sensing techniques have been increasingly used and integrated with climatological data, inclusively to monitor drought impacts on the Caatinga 24 (BARBOSA; KUMAR, 2016; CUNHA et al., 2015; 2018; ERASMI et al., 2014). Nevertheless, methods to forecast seasonal vegetation response are still scarcely used due to the computational effort required to derive pixel-wise forecasting models. Thus, alternative methods should be explored in order to provide researches with new insights on how to monitor and forecast vegetation behavior and response to different meteorological conditions. Objectives of the thesis In this context, despite the wide number of studies developed aiming to investigate and monitor drought dynamics in the NEB region, there are still advancements to be made. By analyzing the currently published studies, several questions arise: how can smaller-scale or watershed-scale studies complement results usually found at the regional scale? How do the well-known effects of large-scale atmospheric patterns on drought over the NEB behave at these smaller scales? What alternatives can be used to assure long-term monitoring of drought in the NEB given the lack of quality observational measurements? Are the currently available indices adequate to provide an in-depth characterization of drought in tropical and semiarid regions? Is it possible to forecast vegetation dynamics in desertification hotspots using remote sensing data? These questionings can be summarized in a single general problematic: how to improve drought characterization in the NEB? In order to contribute to this discussion and to the overall context of drought studies over the NEB, this thesis was developed with the main global objective of characterizing different aspects of drought in the NEB considering different spatial scales, meteorological data characteristics, and remote sensing monitoring alternatives. To reach this objective, this thesis explores different datasets, methods and approaches to characterize drought and improve its assessment in areas with diverse climatic and physical aspects located in the NEB. In fact, three specific objectives were delineated: i) to evaluate the main average and anomalous patterns of incidence of meteorological drought in a multi-climate large-scale watershed using validated gridded datasets; ii) to propose a comprehensive approach to rainfall climatology, including drought assessment, in a small-scale watershed mostly located in the Brazilian semiarid region and using gap-filled observational data; 25 iii) to test and compare different stochastic models derived from remote sensing data in the forecasting of vegetation dynamics over desertification hotspots. Thesis structure This thesis comprises four chapters within a common baseline associated with the aforementioned specific objectives. Apart from the first chapter, each other chapter is a published (or to-be-published) article approaching specific issues at different spatial scales regarding the challenges of drought studies in the NEB, particularly its semiarid region. At the same time, the results retrieved in each study are unprecedent for the specific regions in which they were developed. Since different methodologies were adopted in each study, there is no general material and methods section, which will be presented within each study individually. The first chapter, “Drought and the Northeast Brazil”, comprises an in-depth analysis of the theoretical basis regarding drought assessment studies. In its first section, it covers a detailed literature review on the evolution of the main challenges and scientific discussions regarding drought worldwide and in the NEB. In its second section, the main physical features of the NEB region will be described in order to provide an overall context on which the following chapters will be based. The description will focus on the climate and vegetation aspects of the region. The second chapter is “Meteorological drought in the São Francisco watershed”. In this chapter, meteorological drought is studied in the São Francisco watershed (SFW), which is a large-scale basin with multiple climate types and a strategic role for the development of the NEB. The chapter is divided into two main complementary studies. The first study deals with the lack of consistent, gapless, long-term meteorological data in the basin by thoroughly validating precipitation and potential evapotranspiration (PET) gridded data from the Climate Research Unit Time Series (CRU TS) dataset. In the second study, the validated data is used to characterize meteorological drought in the SFW in the period from 1942 to 2016 (75 years) while also discussing some important methodological aspects regarding the applicability of currently-available drought indices over semiarid regions. In the third chapter, “Climatological drought assessment over small-scale semiarid watersheds: the case of the Piranhas-Açu basin”, we opted for a different approach in comparison to the first chapter. We selected the Piranhas-Açu watershed (PAW) in the Rio Grande do Norte (RN) and Paraíba (PB) states as study area. The PAW is much smaller when 26 compared to the SFW, and is mostly under the influence of semiarid climate. In this study we propose a comprehensive framework to develop climatological studies in watersheds with similar characteristics, that is, small-scale semiarid basins. Instead of using promptly available gridded datasets, we propose the integration of different gap filling methods in order to retrieve a reliable precipitation data series comprising the period from 1962 to 2015. In this study we carried out drought and rainfall climatology analysis based on the Modified Rainfall Anomaly Index (mRAI), which solely depends on rainfall data. In the fourth chapter, “Remote sensing monitoring of vegetation over desertification hotspots: alternative approaches”, we shift the focus of the studies from the climatological perspective to the surface monitoring of vegetation. In this study we explore the use of different stochastic models in the forecasting of vegetation states over six desertification hotspots in the NEB: Cabrobró, Seridó Jaguaribe, Inhamúns, Irauçuba and Gilbués. The performance of Holt-Winters, Box-Jenkins and Box-Jenkins-Tiao models was evaluated in forecasting mean Normalized Difference Vegetation Index (NDVI) and NDVI variance. Data from the Moderate Resolution Imaging Spectroradiometer (MODIS) MOD13A2 product comprising the period from 2000 to 2018 were used. The two modeled parameters (mean NDVI and NDVI variance) were than used in mean-variance plots that represent seasonal vegetation states. Therefore, this thesis adopts what could be called a top-down approach to drought in the NEB (Figure 1). We start from an overall discussion on the methodological drought aspects in a global context and in the context of the NEB (Chapter 1). Then, we assess some of these aspects at the scale of a large continental watershed (Chapter 2). We proceed to carry out a similar study at a small-scale regional watershed, addressing the specific issues that arise due to the specificities of smaller scale study areas in the NEB and its semiarid region (Chapter 3). Finally, we shift the focus of the last study towards specific vulnerable areas in the NEB at the local scale, adopting an approach based on remote sensing monitoring (Chapter 4). 27 Figure 1 – Organizational chart of the thesis. Apart from chapter 1, each other chapter will be introduced by a brief contextualization and followed by the integral published (or to-be-published) articles that derived from each proposed study. After the four chapters, a global conclusion summarizing the main contributions of this work to the scientific community is presented, also providing clues and paths for future researches on drought in the NEB. Context of the development of the thesis The thesis was developed under a co-tutoring agreement signed by the Universidade Federal do Rio Grande do Norte (UFRN), represented by the advisor Prof. Dr. Bergson Guedes 28 Bezerra of the Programa de Pós-graduação em Ciências Climáticas (PPGCC) and attached to the Grupo de Estudos Observacionais e de Modelagem da Interação Biosfera-Atmosfera (GEOMA) research group; and the Université Rennes 2 (UR2), represented by the advisor Prof. Dr. Vincent Dubreuil of the École Doctorale Sociétés, Temps, Territoires (ED-STT) and attached to the Littoral, Environnement, Géomatique, Télédétection (LETG-COSTEL UMR 6554) research unit. The interaction between research groups started in 2014 with the exchange of PhD students between institutions. In 2017, the specific collaboration for the development of the present thesis started, with the signing of the co-tutoring agreement between institution taking place in 2018. The thesis was financed in Brazil by the Coordination for the Improvement of Higher Education Personnel (CAPES). Furthermore, the project also received grants from the Internationalization Program of the CAPES (PRINT/CAPES) and the Eiffel Excellence Scholarship Program of the French Ministry for Europe and Foreign Affairs in order to enable the development of the research in the French laboratory in 2019 and 2020. 29 CHAPTER 1: DROUGHT AND THE NORTHEAST BRAZIL In this first chapter, a brief literature review will be presented on the main theme of the research. In section 1.1 the thesis will be contextualized based on an in-depth theoretical framework regarding drought studies worldwide. Conceptual and operational definitions of drought will be discussed, as well as the Brazilian historical context regarding water and drought policies. Section 1.2, on the other hand, will focus on an overall characterization of the NEB region, focusing on its main physical and socioeconomic aspects associated with the cause and effects of drought in the region. 1.1. Drought: definitions, concepts and approaches Objectively defining drought has been one of the main challenges faced by scientists studying this phenomenon. At a first moment, between the 1960s and the 1980s, the disciplinary compartmentalization of drought studies was already being discussed, with specific drought definitions being adopted depending on the objectives of each study (SUBRAHMANYAM, 1967). For example, meteorologists would study drought through precipitation deficits and its atmospheric characteristics, agronomists would study it considering the reduction in soil moisture, while hydrologists would focus on changes in streamflow rates. However, the limits of this compartmentalization were not clearly defined, and even within the same compartment there were still divergencies on how to operationalize drought studies (WILHITE; GRANTZ, 1985). Despite advancements, this lack of centralization and consensus on the definition of drought lingered throughout the 1980s. In 1985, however, the geographer Donald Wilhite and the engineer Michael Glantz published a theoretical-conceptual study on the overall knowledge about droughts, with fundamental conclusions on the role of its definitions (WILHITE; GLANTZ, 1985). In this study, the authors identified over 150 drought definitions which have been used until that time. The conclusions of said study promoted a paradigm shift regarding studies on this phenomenon. Among them, we highlight: (1) there cannot (and should not) have a single universal definition of drought; (2) most studies focus on the physical aspects of drought while ignoring its social aspects; (3) due to the multiple interests involved when studying this phenomenon, it is useful to divide drought types into four groups: meteorological, agricultural, hydrological and socioeconomic. These conclusions were consolidated in a subsequent study (WILHITE, 2000), which established the conceptual and theoretical framework to be used explicitly or implicitly in most 30 drought studies that followed. The conceptual model by Wilhite (2000) is presented in Figure 1.1. In this model, drought is considered as an event triggered by the natural climate variability, and, as such, it is recurrent, inevitable, and capable of affecting regions under the influence of basically all climate types worldwide (ESFAHANIAN et al., 2017; LI et al., 2017; MISHRA; SINGH, 2010). This natural variability of the climate system may cause precipitation deficits, temperature anomalies, wind intensification, increased solar radiation and reduction in cloud cover (meteorological drought). These conditions have variable durations and intensities and they may lead to an increase in evapotranspiration and a reduction in soil infiltration if they persist. These changes, in turn, will favor the reduction of soil moisture, generating water stress in the vegetation with a consequent reduction in biomass and productivity (agricultural drought). If these conditions are upkeep by the persistence of the anomalous climate and pedological conditions, it will eventually affect the hydrological cycle, with a reduction in streamflow rates and in the recharge of surface and groundwater bodies (hydrological drought). At this stage, the resulting effects on agriculture and livestock production, and on water supply to human activities are expected to impact the population living in the afflicted region, leading to economic, material and environmental losses (socioeconomic drought). Furthermore, once this propagation system is established, a positive feedback process is triggered (ADAMS, 2007; MISHRA; SINGH, 2010). Soil moisture depletion reduces photosynthetic activity and evapotranspiration, which in turn reduces relative air humidity. With a drier air, the probability of occurrence of favorable conditions to reach air saturation reduces and, consequently, so does precipitation. Usually, the end of a well-established drought condition takes place only with the occurrence of external disturbances which transport sufficient moisture to produce rainfall over the afflicted region (BRAVAR; KAVVAS, 1991). 31 Figure 1.1 – Conceptual model for the triggering and development of different drought types. Adapted from Wilhite (2000). One can notice through Figure 1.1 that the different drought typologies are interdependent, occurring as a chained process in time and according to the alterations observed in the compartments of the hydrological cycle (LAMBERT, 1977). On the other hand, the final impacts will be perceived at different intensities and at different time scales depending on the social structure and the type of production system observed in the afflicted region. For example, regions adapted to this kind of phenomenon, with alternative water storage and irrigation systems will probably suffer relevant impacts only during events in which water deficits persist for a long time. Therefore, drought represents a natural hazard with associated risks that are directly related to the degree of exposure of a given region and to the vulnerability of the affected population (DUBREUIL, 1996; WILHITE, 2000; ZARGAR et al., 2011). 32 Throughout the years, the interest of the scientific community in studying the effects of anthropic activities in climate has drastically increased. Currently, one cannot disassociate studies on the climate system and its phenomena from the global climate change context, which is intensified by anthropic activities (IPCC, 2007). Thus, a redefinition of the conceptual model of drought propagation initially described by Wilhite (2000) was proposed, in such a way that human activities should be considered part of the drought propagation system (VAN LOON et al., 2016b). In the original model, mankind plays the role of a passive ‘target’ of the system, located at the end of the drought propagation cascade. For Van Loon et al. (2016b), it is urgent to also consider that water deficit can be caused or modified by changes in the hydrological cycle due to human activities, such as deforestation, urban expansion or interferences on the natural flow of water bodies (Figure 1.2). Figure 1.2 – Hypothetical description of the differentiation between natural drought events, and human-induced or -modified drought events. Simulated values (dashed line) are obtained by neglecting potential anthropic influences and comparing with observed values (solid line). It is important to note that anthropic effects can be either negative or positive. Adapted from Van Loon et al. (2016b). Although the influence of human activities on the propagation of drought events has been implicitly discussed in several studies throughout the years (VAN LOON et al., 2016b), this paradigm shift further highlights the human role on the occurrence of this climate events. In this context, the new adopted approach would refer to drought in the Anthropocene, or drought in the Age of Man. In the present thesis, the impact of human activities and changes in the environment on the occurrence of drought will be a central theme in the discussions, although not in an explicit manner as presented in Figure 1.2. Even though the conceptual model by Wilhite (2000) and Van Loon et al. (2016b) established a structured framework for the definitions and relations between different drought types, providing a dynamic character to this phenomenon, there is still the need to operationalize 33 these definitions. That is, to use objective methods to identify the onset, the duration and the severity of drought events of each typology. To this end, the most commonly used techniques are drought indices, derived from variables that represent each drought typology to be studied. However, the operational definitions of drought also have been through important changes as evidenced by the literature, which will be discussed next. A pioneering and relevant study that innovated on this issue was the study Meteorological Drought by meteorologist Wayne Palmer (PALMER; 1965). The author firstly argued that the operational definitions usually adopted at that time, such as: monthly precipitation below a certain normal threshold; period of consecutive dry days; or a condition where water supply cannot meet the demands, treated drought as an absolute concept and were based on arbitrary generalizations. Thus, Palmer (1965) purposely adopted a conceptual definition that is arbitrary and general: drought consists of an anomalous and prolonged moisture deficit. On the other hand, the author proposed the elaboration of an index – the Palmer Drought Severity Index – (PDSI) that allows an objective calculation of the magnitude and duration of this moisture deficit based on a simplified water balance. Thus, he derived an objective and replicable operational definition with a physical sense from an arbitrary and general conceptual definition. It is important to note that the PDSI was developed through an exclusively meteorological perspective and the agricultural and hydrological aspects of drought should be treated as effects of the observed meteorological conditions (PALMER, 1965). Another important advancement on the operational identification of droughts can be identified in the study by Guerrero-Salazar and Yevjevich (1975). These authors proposed the application of the theory of runs, initially proposed by Mood (1940), in time series of variables representing water availability. The theory of runs consists of the identification of sequences of events (in this case, drought events), preceded or followed by different events (in this case, non- drought events). For example, Figure 1.3 shows the behavior of a time series of a given precipitation index. In a given moment, an inferior threshold (X0) is reached, defining the onset of a drought event (T0). The event persists until its end (Tf) with a duration D as long as the value of the precipitation index remains below the established threshold. For each identified event, one can estimate its total severity (S – sum of the intensities s0 until sn) and magnitude (S/D). 34 Figure 1.3 – Graphical representation of the theory of runs applied to the identification of droughts by using a given index. X0 is the inferior threshold, T0 is the date of the onset of the drought event, Tf is the date of the end of the drought event, D is the duration of the event, s0 is the intensity of the event at its onset, and sn is its intensity at its end. A few years later, Dracup, Lee and Paulson Jr. (1980) proposed a systematic approach to the drought identification process in order to avoid generalizations. In this approach, the research must: (1) define the type of deficit to be studied (meteorological, agricultural or hydrological); (2) define the time scale (week, month, year); (3) define a threshold level (threshold that separates drought events from the rest of the time series); and (4) normalize data in order to allow the comparison between regions. In practical terms, the authors proposed a systematization of the process previously used by Palmer (1965) to create the PDSI, but also taking into account the theory of runs. This systematization was subsequently used in the elaboration of several drought indices. A review carried out by Niemeyer (2008) and expanded by Zargar et al. (2011) identified almost 150 different drought indices. The authors realized that they could be classified according to the drought type they were conceived to assess. For example, indices purely based on meteorological data such as precipitation and temperature are used to identify meteorological drought. Indices based on remote sensing data are normally used to monitor soil moisture conditions, that is, agricultural drought. There are also indices that combine information from different sources (meteorological, pedological and hydrological) to provide a single response value representing global drought conditions over a given region. However, this wide variety of available indices is not necessarily advantageous. In reality, it evidences a certain decentralization regarding drought studies and even a saturation 35 on the availability of indices (NIEMEYER, 2008), since several of those cannot be properly incorporated by the scientific community. As a first solution to this issue, the Lincoln Declaration on Drought Indices was elaborated (HAYES et al., 2011). The declaration was the final product of a meeting promoted by the World Meteorological Organization (WMO) with the participation of 54 drought researchers from 22 different countries. The objective of the meeting was to debate and discuss the different drought indices available, in order to elaborate specific consensual recommendations on the use of standard-indices for the identification and monitoring of meteorological, agricultural and hydrological drought. No consensus was reached for agricultural and hydrological droughts but the Declaration recommended, at that time, the use of the SPI (MCKEE; DOESKEN; KLEIST, 1993) as the standard index for the identification of meteorological drought. Among the final recommendations of the Lincoln Declaration, the development of drought early warning systems was encouraged (HAYES et al., 2011). An example of a successful integrated system which was implemented at the operational level is the Drought Monitor, which monitor the magnitude, the extension and the impacts of drought in the United States through the study of different drought indices, emitting alerts and aiding in the planning of response actions to crisis (SVOBODA et al., 2002). In Brazil, in 2012, several institutions articulated in order to develop and implement a Brazilian version of the Drought Monitor. In 2014, the Monitor de Secas was launched under the coordination of the Ministry of Integration, the National Institute of Meteorology (Instituto Nacional de Meteorologia – INMET) and the ANA (MARTINS et al., 2015). Systems that aggregate different indices are advantageous because they allow the systematic and global evaluation of drought, but usually do not present sufficient information to understand the specific impacts and the correlation between drought types (ESFAHANIAN et al., 2017), as shown in Figure 1.4. Nevertheless, the Monitor de Secas represents an invaluable tool for drought monitoring in Brazil. Indeed, this initiative complies not only with the recommendation of the Lincoln Declaration, but also with the historical need of the semiarid region of the NEB for tools and legal mechanisms that aid in drought management. 36 Figure 1.4 – Example of drought monitoring proposed by the Monitor de Secas for the month of November 2019. Adapted from FUNCEME (2019). The semiarid region of the NEB is historically afflicted by drought events that cause immeasurable losses to agriculture and human losses due to hunger, malnutrition, the dissemination of diseases and the intensification of migratory fluxes (MARENGO; TORRES; ALVES, 2017). Since the 1940s and 1950s, the Brazilian government has been adopting several policies to combat drought in the region, achieving different results (BURITI; BARBOSA, 2018; MARENGO; TORRES; ALVES, 2017). The historically adopted policies include the delineation of strategic zones such as the Drought Polygon (BRASIL, 1946), which was eventually replaced by the Brazilian Semiarid (BRASIL, 1989), which in turn encompasses 1262 municipalities in the NEB and in the Minas Gerais state as of its last update in 2017 (BRASIL, 2017). Furthermore, several administration agencies were created with the objective of developing the NEB and the Brazilian Semiarid, such as the National Drought Prevention Works Department (Departamento Nacional de Obras Contra as Secas – DNOCS) and the Superintendence for the Development of the Northeast (Superintendência de Desenvolvimento do Nordeste – SUDENE). 37 However, the water policies usually adopted such as the construction of dams, barrages, and reservoirs in the region were not followed by management, inclusion and local development policies and, therefore, the expected development and transformation of the Brazilian Semiarid did not occur (BURITI; BARBOSA, 2018). In fact, in the last 70 years, expenses associated with drought combat and economic losses associated with its impacts remain extremely high, as shown in Table 1.1. Table 1.1 – Estimative of expenses and economic losses associated with drought events in the Northeast Brazil. Elaborated with data from Marengo, Torres and Alves (2017). Drought event Expenses/losses (dollars) 1958 803 million 1970 430 million 1976 447 million 1979-1983 7,8 billion 2012-2015 6 billion Only in 1997, with the enactment of the National Water Resources Policy (BRASIL, 1997), specific guidelines for the management of water resources in Brazil were elaborated. With this law, watersheds were now considered minimum planning units for water policies, and Watershed Committees assured the participation of different social agents in water governance (BURITI; BARBOSA, 2018; SIEGMUND-SCHULTZE et al., 2015). This new legal configuration allowed the implementation of large-scale inclusive water and drought policies that contemplated social technologies, such as the One Million Cisterns Program (Programa Um Milhão de Cisternas) (DIAS, 2013; GOMES; HELLER, 2016; LINDOSO et al., 2020). This program allowed the large-scale construction of water cisterns adapted to the Brazilian Semiarid conditions in the property of local subsistence farmers, assuring rainwater harvesting and the access to water even during unfavorable climate periods (BURITI; BARBOSA, 2018). Not only this, but the more social-oriented governments of Brazil during the 2000s and early 2010s were responsible for creating important social policies that also contributed to increase the semiarid population resilience and adaptive capacity to drought. The Bolsa Família cash transfer mechanism and food security programs such as the Brazilian Food Acquisition Program are known to have positively influenced semiarid local farmers living conditions (BEDRAN-MARTINS; LEMOS, 2017; MESQUITA; BURSZTYN, 2017). Nevertheless, the severe drought that occurred in 2012 in the region exposed fragilities in such programs regarding their effectivity during extreme climate events. Indeed, even with improved social conditions, severe drought events still play an important role in determining vulnerability in the 38 semiarid Brazil. This highlights the need to incorporate the environmental sphere into the design of specific social policies to the NEB and its semiarid region (MESQUITA; BURSZTYN, 2017). Furthermore, recent studies have shown that despite the emergence of these water and social policies, there is still an important politic component associated with clientelism and the influence of local political powers (MILHORANCE et al., 2020). For example, despite the undeniable improvements derived from the One Million Cisterns Program, its long-term efficiency proved to be rather inconsistent due to the lingering of clientelist practices (LINDOSO et al., 2014, 2018; MILHORANCE et al., 2020). It does show, however, that the semiarid NEB is ready for the implementation of adaptation strategies to face drought in regions where the political influence is less prominent (LINDOSO et al., 2014), which could drastically reduce future expenses in remediating drought impacts while also improving social conditions. This is the overall context into which this thesis is situated. A context in which there is an urgent need for studies that contemplate the different physical aspects of drought but also for studies that do not exclude sociodemographic aspects from the discussion and the characterization of the different drought types. In the next section, an overall description of the main climate and vegetation features in the NEB will be discussed, which will be crucial for the understanding of the specific studies that compose the thesis. 1.2. Main climate and vegetation features of the Northeast Brazil According to Köppen’s climate classification, the NEB region is under the influence of three major climates: semiarid, tropical and humid subtropical (ALVARES et al., 2014), as shown in Figure 1.5. However, different rainfall regimes are observed throughout its territory, as previously analyzed in different studies (DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017; RODRIGUES et al., 2020; TINÔCO et al., 2018). These regimes are modulated by the acting of different atmospheric systems and the occurrence of drought is strongly related to these mechanisms. The southern and southwestern tropical portions of the NEB and the few neighboring humid subtropical areas (southern Bahia – BA state) are characterized by annual precipitations ranging from 1000 to 1300 mm with the wet season defined between November and March (BEZERRA et al., 2019, DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). Indeed, this is a 39 transition zone between the humid subtropical climate typical of the higher latitudes in Brazil and the semiarid NEB. Figure 1.5 – Scheme representing the main atmospheric systems causing rainfall in the Northeast Brazil: the Intertropical Convergence Zone (ITCZ), easterly wave disturbances (EWD), the South Atlantic Convergence Zone (SACZ) and frontal systems (FS); and its climate types according to Köppen’s classification. Adapted from Cavalcanti (2012) and Alvares et al. (2014). In these regions, rainfall is modulated mainly by the establishment of the South Atlantic Convergence Zone (SACZ) during the summer months (CAVALCANTI, 2012). The SACZ consists of the persistence of an elongated axis of clouds oriented in a northwest-southeast position from the south of the Amazon towards the coast of the Southeast region of Brazil (Figure 1.5) (NOGUÉS-PAEGLE; MO, 1997), being responsible for the occurrence of heavy rainfall over the state of Minas Gerais (MG) and the southern NEB (CAVALCANTI, 2012; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). Frontal systems (FS) also act over this region (Figure 1.5) and are more frequent than the SACZ, although their effects in the occurrence of heavy precipitation rates are weaker (LIMA; SATYAMURTY; FERNÁNDEZ, 2010). Besides occurring separately, FS may also be responsible for the formation of the SACZ when a cold front remains stationary over the region and convective activity is maintained (ANDRADE; CAVALCANTI, 2018; NIETO-FERREIRA; RICKENBACH; WRIGHT, 2011). 40 The inlands of the NEB are characterized by a mostly semiarid climate, with annual rainfall rates ranging from 400 to 800 mm and the wet season usually lasting three to four months between February and May. In this region, rainfall is modulated mainly by the displacement of the Intertropical Convergence Zone (ITCZ) toward the south of the equator (~ 4ºS, Figure 1.5) during said months (MARENGO et al., 2011; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). Annual variations in the precipitated volume over this region are highly dependent on the intensity of the seasonal displacement of the ITCZ (HASTENRATH, 2012; CAVALCANTI, 2012). The western NEB (Maranhão state – MA and part of the Piauí state – PI) and its northern coast usually present higher rainfall rates due to sea breeze circulation and the formation of squall lines, also associated with the ITCZ (DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017; DOS REIS et al., 2020). Similarly, the border between the Ceará (CE), Rio Grande do Norte (RN) and Paraíba (PB) states also represent a slightly wetter region if compared to the neighboring semiarid lands (Figure 1.5). In this particular region, the landscape and topography strongly favor the occurrence of orographic rainfall and local convection, thus characterizing an ‘oasis’ amidst the semiarid dryness (DE ANDRADE et al., 2016). The entire eastern coast of the NEB is marked by a tropical climate with annual rainfall ranging from 1000 to 1300 mm. In the northeastern coast rainfall occurs mainly during the winter, with a dry season established during summer (ALVARES et al., 2014). Besides the effect of sea breezes, heavy rainfall events in the region are mainly associated with easterly wave disturbances (EWD) that propagate from the Atlantic Ocean towards the coast of the NEB (Figure 1.5) (CAVALCANTI, 2012; GOMES et al., 2019; DE OLIVEIRA; LIMA; SANTOS E SILVA, 2013). In the southeastern coast however, there is no dry season due to the recurrent occurrence of upper tropospheric cyclonic vortices (UTCV) during the summer (DE OLIVEIRA; SANTOS E SILVA; LIMA, 2013). Furthermore, the entire NEB is marked by a high interannual variability in rainfall rates, with alternating dry years and years with heavy rainfall. This variability is mainly associated with teleconnection patterns observed between the occurrence of the aforementioned atmospheric systems and large-scale circulation mechanisms. Among these, the most influential to the occurrence of drought in the NEB are the El Niño Southern Oscillation (ENSO), the Atlantic Meridional Mode (AMM) and the Antarctic Oscillation or the Southern Annular Mode (AAO). 41 The ENSO refers to anomalous oceanic and atmospheric patterns associated with sea surface temperature (SST) over the Tropical Pacific. In its El Niño phase, an anomalous warming of the Eastern Tropical Pacific is observed, which leads to a displacement of the ascending branch of the Walker’s cell from the Amazon towards this portion of the ocean (Figure 1.6) (DE SOUZA; AMBRIZZI, 2002; UVO et al., 1998). Therefore, the descending branch of the Walker cell positions itself over the northern Tropical Atlantic (Figure 1.6), weakening the trade winds and influencing the positioning of the ITCZ, which displaces further north (HASTENRATH, 2012). This configuration inhibits convective activity over most of the Brazilian Semiarid and the northern portion of the NEB. Furthermore, Gomes et al. (2019) showed that the El Niño ENSO phase is also associated with a decrease in the propagation of EWD over the coastal NEB, thus also affecting rainfall rates over this region. Figure 1.6 – Effects of the El Niño Southern Oscillation phases on the Walker circulation and rainfall over the Northeast Brazil (NOAA, 2020). During La Niña years, the sea surface of the Tropical Pacific is anomalously cooler, which favors the stationary position of the ascending branch of the Walker cell over the Tropical Atlantic, strengthening trade winds and convective activity over this region (HASTENRATH, 2012). Thus, during La Niña years, the opposite effects of the El Niño phase are observed over the NEB. That is, increased rainfall rates over most part of the semiarid lands and decreased rainfall over the eastern coast. The SST over the Atlantic also plays an important role on rainfall over the NEB, mainly influencing the semiarid region. The inter-hemispheric temperature gradient in the Tropical Atlantic, or AMM, influences the positioning of the ITCZ during the austral summer, which is 42 the wet season over the semiarid NEB (CAVALCANTI, 2012; HASTENRATH, 2012; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). When sea surface over the Northern Tropical Atlantic is anomalously warmer, the ITCZ displaces further north (~ 4ºN a 5ºN), intensifying drought conditions in the NEB (Figure 1.7). Oppositely, when SST over the Southern Tropical Atlantic is anomalously warmer, the ITCZ displaces further south (~ 4ºS), increasing rainfall over the NEB (Figure 1.7). Figure 1.7 – Effects of the sea surface temperature (SST) gradient of the Tropical Atlantic Ocean on the positioning of the Intertropical Convergence Zone (ITCZ). Adapted from da Franca and Mendonça (2016). Finally, the AAO is characterized by the north-south displacement of the westerly winds belt located at middle and higher latitudes of the southern hemisphere (Figure 1.8). In its negative phase, the wind belt displaces further north. The effect of this oscillation is the northern displacement of the SACZ and the reduction of its intensity (CARVALHO; JONES; LIEBMANN, 2004; CAVALCANTI, 2012; ROSSO et al., 2018; ZILLI; CARVALHO; LINTNER, 2019). On the other hand, this northern positioning of the wind belts strengthens the southern subtropical jet, dislocating it further north (CARVALHO; JONES; AMBRIZZI, 2005; REBOITA; AMBRIZZI; DA ROCHA, 2009). This configuration intensifies the propagation of FS and the formation of the SACZ in a northernmost position, increasing the occurrence of rainfall over the BA state and the southeastern coast of the NEB 43 Figure 1.8 – Effects of the positive and negative phases of the Antarctic Oscillation (AAO) on the positioning of the South Atlantic Convergence Zone (SACZ) over Brazil. This climate diversity in the NEB can be further observed by analyzing the surface response, which can be translated as a remarkable biodiversity distributed among four main biomes. The previously mentioned Caatinga, which is a seasonally dry tropical forest occupying approximately 53% of the NEB, and most of its semiarid region; the Cerrado, a savanna-type vegetation occupying roughly 30% of the NEB, mostly its western portions at the transition between the semiarid and more humid climates; the Amazon Forest occupying 7% of the NEB at the northwesternmost MA state; and the Atlantic Forest occupying most of the coastal region (10% of the NEB). The characterization that follows will focus on the Caatinga and Cerrado biomes, which are the most important and relevant for the NEB region. The Caatinga biome is characterized by a highly heterogeneous landscape composed by a mosaic of xerophyte deciduous trees and shrubs (CAMPOS et al., 2019; MUTTI et al., 2019). These plants have adapted structures and mechanisms to survive in arid and semiarid regions where water stress is recurrent (DOMBROSKI et al., 2011). Until the beginning of the 2000s decade, little was known of this ecosystem, which was believed to have a poor variety of species. Nowadays, the region is considered a hotspot of worldwide biodiversity (KOCH; ALMEIDA-CORTEZ; KLEINSCHMIT, 2017; LEAL et al., 2005). Rodrigues (2003), for example, found that a single dune system in the banks of the São Francisco river, representing 0.8% of the total biome area, contained 37% of all lizard and amphibians known to inhabit the biome at that time. This fact shows the tremendous unexplored potential of this ecosystem, and how its discovery is still recent. The Cerrado biome, on the other hand, is also characterized by a mosaic of herbaceous species, shrubs and trees, with the predominance of species typical of savanna vegetations 44 (BATALHA; MONTAVANI, 2000; CAMACHO; VASCONCELOS, 2015). As the Caatinga, the Cerrado is also characterized by a rich biodiversity and currently is the second largest biome in Latin America (BEUCHLE et al., 2015; RATTER; RIBEIRO; BRIDGEWATER, 1997). However, both ecosystems have historically been through numerous transformations, whether by the suppression of native vegetation due to the expansion of agricultural lands (BARBOSA et al., 2019; LEAL et al., 2005; TEIXEIRA, 2010; TOMASELLA et al., 2018) or the intensification of desertification processes caused by the combination of human activities and global climate change (CUNHA et al., 2015; MARIANO et al., 2018). Indeed, recent studies showed that the Cerrado is the Brazilian biome which most suffers from the suppression of native vegetation due to the expansion of arable lands (BEUCHLE et al., 2015), while the Caatinga is the biome most vulnerable to the effects of climate change (MARENGO; TORRES; ALVES, 2017; TOMASELLA et al., 2018). It is worth mentioning that the soil in most of the semiarid and surrounding regions of the NEB presents low to moderate aptitude, with the predominance of shallow, sandy textures with low water retention capacity (CAMPOS et al., 2019). Because of these characteristics, the zones of agriculture expansion in the Cerrado, for example, undergo specific soil preparation practices, with the intense application of fertilizers and liming. These practices represent a risk to the ecosystem and the stability of nearby water bodies (FISCHER et al., 2018). The low aptitude of the soil in the semiarid region demands not only specific soil management practices but also the establishment of irrigation systems. This is usually carried out inadequately, contributing to the increase in soil salinization and desertification, and leading to low productivity (BARROS et al., 2016; COSTA et al., 2016; DE MORAES et al., 2018). The complex diversity of the physical aspects of the NEB associated with its previously discussed sociodemographic particularities highlights the importance of carrying out research in this region of Brazil. Furthermore, the theoretical and historical background established in this chapter reveals the relevant role of droughts in the territorial dynamics of the region. However, while regional scale drought studies are undeniably important, the complexity of the NEB geographical aspects also demands specific, multiple-scale studies in order to better comprehend and detail the dynamics of this phenomenon in the region. 45 CHAPTER 2: METEOROLOGICAL DROUGHT IN THE SÃO FRANCISCO WATERSHED The SFW is the main hydrological system in the NEB, encompassing an area of approximately 640,000 km² that comprises part of the Minas Gerais (MG) and Goiás (GO) states, almost the entire Bahia (BA) state, besides the Pernambuco (PE), Alagoas (AL) and Sergipe (SE) states (Figure 2.1), with an average streamflow of 2100 m³/s (SIMPSON, 1998). It is the largest watershed entirely located in the national territory, presenting four different climate regimes, three main biomes (BEZERRA et al., 2019; SANTOS; DE MORAIS, 2013), eight large capacity water reservoirs besides numerous minor dams (MARQUES; GUNKEL; SOBRAL, 2019), and important socioeconomic contrasts (SIEGMUND-SCHULTZE et al., 2015). Its main water course has its source located at the Southeast region of Brazil, under the influence of a humid subtropical climate (ALVARES et al., 2014), transporting huge volumes of water throughout a considerable portion of the Brazilian Semiarid region (MANETA et al., 2009), and finally reaching its mouth at the coast of the NEB. This configuration makes the SFW a strategic system for the development of the NEB and the Brazilian Semiarid. Among its strategic roles we can highlight power generation and agricultural production. In an average year, for example, approximately 70% of total power provided to the NEB is generated in hydroelectric power stations located in the SFW such as the Sobradinho and Itaparica power plants (DE JONG et al., 2018). Regarding agriculture, the region is considered a strategic irrigation pole, which consumes roughly 76% of total water resources of the basin (CBHSF, 2016). In fact, irrigated agriculture is an activity in expansion in the region. In the last decade an increase of approximately 87% was observed in irrigated lands, especially in the lower- middle portion of the basin, which is under the influence of a semiarid climate (CBHSF, 2016). It is worth highlighting that the importance of the watershed is not limited to the regional scale: its total gross domestic product in 2015 was of approximately R$ 550 billion (U$ 167 billion at the currency rate of that time), which accounted for 9.1% of total national gross product that year (IPEA, 2015). However, the multiple water uses in the basin inevitably lead to recurrent conflicts for the use of its resources (DE JONG et al., 2018; HUNT; DE FREITAS; PEREIRA JR., 2016; MANETA et al., 2009; MARQUES; GUNKEL; SOBRAL, 2019). Furthermore, an increase in consumptive water demands is expected due to the expansion of irrigated lands. This expansion 46 is further motivated and accelerated by the São Francisco Transposition Project, which consists of the construction of artificial channels transporting water from the main river towards semiarid regions outside its watershed (BEZERRA et al., 2019; STOLF et al., 2012). Figure 2.1 – Climate types in the São Francisco watershed and mean climatological behavior for each subregion: Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF. Adapted from Alvares et al. (2014) and Maneta et al. (2009). In fact, this increase in water demands over the SFW has already been associated to a reduction in the frequency of occurrence of high streamflow rates in its water bodies (ANDRADE E SANTOS; POMPEU; KENJI, 2012). Furthermore, recent studies indicated a decrease in precipitation rates over part of the SFW, as well as increasing temperature trends in the region, which leads to more aridity (BEZERRA et al., 2019; DUBREUIL et al., 2019; SANTOS et al., 2018). Studies projecting future climate change scenarios also reported an 47 increase in mean temperature and a considerable reduction in rainfall rates over the SFW until the end of the century (MARENGO et al., 2012). In this context, it is evident that drought studies on the SFW are crucial to provide a better understanding of the relationships between the physical aspects of the basin and its impacts on the local population. Not only this, but the watershed presents itself as an interesting study subject due to its climate diversity and continental dimensions. Indeed, several studies have been developed in the SFW with specific objectives regarding drought assessment: water balance studies at the sub-basin scales (DE ASSIS et al., 2017; DOS SANTOS et al., 2018; LOPES et al., 2017), studies using standardized drought indices at specific sites in the basin (DOS SANTOS et al., 2019; SANTOS et al., 2017, 2019), studies on hydrometeorological aspects of drought (MARTINS et al., 2018; SUN et al., 2016); and studies describing the links between rainfall deficit and large-scale circulation patterns (PAREDES-TREJO et al., 2016). In this second chapter of the thesis, we aim to characterize the incidence of meteorological drought at the entire SFW extension while addressing specific methodological issues. This chapter comprises two complementary studies. In the first part (Chapter 2A): “Assessment of gridded CRU TS data for long-term climatic water balance monitoring over the São Francisco watershed, Brazil” we deal with the lack of consistent, gapless, long-term meteorological data with good spatial coverage in the basin. To that end, we provide a thorough validation of gridded CRU TS rainfall and PET data, exploring its spatial correlations with the available observational data. We also compare the performance of the proposed dataset in detecting and representing observed long-term trends (75 years: 1942-2016). In the following sub-chapter (Chapter 2B): “Drought characterization in the São Francisco watershed using the climatic water balance: methodological aspects and spatiotemporal dynamics”, the previously validated CRU TS data are used in a simplified water balance in order to estimate de deficit of evaporation (DE), which is used as a drought index. However, we first propose a methodological discussion regarding the popular use of standardized drought indices (such as the SPI and the SPEI) to assess drought in semiarid watersheds, showing the advantages of using non-standardized water balance methods such as the DE. The long-term spatial mean monthly behavior of the DE is then evaluated, as well as its anomalous patterns. Trends in the frequency of occurrence of water deficits and in the 48 expansion of drought-afflicted areas are also assessed, as well as the spatial patterns associated with the effects of large-scale circulation systems. In both studies we considered the four subregions of the SFW, which were previously defined by public agencies based on their physiographic characteristics in order to facilitate watershed management (SIEGMUND-SCHULTZE et al., 2015). They are: the Upper São Francisco (USF – 16% of the total basin extension), comprising the source of the river at the Serra da Canastra in Minas Gerais (MG) until the limits of the Brazilian Semiarid region; the Middle São Francisco (MSF – 63% of the total basin extension), encompassing the northern Minas Gerais (MG) state and the entire western portion of the Bahia (BA) state until the limits of the Pernambuco (PE) state; the Lower-middle São Francisco (LMSF – 17% of the total basin extension), comprising the driest portions of the basin until the coast; and the Lower São Francisco (LSF – 4% of the total basin extension) which comprises the coastal portion of the SFW and its mouth between the Alagoas (AL) and Sergipe (SE) states (Figure 2.1). It is worth mentioning that this subdivision was updated in 2016 by the Watershed Committee in such a way that the USF now extends until the limits of the Minas Gerais (MG) state, with the consequent reduction in the area of the MSF (CBHSF, 2016). However, since the analysis in this chapter is based on time series limited at 2016, we opted to keep the original subdivision in order to facilitate discussions and comparison with other studies in the literature. 49 Chapter 2A: Assessment of gridded CRU TS data for long-term climatic water balance monitoring over the São Francisco watershed, Brazil Article published in the Atmosphere journal (BRAZILIAN QUALIS B1; SCIMAGO JR Q3; JCR: 2.046) MUTTI, P. R.; DUBREUIL, V.; BEZERRA, B. G.; ARVOR, D.; DE OLIVEIRA, C. P.; SANTOS E SILVA, C. M. Assessment of Gridded CRU TS Data for Long-Term Climatic Water Balance Monitoring Over the São Francisco Watershed, Brazil. Atmosphere, v.11, n. 1207, p. 1-25, 2020. Doi: https://doi.org/10.3390/atmos11111207. Abstract: Understanding the long-term behavior of rainfall and potential evapotranspiration (PET) over watersheds is crucial for the monitoring of hydrometeorological processes and climate change at the regional scale. The São Francisco watershed (SFW) in Brazil is an important hydrological system that transports water from humid regions throughout the Brazilian semiarid region. However, long-term, gapless meteorological data with good spatial coverage in the region are not available. Thus, gridded datasets, such as the Climate Research Unit TimeSeries (CRU TS), can be used as alternative sources of information, if carefully validated beforehand. The objective of this study was to assess CRU TS (v4.02) rainfall and PET data over the SFW, and to evaluate their long-term (1942–2016) climatological aspects. Point-based measurements retrieved from rain gauges and meteorological stations of national agencies were used for validation. Overall, rainfall and PET gridded data correlated well with point-based observations (r = 0.87 and r = 0.89), with a poorer performance in the lower (semiarid) portion of the SFW (r ranging from 0.50 to 0.70 in individual stations). Increasing PET trends throughout the entire SFW and decreasing rainfall trends in areas surrounding the semiarid SFW were detected in both gridded (smoother slopes) and observational (steeper slopes) datasets. This study provides users with prior information on the accuracy of long-term CRU TS rainfall and PET estimates over the SFW. Keywords: potential evapotranspiration; precipitation; Brazilian Semiarid region; climate change; climate trends 2.1. Introduction Precipitation and PET are among the most important meteorological variables related to the water and energy cycles that regulate climate worldwide. In a global climate change context, evidence indicates that long-term changes in these variables are leading to important impacts 50 in agriculture, water resources management, and ecosystems dynamics in general (ARORA, 2019; FOLBERTH et al., 2020; NELSON et al., 2014; ROSENZWEIG et al., 2014). Therefore, their consistent and reliable monitoring is crucial to understand past, present, and future climate behavior. This is particularly important in regions that are more vulnerable to climate change, such as arid and semiarid regions, which are sensitive to small shifts in precipitation and PET patterns (HUANG et al., 2016). Given the high spatial variability of rainfall, long-term point-based measurements should be carried out with a good spatial coverage in order to account for as many physical factors as possible, such as elevation and the influence of nearby geographical elements (HENN et al., 2018; MICHOT et al., 2018; SHI; LI; WEI, 2017). This is usually less important for PET, since it is generally more spatially homogeneous, but having consistent, long-term, gapless PET data series is still a challenge (ALEMAYEHU; VAN GRIENSVEN; BAUWENS, 2016). Actually, dense meteorological networks with high-quality historical series of data are restricted to only a few regions of the globe (HARRIS et al., 2014). In arid and semiarid regions, the existence of extensive natural areas and the harsh climate conditions impose additional difficulties to the installation and management of measuring networks (CHEN et al., 2008; WAGNER et al., 2012). In this context, several global and regional gridded datasets were developed either by interpolation of existent data and/or the assimilation of satellite data, such as the CRU TS by the Climate Research Unit at the University of East Anglia (HARRIS et al., 2014, 2020), the reanalysis product of the Global Precipitation Climatology Centre (GPCC) operated by the German meteorological service under the auspices of the WMO (SHAMM et al., 2014), the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) by the Climate Hazards Center of the University of Santa Barbara (FUNK et al., 2014), or even satellite- derived databases such as the Tropical Rainfall Measuring Mission (TRMM) products (HUFFMAN; BOLVIN, 2015). However, these datasets still carry uncertainties and should undergo careful validation before being considered for use in climate and environmental studies. In fact, different precipitation and PET gridded datasets have been thoroughly validated in many regions of the globe, presenting a good overall performance in hydrometeorological studies in: China (CRU TS: SHI; LI; WEI, 2017; GPCC precipitation: MA et al., 2009), the Philippines (CRU TS: SALVACION et al., 2018), the European Alps (ERA-Interim: 51 PRÖMMEL et al., 2010), the Caribbean (CRU TS and GPCC: JONES et al., 2016), Central Asia (CRU TS and GPCC: MALSY et al., 2015), and even at the global scale (CRU TS:THORNE et al., 2016). In Brazil, validation results for CHIRPS rainfall data indicated that the product performed worse in semiarid areas of the Northeast region (PAREDES-TREJO; BARBOSA; KUMAR, 2017). TRMM daily data was validated for the Amazon region, with the satellite product performing worse in the dry season (MICHOT et al., 2018), and for the NEB, with overall good results, except in the coastal zones (RODRIGUES et al., 2020). Xavier, King and Scanlon (2016) also developed and validated gridded PET and precipitation data for Brazil. They reported that performance was better in recent years due to the expansion of the measuring network, and also that the gridded product performed worse over basins with lower station density such as the Amazon and the São Francisco. Furthermore, the studies by Paredes-Trejo, Barbosa and Kumar (2017) and Xavier, King and Scanlon (2016) expose one of the main issues regarding point-based surface monitoring of meteorological variables in the semiarid region of Brazil, which is the lack of long-term, gapless, and evenly distributed data, as also reported by Mutti et al. (2020) and Rodrigues et al. (2020). A limitation of CHIRPS, TRMM and Xavier’s databases is that they provide time series starting from the 1980s, which is long enough for climatological studies, but may not be sufficient for robust trend analysis or extreme events assessment (MUDELSEE, 2019). GPCC products provide long-term time series (more than 100 years) but, on the other hand, they consist of reanalysis data rather than interpolation of point-based measurements. In this context, CRU TS products stand out since they provide long-term interpolated data. In Brazil, the SFW is the largest hydrological system located entirely in the national territory, comprising an area of approximately 640,000 km2, with an average discharge of 2850 m3 s−1 (BEZERRA et al., 2019; KOCH et al., 2018; MENDES et al., 2015). Most importantly, it transports water from the Southeast region of the country, characterized by a humid subtropical climate (~1300 to 2000 mm year−1), throughout part of the Northeast region of the country, characterized by a tropical semiarid climate (~500 mm year−1, occurring mostly from February to April). It is, therefore, a major provider of water resources for the semiarid region of Brazil, particularly with the eminent conclusion of the Transposition Project that derives water from the São Francisco River towards other semiarid areas outside the watershed (BEZERRA et al., 2019; MARQUES; GUNKEL; SOBRAL, 2019; STOLF et al., 2012). 52 Changes in rainfall patterns and in the atmospheric demand for water play a key role in defining water management policies in the SFW (CABRAL JÚNIOR et al., 2019; KOCH et al., 2018; MARQUES; GUNKEL; SOBRAL, 2019; MENDES et al., 2015; STOLF et al., 2012) and in predicting potential climate change in the region (AVILA-DIAZ et al., 2020; MARENGO et al., 2012). Thus, long-term, evenly distributed rainfall and PET data are needed in order to ensure the development of high-quality hydrometeorological studies over this region. Studies using point-based observations or satellite data showed mostly nonsignificant decreasing trends in rainfall indices in the SFW (AVILA-DIAZ et al., 2020; BEZERRA et al., 2019; SANTOS et al., 2018), although increasing aridity was observed in the semiarid portion of the basin when considering climate types derived from precipitation and temperature (which can be related to PET) surface data (DUBREUIL et al., 2019). It is important to note that rainfall time series measured by rain gauges in the semiarid portion of the SFW are highly heterogeneous, as previously reported (BEZERRA et al., 2019; COSTA et al., 2020), and yet large areas are not covered by the measuring network. Furthermore, future climate projections indicate increasing temperatures and a considerable reduction in rainfall rates over the SFW until the end of the century (MARENGO et al., 2012). Given the importance of the SFW, and also its climate vulnerability, the use of alternative tools such as gridded datasets should be encouraged in order to compensate the limitations of the existing measuring network and data. However, the accuracy and limitations of such tools in the SFW, and in any other particular region, must be thoroughly evaluated in order to guarantee its proper use and assessment. Thus, the objective of this study was to validate CRU TS precipitation and PET data over the São Francisco river basin, and to assess its main climatological features in comparison with point-based observed data. To this end, statistical assessment, trend analysis, and change-point detection was carried out considering three 25-year periods: 1942–1966; 1967–1991; 1992–2016, and the entire 1942–2016 period (75 years). Furthermore, the validation was conducted in four subregions of the SFW defined according to their geographical characteristics: upper São Francisco (USF), middle São Francisco (MSF), lower-middle São Francisco (LMSF), and lower São Francisco (LSF). 2.2. Material and methods Gridded monthly precipitation and PET data from the CRU TS version 4.02 were compared with observational point-based data measured in rain gauges and meteorological stations located in the SFW. The source, characteristics and quality control of the data will be 53 described in the following sections. Data were compared over the period from 1942 to 2016 (75 years), which was defined based on observed data availability. A longer period could have been considered, but with less spatial representativeness and less data for validation. 2.2.1. CRU TS v4.02 data The CRU TS v4.02 dataset comprises the fourth version of the gridded product developed by the Climate Research Unit of the University of East Anglia. It includes a number of variables (precipitation, mean temperature, vapor pressure, daily temperature range, PET, among others) disposed in a global (excluding Antarctica) 0.5° × 0.5° grid (~56 km) over the period from 1901 to 2017, obtained through the interpolation of monthly data retrieved from the archives of the WMO (HARRIS et al., 2014, 2020). The CRU TS grid for the SFW is shown in Figure 2.2. It is worth mentioning that the CRU TS dataset provides PET data calculated through the Penman–Monteith equation, which takes into consideration elements of the energy balance. Because of this, the proper evaluation of the product would be difficult to carry out, due to the lack of the necessary measured data (vapor pressure, cloud cover and static wind field) in the SFW. Thus, we decided to estimate Thornthwaite’s PET from mean monthly temperature data. It should be noted that we do not intend to discuss the pros and cons of Thornthwaite’s method to estimate PET, as it has been thoroughly discussed in several studies (BEGUERÍA et al., 2014; LOFGREN; HUNTER; WILBARGER, 2011; ZHANG; HE, 2016). Rather, we decided to use it as representation of potential atmospheric demand for water (in opposition to precipitated water) in broader climate terms. Indeed, these studies showed that the Penman– Monteith method undoubtedly retrieves better PET estimates, but temperature-derived PET (such as Thornthwaite’s method) provides reliable estimates for long-term annual cycle studies or drought assessment studies based on the climatic water balance. 54 Figure 2.2 – Location of the measuring stations used in the study and the CRU TS dataset grid (0.5° × 0.5°. USF: Upper São Francisco; MSF: Middle São Francisco; LMSF: Lower- middle São Francisco; LSF: Lower São Francisco. 2.2.2. Point-based measurement data Monthly rainfall and mean air temperature data were obtained from rain gauges monitored by the ANA, meteorological stations of the INMET, and surface stations compiled in the Global Historical Climatology Network (GHCN) database. All stations and gauges located within the boundaries of the SFW plus an additional buffer of up to 60 km (approximately equivalent to one CRU TS grid) were selected. A total of 171 point-based series were selected for the evaluation of rainfall data, and 57 series were selected for the evaluation of PET (Figure 2.2). 2.2.3. Thornthwaite’s potential evapotranspiration PET (mm) was derived from temperature values (°C) through Thornthwaite’s classical method. Monthly PET was estimated based on temperature as follows (THORNTHWAITE, 1948; WILM et al., 1944): 55 16(10 𝑇𝑚⁄𝐼) 𝑎, 0 ≤ 𝑇𝑚 ≤ 26.5 °𝐶𝐸𝑇𝑃 = { (1) −415.85 + 32.24𝑇𝑚 − 0.43𝑇 2 𝑚, 𝑇𝑚 ≥ 26.5 °𝐶 where ETp is the noncorrected standard PET, Tm is mean monthly air temperature (ºC), and the parameters I and a are calculated as: 12 𝐼 = ∑(𝑇 1.514𝑚,𝑖⁄5) (2) 𝑖=1 and, 𝑎 = 6.75 × 10−7𝐼3 − 7.71 × 10−5𝐼2 + 1.79 × 10−2𝐼 + 0.49 (3) Since ETp represents PET that would occur in the thermal conditions of a standard month with 30 days and a photoperiod of 12 h, it needs to be corrected for each month, as follows: 𝑃𝐸𝑇 = 𝐸𝑇𝑃[(𝑁𝐷⁄30)(𝑁⁄12)] (4) where ND is the number of days in the corresponding month, and N is the mean photoperiod in that month. These correction factors can be found in tables presented in several studies, such as the classical Thornthwaite’s work (THORNTHWAITE, 1948). 2.2.4. Observed data quality control Observed data went through careful quality control before being used as reference for the evaluation of the gridded dataset. Each time series was firstly screened by removing all monthly data outside the ±3.5 standard deviation threshold and all duplicated values in adjacent months (CHAPMAN, 1999; PAREDES-TREJO; BARBOSA; KUMAR, 2017). Then, the homogeneity of the time series was assessed by means of the Standard Normal Homogeneity Test (SNHT) developed by Alexandersson (1986). The SNHT was used in order to identify potential rupture points in the series, which could indicate, for example, that the station has been displaced. For each observation i two averages are calculated: one for the k months before the observation i, µ1k; one for the k months after the observation i, µ2k. The statistic of the SNHT (Ti) is then calculated as follows: 𝑇𝑖 = [(𝜇1𝑘,𝑖 − 𝜇𝑖) 2 + (𝜇2𝑘,𝑖 − 𝜇𝑖) 2] 𝑘⁄𝜎2𝑖 (5) 56 where µi is the mean between µ1k,i and µ2k,i, and σi is the standard deviation in the period of k months before and after the observation i. If the peak values of Ti for a given observation are higher than a threshold, these points are marked as potential rupture points in the time series. Although the threshold value of Ti should depend on the number of observations in each evaluated time series, studies show that it varies from 9 to 10.6 for series with 36 to 900 observations with a 95% confidence level (KHALIQ; OUARDA, 2007). Since monthly data in the present study comprises a period of 75 years (maximum of 900 observations), we adopted a Ti threshold value of 10 for each time series. The value of k was set to 60 (5 years) for temperature time series and 120 (10 years) for rainfall time series, since the latter are more heterogeneous and with higher variability. Finally, each time series was visually analyzed with the aid of the Ti statistic, removing nonhomogeneous portions of the series. Overall, we retained the most recent portion of the series, i.e., values measured after the detection of a rupture point, as it is conventionally carried out in similar studies (DOMONKOS, 2013). Figure 2.3 shows an example of the analysis using the SNHT for one rainfall time series. Percentage of gaps was not considered as an exclusion criterion, and therefore, no gap filling procedure was carried out. The objective was to compare CRU TS data with as many reliable and consistent available point-based data as possible. The description and characterization of all observed data used in the study are fully presented in the Supplementary Materials (Table 2.S1, Table 2.S2 and Table 2.S3; Figure 2.S1, Figure 2.S2), including source, latitude, longitude, elevation, total number of observations, percentage of gaps, and data removed in each step of the quality control. 57 Figure 2.3 – Example of the visual analysis of time series with the aid of the standard normal homogeneity test (SNHT) statistic (Ti) to find potential heterogeneities in data. 2.2.5. Statistical assessment 2.2.5.1. Spatial and temporal units of analysis The entire study period comprised 75 years from 1942 to 2016, as previously mentioned. However, CRU TS data were also evaluated in three distinct 25-year periods: 1942–1966, 1967–1991, and 1992–2016. Such a procedure was carried out in order to measure the accuracy of the gridded dataset throughout the whole time series, verifying if data were better or worse represented in a particular period. Intervals of 25 years were chosen because they represent the closest interval to climatological normals, into which the full 75-year time series could be divided. The evaluation was carried out for the entire SFW considering the four previously mentioned subregions: USF, MSF, LMSF, and LSF (Figure 2.2). By doing so, it was possible to identify if there is a particular region under specific climate conditions where CRU TS data is more or less accurate. For that end, Table 2.1 shows the main climate features of each subregion and also the number and density of stations selected for evaluation. 58 Table 2.1 – Rainfall and potential evapotranspiration (PET) climatological characteristics and number of measuring stations selected in the subregions of the São Francisco Watershed. USF: Upper São Francisco; MSF: Middle São Francisco; LMSF: Lower-middle São Francisco; LSF: Lower São Francisco. Mean Annual Mean No. of Rainfall No. of Temperature Köppen’s Rainfall (Wet Subregion Annual Stations/Density × 10−3 (PET) Stations/Density × Climate Type Months) PET (mm) (Station km−2) 10−3 (Station km−2) (mm) USF Cwa/b–humid 1400 (Oct– 1100 48/0.48 13/0.13 subtropical with Mar) dry winter MSF Aw–tropical 1100 (Nov– 1300 82/0.20 28/0.07 with dry winter Mar) LMSF Bsh–dry 600 (Jan–Apr) 1500 28/0.25 10/0.09 semiarid LSF As–tropical with 900 (Mar–Jul) 1400 13/0.51 6/0.24 dry summer Total - - - 171/0.27 57/0.09 Previous studies showed that station’s elevation can be an important source of uncertainty in CRU TS rainfall data (SHI; LI; WEI, 2017), particularly above 3800 m. Since maximum altitude in the SFW is of approximately 1500 m (Figure 2.2) and topography is generally nonheterogeneous, no specific evaluation regarding altitude was carried out. 2.2.5.2. Accuracy measurement The Pearson’s correlation coefficient (r) was calculated between each CRU TS grid and each station within that grid, if any. In some cases, multiple stations were located within the same grid (as seen in Figure 2.2). In these situations, r was calculated for each station individually in order to identify how well CRU TS grids capture small scale nuances and variability in observed data due to spatial heterogeneity. The correlation between all available data was also computed for each subregion and for each period. Furthermore, the monthly root mean square error (RMSE) and percent bias (PBIAS) were calculated considering the mean CRU TS and point-based rainfall and PET time series in each subregion (mean between all grids or stations within each subregion). The monthly reliability of the CRU TS datasets in each 25-year period was also evaluated based on the relative RMSE value. When RMSE is less than 50% of the mean observed value in a given month, the CRU TS estimates are considered reliable in relative terms (ADEYEWA; NAKAMURA, 2003; FRANCHITO et al., 2009). The RMSE and the PBIAS were computed as follows (WILKS, 2006): 59 𝑛 1⁄2 1 𝑅𝑀𝑆𝐸 = ( ∑(𝐸𝐶𝑅𝑈,𝑖 − 𝑂𝑖) 2) (6) 𝑛 𝑖=1 where ECRU refers to CRU TS data, O refers to point-based observed values, n is the total number of compared pairs, and: Σ𝑛𝑖=1(𝐸𝐶𝑅𝑈,𝑖 − 𝑂𝑖) 𝑃𝐵𝐼𝐴𝑆 = 100 𝑛 (7) Σ𝑖=1𝑂𝑖 2.2.5.3. Trend test and change-point detection Trends in CRU TS rainfall and PET data were compared with trends in point-based data, considering the mean time series in each subregion. Series were pre-whitened (12-month moving average) in order to remove autocorrelation, and then linear trends were estimated through the least squares method (SIMMONS et al., 2004). The significance of the trend slopes was calculated through Student’s t-test (p < 0.01). Furthermore, the Pettitt change point detection test (PETTITT, 1979) was used to identify the exact year where a significant shift in the central tendency in the time series occurred, which was also compared between gridded and point-based measurements. When used in climate data, the Pettitt’s test indicates the approximate period when a significant change is observed in the mean behavior of a given meteorological variable. We also compared the slope of the trend in each grid of the CRU TS databases with point-based trends in observational data. In this case, we selected only one station per grid point with a significant trend (p < 0.01), and only stations with consistent data throughout the entire studied period. 2.3. Results 2.3.1. Overall spatial and temporal performance Figure 2.4 shows an overall strong correlation between CRU rainfall data and surface observations during the entire 75 years of the time series (r = 0.87). Individual correlations were mostly higher than 0.80 throughout the USF, MSF, and part of the LMSF. Only the transition zone between the LMSF (semiarid region) and the LSF (tropical coastal region) presented a few stations with which CRU data were weakly or moderately correlated (between 0.50 and 0.70). 60 Figure 2.4 – Point-based correlation between observational precipitation data and CRU TS gridded data considering the 75 years period (1942–2016), and three 25 years periods (1942–1966; 1967–1991; 1992–2016). Overall correlation (r) and total number of stations (n) in each period are shown at the top of each map. USF—upper; MSF—middle; LMSF— lower-middle; LSF—lower São Francisco watershed. Regarding the three 25-year periods, a noticeable increase in available stations can be observed from 1967 onwards. Overall, rainfall was best represented by the CRU TS dataset during the period from 1967–1991, when correlations below 0.80 were observed with only 12 out of the 168 stations. On the other hand, the most recent period (1992–2016) was when CRU 61 TS data presented the weakest correlation (between 0.50 and 0.80) with almost all stations located in the LMSF and the LSF, although overall correlation was high (r = 0.87). Throughout the three periods, CRU TS data was mostly strongly correlated with observed data in the USF and MSF. These results are further detailed in Figure 2.5, which shows the scatter plot of CRU TS rainfall data and observed data in each subregion of the SFW and for each 25-year period. It can be noted that in all plots CRU data seems to frequently underestimate higher monthly rainfall values. Another overall observation is that increasing the sample size did not necessarily lead to a better fit of the linear relationship between datasets. Figure 2.5 – Scatter plot of CRU TS rainfall data versus observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. n indicates sample size and r indicates the correlation coefficient. The blue line indicates the 1:1 rapport and the red line indicates the linear relationship between data. As previously hinted in Figure 2.4, the highest slope coefficients of the scatter plot fit lines were found for the period from 1967–1991 (0.85; 0.90; 0.78; 0.91 for the USF, MSF, LMSF, and LSF, respectively). This means than in this period, the CRU TS rainfall estimates 62 were less biased, and more correlated to observed data as evidenced by the correlation coefficients. The lowest slope coefficients (more frequent underestimations), however, were found in the most recent period, 1992–2016 (0.83; 0.81; 0.60; 0.72 for the USF, MSF, LMSF, and LSF, respectively), despite a larger available sample during these years. It is also worth mentioning that the differences between periods for the USF and MSF are much less prominent than for the LMSF and LSF. Regarding PET values calculated using CRU TS temperature dataset, results appear to be more consistent both spatially and temporally, despite a few isolated poorly correlated stations in 1942–1966 and 1967–1991. Figure 2.6 shows that both in the 75-year period and in the three 25-year subperiods, most of the CRU TS data were mostly strongly correlated with available stations (r > 0.80). Furthermore, good relationships between datasets are found throughout the entire SFW, with no particular subregion presenting a weaker or stronger correlation pattern. This result was expected, since temperature, and therefore Thornthwaite’s PET, is less variable in the region. The scatter plot between CRU TS temperature-derived PET and observed data (Figure 2.7) shows that the gridded dataset tends to more frequently underestimate high PET values, while low PET values are slightly overestimated. This pattern is observed across all analyzed 25-year periods and all subregions. It is also evident that low PET values in the LMSF (semiarid region) are more overestimated than in other regions. For example, the slope coefficient in the LMSF during 1992–2016 was 0.66, while in the USF, MSF, and LSF it was 0.74, 0.68, and 0.72, respectively. 63 Figure 2.6 – Point-based correlation between observational Thornthwaite’s potential evapotranspiration data and CRU TS gridded data considering the 75 years period (1942– 2016), and three 25 years periods (1942–1966; 1967–1991; 1992–2016). Overall correlation (r) and total number of stations (n) in each period are shown at the top of each map. USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. Different to what was observed with rainfall data, correlation is higher during the first 25-year period in the USF and LSF (r = 0.90) with the regression line better fitting to the 1:1 ratio. During the two following periods, the slope of the scatter plot fit line does not seem to 64 have changed much in any subregion, even with the increase in sample size. Nevertheless, the correlation coefficient in the last period (1992–2016) is similar or higher than in the first period (1942–1966) in all subregions, except for the LMSF. Figure 2.7 – Scatter plot of CRU TS Thornthwaite’s potential evapotranspiration calculated using temperature data versus observed data in the: USF—upper; MSF—middle; LMSF— lower-middle; LSF—lower São Francisco watershed. n indicates sample size and r indicates the correlation coefficient. The blue line indicates the 1:1 rapport and the red line indicates the linear relationship between data. 2.3.2. Seasonal performance The seasonal performance of the CRU TS datasets was assessed based on the analysis of monthly RMSE and PBIAS. Regarding rainfall estimates, Figure 2.8 shows that the RMSE generally improves in the most recent periods of the 75-year time series in all subregions. Since the RMSE is scale-dependent, higher values are found during the wet months if compared to dry months. In the LMSF and LSF, RMSE was lower than 68 mm in all months during the period from 1967 until 2016. During dry months, RMSE varied from 4 mm (August 1942–1966 in the MSF) to 34 mm (December 1942–1966 in the LSF). 65 Figure 2.8 – Monthly root mean square error (RMSE) and mean percent bias (PBIAS) of CRU TS rainfall and Thornthwaite’s potential evapotranspiration data in relation to observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. Indeed, by analyzing the PBIAS of CRU TS rainfall data in relation to the observational dataset, one can notice high biases in the dry season. In the USF, mean rainfall is overestimated up to 97.6% (July 1942–1966), 75.9% (August 1942–1966), and 63.7% (July 1992–2016). A similar result is found in the MSF; despite a remarkable improvement in the most recent years, overestimations can reach up to 134.7% in July (1992–2016). In the semiarid region (LMSF), systematic underestimations of up to 47.5% (August 1992–2016) were found despite the low RMSE. Regarding PET estimates, the behavior of monthly RMSE and PBIAS was more consistent and less important in all subregions and during all periods. RMSE ranged from 6.4 mm month−1 in the LSF (August 1942–1966) to 26.8 mm month−1 also in the LSF (October 1992–2016). PBIAS, on the other hand was higher in the LMSF, with overestimations of up to 11.4% in July (1942–1966). Finally, Table 2.A1 (Appendix A, presented at the end of the manuscript before the references) shows the monthly observed and CRU TS rainfall mean (±standard deviation) for each 25-years period and each subregion. Furthermore, it indicates whether CRU TS estimates are reliable or not in each period based on the relative proportion of the RMSE in relation to the observed mean values. Overall, CRU TS rainfall estimates proved to be reliable in the wet season months of the USF and the MSF (approximately from October to March) in all period, except for 1942–1966 in the MSF. For the LMSF and the LSF, the RMSE analysis indicated 66 that the monthly rainfall estimates of the CRU TS dataset are unreliable since the RMSE was higher than 50% of the observed mean values. This result, however, should be taken with care since the relative RMSE value in dry periods are expected to be high, although absolute differences may not be so remarkable. For example, in June 1942–1966, CRU TS rainfall was of 3 ± 7 mm in the MSF, while observed rainfall was of 2 ± 6 mm, which is obviously a reliable estimate despite the relative RMSE lower than 50%. Similarly, Table 2.A2 (Appendix A, presented at the end of the manuscript before the references) shows the monthly observed and CRU TS Thornthwaite’s PET mean (± standard deviation) for each 25-year period and each subregion. In this case, the analysis of the RMSE indicated that the CRU TS product is reliable in all months, during the entire 75-years period and in all subregions. 2.3.3. Trends and change-point comparisons Firstly, we compared the trends and change-points detected in the smoothed monthly mean time series of rainfall and PET derived from the CRU TS dataset and observational data (Figure 2.9). For the USF, the same change-points were identified in both datasets: December 2012 and November 1993 for rainfall and PET, respectively. Furthermore, significant decreasing trends were found (Table 2.2) for both the entire rainfall time series (−1 mm decade−1) and from December 2012 to 2016 (−14 and −17 mm decade−1 for observed and CRU TS data, respectively). Regarding PET data, significant increasing trends were found in both datasets and in all portions of the time series (Figure 2.9 and Table 2.2). Indeed, the change rate in PET increased from 1 mm decade−1 until 1993 to 3 mm decade−1 in the last 25 years of the studied period. The mean PET in the two sections of the time series increased from 1092 mm (observational) and 1036 mm (CRU TS) to 1165 mm (observational) and 1122 mm (CRU TS). 67 Figure 2.9 – Smoothed mean rainfall and potential evapotranspiration derived from point- based observations and the CRU TS dataset in the: USF—upper; MSF—middle; LMSF— lower-middle; LSF—lower São Francisco watershed. The linear trendline and the detected change-points are also indicated. In the MSF, an important change in behavior was identified by Pettitt’s test in the observed time series in March 1964 (Figure 2.9 and Table 2.2). Indeed, rainfall values increased after this year but subsequently presented a significant decreasing trend of −2 mm decade−1. For CRU TS rainfall data, a significant change-point was detected in October 1986, while the decreasing trend was significant only when considering the entire time series (−1 mm decade−1). Regarding PET in the MSF, similar change-points were detected in the two datasets: July 1987 (observations) and July 1986 (CRU TS). Furthermore, significant increasing trends were found in both datasets, although in the observational data the overall change rate was of 8 mm decade−1 compared to only 2 mm decade−1 for the CRU TS time series. 68 Table 2.2 – Summary of the trends and change-point detection analysis for observed and CRU TS rainfall and Thornthwaite’s potential evapotranspiration data in the: USF—upper; MSF— middle; LMSF—lower-middle; LSF—lower São Francisco watershed. Statistically significant trend slopes and change-points (α = 1%) are highlighted in italic. Subregion Change Point Period Annual Mean (mm) Slope (mm/Decade) USF Obs rainfall Dec/2012 Jan/1942–Dec/2016 1339 −1 Jan/1942–Dec/2012 1355 0 Dec/2012–Dec/2016 1056 −14 CRU TS rainfall Dec/2012 Jan/1942–Dec/2016 1383 −1 Jan/1942–Dec/2012 1397 0 Dec/2012–Dec/2016 1120 −17 Obs PET Nov/1993 Jan/1942–Dec/2016 1123 2 Jan/1942–Nov/1993 1092 1 Nov/1993–Dec/2016 1165 3 CRU TS PET Nov/1993 Jan/1942–Dec/2016 1063 2 Jan/1942–Nov/1993 1036 1 Nov/1993–Dec/2016 1122 3 MSF Obs rainfall Mar/1964 Jan/1942–Dec/2016 1006 1 Jan/1942–Mar/1964 913 11 Mar/1964–Dec/2016 1035 −2 CRU TS rainfall Oct/1986 Jan/1942–Dec/2016 1103 −1 Jan/1942–Oct/1986 1138 0 Oct/1986–Dec/2016 1059 0 Obs PET Jul/1987 Jan/1942–Dec/2016 1275 6 Jan/1942–Jul/1987 1171 8 Jul/1987–Dec/2016 1406 2 CRU TS PET Jul/1986 Jan/1942–Dec/2016 1258 2 Jan/1942–Jul/1986 1219 1 Jul/1986–Dec/2016 1312 3 LMSF Obs rainfall Jan/1964 Jan/1942–Dec/2016 557 0 Jan/1942–Jan/1964 520 −4 Jan/1964–Dec/2016 572 −3 CRU TS rainfall Dec/1963 Jan/1942–Dec/2016 528 0 Jan/1942–Dec/1963 475 −1 Dec/1963–Dec/2016 552 −3 Obs PET Jan/1993 Jan/1942–Dec/2016 1484 2 Jan/1942–Jan/1993 1445 0 Jan/1993–Dec/2016 1535 −1 CRU TS PET May/1987 Jan/1942–Dec/2016 1467 1 Jan/1942–May/1987 1444 0 May/1987–Dec/2016 1500 1 LSF Obs rainfall Jul/1979 Jan/1942–Dec/2016 785 −2 69 Jan/1942–Jul/1979 866 5 Jul/1979–Dec/2016 720 1 CRU TS rainfall Jul/1990 Jan/1942–Dec/2016 878 −1 Jan/1942–Jul/1990 928 2 Jul/1990–Dec/2016 803 1 Obs PET May/1995 Jan/1942–Dec/2016 1269 6 Jan/1942–May/1995 1162 −1 May/1995–Dec/2016 1427 11 CRU TS PET Jun/1987 Jan/1942–Dec/2016 1355 1 Jan/1942–Jun/1987 1333 0 Jun/1987–Dec/2016 1387 1 Still regarding Figure 2.9 and Table 2.2, also similar change-points were found in the LMSF rainfall time series: January 1964 and December 1963, for the observations and CRU TS estimates, respectively. Both portions of the time series presented significant decreasing trends: −4 and −3 mm decade−1 for the observational data and −1 and −2 mm decade−1 for CRU TS data. Therefore, rainfall derived from the CRU TS dataset appears to decrease slightly smoother than observed data. Regarding PET data in the LMSF, nonsignificant change-points were found, while a significant overall increasing trend was found in both datasets. Finally, significant change-points at different periods were detected in the LSF (Figure 2.9 and Table 2.2): July 1979 and May 1995, for observed and CRU TS rainfall data, respectively. In these series, significant decreasing trends of −2 and −1 mm decade−1 were found with a mean annual rainfall of 785 mm and 878 mm for observational and CRU TS data, respectively. At the same time, a remarkable significant increasing trend of 6 mm decade−1 was found for observed PET data. This increase reached up to 11 mm decade−1 in the period from May 1995 until the end of the series. CRU TS derived PET also presented significant increasing trends, but with a much less important change rate (1 mm decade−1). The spatial distribution of the significant trend slopes (p < 0.01) of CRU TS data is shown in Figure 2.10. For rainfall, negative trends are found in most part of the MSF, LSF and the northern portion of the USF. Additionally, positive trends are found in the southern portion of the USF. Overall, these results are similar to what was found with observed data, which also indicated negative trends in the same regions of the SFW, although with higher change rates. Furthermore, CRU TS data featured significant rainfall trends in large areas of the basin, while point-based significant trends are more heterogeneous and sparser. Regarding Thonrthwaite’s PET derived from the CRU TS dataset, significant trends were found in the entire SFW, indicating an increase in the atmospheric demand for water. 70 Indeed, observational data was also mostly significant, mainly in the LMSF and LSF while some stations in the MSF and USF presented non-significant trends. One station at the LMSF, near the Sobradinho reservoir, presented negative PET trend, which was not captured by the CRU TS dataset. The magnitude of the change rate is also stronger in observed data than in CRU TS data, particularly in the LMSF and the LSF. Figure 2.10 – Spatial distribution of the slope of the smoothed monthly rainfall and Thornthwaite’s potential evapotranspiration (PET) time series derived from the CRU TS dataset (only the grids with significant trends are plotted) and from selected stations. USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. 2.4. Discussion In the first part of the study we assessed the overall spatial and temporal performance of the CRU TS dataset in representing monthly rainfall and PET (estimated from temperature data) over the SFW. 71 CRU TS rainfall data presented weaker correlation with observed data in the LMSF and the LSF, which is probably associated with the high spatial and temporal variability of precipitation in these regions. The LMSF is entirely located in the semiarid region of Brazil while the LSF comprises a transition zone between a semiarid environment and the coastal zone, with a more humid climate [30]. Thus, it is expected that interpolated datasets with relatively large spatial resolutions such as the CRU TS (~57 km) will perform worse in these regions. Previous studies have already reported a poorer performance of gridded datasets in regions of high rainfall variability such as semiarid zones in Pakistan (AHMED et al., 2019). PET data, on the other hand, presented more consistent results in spatial terms. Indeed, temperature and consequently Thornthwaite’s PET spatial and temporal patterns are less variable (MANETA et al., 2009a, 2009b). The scatter plot between data showed that rainfall CRU TS data generally underestimate higher rainfall and PET values, while low PET values are usually overestimated. This is probably associated with the smoothing of data which is inherent to interpolation techniques used to produce gridded datasets (HENN et al., 2018; PERSAUD et al., 2020). Results showed that the best correlations between data were found in the period from 1967–1991, while the most recent period (1992–2016) presented slightly worst correlations. There are two potential explanations for this fact. Firstly, the density of the contributing stations with at least 75% of observations per decade for the development of the CRU TS product was higher in the second half of the 20th century than in the first decades of the 21st century (HARRIS et al., 2014). Thus, it is expected that a higher density of stations provides better interpolated estimates when comparing with point-based observations. Another explanation is associated with the aforementioned characteristic of gridded datasets of usually performing worse over semiarid and drier regions. The period from 1992–2016 comprised the worst drought event (2012–2016) registered in the semiarid region of Brazil (DE MEDEIROS; DE OLIVEIRA; TORRES, 2020; MARENGO et al., 2017), which may have also contributed to the poorer performance of the algorithm in this 25-years period, especially in the LMSF and LSF. The seasonal performance of the CRU TS datasets was assessed based on the monthly RMSE and PBIAS. Since the RMSE is scale-dependent, it was proportional to the overall magnitude of the assessed variables. Thus, higher rainfall RMSE are found during the wet season, while for PET it is rather stable throughout the year. The PBIAS, on the other hand, is 72 much more sensitive to lower values (dry season) but indicates the direction of the monthly bias (underestimation or overestimation of observed values). Thus, high PBIAS values should be interpreted with caution for the dry season, while high RMSE values should be interpreted with caution for the wet season. Results showed that CRU TS data generally overestimates rainfall during the dry season in the USF and MSF, while during the wet season estimations are reliable (RMSE < 50% of observed mean value). For Thornthwaite’s PET, CRU TS slightly underestimates (maximum PBIAS of −10%) observations during all months in the USF, MSF, and LSF, although the estimates are reliable. Therefore, when using CRU TS data to monitor temperature and temperature-derived PET, one should note that actual atmospheric demand for water is slightly higher in these regions. In the LMSF, however, results are the opposite, with actual PET being slightly slower than CRU TS derived PET. Monthly mean values of rainfall and PET derived from the CRU TS were comparable to those observed in surface stations. In the USF, higher rainfall rates were observed in the austral autumn and summer (October to March), mainly due to the development of the SACZ in the central portion of South America, encompassing the western and southern SFW (BEZERRA et al., 2019; CAVALCANTI, 2012; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). In the MSF, a similar pattern was observed, but with smaller rainfall rates since the region is larger and comprises the transition zone between the humid subtropical climate and the semiarid. In the northern portion of the SFW, the LMSF is under the influence of a semiarid climate, where rainfall is modulated mainly by the position of the ITCZ from February to May (BEZERRA et al., 2019; HASTENRATH, 2012), producing rainfall extreme events in that region (RODRIGUES et al., 2020). The LSF, on the other hand, presents a rainfall regime typical of coastal zones in the NEB (Köppen’s As–tropical with dry summer), with important rain events occurring from March to July, mainly due to the ITCZ, sea breezes, and the propagation of EWD (BEZERRA et al., 2019; DE OLIVEIRA; LIMA; SANTOS E SILVA, 2013; GOMES et al., 2019). Regarding PET, seasonal variations are similar in all subregions of the SFW, although lower values are found in the higher latitudes of the basin (USF and MSF). For the assessment of long-term monitoring of the two components of the climatic water balance, trends and change-points of CRU TS data were compared to observed data. Regarding 73 regional mean time series, a general decreasing trend in rainfall and increasing trend in PET was observed in all subregions. Trends in both datasets presented the same sign, although observed data presented higher slopes, indicating that the actual balance between precipitation and PET may favor higher atmospheric demand for water in the future. In fact, the study by Marengo et al. (2012) already projected an increase of up to 1.5 °C in temperature and a decrease of up to 20% in rainfall over the SFW until 2040. These results were corroborated by De Jong et al. (2018). The smaller slopes in CRU TS data trends may also be related to the smoothing of data due to the interpolation procedure, which can be clearly visualized in Figure 2.9. Remarkable positive trend slopes for PET time series were found in the MSF and LSF (6 mm decade−1). Curiously, these are the regions surrounding the semiarid portion of the basin (LMSF), which may indicate an expansion of arid zones towards regions under more humid climates. Indeed, Dubreuil et al. (2019) recently detected an expansion of the areas under semiarid climate in Brazil. These results are further confirmed by the spatial distribution of CRU TS and point- based trend slopes presented in Figure 2.10. CRU TS data shows clear significant decreasing trends in rainfall over the MSF and LSF and an increase in the atmospheric demand for water throughout the entire watershed. The significance of point-based measures is more heterogeneous, which has been previously evidenced by Bezerra et al. (2019). It is also important to note that the negative rainfall trends observed in the LSF would not necessarily impact water availability in the São Francisco river, since it is located near its mouth. These changes, however, may influence the occurrence of rainfall extreme events in the region, which has been detected in previous studies (DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). Regarding point-based PET trends, a single isolated station presented a decreasing trend. In fact, this station is located downstream of the Sobradinho reservoir and the many irrigated perimeters that developed after the construction of the dam in the early 1980s. Previous studies have already assessed PET trends in this particular region, finding similar results and suggesting that the massive extension of the dam’s lake and the irrigated perimeters might have influenced local climate (CABRAL JÚNIOR et al., 2019). 2.5. Conclusion 74 This study presented a thorough assessment of rainfall and PET (calculated through temperature data) derived from the CRU TS dataset in the SFW, an extremely important basin in Brazil that lacks long-term consistent observed data with good spatial coverage for these variables. The study comprised time series of 75 years (1942–2016), and the evaluation was carried out also considering three 25-years period (1942–1966, 1967–1991, and 1992–2016) and four subregions of the basin (USF, MSF, LMSF, LSF) with different climate characteristics. Overall, CRU TS data was strongly correlated with observational data, with maximum r equal to 0.88 in 1967–1991 for rainfall data and 0.89 in 1992–2016 for PET data. The spatial distribution of correlations indicated that the poorest performances (r ranging from 0.50 to 0.80) occurred in the LMSF and LSF regions, which are characterized by a semiarid climate and the transition to a humid coastal zone. The evaluation of monthly estimates showed that estimations of Thornthwaite’s PET obtained through CRU TS temperature data are reliable (RMSE < 50% of observed mean) in all months and in all subregions. Regarding rainfall data, CRU TS data was reliable for estimations derived during the wet months in the USF and MSF. In either case, results derived from analysis using these products must always be interpreted with caution. In fact, the present study strongly advocates that more in-depth assessment procedures should be carried out before using any gridded product for meteorological variables. Understanding the strengths and limitations of these datasets should be a first step in every study proposing their use. The trends detected in both datasets pointed towards the same direction: increasing PET in all subregions and decreasing rainfall mostly in the MSF and the LSF, the areas surrounding the semiarid portion of the SFW. However, the slope of the trends in observed data was steeper than in CRU TS data. In general, CRU TS data are consistent with observed data and their use should be encouraged, given that the actual conditions of available observed data in the region (as evidenced in Supplementary Figure 2.S1,Figure 2.S2) seriously limits consistent long-term hydroclimatological studies. However, the assessment provided in the present study should always be taken into consideration when using CRU TS rainfall and temperature data over the SFW. 2.6. Supplementary materials - Summary of the observed data characteristics and quality control 75 Table 2.S1 – Summary of the main characteristics of the measuring stations selected in the study. ANA: National Water Agency; INMET: National Institute of Meteorology; GHCN: Global Historical Climatology Network; USF: upper São Francisco; MSF: middle São Francisco; LMSF: lower-middle São Francisco; LSF: lower São Francisco. n refers to the original sample size. The last two columns refer to the number of removed observations in each respective step. ID Source Lat. Long. Alt.[m] Subregion n Gaps ±3.5 sd SNHT (ºS) (ºW) (%) filter PREC_1 ANA -8.3 -39.9 384 LMSF 636 29% 10 - PREC_2 ANA -8.2 -39.7 381 LMSF 650 28% 9 - PREC_3 ANA -8.8 -39.8 355 LMSF 584 35% 9 - PREC_4 ANA -8.6 -40.0 397 LMSF 648 28% 11 19 PREC_5 ANA -9.0 -40.3 366 LMSF 883 2% 7 224 PREC_6 ANA -9.2 -37.1 331 LSF 459 49% 4 - PREC_7 ANA -9.4 -37.3 287 LMSF 455 49% 2 281 PREC_8 ANA -11.3 -42.3 499 MSF 523 42% 5 - PREC_9 ANA -11.9 -45.6 737 MSF 491 45% 6 - PREC_10 ANA -11.9 -45.1 614 MSF 515 43% 5 23 PREC_11 ANA -12.4 -42.6 470 MSF 484 46% 4 - PREC_12 ANA -12.3 -42.8 464 MSF 523 42% 6 - PREC_13 ANA -12.9 -43.4 425 MSF 554 38% 7 - PREC_14 ANA -12.2 -43.2 430 MSF 717 20% 7 420 PREC_15 ANA -12.1 -45.1 558 MSF 521 42% 6 - PREC_16 ANA -12.4 -45.1 543 MSF 524 42% 4 - PREC_17 ANA -12.4 -45.9 731 MSF 559 38% 6 - PREC_18 ANA -15.3 -43.7 484 MSF 648 28% 8 - PREC_19 ANA -15.8 -43.3 533 MSF 568 37% 7 - PREC_20 ANA -15.6 -44.4 459 MSF 497 45% 5 - PREC_21 ANA -15.9 -44.4 650 MSF 485 46% 3 - PREC_22 ANA -16.2 -44.4 766 MSF 481 47% 4 - PREC_23 ANA -16.3 -45.4 476 MSF 604 33% 8 - PREC_24 ANA -16.3 -45.2 495 MSF 714 21% 4 - PREC_25 ANA -16.1 -45.7 512 MSF 599 33% 5 - PREC_26 ANA -16.9 -45.4 489 MSF 666 26% 7 - PREC_27 ANA -16.8 -46.3 564 MSF 637 29% 6 - PREC_28 ANA -16.4 -46.9 651 MSF 601 33% 5 - PREC_29 ANA -16.5 -46.7 595 MSF 524 42% 4 - PREC_30 ANA -16.5 -46.7 595 MSF 476 47% 3 108 PREC_31 ANA -16.2 -47.2 676 MSF 469 48% 4 42 PREC_32 ANA -17.1 -45.9 497 MSF 563 37% 5 - PREC_33 ANA -17.1 -45.4 558 MSF 597 34% 7 - PREC_34 ANA -17.3 -45.5 589 MSF 422 53% 2 - PREC_35 ANA -17.3 -46.1 784 MSF 630 30% 5 - PREC_36 ANA -17.3 -46.5 530 MSF 615 32% 4 - PREC_37 ANA -17.7 -46.4 572 MSF 480 47% 6 86 PREC_38 ANA -17.5 -46.6 537 MSF 525 42% 3 - PREC_39 ANA -17.8 -48.0 806 MSF 510 43% 5 - PREC_40 ANA -18.5 -45.7 777 MSF 569 37% 3 - PREC_41 ANA -18.3 -45.8 774 MSF 672 25% 3 - PREC_42 ANA -18.4 -45.5 794 USF 545 39% 5 - PREC_43 ANA -18.7 -46.4 866 USF 579 36% 4 - PREC_44 ANA -18.5 -46.9 979 USF 637 29% 6 - 76 PREC_45 ANA -18.3 -46.4 802 USF 694 23% 6 - PREC_46 ANA -19.2 -45.4 669 USF 495 45% 2 - PREC_47 ANA -8.1 -39.6 386 LMSF 392 56% 5 - PREC_48 ANA -8.5 -39.6 370 LMSF 469 48% 5 - PREC_49 ANA -9.1 -39.9 378 LMSF 476 47% 7 - PREC_50 ANA -9.5 -40.2 380 LMSF 517 43% 8 - PREC_51 ANA -11.0 -42.3 456 MSF 476 47% 46 - PREC_52 ANA -11.4 -42.3 479 MSF 465 48% 7 - PREC_53 ANA -12.8 -44.0 563 MSF 387 57% 5 - PREC_54 ANA -12.1 -45.8 755 MSF 387 57% 3 - PREC_55 ANA -12.8 -45.9 822 MSF 382 58% 3 - PREC_56 ANA -13.7 -45.4 765 MSF 412 54% 5 254 PREC_57 ANA -15.6 -42.9 577 MSF 407 55% 5 - PREC_58 ANA -15.7 -44.3 504 MSF 397 56% 4 - PREC_59 ANA -16.4 -45.7 509 MSF 388 57% 4 - PREC_60 ANA -17.8 -45.5 825 MSF 355 61% 3 - PREC_61 ANA -17.4 -46.8 597 MSF 380 58% 3 90 PREC_62 ANA -17.9 -47.1 660 MSF 457 49% 5 - PREC_63 ANA -17.6 -46.9 584 MSF 338 62% 4 - PREC_64 ANA -18.5 -44.6 622 MSF 331 63% - - PREC_65 ANA -18.2 -46.8 852 MSF 406 55% 3 - PREC_66 ANA -19.3 -45.3 633 USF 406 55% 1 - PREC_67 ANA -20.1 -46.1 775 USF 760 16% 2 - PREC_68 ANA -19.9 -43.9 827 USF 710 21% 3 66 PREC_69 ANA -20.0 -44.2 817 USF 789 12% 1 - PREC_70 ANA -19.8 -45.2 747 USF 728 19% 8 115 PREC_71 ANA -20.2 -44.1 815 USF 782 13% 2 - PREC_72 ANA -19.9 -43.7 961 USF 792 12% 3 - PREC_73 ANA -20.2 -44.8 777 USF 776 14% 3 - PREC_74 ANA -20.5 -43.8 985 USF 744 17% 2 114 PREC_75 ANA -20.7 -44.1 911 USF 786 13% 2 - PREC_76 ANA -19.9 -44.4 818 USF 769 15% 7 - PREC_77 ANA -18.5 -43.7 1,112 USF 767 15% 6 - PREC_78 ANA -20.0 -44.0 1,013 USF 789 12% 1 - PREC_79 ANA -20.2 -45.7 689 USF 746 17% 2 - PREC_80 ANA -20.5 -45.0 806 USF 780 13% 2 - PREC_81 ANA -20.1 -44.6 874 USF 604 33% 2 - PREC_82 ANA -19.5 -43.7 758 USF 775 14% 3 - PREC_83 ANA -19.7 -45.6 653 USF 770 14% 5 - PREC_84 ANA -17.9 -44.6 589 USF 766 15% 6 - PREC_85 ANA -20.2 -43.9 1,275 USF 788 12% 3 - PREC_86 ANA -19.7 -44.8 812 USF 783 13% 5 - PREC_87 ANA -20.6 -44.4 924 USF 756 16% 1 - PREC_88 ANA -19.6 -44.1 814 USF 784 13% 6 - PREC_89 ANA -18.6 -43.6 1,126 USF 791 12% 6 - PREC_90 ANA -19.9 -43.8 790 USF 776 14% 3 - PREC_91 ANA -18.3 -44.2 606 USF 764 15% 7 - PREC_92 ANA -19.4 -45.9 816 USF 687 24% 3 43 PREC_93 ANA -19.9 -46.0. 675 USF 791 12% 10 - PREC_94 ANA -19.7 -43.7 784 USF 791 12% 6 - PREC_95 ANA -17.4 -43.7 843 USF 783 13% 8 228 PREC_96 ANA -19.7 -43.7 784 USF 767 15% 5 - PREC_97 ANA -11.2 -45.0 647 LMSF 737 18% 9 - 77 PREC_98 ANA -14.3 -44.5 566 MSF 733 19% 6 - PREC_99 ANA -14.8 -43.9 450 MSF 790 12% 10 - PREC_100 ANA -11.3 -43.8 413 MSF 750 17% 9 - PREC_101 ANA -14.3 -44.5 566 MSF 780 13% 7 - PREC_102 ANA -16.7 -43.9 734 MSF 753 16% 10 - PREC_103 ANA -11.6 -43.3 409 MSF 748 17% 6 - PREC_104 ANA -9.6 -40.7 405 MSF 747 17% 16 - PREC_105 ANA -13.4 -44.2 448 MSF 716 20% 10 - PREC_106 ANA -13.3 -44.6 629 MSF 751 17% 14 147 PREC_107 ANA -11.0 -44.5 445 MSF 681 24% 6 - PREC_108 ANA -16.0 -44.9 478 MSF 766 15% 9 - PREC_109 ANA -7.7 -37.7 562 LMSF 580 36% 6 - PREC_110 ANA -8.8 -39.0 313 LMSF 738 18% 11 - PREC_111 ANA -8.1 -37.6 534 LMSF 708 21% 5 248 PREC_112 ANA -8.6 -38.6 335 LMSF 777 14% 10 - PREC_113 ANA -9.4 -40.5 372 LMSF 803 11% 6 247 PREC_114 ANA -7.9 -40.0 438 LMSF 589 35% 13 - PREC_115 ANA -7.8 -38.1 811 LMSF 741 18% 9 - PREC_116 ANA -9.4 -38.0 244 LMSF 752 16% 4 - PREC_117 ANA -9.8 -37.4 65 LMSF 704 22% 5 - PREC_118 ANA -10.3 -36.6 26 LSF 777 14% 8 - PREC_119 ANA -10.4 -36.4 8 LSF 708 21% 7 195 PREC_120 ANA -9.6 -37.8 182 LSF 685 24% 7 121 PREC_121 ANA -10.2 -36.8 6 LMSF 719 20% 6 - PREC_122 ANA -10.0 -37.0 114 LSF 754 16% 6 - PREC_123 INMET -7.9 -40.1 449 LSF 729 19% 7 - PREC_124 INMET -7.8 -38.1 940 LSF 390 57% 5 - PREC_125 INMET -8.5 -39.3 329 LMSF 563 37% 2 - PREC_126 INMET -8.6 -38.6 328 LMSF 554 38% 5 - PREC_127 INMET -8.4 -37.1 641 LMSF 278 69% 5 - PREC_128 INMET -8.4 -36.8 623 LMSF 479 47% 4 - PREC_129 INMET -8.9 -36.5 869 LSF 650 28% 8 - PREC_130 INMET -9.6 -42.1 406 LSF 591 34% 5 - PREC_131 INMET -9.4 -40.5 376 LSF 521 42% 6 - PREC_132 INMET -9.4 -38.2 225 LMSF 513 43% 5 - PREC_133 INMET -9.1 -38.3 302 LSF 515 43% 3 - PREC_134 INMET -9.1 -37.7 506 LMSF 152 83% - - PREC_135 INMET -9.8 -37.4 40 LMSF 121 87% - - PREC_136 INMET -9.5 -36.7 295 LSF 342 62% 1 - PREC_137 INMET -11.0 -44.5 443 LSF 411 54% 4 - PREC_138 INMET -10.2 -36.9 28 MSF 509 43% 6 - PREC_139 INMET -11.3 -41.9 704 LSF 465 48% 6 - PREC_140 INMET -11.2 -41.2 924 LSF 388 57% 4 - PREC_141 INMET -12.4 -46.4 581 MSF 443 51% 5 - PREC_142 INMET -12.2 -45.0 481 MSF 507 44% 4 - PREC_143 INMET -12.7 -43.2 422 MSF 665 26% 6 65 PREC_144 INMET -13.3 -44.6 592 MSF 560 38% 4 - PREC_145 INMET -13.3 -43.4 435 MSF 43 95% 11 - PREC_146 INMET -14.1 -46.4 798 MSF 377 58% 5 - PREC_147 INMET -14.9 -46.3 860 MSF 454 50% 4 - PREC_148 INMET -14.9 -42.9 772 MSF 469 48% 4 3 PREC_149 INMET -14.1 -42.5 966 MSF 422 53% 4 - PREC_150 INMET -15.5 -47.3 863 MSF 477 47% 6 - 78 PREC_151 INMET -15.9 -46.1 527 MSF 544 40% 2 - PREC_152 INMET -16.0 -44.9 472 MSF 583 35% 4 - PREC_153 INMET -15.5 -44.4 476 MSF 446 50% 4 - PREC_154 INMET -15.1 -42.8 865 MSF 260 71% 3 - PREC_155 INMET -15.1 -44.0 451 MSF 530 41% 4 - PREC_156 INMET -16.4 -46.6 693 MSF 444 51% 3 - PREC_157 INMET -16.7 -43.8 635 MSF 377 58% 4 - PREC_158 INMET -17.2 -46.9 724 MSF 438 51% 2 - PREC_159 INMET -17.7 -46.2 727 MSF 516 43% 4 - PREC_160 INMET -17.4 -44.9 522 MSF 460 49% 3 - PREC_161 INMET -18.5 -46.4 963 MSF 591 34% 5 - PREC_162 INMET -18.8 -44.5 665 USF 566 37% 7 - PREC_163 INMET -18.3 -43.6 1,138 USF 613 32% 1 - PREC_164 INMET -19.2 -45.0 683 USF 616 32% 2 - PREC_165 INMET -19.9 -44.4 796 USF 524 42% 3 - PREC_166 INMET -20.0 -46.0 683 USF 481 47% 1 - PREC_167 INMET -19.5 -44.3 783 USF 635 29% 2 - PREC_168 INMET -19.9 -43.9 862 USF 541 40% - - PREC_169 INMET -19.0 -43.4 796 USF 570 37% 1 3 PREC_170 INMET -20.0 -44.1 885 USF 626 30% 3 - PREC_171 INMET -20.7 -44.8 1,017 USF 625 31% 3 - TMP_1 INMET -7.9 -40.1 449 LMSF 347 61% - 94 TMP_2 INMET -7.8 -38.1 940 LMSF 554 38% 1 - TMP_3 INMET -8.5 -39.3 329 LMSF 508 44% - - TMP_4 INMET -8.6 -38.6 328 LMSF 245 73% - - TMP_5 INMET -8.4 -37.1 641 LMSF 412 54% - 93 TMP_6 INMET -8.4 -36.8 623 LSF 570 37% 2 - TMP_7 INMET -8.9 -36.5 869 LSF 504 44% 3 - TMP_8 INMET -9.6 -42.1 406 MSF 398 56% - - TMP_9 INMET -9.4 -40.5 376 LMSF 456 49% - - TMP_10 INMET -9.4 -38.2 225 LMSF 504 44% 1 - TMP_11 INMET -9.1 -38.3 302 LMSF 321 64% 1 - TMP_12 INMET -9.8 -37.4 40 LSF 363 60% 1 - TMP_13 INMET -9.5 -36.7 295 LSF 461 49% - - TMP_14 INMET -11.0 -44.5 443 MSF 361 60% 1 - TMP_15 INMET -10.2 -36.9 28 LSF 365 59% - 40 TMP_16 INMET -11.3 -41.9 704 MSF 382 58% - - TMP_17 INMET -11.2 -41.2 924 LMSF 469 48% - - TMP_18 INMET -12.4 -46.4 581 MSF 617 31% - 131 TMP_19 INMET -12.2 -45.0 481 MSF 489 46% - 57 TMP_20 INMET -12.7 -43.2 422 MSF 306 66% - 55 TMP_21 INMET -13.3 -44.6 592 MSF 445 51% - 278 TMP_22 INMET -13.3 -43.4 435 MSF 416 54% - - TMP_23 INMET -14.1 -46.4 798 MSF 330 63% - - TMP_24 INMET -14.9 -46.3 860 MSF 448 50% - - TMP_25 INMET -14.9 -42.9 772 MSF 470 48% 4 - TMP_26 INMET -14.1 -42.5 966 MSF 530 41% 1 - TMP_27 INMET -15.5 -47.3 863 MSF 406 55% - - TMP_28 INMET -15.9 -46.1 527 MSF 244 73% - - TMP_29 INMET -16.0 -44.9 472 MSF 495 45% 1 - TMP_30 INMET -15.5 -44.4 476 MSF 400 56% - - TMP_31 INMET -15.1 -42.8 865 MSF 346 62% - - TMP_32 INMET -15.1 -44.0 451 MSF 430 52% - - 79 TMP_33 INMET -16.4 -46.6 693 MSF 454 50% - - TMP_34 INMET -16.7 -43.8 635 MSF 393 56% - - TMP_35 INMET -17.2 -46.9 724 MSF 528 41% - - TMP_36 INMET -17.7 -46.2 727 MSF 547 39% - - TMP_37 INMET -17.4 -44.9 522 ASF 609 32% - - TMP_38 INMET -18.5 -46.4 963 MSF 545 39% - - TMP_39 INMET -18.8 -44.5 665 ASF 468 48% 1 - TMP_40 INMET -18.3 -43.6 1,138 ASF 465 48% - - TMP_41 INMET -19.2 -45.0 683 ASF 467 48% - - TMP_42 INMET -19.9 -44.4 796 ASF 471 48% - - TMP_43 INMET -20.0 -46.0 683 ASF 552 39% - - TMP_44 INMET -19.5 -44.3 783 ASF 624 31% - - TMP_45 INMET -19.0 -43.4 862 ASF 605 33% - - TMP_46 INMET -20.0 -44.1 796 ASF 460 49% - - TMP_47 INMET -20.7 -44.8 885 ASF 438 51% - - TMP_48 GHCN -15.9 -47.9 1,017 MSF 645 28% 1 - TMP_49 GHCN -19.6 -46.9 1,047 ASF 504 44% 1 - TMP_50 GHCN -19.9 -43.9 982 ASF 766 15% 5 - TMP_51 GHCN -16.7 -43.9 896 MSF 431 52% - - TMP_52 GHCN -13.3 -43.4 629 MSF 499 45% 3 53 TMP_53 GHCN -14.1 -42.6 434 MSF 549 39% - 257 TMP_54 GHCN -9.6 -42.1 878 MSF 433 52% 1 - TMP_55 GHCN -9.4 -40.5 397 LMSF 323 64% 1 - TMP_56 GHCN -10.9 -37.1 385 LSF 726 19% 8 - 80 Figure 2.S1 – Visual representation of available data (blue squares) and gaps (red squares) in each time series of observational rainfall data after quality control was applied. 81 Figure 2.S2 – Visual representation of available data (blue squares) and gaps (red squares) in each time series of observational temperature (Thornthwaite’s potential evapotranspiration) data after quality control was applied. Table 2.S2 – Summary of observational rainfall data availability after each step of the quality control procedure at the seasonal scale. Total Total Original Original sd filter sd filter SNHT NA SNHT NA Month available available NA (n) NA (%) NA (n) NA (%) (n) (%) data (n) data (%) Jan 4447 34.7% 40 0.3% 263 2.1% 8075 63.0% Feb 4502 35.1% 32 0.2% 265 2.1% 8026 62.6% Mar 4503 35.1% 17 0.1% 260 2.0% 8045 62.7% Apr 4495 35.0% 47 0.4% 264 2.1% 8019 62.5% May 4548 35.5% 91 0.7% 264 2.1% 7922 61.8% Jun 4472 34.9% 145 1.1% 264 2.1% 7944 61.9% Jul 4528 35.3% 150 1.2% 263 2.1% 7884 61.5% Aug 4441 34.6% 161 1.3% 263 2.1% 7960 62.1% Sep 4483 35.0% 82 0.6% 261 2.0% 7999 62.4% Oct 4522 35.3% 59 0.5% 261 2.0% 7983 62.2% Nov 4475 34.9% 33 0.3% 255 2.0% 8062 62.9% Dec 4573 35.7% 44 0.3% 259 2.0% 7949 62.0% Total 53989 35.1% 901 0.6% 3142 2.0% 95868 62.3% Table 2.S3 – Summary of observational temperature (Thornthwaite’s potential evapotranspiration) data availability after each step of the quality control procedure at the seasonal scale. Total Total Original Original sd filter sd filter SNHT NA SNHT NA Month available available NA (n) NA (%) NA (n) NA (%) (n) (%) data (n) data (%) Jan 1998 47.6% 2 0.05% 88 2.1% 2112 50.3% Feb 2011 47.9% 2 0.05% 92 2.2% 2095 49.9% 82 Mar 2030 48.3% 3 0.07% 86 2.0% 2081 49.5% Apr 2043 48.6% 3 0.07% 84 2.0% 2070 49.3% May 2038 48.5% 4 0.10% 89 2.1% 2069 49.3% Jun 2061 49.1% 4 0.10% 82 2.0% 2053 48.9% Jul 2036 48.5% 4 0.10% 87 2.1% 2073 49.4% Aug 2033 48.4% 4 0.10% 90 2.1% 2073 49.4% Sep 2026 48.2% 4 0.10% 90 2.1% 2080 49.5% Oct 2005 47.7% 3 0.07% 91 2.2% 2101 50.0% Nov 2006 47.8% 2 0.05% 90 2.1% 2102 50.0% Dec 2019 48.1% 2 0.05% 89 2.1% 2090 49.8% Total 24306 48.2% 37 0.07% 1058 2.1% 24999 49.6% 83 2.7. Appendix A Table 2.A1 – Monthly mean (µ) ± standard deviation (σ) of CRU TS rainfall estimations and observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. The reliability of CRU TS estimates in each month is also shown based on whether monthly root mean square error is less than 50% of the mean observed value (where ✓ indicates reliable results and  indicates nonreliable results). USF MSF LMSF LSF Month CRU (µ ± σ) Obs (µ ± σ) Reliability CRU (µ ± σ) Obs (µ ± σ) Reliability CRU (µ ±σ) Obs (µ ± σ) Reliability CRU (µ ± σ) Obs (µ ± σ) Reliability 1942–1966 Jan 281(±125) 276(±168) ✓ 190(±116) 191(±162)  63(±54) 68(±69)  49(±41) 45(±48)  Feb 177(±67) 211(±111) ✓ 137(±73) 134(±93)  75(±55) 73(±64)  50(±40) 51(±56)  Mar 151(±66) 160(±104) ✓ 143(±92) 119(±105)  114(±96) 107(±97)  91(±76) 95(±81)  Apr 62(±28) 47(±43)  69(±49) 56(±57)  76(±53) 87(±83)  112(±84) 138(±109)  May 30(±20) 27(±31)  17(±19) 10(±15)  39(±44) 48(±59)  137(±122) 145(±114)  Jun 12(±11) 9(±15)  3(±7) 2(±6)  29(±41) 36(±50)  107(±84) 124(±83)  Jul 13(±14) 7(±15)  5(±11) 2(±7)  22(±35) 30(±48)  96(±76) 96(±72)  Aug 9(±10) 5(±13)  3(±7) 1(±4)  13(±23) 16(±27)  64(±56) 55(±48)  Sep 36(±27) 30(±34)  20(±24) 11(±19)  7(±13) 12(±21)  35(±42) 40(±45)  Oct 116(±58) 110(±76) ✓ 84(±62) 59(±62)  12(±21) 17(±27)  22(±23) 26(±37)  Nov 208(±59) 186(±89) ✓ 190(±88) 156(±102)  37(±46) 42(±63)  38(±37) 36(±39)  Dec 298(±86) 306(±152) ✓ 253(±120) 213(±141) ✓ 48(±55) 58(±73)  38(±44) 41(±50)  1967–1991 Jan 266(±150) 250(±159) ✓ 198(±151) 184(±143) ✓ 79(±70) 74(±76)  49(±42) 40(±47)  Feb 158(±102) 151(±107) ✓ 144(±110) 133(±105) ✓ 90(±73) 83(±76)  69(±55) 57(±61)  Mar 156(±82) 146(±93)  141(±92) 138(±109) ✓ 136(±82) 135(±92) ✓ 105(±67) 97(±70)  Apr 68(±34) 66(±49)  69(±46) 67(±58)  92(±72) 95(±87)  116(±79) 110(±88) ✓ May 30(±25) 28(±27)  18(±23) 14(±19)  43(±47) 48(±62)  127(±117) 114(±113)  Jun 14(±17) 12(±18)  6(±14) 4(±9)  30(±39) 30(±44)  111(±88) 100(±75)  Jul 14(±18) 14(±19)  6(±17) 3(±8)  29(±47) 29(±48)  115(±102) 97(±80)  Aug 14(±17) 13(±18)  6(±13) 4(±10)  11(±18) 14(±25)  60(±51) 53(±45)  84 Sep 47(±33) 44(±38)  22(±22) 21(±25)  9(±15) 11(±18)  40(±43) 37(±41)  Oct 119(±61) 121(±71) ✓ 101(±67) 87(±65)  16(±28) 16(±32)  30(±39) 21(±31)  Nov 218(±76) 218(±101) ✓ 182(±86) 172(±98) ✓ 27(±39) 34(±49)  29(±32) 25(±37)  Dec 279(±89) 273(±112) ✓ 227(±127) 217(±132) ✓ 55(±65) 58(±71)  37(±44) 38(±49)  1992–2016 Jan 275(±120) 260(±138) ✓ 187(±115) 175(±129) ✓ 88(±66) 82(±85)  53(±39) 52(±74)  Feb 140(±68) 151(±101) ✓ 131(±72) 126(±96)  75(±47) 82(±64)  57(±34) 54(±57)  Mar 162(±75) 172(±99) ✓ 147(±83) 148(±107)  106(±67) 105(±82)  74(±45) 69(±71)  Apr 79(±43) 59(±45)  70(±48) 59(±55)  70(±45) 63(±68)  90(±57) 82(±75)  May 29(±23) 28(±27)  18(±17) 14(±20)  41(±36) 38(±52)  113(±88) 96(±92)  Jun 11(±14) 10(±15)  6(±14) 3(±8)  26(±36) 27(±45)  109(±83) 93(±82)  Jul 9(±13) 5(±11)  3(±6) 1(±5)  24(±33) 22(±34)  93(±64) 84(±65)  Aug 11(±11) 10(±14)  6(±9) 3(±9)  9(±13) 12(±23)  54(±43) 53(±47)  Sep 51(±36) 43(±38)  19(±19) 15(±21)  8(±11) 6(±14)  34(±33) 28(±34)  Oct 115(±69) 95(±66)  83(±59) 65(±60)  11(±18) 15(±31)  26(±26) 24(±35)  Nov 213(±55) 214(±90) ✓ 176(±71) 183(±98) ✓ 25(±38) 33(±50)  27(±25) 29(±42)  Dec 295(±106) 299(±126) ✓ 215(±104) 206(±114) ✓ 47(±39) 54(±57)  33(±23) 35(±49)  Table 2.A2– Monthly mean (µ) ± standard deviation (σ) of CRU TS Thornthwaite’s potential evapotranspiration obtained through temperature estimations and observed data in the: USF—upper; MSF—middle; LMSF—lower-middle; LSF—lower São Francisco watershed. The reliability of CRU TS estimates in each month is also shown based on whether monthly root mean square error is less than 50% of the mean observed value (where ✓ indicates reliable results and  indicates nonreliable results). USF MSF LMSF LSF Month CRU (µ ± σ) Obs (µ ± σ) Reliability CRU (µ ± σ) Obs (µ ± σ) Reliability CRU (µ ± σ) Obs (µ ± σ) Reliability CRU (µ ± σ) Obs (µ ± σ) Reliability 1942–1966 Jan 112(±9) 109(±16) ✓ 121(±12) 116(±24) ✓ 147(±14) 142(±30) ✓ 146(±16) 140(±15) ✓ Feb 102(±9) 102(±15) ✓ 111(±12) 105(±21) ✓ 129(±13) 120(±28) ✓ 128(±15) 127(±14) ✓ Mar 100(±9) 104(±15) ✓ 116(±12) 114(±24) ✓ 137(±15) 131(±32) ✓ 138(±15) 134(±15) ✓ Apr 78(±9) 86(±13) ✓ 100(±13) 102(±24) ✓ 120(±13) 112(±29) ✓ 115(±16) 115(±15) ✓ May 63(±7) 64(±11) ✓ 88(±14) 88(±24) ✓ 108(±12) 99(±27) ✓ 105(±10) 101(±13) ✓ 85 Jun 50(±6) 51(±9) ✓ 72(±13) 71(±21) ✓ 89(±10) 77(±18) ✓ 87(±12) 85(±13) ✓ Jul 50(±6) 52(±8) ✓ 70(±12) 71(±21) ✓ 88(±13) 76(±20) ✓ 81(±11) 79(±12) ✓ Aug 66(±8) 71(±16) ✓ 86(±12) 89(±26) ✓ 92(±12) 88(±21) ✓ 85(±12) 81(±10) ✓ Sep 81(±10) 92(±20) ✓ 110(±14) 112(±28) ✓ 112(±16) 110(±26) ✓ 95(±12) 91(±10) ✓ Oct 99(±12) 104(±23) ✓ 124(±15) 128(±31) ✓ 143(±18) 135(±32) ✓ 120(±13) 114(±13) ✓ Nov 97(±10) 108(±22) ✓ 116(±14) 121(±30) ✓ 144(±14) 139(±31) ✓ 132(±16) 123(±13) ✓ Dec 105(±9) 113(±18) ✓ 119(±13) 120(±27) ✓ 153(±16) 146(±31) ✓ 143(±16) 134(±15) ✓ 1967–1991 Jan 114(±11) 120(±20) ✓ 124(±13) 123(±22) ✓ 146(±13) 142(±27) ✓ 145(±16) 138(±23) ✓ Feb 105(±9) 110(±18) ✓ 114(±12) 111(±21) ✓ 127(±13) 123(±25) ✓ 126(±16) 123(±20) ✓ Mar 106(±10) 113(±17) ✓ 120(±12) 118(±20) ✓ 134(±15) 131(±28) ✓ 137(±16) 131(±22) ✓ Apr 83(±10) 89(±17) ✓ 104(±13) 103(±19) ✓ 120(±13) 116(±25) ✓ 115(±17) 116(±23) ✓ May 67(±8) 70(±13) ✓ 92(±14) 92(±20) ✓ 108(±13) 104(±25) ✓ 106(±11) 106(±25) ✓ Jun 53(±7) 56(±10) ✓ 76(±14) 76(±19) ✓ 89(±10) 86(±20) ✓ 88(±13) 89(±25) ✓ Jul 53(±7) 56(±10) ✓ 74(±13) 76(±18) ✓ 88(±14) 82(±19) ✓ 82(±13) 88(±28) ✓ Aug 68(±8) 73(±13) ✓ 90(±13) 96(±22) ✓ 95(±12) 94(±23) ✓ 88(±13) 93(±29) ✓ Sep 77(±9) 87(±17) ✓ 110(±15) 114(±24) ✓ 116(±17) 115(±27) ✓ 97(±13) 103(±31) ✓ Oct 98(±11) 106(±19) ✓ 126(±16) 129(±25) ✓ 144(±17) 141(±29) ✓ 123(±14) 123(±27) ✓ Nov 103(±9) 108(±18) ✓ 120(±14) 120(±23) ✓ 145(±14) 146(±27) ✓ 134(±16) 130(±23) ✓ Dec 109(±10) 115(±18) ✓ 123(±14) 121(±22) ✓ 154(±16) 152(±28) ✓ 144(±16) 137(±23) ✓ 1992–2016 Jan 120(±11) 126(±21) ✓ 131(±13) 136(±22) ✓ 150(±13) 147(±27) ✓ 149(±17) 150(±23) ✓ Feb 110(±11) 115(±19) ✓ 119(±13) 123(±21) ✓ 131(±13) 130(±23) ✓ 130(±16) 134(±20) ✓ Mar 109(±11) 114(±18) ✓ 125(±13) 127(±22) ✓ 140(±14) 140(±26) ✓ 142(±15) 145(±22) ✓ Apr 90(±10) 95(±15) ✓ 112(±13) 116(±21) ✓ 125(±14) 126(±27) ✓ 120(±18) 131(±22) ✓ May 68(±8) 70(±12) ✓ 97(±15) 101(±23) ✓ 113(±13) 115(±28) ✓ 110(±12) 120(±25) ✓ Jun 54(±6) 56(±11) ✓ 79(±15) 82(±19) ✓ 91(±10) 90(±23) ✓ 90(±15) 101(±26) ✓ Jul 57(±6) 59(±12) ✓ 81(±14) 83(±19) ✓ 93(±14) 87(±23) ✓ 85(±15) 99(±30) ✓ Aug 72(±9) 75(±16) ✓ 96(±14) 102(±21) ✓ 97(±12) 97(±27) ✓ 90(±14) 106(±34) ✓ Sep 85(±12) 96(±21) ✓ 120(±16) 130(±23) ✓ 120(±18) 119(±28) ✓ 101(±14) 118(±33) ✓ 86 Oct 108(±14) 118(±24) ✓ 138(±16) 149(±24) ✓ 148(±17) 144(±27) ✓ 128(±14) 138(±31) ✓ Nov 106(±10) 113(±22) ✓ 126(±15) 133(±24) ✓ 149(±14) 149(±26) ✓ 138(±16) 144(±25) ✓ Dec 117(±11) 122(±20) ✓ 131(±14) 136(±23) ✓ 159(±15) 154(±27) ✓ 149(±15) 154(±23) ✓ 87 Chapter 2B: Drought characterization in the São Francisco watershed using the climatic water balance: methodological aspects and spatiotemporal dynamics This study is currently being organized as a full article in order to be submitted to an international journal. Abstract: Partially located in the Brazilian Semiarid region, the São Francisco watershed (SFW) is one of the main hydrological systems in Brazil, representing a strategic asset for power generation, agricultural development and the integration of the national territory. Drought is a recurrent phenomenon in the Brazilian Semiarid, but climate change projected scenarios indicate an intensification of these events in the SFW region, which may amplify their impacts on the population that depends on its natural resources. The objective of this study is to identify spatial and temporal patterns of meteorological drought incidence over the SFW. We used gridded Climate Research Unit Time Series (CRU TS) rainfall and potential evapotranspiration data for the period from 1942 to 2016 (75 years of data). We firstly compared the Standardized Precipitation-Evapotranspiration Index with the deficit of evaporation (DE) derived from the simplified Thornthwaite and Mather climatological water balance. Both indices retrieved similar results considering 12-month accumulated data. For the monthly monitoring, however, the DE appears to be more reliable since it is not affected by skewed rainfall data distribution during dry months while also retaining a physical sense in its proposed monthly classification. Results showed that water deficit periods are becoming more frequent and more intense, afflicting larger areas in the middle and lower portions of the SFW. Months with water surplus in these regions are also becoming less frequent. In the upper SFW, evidences show that water deficit periods are also becoming more frequent, although the wet season still provides important water surpluses. Keywords: SPEI, climatic water balance, evapotranspiration, climate extremes, drought index, semiarid 2.8. Introduction The theoretical models and the discussions on the role of drought definitions presented by Wilhite and Glantz (1985) and Wilhite (2000) as commented in the first chapter of this thesis established the basis for the investigation of this phenomenon under different perspectives. These authors highlighted that, inevitably, the main interest to society resides on the studies of the effects and impacts of drought: agricultural, hydrological and socioeconomic drought. 88 However, it is not possible to fully understand these impacts without investigating in detail their direct causes. In the drought propagation model proposed by Wilhite (2000), meteorological drought characterizes the immediate effect of the natural variability of the climate system on local atmospheric and meteorological conditions. Therefore, its characterization is the first fundamental step in order to better understand the global behavior of this phenomenon. There are different methods to assess meteorological drought, usually considering alterations and variations in climatological aspects in relation to a normal behavior. At a first moment, the main indices and techniques used were based on the fact that meteorological drought is established with the occurrence and persistence of negative anomalous rainfall amounts (MISHRA; SINGH, 2010). Among the indices used to determine these precipitation deficits we can highlight: the PDSI (PALMER, 1965); the SPI (MCKEE; DOESKEN; KLEIST, 1993), which is indicated by the Lincoln Declaration as the standard index for meteorological drought assessment (HAYES et al., 2011); the Rainfall Anomaly Index – RAI (VAN ROOY, 1965); the Effective Drought Index (BYUN; WILHITE, 1999); besides the use of percentiles, deciles, and other deviations from the mean. However, in the current climate change context, it is believed that the characterization of meteorological drought cannot be restrained to the assessment of rainfall anomalies (NAUMANM et al., 2018; VICENTE-SERRANO; BEGUERÍA; LÓPEZ-MORENO, 2010). Even in arid and semiarid regions, where rainfall is indeed the main driver of water deficit, the projected increases in temperature might significantly impact atmospheric demand for water. Several studies have already shown that this increase may be associated with the triggering and intensification of drought events (CIAIS et al., 2005; OTKIN et al., 2016; VICENTE- SERRANO et al., 2018). The SPEI, developed by Vicente-Serrano, Beguería and López- Moreno (2010) and based on the popular SPI, is being increasingly used precisely because it characterizes drought based on a simplified water balance between precipitation and PET. Therefore, the index takes into account the potential effects of increasing temperatures due to global climate change on drought characterization. Furthermore, the SPEI can be calculated at different time scales, which is an advantage if compared to other classical indices based on the water balance, such as the PDSI. The SPEI, on the other hand, presents some limitations that are yet to be properly dealt with. One of these limitations is associated with its estimation for series with skewed distribution due to the inflation of zero precipitation values during the dry season, which is 89 particularly common in arid and semiarid regions (KUMAR et al., 2009). In these cases, the water balance standardization process proposed by the SPEI is greatly jeopardized (BEGUERÍA et al., 2014; STAGGE et al., 2015). It is observed that, in these regions, the use of the SPEI is only truly consistent for larger time scales, considering the accumulated water balance in six or more months (STAGGE et al., 2015; WU et al., 2007). Since in arid and semiarid regions the occurrence of drought during the dry season is in reality the normal observed behavior, meteorological drought characterization by the SPEI at time scales of six months or more should not be a problem because the duration of the drought is more important than its intensity in these cases (CLARK, 1993; WU et al., 2007). However, the monthly scale also plays a fundamental role in the determination of the onset of water deficit periods and also in the detection of short-term shifts in drought conditions. Because part of the SFW is located in a semiarid climate region, the aforementioned challenge must be taken into consideration when evaluating the climatological aspects of drought in the basin. Furthermore, even regions under the influence of humid subtropical climate in the SFW are marked by dry seasons with recurrent zero or near-zero monthly precipitation. Therefore, in order to characterize meteorological drought in the SFW we must first discuss the methods to be used. Methods that allow the detection not only of persistent drought and accumulated water deficits but also small shifts in drought incidence patterns should be privileged. In this context, this study proposes the use of the monthly DE estimated through the Thornthwaite and Mather Climatological Water Balance – TM-CWB (THORNTHWAITE; MATHER, 1955). The DE expresses the monthly difference between PET and actual evapotranspiration (AET), being easily calculated and used as a drought index in several studies (DUBREUIL, 1996, 1997; LAMY, 2013; LAMY; DUBREUIL; MELLO-THERY, 2014; MOUNIER, 1977). In the TM-CWB, AET depends on the relationship between precipitation and PET, and on the depletion of water stored in the soil, which is usually expressed as a function of the available water capacity (AWC). The AWC, in its turn, is a surface parameter and would distort the use of the DE as a meteorological drought index. Thus, we propose the use of a homogeneous value for AWC in the entire extension of the SFW, independent of the actual surface characteristics. Thus, DE can be considered an index that will depend exclusively on the variations of meteorological parameters: precipitation and PET. 90 The use of the DE as a meteorological drought index complies with the main requirements for the adequate characterization of this type of event: it allows the comparison between different climate regimes (since surface characteristics are considered homogeneous); it can be calculated at different time scales (by simply accumulating the elements of the TM- CWB in three, six, nine, 12 or more months); and considers that both precipitation and PET can influence water deficit conditions (agreeing with recent drought studies on the context of climate change). The DE can, therefore, be used to characterize the spatial and temporal patterns of meteorological drought in the SFW, considering the frequency, intensity and duration of the identified events. This characterization, on the other hand, needs to take into consideration another fundamental climatological aspect, which is the influence of teleconnection patterns between large-scale circulation mechanisms and drought in the SFW. It is known, for example, that the ENSO phenomenon oppositely influences the rainfall regime in the USF and the LMSF (DOS SANTOS et al., 2011; GALVÍNCIO; SOUSA, 2002; SOUTO; BELTRÃO; TEODORO, 2019). It is also known that the SST gradient in the Tropical Atlantic Ocean strongly influences rainfall over a considerable extension of the NEB, particularly its semiarid region (ANDREOLI; KAYANO, 2006; CAVALCANTI, 2012; MARENGO; TORRES; ALVES, 2017). Furthermore, there are evidences indicating the phase of the AAO influences on the formation and positioning of the SACZ, which is responsible for considerable amounts of precipitation during the summer and autumn in the Southeast region of Brazil (REBOITA; AMBRIZZI; DA ROCHA, 2009; ROSSO et al., 2018; VASCONCELLOS; CAVALCANTI, 2010). Thus, it is crucial to understand what are the individual and combined effects of these phenomena on drought incidence over the SFW. Although the relationship between the ENSO and precipitation in the basin has already been studied (DOS SANTOS et al., 2011; GURJÃO et al., 2012; SANTOS et al., 2019; SOUTO; BELTRÃO; TEODORO, 2019; SUN et al., 2016), more comprehensive studies on drought characterization over the SFW considering not only the ENSO but its combined effect with other large-scale circulation mechanisms are needed. Knowing the spatial patterns of drought incidence in the SFW according to these teleconnections is of the uttermost importance for all agents involved in the management and use of its water resources. In this context, the main objective of this study is to identify spatial and temporal patterns of meteorological drought incidence over the SFW. Specifically, this study aims to 91 firstly discuss the methodological aspects of meteorological drought characterization in semiarid regions or, more broadly, regions with dry season marked by the absence of rainfall. To this end, DE obtained through the TM-CWB was compared to the SPEI at the monthly and 12-month scales. Secondly, we effectively characterized meteorological drought in the watershed by considering its normal behavior, but also investigating patterns observed under the influence of large-scale mechanisms: the ENSO, the AMM and the AAO. 2.9. Material and methods 2.9.1. Data In this study, the determination of meteorological drought considered monthly precipitation and PET (estimated through mean temperature) data. In fact, the time series of the CRU TS v4.02 (subsequently referred to simply as CRU data) which were thoroughly validate in Chapter 2A were used. They comprehend the period from 1942 to 2016 (75 years of data), and refer to interpolated observational data over a horizontal grid with a 0.5º x 0.5º (~56 km) resolution (HARRIS et al., 2014). 2.9.2. Thornthwaite and Mather climatological water balance Meteorological drought was evaluated based on the elements of the simplified water balance firstly proposed by Thornthwaite (1948) and subsequently detailed by Thornthwaite and Mather (1955). This water balance is based on the relationship between precipitation (P) and PET, therefore considering input water fluxes, the atmospheric demand for water and the effects of temperature. Specific meteorological drought conditions were assessed through the monthly DE (mm), which consists of the difference between monthly PET and AET as follows: 𝐷𝐸 = 𝑃𝐸𝑇 − 𝐴𝐸𝑇 (1) and: 𝑃𝐸𝑇, 𝑃 ≥ 𝑃𝐸𝑇 𝐴𝐸𝑇 = { (2) 𝑃 + 𝐴𝐿𝑇, 𝑃 < 𝑃𝐸𝑇 The AET obtained through the TM-CWB depends on the relationship between P and PET. If P > PET than water availability is maximum, surface water evapotranspirates at potential conditions and there is no DE. On the other hand, if P < PET, AET is equal to P plus a given amount of water stored in the soil that evaporates (ALT – mm), and in this case DE 92 represents the atmospheric water deficit. The value of ALT consists of an empirical estimate of the difference between water stored in the soil in the previous month (WSSi-1 – mm) and water stored in the soil in the actual month (WSSi – mm): 𝐴𝐿𝑇𝑖 = 𝑊𝑆𝑆𝑖−1 − 𝑊𝑆𝑆𝑖 (3) where: 𝑊𝑆𝑆 = 𝐴𝑊𝐶𝑒[∑(𝑃−𝑃𝐸𝑇)⁄𝐴𝑊𝐶] (4) and ∑(P - PET) is the accumulated sum of the precipitation deficit (when P < PET) and AWC is the available water capacity (mm) or the content of water in the saturated soil that can be used by plants. The AWC is the TM-CWB parameter that is actually highly dependent on soil and vegetation characteristics, and normally ranges from 75 to 150 mm. Since AET is indeed a surface parameter, some assumptions must be made in order to assure the use of the DE as a meteorological drought index. Therefore, we opted to use a single AWC value equal to 100 mm in the entire study area, independent of the surface, soil or climate type. Thus, with a homogeneous AWC value throughout the entire SFW, DE becomes an index solely based on the rainfall and PET characteristics in each region. An AWC of 100 mm was previously adopted in other water balance studies over the SFW (DOS SANTOS et al., 2018; LOPES et al., 2017), and agrees with recommendations by Pereira, Angelocci and Sentelhas (2002). The TM-CWB was calculated for each grid point of the CRU dataset and also for the mean rainfall and PET time series in each subregion of the watershed: USF, MSF, LMSF and LSF. The water balance is calculated sequentially starting at the last month of the first wet season of each series (year 1942), when water available in the soil for evaporation is considered maximum (WSS = AWC). The complete and detailed description of the TM-CWB method can be consulted in Pereira, Angelocci and Sentelhas (2002). The classification of the months regarding meteorological drought according to the water balance and the DE is shown in Table 2.3. The limits of each class were selected based on a simple arithmetic progression. 93 Table 2.3 – Classification of months according to the water balance and the intensity of the deficit of evaporation (DE). Adated from Mounier (1977). Water balance Monthly classification P > PET, excess water Very wet P > PET, AWC recovery Wet P < PET, DE < 40 mm Low water deficit P < PET, 40 ≤ DE < 80 mm Dry P < PET, 80 ≤ DE < 120 mm Very dry P < PET, DE ≥ 120 mm Extremely dry Through a strictly climatological perspective, the interpretation of DE values allows the identification of the following scenarios at the monthly scale: a) Months in which P is always higher than PET (DE = 0). Even when below-average precipitation occurs, it surpasses the atmospheric demand for water, thus contributing to the recovery of soil moisture and/or generating excess water (EXC), thus attenuating drought conditions; b) Months in which P is always lower than PET (DE > 0). These months compose the dry season, when the complete absence of rainfall is usually observed. Higher magnitude deficits intensify the drought conditions established in previous months or that might be established in future months; c) Months in which P may be higher or lower than PET (DE = 0 or DE > 0). These are the key months for the determination of the persistence of drought conditions. In semiarid regions they correspond to roughly three to four months in the year. The water deficit observed (or not) in these months will intensify (or attenuate) the persistence of the original water deficit observed during the dry season months. Additionally, DE can be accumulated at different time scale, representing the accumulated deficit throughout a specific period. In these cases, the classification proposed at Table 2.3 loses its sense, and thus the analysis should be based on deviations from the mean. Even in these cases, the normalization of accumulated values should be taken with care, especially when comparing regions with different climate types. This is because normalized deviations of the same statistical magnitude might represent actual physical drought condition of completely different natures. This is one of the aspects to be discussed when comparing the DE and the SPEI for the characterization of drought events in semiarid regions. 2.9.3. Standardized Precipitation-Evapotranspiration Index (SPEI) 94 Apart from the DE, the mean SPEI in each subregion of the SFW was also calculated. The SPEI is currently one of the most used indices for the characterization of drought, with its main advantage being its multi-temporal scalar nature and the fact that it does not consider only precipitation, but the balance between P and PET. It consists of a standardized index, which allows comparison between regions with different climate characteristics. However, one recurrent issue associated with the use of standardized indices based on anomalies over arid and semiarid lands is the occurrence of months with skewed distribution of data due to zero-value precipitation amounts (BRITO et al., 2018; MUTTI et al., 2020b; STAGGE et al., 2015). In these cases, the best results are found only at larger time scale, through the accumulation of P – PET over periods of six months or more. As a consequence, this index is a robust long-term meteorological drought monitor tool, but presents serious limitations when identifying short- term patterns and shifts (at the monthly scale, for example). The SPEI is calculated analogously to the SPI, considering, however, monthly series of the difference between P and PET (Dif – mm), while the SPI considers only precipitation: 𝐷𝑖𝑓𝑖 = 𝑃𝑖 − 𝑃𝐸𝑇𝑖 (5) In the case of other time scales (three, six, nine, 12 or more months), we simply accumulate the values of Dif over the target period. Monthly difference series are then adjusted to a three-parameter log-logistic distribution. The probability distribution function of the Dif series according to the log-logistic distribution is: 𝜃 −1 𝐹(𝐷𝑖𝑓) = [1 + ( )𝛽] 𝐷𝑖𝑓 − 𝛾 (6) where θ, β and γ are the scale, shape and location parameters, respectively (for γ < Dif < ∞). These parameters are estimated through the L-moments method described by Singh, Guo and Yu (1993). The SPEI value in each month can be obtained through the standardized values of F(Dif). According to the classical approximation of Abramowitz and Stegun (1972), we have: 𝑆𝑃𝐸𝐼 = 𝑊 − (𝐶0 + 𝐶1𝑊 + 𝐶2𝑊 2)⁄(1 + 𝑑1𝑊 + 𝑑 2 3 2𝑊 + 𝑑3𝑊 ) (7) where √−2 ln(𝑝) , 𝑝 ≤ 0.5 𝑊 = { (8) √−2 ln(1 − 𝑝) , 𝑝 > 0.5 95 and p is the probability that a given Dif value is surpassed at a given month: p = 1-F(Dif). When p > 0.5 the sign of the SPEI calculated through equation 8 is inverted. The constants are C0 = 2.515517; C1 = 0.802853; C2 = 0.010328; d1 = 1.432788; d2 = 0.189269; and d3 = 0.001308. Finally, the SPEI represents standardized anomalies of the Dif series and its values can be classified according to the degree of intensity of the drought as shown in Table 2.4. This classification is equal to that usually used with the SPI. Table 2.4 – Classification of anomalies identified by the Standardized Precipitation- Evapotranspiration Index. Adapted from Loukas and Vasiliades (2010). SPEI value Anomaly category 2,00 or more Extremely wet 1,00 to 1,99 Severely wet 0,00 to 1,00 Normal (wet) -1,00 to 0,00 Normal (dry) -1,99 to -1,00 Severely dry -2,00 or less Extremely dry In the present study the SPEI was calculated at the 1-month and 12-month scales, considering the mean Dif series in each subregion of the SFW. 2.9.4. Characterizing meteorological drought The first step of this study consisted in the comparison between results retrieved using the SPEI and the DE at the 1-month and 12-month time scales in the driest and wettest regions of the basin. The second step consisted of the effective characterization of meteorological drought through the TM-CWB and the DE. This characterization was carried out through the analysis of the monthly climatological aspects of the water balance and the incidence of water deficit periods in the subregions of the basin. We assessed the temporal evolution of the frequency of occurrence of water deficits or surpluses in each month and in each subregion. We also verified linear trends regarding the number of deficit months per year (Student’s t-test at the 5% significance level). Regarding the spatial aspects, we verified the extension of the incidence of meteorological drought in relation to its intensity and frequency considering the entire 1942- 2016 period (75 years) and also subperiods of 25 years: 1942-1966, 1967-1991, and 1992-2016. We also detected trends in the extension of the area afflicted by drought in each subregion (Student’s t-test at the 5% significance level). 96 Finally, the behavior of the incidence of meteorological drought was characterized as a function of the observed phase of large-scale circulation mechanisms which are known to influence the rainfall regime over the SFW. We analyzed the ENSO, AMM and AAO patterns through the following indices: a) Bivariate ENSO Timeseries index – BEST (SMITH; SARDESHMUKH, 2000): consists of the combination of the TSM Niño 3.4 anomaly index and the atmospheric Southern Oscillation Index determined as the difference between pressure at the Tahiti and Darwin stations. Months were classified as El Niño or La Niña when the 5-months moving average was among the 20% highest/lowest values for both indices. b) AMM index (CHIANG; VIMONT, 2004): defined through the projection of Tropical Atlantic SST series, indicating an anomalous warming pattern over the Northern Tropical Atlantic (positive AMM index) or the Southern Tropical Atlantic (negative AMM index). Representative months were considered when the AMM index was among the 20% highest/lowest values for at least three consecutive months. c) AAO index (GONG; WANG, 1999): defined based on the standardization of the monthly sea level pressure difference between the 40ºS and 65ºS latitudes. The index indicates the non-seasonal displacement of westerlies circulating between middle and higher latitudes in the southern hemisphere. When the AAO index is positive, the wind belt displaces to the south and when the AAO index is negative it displaces further north. Representative months were considered when the AAO index was among the 20% highest/lowest values for at least three consecutive months. All indices are available at the Earth System Research Laboratory website of the National Oceanic and Atmospheric Administration (ESRL-NOAA)1. The AAO index data provided by the ESRL were complemented with series provided by the Climate Prediction Center, also part of the NOAA2. Table 2.5 summarizes the expected combined effects of the different large-scale circulation mechanisms on drought conditions over the SFW. 1 Available at: 2 Available at: 97 Table 2.5 – Summary of the general expected effects of the different large-scale circulation mechanisms on drought over the São Francisco watershed (SFW). USF: Upper São Francisco, MSF: Middle São Francisco, LMSF: Lower-middle São Francisco, and LSF: Lower São Francisco. Pattern Expected effect on the SFW ENSO+ (La Niña) Increased rainfall over the northern portion of the MSF, LMSF and LSF; intensification of drought over the southern USF ENSO- (El Niño) Intensification of drought over the northern portion of the MSF, LMSF and LSF; increased rainfall over the southern USF AMM+ (warmer STT over Intensification of drought over the northern portion of the the North Atlantic) MSF, LMSF and LSF AMM- (warmer STT over Increased rainfall over the northern portion of the MSF, LMSF the South Atlantic) and LSF AAO+ (westerlies further Increased rainfall over the USF and the western portion of the south) MSF AAO- (westerlies further Intensification of drought over the USF and the western north) portion of the MSF Additionally, Table 2.6 shows the scenarios obtained through the combination of the sign of the studied patterns. These scenarios were defined based on the expected effects presented at Table 2.5. Overall, we aimed to identify spatial drought characteristics under: normal conditions, without the influence of large-scale patterns (Normal scenario – NOR); the influence of La Niña but without the influence of the AAO (La Niña Pure – LN_P); the influence of El Niño but without the influence of the AMM (El Niño Pure – EN_P); the combined influence of La Niña and negative AAO (LN_INT); the combined influence of El Niño and positive AMM (EN_INT); the influence of negative AAO but without the influence of the ENSO (AAO_P); and finally the influence of positive AMM but without the influence of ENSO (AMM_P). The years were classified according to these scenarios when the indices reached the aforementioned thresholds for at least three consecutive months. Furthermore, for the AMM we only considered austral summer and autumn months, when the ITCZ is expected to displace south of the Equator. For the AAO, we only considered the austral spring and summer months, which represent the onset and peak of the wet season over the southern portion of the SFW. 98 Table 2.6 – Characterization of the different scenarios related to the acting of large-scale circulation patterns and the respective representative years selected in the 1942-2016 period (+ positive, - negative, or n neutral). Scenario ENSO AMM AAO Years Normal (NOR) n n n 1952, 1954, 1955, 1959, 1960, 1961, 1964, 1971, 1978, 1990, 2000, 2003, 2006 and 2007 La Niña Pure (LN_P) + -/n/+ n 1974, 1975, 1988, 2009, 2011 and 2012 El Niño Pure (EN_P) - n -/n/+ 1965, 1982, 1992 and 2016 La Niña + AAO negative + -/n/+ - 1950 and 1970 (LN_INT) El Niño + AMM positive - + -/n/+ 1958, 1997 and 2010 (EN_INT) AAO negative (AAO_P) n -/n/+ - 1951, 1957, 1967, 1968, 1969 and 1977 AMM positive (AMM_P) n + -/n/+ 1951, 1953, 1956, 1962, 1962, 1966, 1969, 1979, 1980, 1981, 1996, 2004, 2005 and 2013 The spatial characteristics of drought incidence were identified according to each of the defined scenarios in comparison to the normal climatology in the studied period. 2.10. Results and discussion 2.10.1. Comparison between the DE and the SPEI Before characterizing meteorological drought in the SFW, the reasons why the DE may be a more interesting index alternative for semiarid regions should be discussed. Thus, we compared it with the widely used SPEI in assessing drought over the USF and the LMSF, regions with contrasting climate features in the SFW (as previously seen in this chapter). The USF has a well-defined dry season and a wet season marked by intense precipitation that usually surpasses atmospheric demand for water (humid subtropical climate). The LMSF, on the other hand, is a typical semiarid region, with the wet season concentrated in three to four months and rainfall volumes that are usually lower than the PET. The first step in this comparison consists of showing that the DE is indeed comparable to the SPEI, that is, both indices meet the main requirements to the identification of drought events. To this end, Figure 2.11 shows the 12-month SPEI time series (normally used for the characterization of drought in arid and semiarid regions) in comparison with the DE and the water surplus, also accumulated in 12 months. The series are strongly correlated both in the USF (r = 0.99) and the LMSF (r = 0.98). In fact, this is an obvious and expected result since DE and EXC are determined by the water balance through the relationship between P and PET, 99 which is also the base for the SPEI calculation. Differences between indices derive precisely from the fact that the TM-CWB takes into account the influence of water stored in the soil on the water balance. Figure 2.11 – Time series of the 12-month SPEI and the 12-month accumulated difference between excess water (EXC) and deficit of evaporation (DE) in: (a) Upper São Francisco – USF; and (b) Lower-middle São Francisco – LMSF. The adjustment of the distribution of 12-month accumulated P – PET values in the SPEI calculation process do not incur in the issues usually observed at smaller time scales resulting from the inflated occurrence of precipitation values equal to 0. Figure 2.11 highlights an important issue to be considered in the interpretation of SPEI results. On the one hand, its interpretation allows the comparison of its values between regions with different climates. On the other hand, the classification proposed by the SPEI (Table 2.4) lacks physical sense because the normalized values are always associated with the specific P – PET distribution in each region. For example, in the USF (Figure 2.11a) the average behavior corresponds to the occurrence of water surplus of the order of 270.8 mm (SPEI = 0). In the LMSF (Figure 2.11b), however, the average behavior is of water deficit (-897.0 mm). In reality, even when SPEI = 3 in the LMSF the actual atmospheric situation is of water deficit (-404.9 mm). That is because in this region, PET is systematically higher than precipitation, even during the wet season. Likewise, in the USF, truly important water deficit conditions are only observed when SPEI ≈ -1.2 (dashed line in Figure 2.11a). In the LMSF, these SPEI values correspond to a situation of atmospheric water deficit of the order of 1000 mm. These results highlight that although the SPEI allows the comparison of drought by representing the intensity of deviations from a normal condition, the actual observed physical condition can only be truly assessed by 100 observing the values of the water balance. In this sense, using the DE is more advantageous than the SPEI, since it describes the actual observed conditions in each site. However, it is worth mentioning that these conclusions refer strictly to the meteorological aspects of drought, and do not consider the practical effects on the vegetation soil and surface water availability. The use of the SPEI or the DE to identify anomalously dry years also retrieved similar results (Figure 2.12). In the USF, the driest identified years were 1963, 1984, 1990, 2007 and 2014, which corroborates with previous studies in the region (SANTOS et al., 2017). In the LMSF, both indices identified the years 1958, 1993, 1998, 2012 and 2016 as the driest in the series, although the SPEI (but not the DE) also indicated 1990 as a very dry year. These results agree with previous studies on the historical aspects of drought in the Northeast region of Brazil (MARENGO; TORRES; ALVES, 2017; RAO; HADA; HERDIES, 1995). Figure 2.12 – Identification of the driest years in the: (a) Upper São Francisco – USF; and the (b) Lower-middle São Francisco – LMSF, considering 12-month SPEI in December of each year and the yearly accumulated difference between excess water (EXC) and deficit of evaporation (DE). The scale of the graphics is not comparable. However, using 12-month SPEI or accumulated DE presents an important limitation when defining the onset and the ending of drought events. Furthermore, this time scale does not allow the identification of short-term and short-duration shifts and patterns in dry conditions, which may be important to the attenuation of its impacts on economic activities. Figure 2.13a, for example, shows 12-month SPEI and the monthly DE (or EXC) at the USF in the period from January 2005 to January 2011. One can notice that the SPEI indicates a persistent dry period starting in September 2007 (detail in red in Figure 2.13a). Since it refers to a 12-month accumulated value, the true water deficit is expected to have started earlier, but 101 the index at this time scale does not allow the identification of the true onset. Monthly DE, on the other hand, shows that the water deficit effectively began in March 2007, six months earlier than what was indicated by the SPEI. Similarly, when observing 12-month SPEI one cannot observe that from January to April 2008 (detail in red in Figure 2.13a) important water surpluses occurred in the USF, which certainly somewhat attenuated drought effects that year. Figure 2.13 – Difference between monthly excess water (EXC) and deficit of evaporation (DE) (colored bars) and: (a) 12-month SPEI (black line); (b) 1-month SPEI (black line) in the Upper São Francisco – USF region during the period from Jan/2005 to Jan/2011. These analyses show that 12-month SPEI is not sufficient to fully characterize meteorological drought in the studied region. One might ask why not use, for example, 1-month SPEI to identify short-term changes in drought conditions? Figure 2.13b compares 1-month SPEI with monthly DE (or EXC). In this case, the issue with using the SPEI lies on the existence of dry months in which almost no precipitation occurs throughout the entire time series, and therefore have little influence on attenuating dry conditions. There can also be months in which registered precipitation is still much higher than PET even in unfavorable years for the occurrence of rainfall. For example, in December 2006 (detail in red in Figure 2.13b) the SPEI is slightly negative while the actual condition was of a remarkable water surplus. Similarly, Figure 2.14a shows that for the semiarid (LMSF) region this is also true. The SPEI indicates a dry period starting in April 2015 that peaks in 2016, while in reality water deficit started accumulating in August 2014, as revealed by the monthly DE series. In Figure 2.14b one can also notice that in this same period 1-month SPEI presents a high variability, and it poorly represents the actually observed hydrometeorological conditions. For example, in the dry months of 2014 the SPEI retrieved some positive values, while in reality these are known 102 to be the dry months in the region. In September, monthly SPEI was of 0.55 which is classified as normal (wet) according to Table 2.4 while in reality total precipitation this month was of 15 mm, which is insufficient to attenuate any dry condition in the region. Figure 2.14 – Difference between monthly excess water (EXC) and deficit of evaporation (DE) (colored bars) and: (a) 12-month SPEI (black line); (b) 1-month SPEI (black line) in the Lower-middle São Francisco – LMSF region during the period from Jan/2011 to Dec/2016. In order to strengthen the discussion on the limitations of using monthly SPEI for the identification of short-term changes in drought conditions we propose a simple analysis of its relationship with the original distribution of P – PET values. Figure 2.15 shows this relationship for two key-months: December in the USF, marked by a high variability of rainfall, but in which high rainfall volumes always occur, and September in the LMSF, which is the month representing the peak of the annual dry season in the region, with almost no rainfall variability and consistent zero or near-zero precipitation. 103 Figure 2.15 – Relationship between 1-month SPEI and the difference between precipitation (P) and potential evapotranspiration (PET) in December for the Upper São Francisco – USF and September for the Lower-middle São Francisco – LMSF. For September in the LMSF, the figure shows that the water deficit condition is nearly constant throughout the entire time series (approximately -100 mm). Thus, the monthly classification proposed by the SPEI makes no physical sense, because even when the index assumes high positive values, the actual observed condition is of intense water deficit. For December in the USF the analysis can be carried out analogously. Although indeed there is a great variability in P – PET values, ranging from -16 to 410 mm, there is a noteworthy predominance of positive values, that is, water surplus. Thus, even negative SPEI values are associated, in reality, with situations of water surplus in the USF. Table 2.7, on the other hand, highlights how the use of the SPEI may lead to imprecise conclusions in relation to the frequency and intensity of occurrence of drought events in regions with similar characteristics to the studied region. The SPEI indicates that, in the 75 years of study, 17.2% of the December months in the USF were severely dry and only 17.4% were severely or extremely wet. Although these results describe the behavior of the P – PET balance around its own mean, in physical terms the DE indicates that 93.3% of the months presented a positive water balance. Equivalently, for September in the LMSF, SPEI indicated that 22.7% of occurrences were severely or extremely wet months, and only 14.7% were severely or extremely dry months. In reality, in 92% of the occurrences there was a DE between 80 and 120 mm, which represents a severe water deficit. 104 Table 2.7 – Frequency of occurrence of Standardized Precipitation-Evapotranspiration Index (SPEI) and deficit of evaporation (DE) classes for the months of December in the Upper São Francisco – USF and September in the Lower-middle São Francisco – LMSF. SPEI Dec. USF Sep. LMSF DE Dec. USF Sep. LMSF Classification (%) (%) Classification (%) (%) > 2,00 2.7 2.7 P > PET, EXC 93.3 - 1,00 to 1,99 14.7 20.0 P > PET* 4.0 - 0,00 to 1,00 34.7 25.3 < 40 mm 2.7 - -1,00 to 0,00 30.7 37.3 40 to 80 mm - 1.3 -1,99 to -1,00 17.2 13.3 80 to 120 mm - 92.0 < -2,00 - 1.4 ≥ 120 mm - 6.7 *recovery of the available water capacity (AWC) Finally, one can observe through Figure 2.16 the spatial consequences of the present discussion. The figure shows the monitoring of the spatial evolution of a drought event in the SFW through monthly SPEI and DE, focusing on the period from July 2011 to June 2012. The period from July to September marks the dry season in basically the entire watershed (with the exception of the LSF), with the increase in water deficits starting on the northwestern portion of the MSF and western LMSF, as illustrated by the DE. The SPEI, on the other hand, does not properly represents this evolution and even assumes positive values in August in the LMSF. In October, drought peaks at the LMSF while the first rains of the wet season occur in the USF and the MSF, recovering the AWC and generating excess water. This pattern persists until January 2012 in these regions and both the DE and the SPEI accurately represent it. During this period in the LMSF the SPEI indicates an overall normal condition. However, in this transition between the dry and wet season over the Brazilian Semiarid the DE indicates a persistence of water deficit conditions. During the semiarid wet season (March to May), the SPEI accurately represents the occurrence of negative anomalies that contributed to make 2012 an extremely dry year. Similarly, the DE also represents this pattern, with an expansion of areas in the MSF and the LMSF with deficits above 80 mm. 105 Figure 2.16 – Spatial pattern of the evolution of drought in the period from July 2011 until June 2012 according to the 1-month Standardized Precipitation-Evapotranspiration Index (SPEI) and the monthly deficit of evaporation (DE). These results show that the SPEI can be properly used at larger time scales for the characterization of persistent drought events in regions with a climate regime similar to the SFW. It can also be used to accurately determine dry years and to map the evolution of a drought event. However, in these regions, the SPEI cannot represent short-term monthly changes due to the skewed distribution of rainfall in certain months. The DE index, on the other hand, can be used in a similar fashion while also retaining the physical sense of its interpretation. Furthermore, it can be used at the monthly scale to monitor drought conditions while retrieving accurate results. 2.10.2. Meteorological drought incidence patterns Calculating the sequential water balance at the monthly scale allows the monitoring of the evolution of water deficit in the SFW. Figure 2.17 shows this monthly classification for mean time series over the USF, MSF, LMSF and LSF. A purely visual analysis reveals, for example, that the months from December to March in the USF are marked by water surpluses. 106 This means that, in this region, rainfall during the wet season is usually sufficient to meet atmospheric demands and recover the AWC of the soil. In the MSF, which is the transition zone between the more humid climate typical of the USF and the semiarid LMSF, the wet season occurs approximately at the same period, but is not strong enough to assure the occurrence of frequent water surplus. Indeed, one can notice an intensification of deficits especially over the last decades of the time series. Figure 2.17 – Interannual variability of the sequential water balance in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF, in the period from 1942 to 2016. In the LMSF, the drought pattern is marked by the almost permanent persistence of water deficit conditions. Rainfall is concentrated in the months from February to May and not 107 always is enough to surpass atmospheric demands, especially because of high temperatures in this region. In the LSF, the wet season (May to July) normally registers positive water balance values although it is possible to observe a clear reduction in the occurrence of months with water surplus in the last decades. Results from Figure 2.17 can be better discussed through a statistic and probabilistic perspective. Therefore, Figure 2.18 shows the monthly frequency of occurrence of each water balance class in each subregion of the SFW in each 25-year period of the studied time series. It is possible to observe that, in the USF, positive water balances were usually assured by the peak of the wet season in December in the first 50 years of the series. In the period from 1992-2016, however, low deficits (< 40 mm) have occurred in 8.0% of the occasions. Similarly, the frequency of deficits in February increased from 16.7% in 1942-1966 to 36.0% in the last 25 years. In relation to the dry season, the frequency of occurrence of deficits between 40 and 80 mm in July also increased from 16.0% to 64.0% throughout the studied period, indicating that the dry season may be starting earlier in the region. This is reinforced by the fact that deficits between 40 and 80 mm could also be observed, even though timidly, since February in the period from 1992 to 2016, which was not observed in the first half of the time series. In the MSF, one can notice a reduction in the frequency of months with water surplus at the onset of the wet season: from 32.0% to 8.0% in November and from 72.0% to 44.0% in December. Another remarkable change in the MSF refers to the increase in the frequency of months with extreme deficit (> 120 mm) in the dry season, more precisely in September and October. While deficits with this intensity were nearly inexistent in the period from 1942 to 1966, they represent 32.0% of the months of September and October in the last 25 years. In the semiarid portion of the SFW, the LMSF, there are no latent changes in the frequency of occurrence of water deficit months during the dry season, which remained roughly with the same duration and intensity during the studied period. However, the wet season presented drastic changes. In the first 50 years of the series it is possible to note water surpluses in all months of the wet season (mainly February to April), even if with a low frequency (4 to 12.5%). However, in the period from 1992 to 2016 only March presented water surplus and in only one year. In all other years of this last 25-year period the rainfall occurring during the wet season was not capable of surpassing the atmospheric demand for water in order to recover the AWC. 108 Figure 2.18 – Monthly frequency of the sequential water balance in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF, in the periods from: 1942 to 1966, 1967 to 1991 and 1992 to 2016. Finally, Figure 2.18 shows that in the LSF the dry period usually extended until March with deficits higher than 40 mm in 66.7% of the cases from 1942 to 1966. In the last 25-year periods, however, this frequency accounted for 96.0% of the cases. As in the LMSF, the wet season seems to have been through important changes. In the first 50 years of the series, water surpluses could be observed from March to August, with frequencies of up to 41.6% during June and July. In the period from 1992 to 2016, the occurrence of excess water was limited to the months from May to July, with a maximum frequency of 28.0% in July. 109 Regarding the spatial characteristics of meteorological drought, deviations from the mean behavior throughout the 75-year period can be observed, month by month, in Figure 2.19 and Figure 2.20. January is the second wettest month in the USF and the western MSF, characterized by summer rains associated with the establishment of the SACZ (CAVALCANTI, 2012; DE OLIVEIRA; SANTOS E SILVA; LIMA, 2017). The zone where water surpluses occur in these regions diminishes with the end of summer and the arrival of spring, and in June water deficit conditions can be observed in almost the entirety of these portions of the basin (Figure 2.19). The coastal region (LSF) presents an opposite behavior, with the peak of the dry season occurring in January and the wet season establishing during the austral winter. In the first half of the year, water deficit conditions behave as a spot (Figure 2.19) extending over the LSF, the LMSF and the northeastern MSF. Until April, this spot maintains basically the same proportions, but with reducing intensities over the LSF. From April on, the spot drastically expands, with deficits ranging from 40 to 80 mm and encompassing most of the watershed in June, with the exception of the USF and the LSF. In this period, important deficits (between 80 and 120 mm) are observed over the border between the MSF and the LMSF. The months of May and June presented low variability in the first three 25-year periods in relation to the mean behavior. April presented deficits 1.2 to 1.6 times higher in the last 25 years in the LSF, in the northern LMSF and mainly in the western MSF. The maps in Figure 2.19 also show that in the period from 1967 to 1991 the southwestern portion of the MSF presented deficits 1.6 times more intense than normal, while in the period from 1992 to 2016 they were lower over this region. On the other hand, the last 25 years in the USF also presented more severe deficits during the first three months of the year. In the second half of the year (Figure 2.20) the drought spot expands until reaching basically the entire SFW in September. In these months, deficits between 80 and 120 mm are observed throughout the entire MSF and LMSF. From October to December the spot begins to retract due to the beginning of the wet season in the USF and western MSF during the austral autumn. However, it is precisely during this period that dry conditions are more intense (> 120 mm) over the LMSF and the LSF. 110 Figure 2.19 – Spatial distribution of the mean deficit of evaporation (DE) over the São Francisco watershed in the months from January to June in the period from 1942 to 2016. The smaller maps show the proportion of the mean DE in each 25-year period: 1942 to 1966, 1967 to 1991 and 1992 to 2016. Regarding the variability of drought incidence throughout the 75 years of study, Figure 2.20 shows that from July to September no remarkable variability patterns occurred. In October, however, more water deficits occurred in the last 25 years in the USF and the western MSF. 111 Figure 2.20 – Spatial distribution of the mean deficit of evaporation (DE) over the São Francisco watershed in the months from July to December in the period from 1942 to 2016. The smaller maps show the proportion of the mean DE in each 25-year period: 1942 to 1966, 1967 to 1991 and 1992 to 2016. The months from November to December presented a curious behavior in the USF and the western MSF during the period from 1966-1991 and 1992-2016. While generally 1.6 times more water deficits occurred in November (December) in 1966-1991 (1992-2016) in this region, December (November) presented up to 60.0% less deficits in the same period. These results indicate that the period from 1966 to 1991 was possibly characterized by a prolonged dry season over this region, with deficits persisting until November. On the other hand, from 112 1992 to 2016 more important deficits were observed in December, indicating a potential weakening of rains or an increase in temperatures during the wet season. Results on the monthly behavior of meteorological drought as described by the water balance are consolidated in Figure 2.21 and Figure 2.22. Figure 2.21 shows that the LMSF and the surrounding portions of the MSF and LSF usually do not present months with excess water. These are regions that, even during the wet season when precipitation surpasses PET, the positive water balance may recover the AWC but may not be sufficient to generate water surplus. Evidently, these results should be taken with care, since we used a homogeneous AWC for the entire surface of the basin. Through a meteorological perspective, this result corroborates with the knowledge that the wet season over this region is usually short and weak while PET is usually high. Figure 2.21 – Mean and lower quartile (Q1) of the number of months with water surplus in the São Francisco watershed in the period from 1942 to 2016. The variation in the number of months in relation to the mean is shown in the smaller maps: 1 – from 1942 to 1966, 2 – from 1967 to 1991 and 3 – from 1992 to 2016. The coastal portion of the LSF, the western MSF and the USF, on the other hand, present an average of two to six months with water surplus per year. This is particular important if we consider that some of these regions encompass the watershed portion near the source of the São 113 Francisco river. Through a climatological perspective, it is a region with plenty of water surplus, which leads to reliable conditions for the recharge of the main water body. In fact, Figure 2.21 also shows that even when considering unfavorable scenarios (Q1 – first quartile) the USF still presents from two to four months with water surplus. This characterization reinforces the strategic importance of the São Francisco river to the semiarid region of Brazil, since even in unfavorable years the water balance over the USF is positive during a third of the year. By comparing the variability of the number of months with water surplus in the SFW in each 25-year period (Figure 2.21) one can notice an evident unfavorable scenario in the period from 1992 to 2016. In fact, what is shown is not a deterioration of drought conditions in the semiarid region, but rather an expansion of areas with zero to two months of water surplus per year. This expansion is noteworthy mainly in the central MSF and the LSF. As a complement to Figure 2.21, Figure 2.22 shows the mean number of water deficit months per year in the SFW. It can be observed that most of the LMSF and the surrounding MSF and LSF present from 10 to 11 months per year with precipitation rates lower than PET. In the western MSF and the USF the number of deficit months ranges from five to a maximum of eight. The map that shows the number of deficit months in unfavorable scenarios (Q3 – third quartile), however, does not reveal any remarkable difference from the normal behavior, especially over the USF. In this region, the combined analysis of Figure 2.21 and Figure 2.22 reveals that there is not necessarily an enlargement of the water deficit period during unfavorable years, but rather a reduction in the number of months in which the positive water balance is relevant to the point of generating water surplus. Figure 2.22 also reveals an increase in the number of deficit months in the central portion of the MSF and the LSF in the period from 1992 to 2016 in relation to the normal behavior. This result corroborates with what was shown in the previous figure, in which the region with persistent negative water balance appears to be expanding from the LMSF towards the LSF and the MSF. 114 Figure 2.22 – Mean and upper quartile (Q3) of the number of months with water deficit in the São Francisco watershed in the period from 1942 to 2016. The variation in the number of months in relation to the mean is shown in the smaller maps: 1 – from 1942 to 1966, 2 – from 1967 to 1991 and 3 – from 1992 to 2016. The maps shown in the previous figures allow the identification of the spatial variability of the water balance over the SFW and also considering periods of 25 year within the total studied time series. Regarding the mean behavior in each subregion, it is also possible to evaluate the behavior and trends in drought incidence and extension. In this sense, Figure 2.23 shows that in the USF there is a non-significant reduction trend in the number of months with positive water balance. However, there is a positive increasing trend in the number of deficit months, observed especially from the 1990s. The same behavior can be observed for the MSF. This result indicates that in these two regions the water deficit periods are becoming longer in average, while periods of positive water balance are becoming shorter. In the LMSF, no significant trend was found. This means that at the interannual scale, there is no increasing or decreasing trends for periods of water deficit. This does not mean that the magnitude of dry periods is not intensifying, because this analysis refers to the occurrence of deficit months per year. In the LSF, more significant trends are detected. At the beginning of the studied period there were six deficit months per year on average while in the last portion 115 of the time series this number increased by two months. Analogously, the number of months in which rainfall surpasses PET is reducing significantly. Figure 2.23 – Annual behavior of the number of months in which P > PET and the number of months with deficit of evaporation (DE) higher than 40 mm in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF. Significant (α = 5%) linear trends (black line) are marked with a star (*). Figure 2.24 shows the linear trends of expansion or reduction in the mean monthly extension of the water balance in each subregion of the SFW. In the USF, even if a rupture can be identified in the series in 2012, it was not sufficient to lead to a significant increasing or decreasing trend in the area affected by water deficit or surplus. In the MSF, there is a significant increasing trend in the extension of water deficit incidence and a decreasing trend in the extension of water surplus incidence. This results also agrees with what was spatially observed in Figure 2.21 and Figure 2.22, which showed an apparent expansion of semiarid regions where PET is higher than precipitation throughout most part of the year from the LMSF toward the central portion of the MSF. 116 Figure 2.24 – 12-month moving average behavior of the area with water deficit or surplus in the Upper São Francisco – USF; Middle São Francisco – MSF; Lower-middle São Francisco – LMSF; and Lower São Francisco – LSF. Significant (α = 5%) linear trends (black line) are marked with a star (*). Since it consists of a semiarid region, the LMSF already has a pretty significant mean area afflicted by water deficit, but a statistically significant increasing trend was found for this subregion. In the LSF, results are once again revealing, with significant increasing trends for the areas afflicted by water deficits and significant decreasing trends for the area with water surplus. This is also coherent with what was observed in previous results indicating an expansion of semiarid areas from the LMSF towards the coastal SFW and the central MSF. The results presented so far must be analyzed in the context of recent studies that evaluated climate change scenarios in the SFW. Marengo et al. (2012) projected, until 2040, an increase of approximately 1.5 ºC in mean annual temperature over the basin. In the same study, the authors also projected a reduction of approximately 20.0% in summer rainfall over the SFW, which was corroborated by the study by De Jong et al. (2018). Under this perspective, the evolution of the water balance in the watershed points towards an increase in the frequency of 117 scenarios in which P < PET due to the reduction in precipitation or the increase in temperature (PET). Previous studies analyzed trends in extreme precipitation indices in the SFW and found that the intensity of precipitation events is reducing and that these events are occurring in shorter periods of time in the entire basin (BEZERRA et al., 2019). Simlarly, Souto, Beltrão and Teodoro (2019) reported that approximately 69.0% of the extension of the SFW presents a decreasing trend in annual rainfall amount, although no statistical significance was found. Overall, results found in the present study corroborates and reinforce such findings. The analysis of the water balance showed a clear reduction in the occurrence of months with water surplus in the entire basin in the period from 1992 to 2016. Furthermore, results detected significant increasing trends in the number of deficit months in the USF, MSF and LSF. By analyzing each subregion individually, results showed that the LSF and the central- eastern portion of the MSF are those in which more changes were detected in the period from 1992-2016 in relation to the mean climatological behavior. Furthermore, these regions present significant trends of expansion of areas under the influence of water deficits, while decreasing trends were found for areas under the influence of water surplus. This result provides a finer- scale perspective to what was previously found in the study by Dubreuil et al. (2019), in which an expansion of regions under semiarid climate was detected in Brazil over the last years. Indeed, in the SFW this expansion would be perceived by the expansion of the area under water deficit and their more frequent occurrence over the MSF and the LSF. Although no trends were found for the reduction of water surplus months in the USF, an increase in the frequency of occurrence of water deficit months was found in this region. As previously discussed, these results might indicate that the dry season is starting earlier or becoming longer. Indeed, previous specific studies over the USF showed a reduction in total precipitated amount and an increase in the number of consecutive dry days in most part of this subregion (SANTOS et al., 2017, 2018). Finally, in order to complement the drought analysis proposed in this study, the behavior of meteorological drought incidence in the SFW will be evaluated based on the occurrence of large-scale teleconnection patterns. Figure 2.25 shows the mean annual anomalies of the water deficit in the SFW in each scenario previously described in Table 2.6 in relation to the normal scenario (NOR). The figure also shows the characterization of the mean anomalies of the NOR scenario in relation to the total mean pattern obtained from the entire 75 years of study. One 118 can observe that, despite the NOR scenario refers to years when large-scale mechanisms supposedly were in their neutral phases, it is a rather dry NOR scenario in comparison to the entire period, especially in the MSF. This information should be taken into consideration when analyzing the other scenarios, since it represents the baseline scenario. Figure 2.25 – Annual anomaly of the accumulated deficit of evaporation (DE) during the wet season in each teleconnection scenario in relation to the normal scenario (NOR): for EN_P, AMM_P and EN_INT: February to May; for LN_P, AAO_P and LN_INT: November to March. The anomalies in the NOR scenario in relation to the mean of the entire 1942-2016 period is also shown for comparison. In the EN_P scenario (influence of the El Niño alone) the regions of the LMSF and the LSF presented positive anomalies of more than 50 mm in basically all their extension. Furthermore, anomalies between 10 and 50 mm were observed in the central portion of the MSF and the USF. Negative anomalies are expected in the northern portion of the basin due to the displacement of the Walker cell towards the central Equatorial Pacific during El Niño events. This displacement makes the descending branch of the Walker cell to remain over the NEB, inhibiting convection and cloud formation and, consequently, precipitation (ANDREOLI; KAYANO, 2006; CHIANG; VIMONT, 2004; KANE, 1997; MEDEIROS, 2019). In these cases, the occurrence of more intense rainfall is also expected in the Southeast 119 and South of Brazil, which encompass the southernmost portion of the USF. The figure, however, do not show negative anomalies in the water deficit over this region. In fact, Grimm and Tedeschi (2009) showed that the response observed in precipitation extremes during El Niño episodes is less consistent than during La Niña events in a considerable portion of South America. In years when the La Niña acted isolated (LN_P) the opposite was observed. Negative DE anomalies were detected over the LMSF, part of the LSF and the MSF. In the USF, positive anomalies were found, although with a low magnitude. In the MSF a nucleus of positive anomalies surrounded by the expected negative anomalies can be observed. A possible explanation for this pattern might be associated with the positioning of the cores of UTCV over this region during austral summer. These high-pressure cores inhibit precipitation and transport moisture to the borders of the system, where intense rainfall occur. The influence of the UTCV on the MSF region inhibiting precipitation during La Niña years was previously studied by Gurjão et al. (2012). In the AMM_P scenario one can observe the isolated influence of the displacement of the ITCZ towards the higher latitudes of the northern hemisphere during the wet season in the northern portion of NEB. This effect, caused by the anomalous warming of the Tropical North Atlantic Ocean is consistent and was investigated in several studies in the region (ANDREOLI; KAYANO, 2006; HASTENRATH, 2006; KAYANO et al., 2018; MEDEIROS, 2019; UTIDA et al., 2019). It leads to important positive water deficit anomalies (higher than 50 mm) in the northern portion of the MSF, the LMSF and the LSF. It is curious to note, however, that reductions in the annual water deficit were also observed in the central-south portion of the MSF and the northern USF in this scenario. The spatial pattern observed in the AMM_P scenario is basically the same as in the EN_INT (El Niño combined with warmer SST over the Northern Tropical Atlantic Ocean) scenario. Regions with positive DE anomalies are limited to the northern MSF, but they have important magnitudes. Similarly, reductions in the water deficit are observed more intensively in the central-southern MSF and the USF. Previous studies showed that the changes in the Walker cell and the Hadley cell (associated with the ITCZ) during El Niño events affect precipitation in the northern portion of the NEB, which is further intensified by the positive AMM (ANDREOLI; KAYANO, 2006; DE SOUZA; AMBRIZZI, 2002; MEDEIROS, 2019). 120 The AAO_P scenario (southern hemisphere westerlies displaced further north) drives two effects over the SFW. On the one hand, the negative AAO phase is associated with the reduction in the SACZ activity or its displacement further north (CARVALHO; JONES; LIEBMANN, 2004; CAVALCANTI, 2012; ROSSO et al., 2018; ZILLI; CARVALHO; LINTNER, 2019). This effect can be perceived by observing the positive DE anomalies in the entire USF and the western MSF. On the other hand, the northernmost positioning of the westerlies strengthens the southern subtropical jet and displaces it further north (CARVALHO; JONES; AMBRIZZI, 2005; REBOITA; AMBRIZZI; DA ROCHA, 2009). These effects increase cyclonic activity in the northern portions of the southern hemisphere, which may favor the formation of the SACZ in a dislocated position or may intensify the propagation of FS towards the central-western Bahia state (mostly the MSF). Thus, negative DE anomalies can be observed in the eastern MSF extending towards the LMSF. Finally, the LN_INT scenario, which combines La Niña with the negative phase of the AAO, features the same spatial patterns observed in the AAO_P scenario. In this case, however, the pattern is more intense, but less consistent. It should be noted that this scenario is composed by only two years of data (1950 and 1970), which is a rather small sample to delineate a reliable pattern. The study by Carvalho, Jones and Ambrizzi (2005) had already shown that there is a relationship between La Niña and the negative phase of the AAO, although the extreme occurrence of these events will not always be associated with one another. Results shown in Figure 2.25 mostly agree with the expected responses in each subregion of the SFW based on previous regional, global and watershed scale studies. For example, Galvíncio and Sousa (2002) showed that the El Niño (La Niña) would be associated with negative (positive) precipitation anomalies in the LMSF and positive (negative) anomalies in the USF. However, said study was based on only one representative El Niño and La Niña year. The present study showed that, when considering multiple representative El Niño and La Niña years, this pattern indeed seems to be confirmed for the LMSF but for the USF the effects of the ENSO phase seem to be less consistent. Other studies have assessed the relationship between rainfall in the SFW (or its subregions) and the ENSO, finding similar results (GURJÃO et al., 2012; DOS SANTOS et al., 2011; SANTOS et al., 2019; SUN et al., 2016). Specific results related to the AMM and AAO influence over the SFW were previously reported by Paredes-Trejo et al. (2016), but focusing on the USF. Results reported in the present study corroborate with the study by these authors. It is worth mentioning that the present study is 121 unprecedent regarding the description of the average spatial pattern of the effects of different large-scale mechanisms that influence drought incidence over the SFW. 2.11. Conclusion The objective of this study was to characterize meteorological drought in the SFW, considering its intensity, frequency and extension. A critical discussion was carried out regarding the use of the SPEI for the identification of droughts in the semiarid and subtropical subregions of the basin, comparing it with the DE obtained through the TM-CWB. The water balance and the DE were then used to characterize drought in the SFW, also considering scenarios influenced by different large-scale circulation mechanisms. The comparison between the SPEI and the water balance showed that both indices retrieve similar results when applied to large time scales, such as 12-month periods. At this scale, the recurrent problem associated with the skewed rainfall distribution during dry months in semiarid and subtropical climate regions does not exist. However, for the monthly monitoring of drought conditions, the use of the DE was more advantageous because it better represented short-term (monthly scale) shifts in water availability conditions. Furthermore, DE is not affected by the skewed distribution of data in dry months, categorizing months based on the physical aspects of drought and not only deviations from the mean. On the other hand, using the DE requires caution, especially when comparing regions under the influence of different climate types since the results found will probably have discrepant magnitudes. Meteorological drought characterization in the SFW showed that the phenomenon is recurrent in the LMSF and in the surrounding areas of the LSF and the MSF. In these regions, high DE are frequent and water deficit conditions are observed in 10 to 11 months per year. Even in the wet season, rainfall is usually sufficient to recover AWC, but important water surpluses are rare. Drought in this region is also greatly intensified by the influence of El Niño, the occurrence of the positive AMM phase and the interaction between these two phenomena. Results indicated that there is an increasing trend for the expansion of drought-affected areas from the LMSF toward the MSF and the LSF. In these regions, water deficits are expected to become more frequent, while months of water surplus will occur more scarcely. The results of this study also showed that the USF is the subregion that have been suffering less changes in climatological drought patterns. No significant trends were identified for the expansion of areas affected by water deficits in this region, although the period from 122 2012 to 2016 was marked by the persistence of negative water balance conditions. In the USF, the only significant trend found regarded the increase in the number of months with water deficit per year, which might indicate a shortening of the wet season. Regarding the influence of large-scale mechanisms, results showed that the region is less affected than the remaining of the basin, but is susceptible to unfavorable drought conditions during La Niña and negative AAO episodes. 123 CHAPTER 3: CLIMATOLOGICAL DROUGHT ASSESSMENT OVER SMALL- SCALE SEMIARID WATERSHEDS: THE CASE OF THE PIRANHAS-AÇU BASIN The discussions addressed in the second chapter regarding methodological issues to be dealt with before actually characterizing drought over semiarid environments should always be considered when carrying out such studies. However, the SFW presented specific characteristics in the NEB regarding its spatial dimension and the availability of observational data. These specificities allowed the conduction of a detailed validation of alternative datasets and the use of methods based on the water balance. However, when studying smaller-scale local watersheds the overall scenario is slightly different. The already sparsely and heterogeneously distributed measuring network can become a serious limitation to the development of long-term spatial drought assessment based on the water balance. Therefore, in this third chapter we propose a different approach to the characterization of drought in the NEB, with a special focus on the particular characteristics of small-scale basins regarding data availability. Among the many potential small-scale water basins to be selected, we opted to study the Piranhas-Açu river watershed, which is entirely located in the Brazilian Semiarid and encompasses an area of 43,681 km² (approximately the size of the USF) in the Rio Grande do Norte (RN) and Paraíba (PB) states. Like many other semiarid basins in the NEB, the PAW comprises at least 52 dams with the capacity to store more than 10 hm³ of water, which are strategic to the water supply scheme of the region. Among these, the Coremas Mãe-d’Água reservoir in the Paraíba (PB) state and the Armando Ribeiro Gonçalves reservoir in the Rio Grande do Norte (RN) state account for approximately 70% of total available surface water in the PAW (ANA, 2014). These reservoirs assure water supply to part of the population in both states, and are crucial to the development of mining, aquaculture and irrigated agriculture in the region (ANA, 2014). In 2014, the gross domestic product of the basin was of approximately R$ 16.5 billion (U$ 7.7 billion at the currency of that time) which represented roughly 18% of total gross product of the RN and PB states that year (IBGE, 2016). Despite being entirely located in the Brazilian Semiarid and having rather small proportions, the PAW still presents some physical geographical aspects that cause some variability in its climate and landscapes (Figure 3.1). For example, the Borborema Plateau located at the southeastern PAW with altitudes varying from 800 to 1000 m is known to influence on the inhibition of rainfall over the adjacent depression (Depressão Sertaneja). Reboita et al. (2016) showed that, indeed, the Borborema Plateau orographic effect causes 124 subsidence over the Depressão Sertaneja, increasing its temperature and reducing relative humidity. Therefore, the central portions of the PAW are usually drier than its upper portions, which have higher altitude and therefore are slightly wetter (Figure 3.1). Figure 3.1 – Main geographical and climate features of the Piranhas-Açu watershed. Adapted from Mutti (2018) and Mutti et al. (2019). However, to this moment only a few relevant studies have been developed with the objective of detailing the climatological aspects of rainfall or drought over the PAW (FELIX, 2015; MUTTI et al., 2019). Thus, this chapter provides the first in-depth study on rainfall climatology over this watershed while also characterizing drought incidence and its relationship with large-scale atmospheric systems. In the case of the PAW, the lack of consistent, gapless, long-term meteorological data was addressed differently if compared to the study over the SFW presented in chapter 2. We now proposed the gap-filling of all available monthly time series with up to 30% of gaps. For 125 this end, we used three different gap-filling methods, selecting the best-performing ones for each missing data. Additionally, since temperature data in the region is even more scarce (only seven heterogeneously distributed measuring stations, as evidence by Mutti et al. 2019) than in the SFW, we used a drought index derived only from rainfall data: the mRAI. The framework proposed in this study also comprises the use of statistical tools such as principal component analysis and cluster analysis in order to identify potential differences in rainfall regimes within the basin. For example, the physical influence of geographical barriers on rainfall over the PAW creates different patterns within the base although it is entirely located in a semiarid region. Identifying the spatial limits of these patterns may evidence different specific drought characteristics in relation to the influence of large-scale atmospheric mechanisms, for example. We expect that the framework proposed in this chapter can be replicated and used in other watersheds within or without the NEB that present similar climate and data availability characteristics. 126 A detailed framework for the characterization of rainfall climatology in semiarid watersheds Article published in the Theoretical and Applied Climatology journal (BRAZILIAN QUALIS B1; SCIMAGO JR Q2; JCR: 2.882) and fully available at: https://link.springer.com/article/10.1007/s00704-019-02963-0 MUTTI, P. R.; DE ABREU, L. P.; ANDRADE, L. de M. B.; SPYRIDES, M. H. C.; LIMA, K. C.; DE OLIVEIRA, C. P.; DUBREUIL, V.; BEZERRA, B. G. A detailed framework for the characterization of rainfall climatology in semiarid watersheds. Theoretical and Applied Climatology, v. 139, n.-, p. 109-125, 2020. Doi: https://doi.org/10.1007/s00704-019- 02963-0. 127 CHAPTER 4: REMOTE SENSING MONITORING OF VEGETATION OVER DESERTIFICATION HOTSPOTS: ALTERNATIVE APPROACHES The discussions and results presented in Chapter 2 and Chapter 3 are of an undeniable importance since meteorological drought is the precursor event on drought propagation models. On the other hand, studying the effects of these events on the vegetation and hydrological systems is also needed. In vulnerable regions such as the NEB, both climate change and human activities are responsible for the intensification of land degradation and for exerting pressure over natural ecosystems. In this context, remote sensing techniques are invaluable tools for the monitoring of surface dynamics, providing evidences on the response of vegetated lands to climate and/or anthropic influence. To explore the use of remote sensing on the monitoring of vegetation response to drought over the NEB, we opted for a different spatial unit of analysis in this fourth chapter. Instead of watersheds, this study focuses on six desertification hotspots in the NEB: Irauçuba (IRA), Jaguaribe (JAG), Cabrobró (CAB), Inhamúns (INH), Seridó (SER) and Gilbués (GIL), as defined by the National Institute of the Semiarid. All these hotspots are entirely located in the Brazilian Semiarid region, except for GIL, which is only partially situated in it. Therefore, the main vegetation type on these hotspots is the Caatinga, with the exception of the GIL hotspot which is mainly covered by the Cerrado vegetation (Figure 4.1) These areas have common characteristics that classify them as desertification hotspots and which have been previously discussed in Chapter 1. They refer to areas with a rich native biodiversity that are under the influence of both harsh climate conditions and the expanding influence of human activities. Particularly, the hotspots located in the semiarid NEB have been impacted by the extraction of firewood, poor soil management practices, soil salinization and extensive livestock farming (TOMASELLA et al., 2018). The GIL hotspot, on the other hand, is located at what is currently considered one of the largest expanding agriculture frontiers in the world: the MATOPIBA region comprising the states of Maranhão (MA), Tocantins (TO), Piauí (PI) and Bahia (BA) (DOS REIS et al., 2020; REIS et al., 2020). There is a huge amount of drought indices derived from remote sensing, and even promptly available land-use change tools such as the MapBiomas (https://mapbiomas.org/) initiative, that can aid in the monitoring of such fragile areas. Nevertheless, specific studies on vegetation monitoring by remote sensing over these desertification hotspots are still scarce. The main recent contribution is the study by Tomasella et al. (2018), which proposed a NDVI-based 128 index to monitor bare soil areas in the NEB. The authors reported land degradation in the Brazilian Semiarid desertification hotspots is increasing due to the occurrence of persistent drought events, particularly the 2012-2016 drought. Figure 4.1 – Land cover classification and location of Brazilian desertification hotspots. The preliminary study by Mutti and Bezerra (2018) also provided initial evidence regarding an increase in land degradation over the aforementioned Brazilian desertification hotspots. These authors detected significant negative trends in NDVI over almost the entire extension of all hotspots over the period from 2000 to 2018 (Figure 4.2). Curiously, the analysis also showed positive NDVI trends in some regions of the CAB hotspot, which is located in the SFW. These trends are associated with the development of irrigated districts near the São Francisco river. The GIL region seems to be the most intensely affected hotspot, with negative trends of up to -0.3 (given that NDVI ranges from -1 to 1) in several portions of its area. 129 Figure 4.2 – Magnitude of trends in NDVI (2000-2018) over six desertification hotspots in Brazil: Irauçuba – IRA; Jaguaribe – JAG; Cabrobró – CAB; Seridó – SER; Inhamúns – INH; and Gilbués – GIL. Figure extracted from Mutti and Bezerra (2018). All these results point towards the importance of studies specifically aimed at these vulnerable regions. Despite the wide variety of available tools, remote sensing monitoring demands high computational effort and thus the use of alternative methods regarding this type of data should be encouraged. In this chapter, we explore the use of stochastic models for the forecasting of mean NDVI and NDVI variance over the six aforementioned desertification hotspots. Despite losing the spatial character of pixel-wise models, this approach allows the application of mean NDVI and NDVI variance time series in mean-variance plots. These plots provide simple but useful information on the seasonal vegetation state by indicating the greenness and heterogeneity level of the vegetation at each seasonal step. When testing the models, we also considered the influence of rainfall as exogenous predictive variable, since there is a clear relationship between these variables (BARBOSA; KUMAR, 2016). 130 NDVI time series stochastic models for the forecast of vegetation dynamics over desertification hotspots Article published in the International Journal of Remote Sensing journal (BRAZILIAN QUALIS B1; SCIMAGO JR Q1; JCR: 2.976) and fully available at: https://www.tandfonline.com/doi/full/10.1080/01431161.2019.1697008 MUTTI, P. R.; LÚCIO, P. S.; DUBREUIL, V.; BEZERRA, B. G. NDVI time series stochastic models for the forecast of vegetation dynamics over desertification hotspots. International Journal of Remote Sensing, v. 41, n. 7, p. 2759-2788, 2020. Doi: https://doi.org/10.1080/01431161.2019.1697008. 131 FINAL CONSIDERATIONS Drought is a vastly studied subject, since it is the natural disaster that impacts most people worldwide. In Brazil, the Northeast region is historically afflicted by severe drought events, which amplify the already harsh and dry conditions typical of the semiarid environment that encompasses most of its lands. Because of this, there is a wide variety of drought studies in the NEB, aimed at identifying and characterizing its meteorological, hydrological, surface and socioeconomic aspects. Nevertheless, there is still much too be discussed and improved in terms of drought assessment regarding methodological aspects, finer spatial scale results and alternatives to the monitoring of surface conditions based on remote sensing data. Therefore, the objective of this thesis was to provide a set of specific studies aimed at discussing some of these elements, contributing to the overall understanding of drought dynamics in the NEB. At a first moment, the climatological aspects of drought were assessed in two watersheds with different scales and with different climate characteristics: the SFW, which is a complex, large-scale basin under the influence of multiple climates (Chapter 2); and the PAW, which is a small-scale basin mostly inserted in the semiarid portion of the NEB (Chapter 3). The results found in the proposed studies largely agrees with recent findings and with studies developed at the regional scale, also revealing the potential effects of ongoing climate change. In the SFW, drought assessment over 75 years evidenced an apparent expansion of aridity from the semiarid portion of the basin towards the coast and the southern NEB. This result corroborates an important conclusion that has been hinted by previous studies at the regional scale: the Brazilian Semiarid is expanding, with water deficits becoming increasingly more frequent in surrounding regions that are usually wetter and cooler. At the even smaller scale of the PAW, this result was further corroborated. Significant trends over 54 years were found indicating an increase in the occurrence of negative rainfall anomalies mostly over the upper portions of the basin, located in higher altitude regions which are historically wetter. Additionally, the study on the SFW is unprecedent regarding the characterization of the spatial patterns of drought incidence when considering teleconnection patterns with different large-scale atmospheric circulation systems. While the dynamic and atmospheric links of the ENSO, the AMM and the AAO on rainfall over the NEB (and the SFW) are well known, it is the first time their combined and individual effects on drought incidence was assessed in the SFW. Similarly, we also identified the probability of occurrence of negative or positive rainfall anomalies over different portions of the PAW based on the observed ENSO and AMM phase. 132 For example, the combined effect of the El Niño with the occurrence of warmer SST over the Northern Tropical Atlantic Ocean leads to a drastic intensification of water deficits over the northern SFW and the NEB. However, this configuration also provides an increase in rainfall and water surplus over the southern NEB and the Southeast Brazil. At the watershed scale, this means that even if dry conditions are intensified in the semiarid lower portions of the São Francisco basin, its upper portions (near the river source) may receive copious amounts of rainfall, inducing the recharge of surface waters and maintaining the balance of the hydrological system. These results are of the uttermost importance to the actors involved in developing water management policies at the watershed scale, especially in such hydrological systems marked by water conflicts. Apart from the results obtained through the meteorological drought assessment in both watersheds, important methodological aspects were also discussed. For instance, the lack of consistent, gapless, long-term meteorological data in the NEB is a well-known issue. As such, it is difficult to develop long-term studies on the climatological aspects of drought over this region without relying on the adoption of several assumptions. Therefore, we proposed two different approaches to deal with this issue: a thorough validation of promptly available gridded datasets (Chapter 2) and a comprehensive methodology to fill observational data time series with up to 30% of gaps (Chapter 3). In the large-scale SFW, we opted to use the available network of measuring stations to validate CRU TS monthly rainfall and PET interpolated data (~ 56 km of spatial resolution). 171 rain gauges and 57 meteorological stations were used to validate the gridded dataset over 75 years (1942-2016). We found that CRU TS data generally correlates well with observed data. However, when used to detect trends in the assessed variables, CRU TS data retrieves much smoother slopes than observational data, largely due to the inevitable smoothing process inherent to interpolation techniques. It is important to note that this characteristic is probably shared by most of the other available gridded products for meteorological data. This means that studies using similar datasets as alternatives to the lack of high-quality observational data in the NEB should always consider the potential effects of smoothing. In fact, results obtained in this thesis (Chapter 2A) indicate that trends detected in observational data are more significant and steeper than the ones detected using the interpolated dataset. In the PAW a different approach was proposed. In smaller scale basins the amount of available data may not be enough to locally validate long-term gridded or satellite-derived 133 products. Thus, we opted to select all 56 available measuring stations with up to 30% of gaps, which led to a total study period of 54 years (1962-2015). In our study we filled all gaps using three different methods. For each gap filled we derived a statistical indicator of the quality of the method and kept the estimation obtained with the method presenting the best statistical indicator. Such process improved the quality of the gap-filled estimates from 3 to 20% when compared to using only one method. Both approaches showed that even when consistent, homogeneously distributed data is unavailable, reliable long-term time series can be derived if data quality and control is carried out rigorously. Results also showed that, by doing so, comprehensive meteorological drought characterizations can be obtained in both large-scale and small-scale watersheds. Another important issue discussed in this thesis regards the use of standardized drought indices (such as the SPEI and the SPI) over semiarid, tropical or subtropical regions where the dry season is marked by months with the recurrence of zero precipitation values. Despite being thoroughly used on drought studies over these regions (such as the NEB), assessment is usually carried out at 6-month or 12-month scales, when the zero-inflation problem ceases to be an issue. We then compared the SPEI with the DE, which is based on a simplified water balance (difference between precipitation and PET), on a 12-month scale but also at the monthly scale and over a semiarid region and a subtropical humid region in the SFW (Chapter 2). Our results showed that both indices correlate well when identifying persistent droughts at the 12-month scale. However, at the monthly scale, using the DE is much more advantageous than the standardized index. The standardization process is hindered by the skewed distribution of rainfall data typical of dry months, and therefore the SPEI monthly classification based on deviations retrieve results that are far from the physical observed reality. Therefore, the DE is not only perfectly capable of identifying persistent long-term droughts, but also excels in identifying short-term monthly shifts in drought patterns over such regions, while also maintaining a physical meaning to its interpretation. This discussion is extremely important for the selection of the correct methodological approaches when considering different indices for the monitoring of meteorological drought over the NEB. Apart from the meteorological aspects of drought, this thesis also presented an alternative approach for the monitoring of vegetation states in desertification hotspots in the NEB by using remote sensing data (Chapter 4). While the assessment of satellite surface data time series has been thoroughly used to monitor land use change and vegetation response to 134 drought in the NEB, they usually demand high computational effort and processing capacities. While some research centers worldwide (and even in Brazil) might have access to the required infrastructure, the development of alternative simpler and efficient techniques should be encouraged in order to decentralize monitoring studies using remote sensing data. In the fourth chapter of this thesis, we derived mean NDVI and NDVI variance time series (2000-2018) over six desertification hotspots in the NEB: Cabrobró, Jaguaribe, Inhamúns, Irauçuba, Seridó and Gilbués. Data from the MOD13A2 product (1 km spatial resolution) of the MODIS sensor aboard the Terra satellite were used. Different stochastic models were tested in the forecasting of these variables, and performed quite satisfactorily for that end. We then applied the forecasted values in a mean-variance plot, which is a simple yet efficient method to visualize the evolution of vegetation state throughout time. Results showed that the chosen models performed better in forecasting dry and degraded vegetation states than in forecasting robust heterogeneous vegetation (typical of the Caatinga during wet periods). While pixel-wise models have a clear advantage regarding spatial monitoring, the alternative method tested in this thesis may also be of valuable use when the required infrastructure and processing capacities are not available. Regarding future perspectives, the conclusions drawn from the study presented in chapter four lead directly to a new research question: how would the tested models perform if applied to pixel-wise NDVI time series? Despite the inevitable required computational effort, this could lead to the development of an interesting tool for the forecasting and monitoring of vegetation states and responses to drought. Similarly, the meteorological studies performed in this thesis (Chapter 2 and 3) were carried out considering monthly time series of data. Exploring data at finer time scales, such as weekly or daily time scales, may lead to new interesting approaches to drought characterization, with the possibility of identifying changes in spatial and temporal aspects of the incidence of consecutive dry days, for example. On the other hand, such new perspectives would also be associated with additional difficulties regarding the quality of the available data for that end. Perspectives for future researches deriving from this thesis are also associated with the theoretical background presented in Chapter 1. It was seen that the most recent drought definitions situate mankind not only as passive targets of the effects of these phenomena, but as an active actor. Therefore, a next important step to drought studies in the NEB should contemplate the direct and indirect effects of human activities on drought propagation. For 135 example, given a certain drought event that greatly impacted vegetation over an area at risk of desertification, how can we account for the damage that could have been avoided if human activity had not previously contributed to the degradation of that land? It is indeed a complex question, but a question that needs to be posed because knowing and reporting its answer may promote public awareness and drive political forces to push for more rigid laws for the preservation and recovery of vulnerable ecosystems such as the Caatinga. Pushing this concept even further, one can also reflect upon different climate change scenarios. How will drought incidence and intensification in the NEB behave in future scenarios with the expansion of agricultural lands for example? Or to which magnitude can we reduce the future impacts of drought if immediate action is taken and inclusive water policies such as the One Million Cistern Programs are implemented at even larger scales in the NEB? This directly leads to another important issue discussed in Chapter 1, which refers to the interrelation between physical and socioeconomic drought aspects. Although comprehensive physical drought studies such as this thesis are of undeniable importance, it lacks integration with more practical social aspects that are crucial to the region. Future studies and discussions should explore this complementary relation between the environmental and socioeconomic spheres. Furthermore, the conceptual drought propagation models studied in Chapter 1 also highlight the importance of studies based on global, systematic approaches to drought. In other words, studies that evaluate not only meteorological drought, but its direct relation to the subsequent impacts on vegetation, the hydrological system and the socioeconomic aspects of a given location. In this case, watershed studies including drought assessment in all these compartments and the characterization of their relationships should be strongly encouraged. Indeed, studying drought in the Anthropocene is the next great challenge regarding this phenomenon, and the NEB will remain as one of the key global regions to the development of new innovative studies regarding past, present and future drought incidence patterns. 136 REFERENCES ABRAMOWITZ, M.; STEGUN, I. Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. 10. ed. Washington, DC, USA: United States Department of Commerce, 1972. ADAMS, J. The desert makes the desert: Climate feedbacks from the vegetation of arid zones. In: Vegetation-Climate Interaction: how vegetation makes the global environment. Chichester, UK: Springer-Praxis, 2007. p. 4–274. ADEYEWA, Z. D.; NAKAMURA, K. Validation of TRMM radar rainfall data over major climatic regions in Africa. Journal of Applied Meteorology, [s. l.], v. 42, n. 2, p. 331– 347, 2003. AHMED, K. et al. Evaluation of gridded precipitation datasets over arid regions of Pakistan. Water, [s. l.], v. 11, n. 210, p. 1–22, 2019. ALEMAYEHU, T.; VAN GRIENSVEN, A.; BAUWENS, W. Evaluating CFSR and WATCH data as input to SWAT for the estimation of the potential evapotranspiration in a data- scarce Eastern-African catchment. Journal of Hydrologic Engineering, [s. l.], v. 21, n. 3, p. 1–16, 2016. ALEXANDERSSON, H. A homogeneity test applied to precipitation data. Journal of Climatology, [s. l.], v. 6, n. 6, p. 661–675, 1986. ALHAMAD, M. N.; STUTH, J.; VANNUCCI, M. Biophysical modelling and NDVI time series to project near-term forage supply: Spectral analysis aided by wavelet denoising and ARIMA modelling. International Journal of Remote Sensing, [s. l.], v. 28, n. 11, p. 2513–2548, 2007. ALMAZROUI, M. et al. Principal components-based regionalization of the Saudi Arabian climate. International Journal of Climatology, [s. l.], v. 35, n. 9, p. 2555–2573, 2015. ALVARES, C. A. et al. Köppen’s climate classification map for Brazil. Meteorologische Zeitschrift, [s. l.], v. 22, n. 6, p. 711–728, 2014. ANA. Plano de Recursos Hídricos da Bacia Hidrográfica do Rio Piranhas-Açu. Brasília, DF, Brasil. Disponível em: . ANDERBERG, M. R. Cluster Analysis for Applications. 1ed. ed. [s.l.] : Academic Press, 1973. ANDERSON, M. C. et al. The Evaporative Stress Index as an indicator of agricultural drought in Brazil: An assessment based on crop yield impacts. Remote Sensing of Environment, [s. l.], v. 174, p. 82–99, 2016. Disponível em: ANDRADE E SANTOS, H. De; POMPEU, P. dos S.; KENJI, D. O. L. Changes in the flood regime of São Francisco River (Brazil) from 1940 to 2006. Regional Environmental Change, [s. l.], v. 12, n. 1, p. 123–132, 2012. 137 ANDRADE, K. M.; CAVALCANTI, I. F. A. Atmospheric characteristics that induce extreme precipitation in frontal systems over Southeastern Brazil during summer: Observations and atmospheric model simulation. International Journal of Climatology, [s. l.], v. 38, n. 14, p. 5368–5385, 2018. ANDREOLI, R. V.; KAYANO, M. T. Tropical Pacific and South Atlantic Effects on rainfall variability over Northeast Brazil. International Journal of Climatology, [s. l.], v. 26, p. 1895–1912, 2006. ARELLANO-SOTA, C. et al. UNEP - Program on desertification control in Latin America and the Caribbean. Desertification control bulletin, Nairobi, v. 29, p. 56–62, 1996. ARORA, N. K. Impact of climate change on agriculture production and its sustainable solutions. Environmental Sustainability, [s. l.], v. 2, n. 2, p. 95–96, 2019. Disponível em: ARVOR, D. et al. Classification of MODIS EVI time series for crop mapping in the state of Mato Grosso, Brazil. International Journal of Remote Sensing, [s. l.], v. 32, n. 22, p. 7847–7871, 2011. Disponível em: ARVOR, D. et al. Land use sustainability on the South-Eastern Amazon agricultural frontier: Recent progress and the challenges ahead. Applied Geography, [s. l.], v. 80, p. 86–97, 2017. Disponível em: AVILA-DIAZ, A. et al. Assessing current and future trends of climate extremes across Brazil based on reanalyses and earth system model projections. Climate Dynamics, [s. l.], v. 55, n. 5–6, p. 1403–1426, 2020. Disponível em: BARBOSA, H. A. et al. Assessment of Caatinga response to drought using Meteosat-SEVIRI Normalized Difference Vegetation Index (2008–2016). ISPRS Journal of Photogrammetry and Remote Sensing, [s. l.], v. 148, n. December 2018, p. 235–252, 2019. Disponível em: BARBOSA, H. A.; HUETE, A. R.; BAETHGEN, W. E. A 20-year study of NDVI variability over the Northeast Region of Brazil. Journal of Arid Environments, [s. l.], v. 67, n. 2, p. 288–307, 2006. BARBOSA, H. A.; LAKSHMI KUMAR, T. V. Influence of rainfall variability on the vegetation dynamics over Northeastern Brazil. Journal of Arid Environments, [s. l.], v. 124, p. 377–387, 2016. Disponível em: BARKER, L. J. et al. From meteorological to hydrological drought using standardised indicators. Hydrology and Earth System Sciences, [s. l.], v. 20, n. 6, p. 2483–2505, 2016. BARRODALE, I.; ROBERTS, F. D. K. An Improved Algorithm for Discrete L1 Linear Approximation. SIAM Journal, [s. l.], v. 10, n. 5, p. 839–848, 1973. BARROS, E. de S. et al. Eficiência na produção agrícola do Vale São Francisco: mensuração de escores e análise de fatores correlacionados. Rev. de Economia Agrícola, [s. l.], v. 63, n. 2, p. 35–50, 2016. 138 BATALHA, M. A.; MANTOVANI, W. Reproducive Phenological Patterns of Cerrado Plant Species At the Pé-De-Gigante Reserve ( Santa Rita Do Passa Quatro , Sp , Brazil ): a Comparision Between the Herbaceous and Woody Floras. Revista Brasileira de Biologia, [s. l.], v. 60, n. 1, p. 129–145, 2000. BEDRAN-MARTINS, A. M.; LEMOS, M. C. Politics of drought under Bolsa Família program in Northeast Brazil. World Development Perspectives, [s. l.], v. 7–8, n. March, p. 15–21, 2017. Disponível em: BEGUERÍA, S. et al. Standardized precipitation evapotranspiration index (SPEI) revisited: Parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. International Journal of Climatology, [s. l.], v. 34, n. 10, p. 3001–3023, 2014. BERMÚDEZ, J. D.; SEGURA, J. V.; VERCHER, E. A decision support system methodology for forecasting of time series based on soft computing. Computational Statistics and Data Analysis, [s. l.], v. 51, n. 1, p. 177–191, 2006. BESTELMEYER, B. T. et al. Desertification, land use, and the transformation of global drylands. Frontiers in Ecology and the Environment, [s. l.], v. 13, n. 1, p. 28–36, 2015. BEUCHLE, R. et al. Land cover changes in the Brazilian Cerrado and Caatinga biomes from 1990 to 2010 based on a systematic remote sensing sampling approach. Applied Geography, [s. l.], v. 58, p. 116–127, 2015. BEZERRA, B. G. et al. Changes of precipitation extremes indices in São Francisco River Basin, Brazil from 1947 to 2012. Theoretical and Applied Climatology, [s. l.], v. 135, n. 1–2, p. 565–576, 2019. BEZERRA, B. G.; FIDELES FILHO, J. Growth analysis of a cotton crop irrigated with wastewater. Revista de Ciências Agronômicas, [s. l.], v. 40, n. 3, p. 339–345, 2009. BOX, G. E. P.; JENKINS, G. M. Time series analysis: forecasting and control. San Francisco, CA: Holden-Day, 1970. BOX, G. E. P.; TIAO, G. C. Intervention analysis with applications to economic and environmental problems. Journal of the American Statistical Association, [s. l.], v. 70, n. 349, p. 70–79, 1975. BRASIL. Decreto-Lei no 9.857, de 13 de setembro de 1946. 1946. Disponível em: . Acesso em: 12 fev. 2020. BRASIL. Lei no 7.827, de 27 de setembro de 1989. 1989. Disponível em: . Acesso em: 12 fev. 2020. BRASIL. Lei no 9.433, de 8 de janeiro de 1997. 1997. Disponível em: . Acesso em: 15 maio. 2017. BRASIL. Resolução no 115, de 23 de novembro de 2017. 2017. Disponível em: . Acesso em: 12 fev. 2020. 139 BRAVAR, L.; KAVVAS, M. L. On the physics of droughts. I. A conceptual framework. Journal of Hydrology, [s. l.], v. 129, n. 1–4, p. 281–297, 1991. BRITO, S. S. B. et al. Frequency, duration and severity of drought in the Semiarid Northeast Brazil region. International Journal of Climatology, [s. l.], v. 38, n. 2, p. 517–529, 2018. BURITI, C. de O.; BARBOSA, H. A. Um século de secas: Por que as políticas hídricas não transformaram o Semiárido Brasileiro? 1. ed. Portugal: Editora Chiado, 2018. BYUN, H.-R.; WILHITE, D. A. Daily Quantification of Drought Severity and Duration. Journal of Climate, [s. l.], v. 12, n. September, p. 2747–2756, 1999. CABRAL JÚNIOR, J. B. et al. Detecting linear trend of reference evapotranspiration in irrigated farming areas in Brazil’s semiarid region. Theoretical and Applied Climatology, [s. l.], v. 138, n. 1–2, p. 215–225, 2019. CAMACHO, G. P.; VASCONCELOS, H. L. Ants of the Panga Ecological Station, a Cerrado reserve in Central Brazil. Sociobiology, [s. l.], v. 62, n. 2, p. 281–295, 2015. CAMELO, H. do N. et al. Innovative hybrid models for forecasting time series applied in wind generation based on the combination of time series models with artificial neural networks. Energy, [s. l.], v. 151, n. June 2017, p. 347–357, 2018. CAMPOS, S. et al. Closure and partitioning of the energy balance in a preserved area of a Brazilian seasonally dry tropical forest. Agricultural and Forest Meteorology, [s. l.], v. 271, n. March, p. 398–412, 2019. Disponível em: CARVALHO, L. M. V.; JONES, C.; AMBRIZZI, T. Opposite Phases of the Antarctic Oscillation and Relationships with Intraseasonal to Interannual Activity in the Tropics during the Austral Summer. Journal of Climate, [s. l.], v. 18, n. March, p. 702–718, 2005. CARVALHO, L. M. V.; JONES, C.; LIEBMANN, B. The South Atlantic convergence zone: Intensity, form, persistence, and relationships with intraseasonal to interannual activity and extreme rainfall. Journal of Climate, [s. l.], v. 17, n. 1, p. 88–108, 2004. CAVALCANTI, I. F. A. Large scale and synoptic features associated with extreme precipitation over South America: A review and case studies for the first decade of the 21st century. Atmospheric Research, [s. l.], v. 118, p. 27–40, 2012. Disponível em: CBHSF. Resumo Executivo do Plano de Recursos Hídricos da Bacia Hidrográfica do Rio São Francisco 2016-2025. Maceió, Alagoas, BR: Comitê da Bacia Hidrográfica do Rio São Francisco, 2016. Disponível em: CERONI, M. et al. Estudio del NDVI mediante análisis multiescalar y series temporales utilizando imágenes SPOT, durante el período 1998-2012 en el Uruguay. Revista de Teledeteccion, [s. l.], n. 43, p. 31–42, 2015. CHAPMAN, A. D. Quality Control and Validation of Point-sourced Environmental Resource 140 Data. In: LOWELL, K.; JATON, A. (Eds.). Spatial Accuracy Assessment: Land Information Uncertainty in Natural Resources. 1. ed. Chelsea, Michigan, USA: Sleeping Bear Press, Inc., 1999. p. 409. CHARRAD, M. et al. NbClust: An R Package for Determining the Relevant Number of Clusters in a Data Set. Journal of Statistical Software, [s. l.], v. 61, n. 6, p. 1–36, 2014. CHAVES, M. E. D.; MATAVELI, G. A. V.; JUSTINO, R. C. Uso da modelagem estatística para monitoramento da vegetação no Parque Nacional da Serra da Canastra, Minas Gerais. Caderno de Geografia, [s. l.], v. 24, n. 1, p. 120–132, 2014. CHEN, F. W.; LIU, C. W. Estimation of the spatial rainfall distribution using inverse distance weighting (IDW) in the middle of Taiwan. Paddy and Water Environment, [s. l.], v. 10, n. 3, p. 209–222, 2012. CHEN, M. et al. Assessing objective techniques for gauge-based analyses of global daily precipitation. Journal of Geophysical Research Atmospheres, [s. l.], v. 113, n. 4, p. 1– 13, 2008. CHEN, Y. et al. Mapping croplands, cropping patterns, and crop types using MODIS time- series data. International Journal of Applied Earth Observation and Geoinformation, [s. l.], v. 69, n. March, p. 133–147, 2018. Disponível em: CHIANG, J. C. H.; VIMONT, D. J. Analogous Pacific and Atlantic meridional modes of tropical atmosphere-ocean variability. Journal of Climate, [s. l.], v. 17, n. 21, p. 4143– 4158, 2004. CIAIS, P. et al. Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature, [s. l.], v. 437, n. 7058, p. 529–533, 2005. CLARK, C. How dry is a drought ? Crossosoma, [s. l.], v. 19, n. 2, p. 37–48, 1993. COPPIN, P. et al. Digital change detection methods in ecosystem monitoring: A review. International Journal of Remote Sensing, [s. l.], v. 25, n. 9, p. 1565–1596, 2004. CORREIA FILHO, W. L. F.; LUCIO, P. S.; SPYRIDES, M. H. C. Caracterização dos extremos de precipitação diária no nordeste do Brasil. Boletim Goiano de Geografia, [s. l.], v. 36, n. 3, p. 539–554, 2016. COSTA, A. R. S. et al. Desertificatoin in semiarid northeast of Brazil. Revista GEAMA, [s. l.], v. 2, n. 4, p. 427–445, 2016. COSTA, D. D. et al. Understanding Drought Dynamics during Dry Season in Eastern Northeast Brazil. Frontiers in Earth Science, [s. l.], v. 4, n. 69, p. 1–11, 2016. COSTA, R. L. et al. Analysis of climate extremes indices over northeast Brazil from 1961 to 2014. Weather and Climate Extremes, [s. l.], v. 28, n. 100254, 2020. COSTA, T. C. e C. Da et al. Phytomass mapping of the “seridó caatinga” vegetation by the plant area and the normalized difference vegetation indeces. Scientia Agricola, [s. l.], v. 59, n. 4, p. 707–715, 2002. Disponível em: 141 CUI, X. et al. Using Remote Sensing to Quantify Vegetation Change and Ecological Resilience in a Semi-Arid System. Land, [s. l.], v. 2, n. 2, p. 108–130, 2013. Disponível em: CUNHA, A. P. M. A. et al. Changes in the spatial–temporal patterns of droughts in the Brazilian Northeast. Atmospheric Science Letters, [s. l.], v. 19, n. 10, p. 1–8, 2018. CUNHA, A. P. M. A. et al. Extreme Drought Events over Brazil from 2011 to 2019. Atmosphere, [s. l.], v. 10, n. 642, p. 1–20, 2019. Disponível em: CUNHA, A. P. M. A. et al. Impacts of land use and land cover changes on the climate over Northeast Brazil. Atmospheric Science Letters, [s. l.], v. 16, n. 3, p. 219–227, 2015. Disponível em: CUNHA, A. P. M. et al. Monitoring vegetative drought dynamics in the Brazilian semiarid region. Agricultural and Forest Meteorology, [s. l.], v. 214–215, p. 494–505, 2015. Disponível em: . Acesso em: 11 set. 2018. D’ODORICO, P. et al. Global desertification: Drivers and feedbacks. Advances in Water Resources, [s. l.], v. 51, p. 326–344, 2013. Disponível em: DA FRANCA, R. R.; MENDONÇA, F. de A. A pluviosidade na Amazônia meridional: variabilidade e teleconexões extra-regionais. Confins, [s. l.], n. 29, p. 1–11, 2016. DA SILVA, D. F.; DE SOUSA, F. de A. S.; KAYANO, M. T. Uso de IAC e ondeletas para análise da influência das multi-escalas temporais na precipitação da bacia do Rio Mundaú. Engenharia Ambiental - Espírito Santo do Pinhal, [s. l.], v. 6, n. 1, p. 180–195, 2009. DA SILVA, P. E. et al. Precipitation and air temperature extremes in the Amazon and northeast Brazil. International Journal of Climatology, [s. l.], n. January, p. 1–17, 2018. DA SILVA, R. M. et al. Rainfall and river flow trends using Mann – Kendall and Sen’s slope estimator statistical tests in the Cobres River basin. Natural Hazards, [s. l.], v. 77, p. 1205–1221, 2015. DAULTREY, S. Principal components analysis. Concepts and techniques in modern geography, [s. l.], v. 930, p. 527–47, 1976. Disponível em: DE AMORIM, A. L.; RIBEIRO, M. M. R.; BRAGA, C. F. C. Conflitos em bacias hidrográficas compartilhadas: o caso da bacia do rio Piranhas-Açu/PB-RN. Revista Brasileira de Recursos Hídricos, [s. l.], v. 21, n. 1, p. 36–45, 2016. DE ANDRADE, E. M. et al. Uncertainties of the rainfall regime in a tropical semi-arid region: the case of the State of Ceará. Revista Agro@mbiente On-line, [s. l.], v. 10, n. 2, p. 88– 142 95, 2016. DE ARAÚJO FILHO, J. A. Manejo Pastoril Sustentável da Caatinga. 1. ed. Recife, PE: Projeto Dom Helder Camara, IICA, 2013. DE ARAÚJO, M. L. S. et al. Spatiotemporal dynamics of soybean crop in the Matopiba region, Brazil (1990–2015). Land Use Policy, [s. l.], v. 80, n. October 2018, p. 57–67, 2019. DE ASSIS, J. M. O. et al. Scenarios of projection of water availability in the stretch sub-middle São Francisco river basin. In: XXII SIMPÓSIO BRASILEIRO DE RECURSOS HÍDRICOS 2017, Anais... [s.l: s.n.] DE JONG, P. et al. Hydroelectric production from Brazil’s São Francisco River could cease due to climate change and inter-annual variability. Science of the Total Environment, [s. l.], v. 634, p. 1540–1553, 2018. Disponível em: DE JONG, R. et al. Analysis of monotonic greening and browning trends from global NDVI time-series. Remote Sensing of Environment, [s. l.], v. 115, n. 2, p. 692–702, 2011. Disponível em: DE MEDEIROS, F. J. et al. Numerical simulation of the circulation and tropical teleconnection mechanisms of a severe drought event (2012–2016) in Northeastern Brazil. Climate Dynamics, [s. l.], v. 54, n. 9–10, p. 4043–4057, 2020. Disponível em: DE MEDEIROS, F. J.; DE OLIVEIRA, C. P.; TORRES, R. R. Climatic aspects and vertical structure circulation associated with the severe drought in Northeast Brazil (2012–2016). Climate Dynamics, [s. l.], v. 16, p. 1–15, 2020. Disponível em: DE MORAES, M. M. G. A. et al. The impact of global change on economic values of water for Public Irrigation Schemes at the São Francisco River Basin in Brazil. Regional Environmental Change, [s. l.], v. 18, n. 7, p. 1943–1955, 2018. DE OLIVEIRA, P. T.; LIMA, K. C.; SANTOS E SILVA, C. M. Synoptic environment associated with heavy rainfall events on the coastland of Northeast Brazil. Advances in Geosciences, [s. l.], v. 35, p. 73–78, 2013. DE OLIVEIRA, P. T.; SANTOS E SILVA, C. M.; LIMA, K. C. Climatology and trend analysis of extreme precipitation in subregions of Northeast Brazil. Theoretical and Applied Climatology, [s. l.], v. 130, n. 1–2, p. 77–90, 2017. DE OLIVEIRA, P. T.; SANTOS E SILVA, C. M.; LIMA, K. C. Linear trend of occurrence and intensity of heavy rainfall events on Northeast Brazil. Atmospheric Science Letters, [s. l.], v. 15, n. 3, p. 172–177, 2014. DE SOUZA, D. C.; OYAMA, M. D. Climatic consequences of gradual desertification in the semi-arid area of Northeast Brazil. Theoretical and Applied Climatology, [s. l.], v. 103, n. 3–4, p. 345–357, 2011. DE SOUZA, E. B.; AMBRIZZI, T. ENSO impacts on the South American rainfall during 143 1980s: Hadley and Walker circulation. Atmósfera, [s. l.], v. 15, p. 105–120, 2002. DIAS, R. D. B. Tecnologia social e desenvolvimento local : reflexões a partir da análise do Programa Um Milhão de Cisternas. Revista Brasileira de Desenvolvimento Regional, [s. l.], v. 1, n. 2, p. 173–189, 2013. DIDAN, K. MOD13A2 MODIS/Terra Vegetation Indices 16-Day L3 Global 1km SIN Grid V006. 2015. Disponível em: . DOMBROSKI, J. L. D. et al. Water relations of Caatinga trees in the dry season. South African Journal of Botany, [s. l.], v. 77, n. 2, p. 430–434, 2011. Disponível em: DOMONKOS, P. Measuring performances of homogenization methods. Quaterly Journal of the Hungarian Meteorological Service, [s. l.], v. 117, n. 1, p. 91–112, 2013. DOS REIS, L. C. et al. Analysis of Climate Extreme Indices in the MATOPIBA Region, Brazil. Pure and Applied Geophysics, [s. l.], p. 1–22, 2020. DOS SANTOS, D. P. et al. Dinâmica de vegetação, balanço climatológico, índices climáticos, na sub-bacia do Rio São Francisco. Sustentare, [s. l.], v. Edição Esp, n. II SBHSF, p. 44– 55, 2018. DOS SANTOS, E. A. B. et al. Application of Markov chains to Standardized Precipitation Index (SPI) in São Francisco River Basin. Revista Ambiente & Água, [s. l.], v. 14, n. 3, p. 1–15, 2019. DOS SANTOS, E. P. et al. Eventos Extremos de Chuva e Alterações no Regime Hidrológico da Bacia Hidrográfica do Rio São Francisco: Uma Aplicação do Índice RAI (Rainfall Anomaly Index). Engenharia Ambiental - Espirito Santo do Pinhal, [s. l.], v. 8, n. 2, p. 315–330, 2011. DRACUP, J. A.; LEE, K. S.; PAULSON JR., E. G. On the Definition of Droughts. Water Resources Research, [s. l.], v. 16, n. 2, p. 297–302, 1980. Disponível em: DUBREUIL, V. et al. Climate change evidence in Brazil from Köppen’s climate annual types frequency. International Journal of Climatology, [s. l.], v. 39, n. 3, p. 1446–1456, 2019. DUBREUIL, V. La sécheresse dans la France de l’Ouest: une contrainte climatique trop souvent oubliée. Sécheresse, [s. l.], v. 8, n. 1, p. 47–55, 1997. DUBREUIL, V. Synthèse géographique de la sécheresse dans les régions océaniques. La Météorologie, [s. l.], v. 8, n. 15, p. 22–34, 1996. Disponível em: ECKERT, S. et al. Trend analysis of MODIS NDVI time series for detecting land degradation and regeneration in Mongolia. Journal of Arid Environments, [s. l.], v. 113, p. 16–28, 2015. Disponível em: EISCHEID, J. K. et al. Creating a Serially Complete, National Daily Time Series of Temperature and Precipitation for the Western United States. Journal of Applied 144 Meteorology, [s. l.], v. 39, n. 9, p. 1580–1591, 2000. EKLUNDH, L.; PILESJÖ, P. Regionalization and spatial estimation of ethiopian mean annual rainfall. International Journal of Climatology, [s. l.], v. 10, n. 5, p. 473–494, 1990. ERASMI, S. et al. Vegetation greenness in northeastern brazil and its relation to ENSO warm events. Remote Sensing, [s. l.], v. 6, n. 4, p. 3041–3058, 2014. ESFAHANIAN, E. et al. Development and evaluation of a comprehensive drought index. Journal of Environmental Management, [s. l.], v. 185, p. 31–43, 2017. Disponível em: FAZEL, N. et al. Regionalization of precipitation characteristics in Iran’s Lake Urmia basin. Theoretical and Applied Climatology, [s. l.], v. 132, n. 1–2, p. 363–373, 2018. FELIX, V. de S. Análise de 40 anos de precipitação pluviométrica da Bacia Hidrográfica do Rio Espinharas-PB. Revista Brasileira de Geografia, [s. l.], v. 8, n. 5, p. 1347–1358, 2015. FERNÁNDEZ-MANSO, A.; QUINTANO, C.; FERNÁNDEZ-MANSO, O. Forecast of NDVI in coniferous areas using temporal ARIMA analysis and climatic data at a regional scale. International Journal of Remote Sensing, [s. l.], v. 32, n. 6, p. 1595–1617, 2011. FERREIRA, T. R. et al. Identification of Desertification- Sensitive Areas in the Brazilian Northeast Through Vegetation Indices. Eng. Agric., Jaboaticabal, [s. l.], v. 37, n. 6, p. 1190–1202, 2017. Disponível em: FERREIRA, V. G. et al. Space-based observations of crustal deflections for drought characterization in Brazil. Science of The Total Environment, [s. l.], v. 644, p. 256–273, 2018. Disponível em: FISCHER, P. et al. Phosphorus saturation and superficial fertilizer application as key parameters to assess the risk of diffuse phosphorus losses from agricultural soils in Brazil. Science of the Total Environment, [s. l.], v. 630, p. 1515–1527, 2018. Disponível em: FOLBERTH, C. et al. The global cropland-sparing potential of high-yield farming. Nature Sustainability, [s. l.], v. 3, n. 4, p. 281–289, 2020. Disponível em: FORKEL, M. et al. Trend Change detection in NDVI time series: Effects of inter-annual variability and methodology. Remote Sensing, [s. l.], v. 5, n. 5, p. 2113–2144, 2013. FRANCHITO, S. H. et al. Validation of TRMM precipitation radar monthly rainfall estimates over Brazil. Journal of Geophysical Research, [s. l.], v. 114, n. 2, p. 1–9, 2009. FRANCHITO, S. H.; FERNANDEZ, J. P. R.; PAREJA, D. Surrogate Climate Change Scenario and Projections with a Regional Climate Model: Impact on the Aridity in South America. American Journal of Climate Change, [s. l.], v. 03, n. 05, p. 474–489, 2014. FRAZIER, A. G. et al. Comparison of geostatistical approaches to spatially interpolate month- 145 year rainfall for the Hawaiian Islands. International Journal of Climatology, [s. l.], v. 36, n. 3, p. 1459–1470, 2016. FUNCEME. Monitor de Secas. 2019. Disponível em: . Acesso em: 12 fev. 2020. FUNK, C. C. et al. A quasi-global precipitation time series for drought monitoring. 2014. Disponível em: . FUNK, C. C.; BROWN, M. E. Intra-seasonal NDVI change projections in semi-arid Africa. Remote Sensing of Environment, [s. l.], v. 101, n. 2, p. 249–256, 2006. GADGIL, S.; IYENGAR, R. N. Cluster analysis of rainfall stations of the Indian peninsula. Quarterly Journal of the Royal Meteorological Society, [s. l.], v. 106, n. 450, p. 873– 886, 1980. GALVÍNCIO, J. D.; SOUSA, F. de A. S. De. Relação Entre o Fenômeno El Niño e os Totais Mensais Precipitados na Bacia Hidrográfica do Rio São Francisco. Revista Brasileira de Recursos Hídricos, [s. l.], v. 7, n. 4, p. 113–119, 2002. GARRIGUES, S. et al. Quantifying spatial heterogeneity at the landscape scale using variogram models. Remote Sensing of Environment, [s. l.], v. 103, n. 1, p. 81–96, 2006. GIAMBELLUCA, T. W. et al. Online rainfall atlas of Hawai’i. Bulletin of the American Meteorological Society, [s. l.], v. 94, n. 3, p. 313–316, 2013. GOCIC, M.; TRAJKOVIC, S. Spatio-temporal patterns of precipitation in Serbia. Theoretical and Applied Climatology, [s. l.], v. 117, n. 3–4, p. 419–431, 2014. GOMES, H. B. et al. Climatology of easterly wave disturbances over the tropical South Atlantic. Climate Dynamics, [s. l.], v. 53, n. 3–4, p. 1393–1411, 2019. Disponível em: GOMES, U. A. F.; HELLER, L. Acesso à água proporcionado pelo Programa de Formação e Mobilização Social para Convivência com o Semiárido : Um Milhão de Cisternas Rurais : combate à seca ou ruptura da vulnerabilidade ? Eng. Sanit. Ambient., [s. l.], v. 21, n. 3, p. 623–633, 2016. GONG, D.; WANG, S. Definition of Antarctic oscillation index. Geophysical Research Letters, [s. l.], v. 26, n. 4, p. 459–462, 1999. GONG, X.; RICHMAN, M. B. On the application of cluster analysis to growing season precipitation data in North America east of the Rockies, 1995. GOOSSENS, C.; BERGER, A. Annual and seasonal climatic variations over the northern hemisphere and Europe during the last century. Annales Geophysicae, [s. l.], v. 4, n. 4, p. 385–400, 1986. GRIMM, A. M.; TEDESCHI, R. G. ENSO and Extreme Rainfall Events in South America. Journal of Climate, [s. l.], v. 22, n. 4, p. 1589–1609, 2009. GUERRERO-SALAZAR, P.; YEVJEVICH, V. Analysis of Drought Characteristics by The 146 Theory of Runs. Hydrology Papers 80, [s. l.], n. 80, p. 1–44, 1975. GURJÃO, C. D. de S. et al. Influência do Enos (El Niño-Oscilação Sul) no Regime Hidrológico do Rio São Francisco: uma Análise em Regiões com Fortes Pressões Antrópicas. Revista Brasileira de Geografia Física, [s. l.], v. 4, p. 774–790, 2012. GUTIÉRREZ, A. P. A. et al. Drought preparedness in Brazil. Weather and Climate Extremes, [s. l.], v. 3, p. 95–106, 2014. Disponível em: HAN, P. et al. Drought forecasting based on the remote sensing data using ARIMA models. Mathematical and Computer Modelling, [s. l.], v. 51, n. 11–12, p. 1398–1403, 2010. Disponível em: HÄNSEL, S.; SCHUCKNECHT, A.; MATSCHULLAT, J. The Modified Rainfall Anomaly Index ( mRAI ) — is this an alternative to the Standardised Precipitation Index ( SPI ) in evaluating future extreme precipitation characteristics ? Theoretical and Applied Climatology, [s. l.], v. 123, p. 827–844, 2016. HARRIS, I. et al. Updated high-resolution grids of monthly climatic observations - the CRU TS3.10 Dataset. International Journal of Climatology, [s. l.], v. 34, n. 3, p. 623–642, 2014. HARRIS, I. et al. Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Scientific data, [s. l.], v. 7, n. 1, p. 109, 2020. HASTENRATH, S. Circulation and teleconnection mechanisms of Northeast Brazil droughts. Progress in Oceanography, [s. l.], v. 70, n. 2–4, p. 407–415, 2006. HASTENRATH, S. Exploring the climate problems of Brazil’ s Nordeste : a review. Climatic Change, [s. l.], v. 112, p. 243–251, 2012. HAYES, M. et al. The Lincoln declaration on drought indices: Universal meteorological drought index recommended. Bulletin of the American Meteorological Society, [s. l.], v. 92, n. 4, p. 485–488, 2011. HENN, B. et al. An assessment of differences in gridded precipitation datasets in complex terrain. Journal of Hydrology, [s. l.], v. 556, p. 1205–1219, 2018. Disponível em: HEUMANN, B. W. et al. AVHRR derived phenological change in the Sahel and Soudan, Africa, 1982-2005. Remote Sensing of Environment, [s. l.], v. 108, n. 4, p. 385–392, 2007. HEWITT, K. Regions of risk: A geographical introduction to disasters. London, UK: Addison Wesley Longman Limited, 1997. HOLT, C. C. Forecasting seasonals and trends by exponentially weighted moving averages. International Journal of Forecasting, [s. l.], v. 20, n. 1, p. 5–10, 2004. HUANG, J. et al. Global semi-arid climate change over last 60 years. Climate Dynamics, [s. l.], v. 46, n. 3–4, p. 1131–1150, 2016. 147 HUFFMAN, G. J.; BOLVIN, D. T. TRMM and Other Data Precipitation Data Set Documentation. 2015. Disponível em: . HUNT, J. D.; FREITAS, M. A.; PEREIRA JR., A. O. Usinas Hidrelétricas Reversíveis Sazonais no Rio São Francisco: aumentando o armazenamento energético e diminuindo a evaporação. Sustentabilidade em Debate, [s. l.], v. 7, n. 3, p. 18, 2016. Disponível em: IBGE. Instituto Brasileiro de Geografia e Estatística - IBGE Cidades. 2016. Disponível em: . Acesso em: 16 maio. 2017. IPCC. Central and South America. In: BARROS, V. (Ed.). Climate Change 2014: Impacts, Adaptation and Vulnerability. Part B: Regional Aspects, Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge University Press, 2014. p. 1499–1566. IPCC. Climate Change 2007: the physical science basis. contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge University Press, 2007. IPEA. Intituto de Pesquisa Econômica Aplicada - Ipeadata. 2015. Disponível em: . Acesso em: 20 set. 2011. JAMALI, S. et al. Detecting changes in vegetation trends using time series segmentation. Remote Sensing of Environment, [s. l.], v. 156, p. 182–195, 2015. Disponível em: JIA, G. J.; EPSTEIN, H. E.; WALKER, D. A. Spatial heterogeneity of tundra vegetation response to recent temperature changes. Global Change Biology, [s. l.], v. 12, n. 1, p. 42– 55, 2006. JIANG, Z.; LIAN, Y.; QIN, X. Rocky desertification in Southwest China: Impacts, causes, and restoration. Earth-Science Reviews, [s. l.], v. 132, p. 1–12, 2014. Disponível em: . Acesso em: 20 set. 2018. JONES, P. D. et al. Long-term trends in precipitation and temperature across the Caribbean. International Journal of Climatology, [s. l.], v. 36, n. 9, p. 3314–3333, 2016. KANE, R. P. Prediction of droughts in north-east Brazil: Role of ENSO and use of periodicities. International Journal of Climatology, [s. l.], v. 17, p. 655–665, 1997. KANNAN, S.; GHOSH, S. Prediction of daily rainfall state in a river basin using statistical downscaling from GCM output. Stochastic Environmental Research and Risk Assessment, [s. l.], v. 25, n. 4, p. 457–474, 2011. KAUFMAN, L.; ROUSSEEUW, P. J. Finding groups in data. 1ed. ed. New Jersey: John Wiley & Sons, 1990. KAYANO, M. T. et al. How the two nodes of the tropical Atlantic sea surface temperature dipole relate the climate of the surrounding regions during austral autumn. International Journal of Climatology, [s. l.], v. 38, n. 10, p. 3927–3941, 2018. 148 KENDALL, M. G. A New Measure of Rank Correlation. Biometrika, [s. l.], v. 30, n. 1/2, p. 81–93, 1938. KHALIQ, M. N.; OUARDA, T. B. M. J. On the critical values of the standard normal homogeneity tes (SNHT). International Journal of Climatology, [s. l.], v. 27, p. 681– 687, 2007. Disponível em: KOCH, H. et al. Reservoir operation and environmental water demand: Scenarios for the Sub- Middle and Lower São Francisco River basin, Brazil. Ecohydrology, [s. l.], v. 11, n. 8, 2018. KOCH, R.; ALMEIDA-CORTEZ, J. S.; KLEINSCHMIT, B. Revealing areas of high nature conservation importance in a seasonally dry tropical forest in Brazil: Combination of modelled plant diversity hot spots and threat patterns. Journal for Nature Conservation, [s. l.], v. 35, p. 24–39, 2017. Disponível em: KUMAR, M. N. et al. On the use of Standardized Precipitation Index (SPI) for drought intensity assessment. Meteorological Applications, [s. l.], v. 16, n. April 2009, p. 381–389, 2009. LACERDA, F. et al. Long-term Temperature and Rainfall Trends over Northeast Brazil and Cape Verde. Earth Science & Climatic Change, [s. l.], v. 6, n. 8, p. 1–8, 2015. LAMBERT, R. Géographie et budget hydrologique. Annales de Géographie, [s. l.], v. 86, n. 477, p. 513–521, 1977. LAMY, C. Impact du changement climatique sur la fréquence et l’intensité des sécheresses en Bretagne. 2013. Université Rennes 2, [s. l.], 2013. LAMY, C.; DUBREUIL, V.; MELLO-THERY, N. A. De. Mudanças climáticas e secas: modelagem espacial e desafios para a agricultura na Bretanha. Revista Brasileira de Climatologia, [s. l.], v. 14, n. 10, p. 97–113, 2014. Disponível em: LEAL, I. R. et al. Mudando o curso da conservação da biodiversidade na Caatinga do Nordeste do Brasil. MEGADIVERSIDADE, [s. l.], v. 1, n. 1, p. 139–146, 2005. LEE, S.; CHO, S.; WONG, P. M. Rainfall Prediction Using Artificial Neural Networks. Journal of Geographic Information and Decision Analysis, [s. l.], v. 2, n. 2, p. 233–242, 1998. LEVENE, H. Robust Tests for Equality of Variances. In: OLKIN, I. (Ed.). Contributions to Probability and Statistics: Essay in HOnor of Harold Hotelling. Palo Alto: Stanford University Press, 1960. p. 278–292. LI, P. et al. Evaluation of drought implications on ecosystem services: Freshwater provisioning and food provisioning in the upper Mississippi river basin. International Journal of Environmental Research and Public Health, [s. l.], v. 14, n. 5, p. 1–23, 2017. LIMA, K. C.; SATYAMURTY, P.; FERNÁNDEZ, J. P. R. Large-scale atmospheric conditions associated with heavy rainfall episodes in Southeast Brazil. Theoretical and Applied Climatology, [s. l.], v. 101, n. 1, p. 121–135, 2010. 149 LINDOSO, D. P. et al. Integrated assessment of smallholder farming’s vulnerability to drought in the Brazilian Semi-arid: a case study in Ceará. Climatic Change, [s. l.], v. 127, n. 1, p. 93–105, 2014. LINDOSO, D. P. et al. Harvesting water for living with drought: Insights from the Brazilian Human Coexistence with Semi-Aridity approach towards achieving the sustainable development goals. Sustainability (Switzerland), [s. l.], v. 10, n. 622, p. 1–16, 2018. LIU, W. T.; NEGRÓN JUÁREZ, R. I. ENSO drought onset prediction in northeast Brazil using NDVI. International Journal of Remote Sensing, [s. l.], v. 22, n. 17, p. 3483–3501, 2001. LJUNG, G. M.; BOX, G. E. P. On a measure of lack of fit in time series models. Biometrika, [s. l.], v. 65, n. 2, p. 297–303, 1978. LOFGREN, B. M.; HUNTER, T. S.; WILBARGER, J. Effects of using air temperature as a proxy for potential evapotranspiration in climate change scenarios of Great Lakes basin hydrology. Journal of Great Lakes Research, [s. l.], v. 37, n. 4, p. 744–752, 2011. Disponível em: LOPES, I. et al. Balanço hídrico em função de regimes pluviométricos na região de Petrolina- PE. Irriga, [s. l.], v. 22, n. 3, p. 443–457, 2017. LOPES, L. S. de O.; SOARES, R. C. Desertification susceptibility of Gilbués (Piauí State) and Cabrobró (Pernambuco State) drylands, in Northeastern Brazil. Revista Geográfica de América Central, [s. l.], v. 1, n. 56, p. 307–323, 2016. LOUKAS, A.; VASILIADES, L. Probabilistic analysis of drought spatiotemporal characteristics in Thessaly region, Greece. Natural Hazards and Earth System Science, [s. l.], v. 4, n. 5/6, p. 719–731, 2010. LYRA, G. B.; OLIVEIRA-JÚNIOR, J. F.; ZERI, M. Cluster analysis applied to the spatial and temporal variability of monthly rainfall in Alagoas state, Northeast of Brazil. International Journal of Climatology, [s. l.], v. 34, n. 13, p. 3546–3558, 2014. MA, L. et al. Evaluation of precipitation from the ERA-40, NCEP-1, and NCEP-2 reanalyses and CMAP-1, CMAP-2, and GPCP-2 with ground-based measurements in China. Journal of Geophysical Research Atmospheres, [s. l.], v. 114, n. 9, p. 1–20, 2009. MAHALANOBIS, P. C. On the generalized distance in statistics. Proceedings of the National Institute of Sciences of India, [s. l.], v. 2, n. 1, p. 50–55, 1936. MALSY, M.; AUS DER BEEK, T.; FLÖRKE, M. Evaluation of large-scale precipitation data sets for water resources modelling in Central Asia. Environmental Earth Sciences, [s. l.], v. 73, n. 2, p. 787–799, 2015. MANETA, M. P. et al. A parsimonious crop-water productivity index: an application to Brazil. Area, [s. l.], v. 41, n. 1, p. 94–106, 2009. a. MANETA, M. P. et al. Water demand and flows in the São Francisco River Basin (Brazil) with increased irrigation. Agricultural Water Management, [s. l.], v. 96, n. 8, p. 1191–1200, 2009. b. 150 MANN, H. B. Nonparametric tests against trend. Econometrica, [s. l.], v. 13, n. 3, p. 245–259, 1945. MARENGO, J. A. et al. Climatic characteristics of the 2010-2016 drought in the semiarid Northeast Brazil region. Anais da Academia Brasileira de Ciências, [s. l.], p. 1–13, 2017. Disponível em: MARENGO, J. A. et al. Climatic characteristics of the 2010-2016 drought in the semiarid Northeast Brazil region. Anais da Academia Brasileira de Ciências, [s. l.], v. 90, p. 1973– 1985, 2018. MARENGO, J. A. et al. Development of regional future climate change scenarios in South America using the Eta CPTEC/HadCM3 climate change projections: Climatology and regional analyses for the Amazon, São Francisco and the Paraná River basins. Climate Dynamics, [s. l.], v. 38, n. 9–10, p. 1829–1848, 2012. MARENGO, J. A. et al. Variabilidade e mudanças climáticas no semiárido brasileiro. In: Recursos hídricos em regiões áridas e semiáridas. Campina Grande: Instituto Nacional do Semiárido, 2011. p. 383–422. MARENGO, J. A.; BERNASCONI, M. Regional differences in aridity / drought conditions over Northeast Brazil : present state and future projections. Climatic Change, [s. l.], v. 129, p. 103–115, 2015. MARENGO, J.; TORRES, R. R.; ALVES, L. M. Drought in Northeast Brazil—past, present, and future. Theoretical and Applied Climatology, [s. l.], v. 129, n. 3–4, p. 1189–1200, 2017. Disponível em: MARIANO, D. A. et al. Use of remote sensing indicators to assess effects of drought and human-induced land degradation on ecosystem health in Northeastern Brazil. Remote Sensing of Environment, [s. l.], v. 213, n. May, p. 129–143, 2018. Disponível em: MARQUES, É. T.; GUNKEL, G.; SOBRAL, M. C. Management of tropical river basins and reservoirs under water stress: experiences from Northeast Brazil. Environments, [s. l.], v. 6, n. 62, p. 1–22, 2019. MARQUES, T. V et al. Environmental and biophysical controls of evapotranspiration from Seasonally Dry Tropical Forests (Caatinga) in the Brazilian Semiarid. Agricultural and Forest Meteorology, [s. l.], v. 287, n. February 2019, p. 107957, 2020. Disponível em: MARTINS, E. S. P. R. et al. A multimethod attribution analysis of the prolonged Northeast Brazil hydrometeorological drought (2012-16). Bulletin of the American Meteorological Society, [s. l.], v. 99, n. 1, p. S65–S69, 2018. MARTINS, E. S. P. R. et al. Monitor de Secas do Nordeste, em busca de um novo paradigma para a gestão de secas. In: Série Água Brasil. 1a. ed. Brasília, DF, Brasil: Banco Mundial, 2015. p. 124. MCKEE, T. B.; DOESKEN, N. J.; KLEIST, J. The relationship of drought frequency and 151 duration to time scales. AMS 8th Conference on Applied Climatology, [s. l.], n. January, p. 179–184, 1993. Disponível em: MEDEIROS, F. J. Aspectos dinâmicos da atmosfera associados a seca de 2012- 2016 no Nordeste do Brasil. 2019. Universidade Federal do Rio Grande do Norte, [s. l.], 2019. MELO, D. D. C. D. et al. Reservoir storage and hydrologic responses to droughts in the Paraná River basin, south-eastern Brazil. Hydrology and Earth System Sciences, [s. l.], v. 20, n. 11, p. 4673–4688, 2016. MELO, M. M. M. S. et al. Trends in Temperature and Rainfall Extremes near the Artifical Sobradinho Lake, Brazil. Revista Brasileira de Meteorologia, [s. l.], v. 33, n. 3, p. 426– 440, 2018. MENDES, K. R. et al. Seasonal variation in net ecosystem CO2 exchange of a Brazilian seasonally dry tropical forest. Scientific Reports, [s. l.], v. 10, n. 9454, p. 1–16, 2020. MENDES, L. A. et al. Trade-off analysis among multiple water uses in a hydropower system: Case of São Francisco River Basin, Brazil. Journal of Water Resources Planning and Management, [s. l.], v. 141, n. 10, p. 1–10, 2015. MESQUITA, P. S.; BURSZTYN, M. Food acquisition programs in the Brazilian semi-arid region: benefits to farmers and impacts of climate change. Food Security, [s. l.], v. 9, n. 5, p. 1041–1051, 2017. MICHOT, V. et al. Performance of TRMM TMPA 3B42 V7 in replicating daily rainfall and regional rainfall regimes in the Amazon basin (1998-2013). Remote Sensing, [s. l.], v. 10, n. 12, p. 1–20, 2018. MILHORANCE, C. et al. Unpacking the policy mix of adaptation to climate change in Brazil’s semiarid region: enabling instruments and coordination mechanisms. Climate Policy, [s. l.], v. 20, n. 5, p. 593–608, 2020. Disponível em: MILITINO, A. F.; UGARTE, M. D.; PÉREZ-GOYA, U. Stochastic Spatio-Temporal Models for Analysing NDVI Distribution of GIMMS NDVI3g Images. Remote Sensing, [s. l.], v. 9, n. 76, p. 1–17, 2017. MISHRA, A. K.; SINGH, V. P. A review of drought concepts. Journal of Hydrology, [s. l.], v. 391, n. 1–2, p. 202–216, 2010. Disponível em: MONDAL, A.; KUNDU, S.; MUKHOPADHYAY, A. Rainfall trend analysis by Mann- Kendall test : a case study of North-Eastern part of Cuttack district , Orissa. International Journal of Geology, Earth and Environmental Sciences, [s. l.], v. 2, n. 1, p. 70–78, 2012. MOOD, A. M. The distribution theory of runs. The Annals of Mathematical Statistics, [s. l.], v. 11, n. 4, p. 367–392, 1940. MOSCATI, M. C. de L.; GAN, M. A. Rainfall variability in the rainy season of semiarid zone of Northeast Brazil (NEB) and its relation to wind regime. International Journal of 152 Climatology, [s. l.], v. 27, p. 493–512, 2007. Disponível em: MOUNIER, J. Aspects et fréquences de la sécheresse en Bretagne : essai de définition de la sécheresse en Europe océanique. Revue de géographie de Lyon, [s. l.], v. 52, n. 2, p. 167– 176, 1977. MOURA, A. D.; SHUKLA, J. On the Dynamics of Droughts in Northeast Brazil: Observations, Theory and Numerical Experiments with a General Circulation Model. Journal of the Atmospheric Sciences, [s. l.], v. 38, p. 2653–2675, 1981. MUDELSEE, M. Trend analysis of climate time series: A review of methods. Earth-Science Reviews, [s. l.], v. 190, n. October 2018, p. 310–322, 2019. Disponível em: MUTTI, P. R. Estimativa da evapotranspiração sobre a bacia hidrográfica do Rio Piranhas-Açu utilizando dados de sensoriamento remoto. 2018. Federal University of Rio Grande do Norte, [s. l.], 2018. MUTTI, P. R. et al. A detailed framework for the characterization of rainfall climatology in semiarid watersheds. Theoretical and Applied Climatology, [s. l.], v. 139, n. 1–2, p. 109– 125, 2020. a. Disponível em: MUTTI, P. R. et al. Basin scale rainfall-evapotranspiration dynamics in a tropical semiarid environment during dry and wet years. International Journal of Applied Earth Observation and Geoinformation, [s. l.], v. 75, n. October 2018, p. 29–43, 2019. Disponível em: MUTTI, P. R. et al. NDVI time series stochastic models for the forecast of vegetation dynamics over desertification hotspots. International Journal of Remote Sensing, [s. l.], v. 41, n. 7, p. 2759–2788, 2020. b. Disponível em: MUTTI, P. R. et al. Assessment of Gridded CRU TS Data for Long-Term Climatic Water Balance Monitoring over the São Francisco Watershed, Brazil. Atmosphere, [s. l.], v. 11, n. 1207, p. 1–25, 2020. c. MUTTI, P. R.; BEZERRA, B. G. Tendências de NDVI nos núcleos de desertificação do Semiárido Brasileiro no período 2000-2018. In: I CONGRESSO NACIONAL DA DIVERSIDADE DO SEMIÁRIDO 2018, Natal, Brazil. Anais... Natal, Brazil NAUMANN, G. et al. Global Changes in Drought Conditions Under Different Levels of Warming. Geophysical Research Letters, [s. l.], v. 45, n. 7, p. 3285–3296, 2018. NELSON, G. C. et al. Climate change effects on agriculture: Economic responses to biophysical shocks. Proceedings of the National Academy of Sciences of the United States of America, [s. l.], v. 111, n. 9, p. 3274–3279, 2014. NIEMEYER, S. New drought indices. Options Méditerranéennes, [s. l.], v. 80, n. 80, p. 267– 274, 2008. NIETO-FERREIRA, R.; RICKENBACH, T. M.; WRIGHT, E. A. The role of cold fronts in the 153 onset of the monsoon season in the South Atlantic convergence zone. Quarterly Journal of the Royal Meteorological Society, [s. l.], v. 137, n. 657, p. 908–922, 2011. NIMER, E. Desertificação: realidade ou mito? Revista Brasileira de Geografia, [s. l.], v. 50, n. 1, p. 7–40, 1988. NOAA. El Niño & La Niña (El Niño-Southern Oscillation). 2020. Disponível em: . NOGUÉS-PAEGLE, J.; MO, K. C. Alternating wet and dry conditions over South America during summer. Monthly Weather Review, [s. l.], v. 125, n. 2, p. 279–291, 1997. NOOJIPADY, P. et al. Forest carbon emissions from cropland expansion in the Brazilian Cerrado biome. Environmental Research Letters, [s. l.], v. 12, n. 2, p. 1–11, 2017. OKIN, G. S. et al. The Impact of Drought on Native Southern California Vegetation : Remote Sensing Analysis Using MODIS-Derived Time Series. Journal of Geophysical Research: Biogeosciences, [s. l.], v. 123, p. 1927–1939, 2018. OTKIN, J. A. et al. Assessing the evolution of soil moisture and vegetation conditions during the 2012 United States flash drought. Agricultural and Forest Meteorology, [s. l.], v. 218–219, p. 230–242, 2016. Disponível em: OYAMA, M. D.; NOBRE, C. A. Climatic consequences of a large-scale desertification in northeast Brazil: A GCM simulation study. Journal of Climate, [s. l.], v. 17, n. 16, p. 3203–3213, 2004. PALMER, W. . Meteorological Drought. Research Paper no. 45, [s. l.], n. February 1965, 1965. PAREDES-TREJO, F. et al. Main features of large-scale oceanic-atmospheric circulation related to strongest droughts during rainy season in Brazilian São Francisco River Basin. International Journal of Climatology, [s. l.], v. 36, n. 12, p. 4102–4117, 2016. PAREDES-TREJO, F. J.; BARBOSA, H. A.; KUMAR, T. V. L. Validating CHIRPS-based satellite precipitation estimates in Northeast Brazil. Journal of Arid Environments, [s. l.], v. 139, p. 26–40, 2017. Disponível em: PASSOS, M. M. Dos; DUBREUIL, V. A relação clima-vegetação no semi-árido brasileiro. Boletim de Geografia, [s. l.], v. 22, n. 1, p. 35–48, 2004. PEREIRA, A. R.; ANGELOCCI, L. R.; SENTELHAS, P. C. Agrometeorologia Fundamentos e Aplicações Práticas. Guaíba, RS: Livraria e Editora Agropecuária Ltda., 2002. PEREZ-MARIN, A. M. et al. Núcleos de desertificação no semiárido brasileiro : ocorrência natural ou antrópica ? Parcerias Estratégicas, [s. l.], v. 17, n. 34, p. 87–106, 2012. PERSAUD, B. D. et al. Evaluating the suitability of three gridded-datasets and their impacts on hydrological simulation at Scotty Creek in the southern Northwest Territories, Canada. Hydrological Processes, [s. l.], v. 34, n. 4, p. 898–913, 2020. 154 PETTITT, A. N. A Non-Parametric Approach to the Change-Point Problem. Journal of the Royal Statistical Society, [s. l.], v. 28, n. 2, p. 126–135, 1979. PICOLI, M. C. A. et al. Big earth observation time series analysis for monitoring Brazilian agriculture. ISPRS Journal of Photogrammetry and Remote Sensing, [s. l.], v. 145, n. August, p. 328–339, 2018. Disponível em: PLACKETT, R. L. Karl Pearson and the Chi-squared Test. International Statistical Review, [s. l.], v. 51, n. 1, p. 59–72, 1983. PRÖMMEL, K. et al. Evaluation of the skill and added value of a reanalysis-driven regional simulation for Alpine temperature. International Journal of Climatology, [s. l.], v. 30, n. 5, p. 760–773, 2010. RAO, V. B. et al. An update on the rainfall characteristics of Brazil: Seasonal variations and trends in 1979-2011. International Journal of Climatology, [s. l.], v. 36, n. 1, p. 291– 302, 2016. RAO, V. B.; HADA, K. Characteristics of rainfall over Brazil: Annual variations and connections with the Southern Oscillation. Theoretical and Applied Climatology, [s. l.], v. 42, n. 2, p. 81–91, 1990. RAO, V. B.; HADA, K.; HERDIES, D. L. On the severe drought of 1993 in north‐east Brazil. International Journal of Climatology, [s. l.], v. 15, n. 6, p. 697–704, 1995. RATTER, J. A.; RIBEIRO, J. F.; BRIDGEWATER, S. The Brazilian cerrado vegetation and threats to its biodiversity. Annals of Botany, [s. l.], v. 80, n. 3, p. 223–230, 1997. RAU, P. et al. Regionalization of rainfall over the Peruvian Pacific slope and coast. International Journal of Climatology, [s. l.], v. 37, n. March 2016, p. 143–158, 2017. REBOITA, M. S. et al. Causas da semi-aridez do Sertão Nordestino. Revista Brasileira de Climatologia, [s. l.], v. 19, n. 12, p. 254–277, 2016. REBOITA, M. S.; AMBRIZZI, T.; DA ROCHA, R. P. Relationship between the southern annular mode and southern hemisphere atmospheric systems. Revista Brasileira de Meteorologia, [s. l.], v. 24, n. 1, p. 48–55, 2009. REICHLE, L. M. et al. Spatial Heterogeneity of the Temporal Dynamics of Arctic Tundra Vegetation. Geophysical Research Letters, [s. l.], v. 45, n. 17, p. 9206–9215, 2018. REIS, L. et al. Influence of Climate Variability on Soybean Yield in MATOPIBA, Brazil. Atmosphere, [s. l.], v. 11, n. 1130, p. 1–24, 2020. RODELL, M. et al. Emerging trends in global freshwater availability. Nature, [s. l.], v. 557, n. 7707, p. 651–659, 2018. RODRIGUES, D. T. et al. Spatial and temporal assessment of the extreme and daily precipitation of the Tropical Rainfall Measuring Mission satellite in Northeast Brazil. International Journal of Remote Sensing, [s. l.], v. 41, n. 2, p. 549–572, 2020. Disponível em: 155 RODRIGUES, M. T. Herpetofauna da Caatinga. In: LEAL, I. R.; TABARELLI, M.; DA SILVA, J. M. C. (Eds.). Ecologia e Conservação da Caatinga. 1. ed. Recife, PE, Brasil: Ed. Universitária da UFPE, 2003. p. 181–237. RODRIGUEZ-PUEBLA, C. et al. Spatial and temporal patterns of annual precipitation variability over the Iberian peninsula. International Journal of Climatology, [s. l.], v. 18, p. 299–316, 1998. ROSENZWEIG, C. et al. Assessing agricultural risks of climate change in the 21st century in a global gridded crop model intercomparison. Proceedings of the National Academy of Sciences of the United States of America, [s. l.], v. 111, n. 9, p. 3268–3273, 2014. ROSSO, F. V. et al. Influence of the Antarctic Oscillation on the South Atlantic convergence zone. Atmosphere, [s. l.], v. 9, n. 11, 2018. SA’ADI, Z. et al. Trends analysis of rainfall and rainfall extremes in Sarawak , Malaysia using modified Mann – Kendall test. Meteorology and Atmospheric Physics, [s. l.], n. 0123456789, 2017. SAADI, S. et al. Assessment of actual evapotranspiration over a semiarid heterogeneous land surface by means of coupled low-resolution remote sensing data with an energy balance model: Comparison to extra-large aperture scintillometer measurements. Hydrology and Earth System Sciences, [s. l.], v. 22, n. 4, p. 2187–2209, 2018. SALVACION, A. R. et al. Evaluation and spatial downscaling of CRU TS precipitation data in the Philippines. Modeling Earth Systems and Environment, [s. l.], v. 4, n. 3, p. 891–898, 2018. Disponível em: SANTOS, C. A. G. et al. Drought assessment using a TRMM-derived standardized precipitation index for the upper São Francisco River basin, Brazil. Environmental Monitoring and Assessment, [s. l.], v. 189:250, n. 6, 2017. SANTOS, C. A. G. et al. Integrated spatiotemporal trends using TRMM 3B42 data for the Upper São Francisco River basin, Brazil. Environmental Monitoring and Assessment, [s. l.], v. 190, n. 3, 2018. SANTOS, C. A. G.; DE MORAIS, B. S. Identification of precipitation zones within São Francisco River basin (Brazil) by global wavelet power spectra. Hydrological Sciences Journal, [s. l.], v. 58, n. 4, p. 789–796, 2013. Disponível em: SANTOS, M. S. et al. Time-space characterization of droughts in the São Francisco river catchment using the Standard Precipitation Index and continuous wavelet transform. Revista Brasileira de Recursos Hidricos, [s. l.], v. 24, n. 28, p. 1–12, 2019. SCHAMM, K. et al. Global gridded precipitation over land: A description of the new GPCC First Guess Daily product. Earth System Science Data, [s. l.], v. 6, n. 1, p. 49–60, 2014. SEARCY, J. K.; HARDISON, C. H.; LANGBEIN, W. B. Double-Mass CurvesGeological Surbey Water Supply Paper 1541-B. Washington, DC. SHAPIRO, S. S.; WILK, M. B. An Analysis of Variance Test for Normality ( Complete 156 Samples ). Biometrika, [s. l.], v. 52, n. 3, p. 591–611, 1965. SHI, H.; LI, T.; WEI, J. Evaluation of the gridded CRU TS precipitation dataset with the point raingauge records over the Three-River Headwaters Region. Journal of Hydrology, [s. l.], v. 548, p. 322–332, 2017. Disponível em: SIEGMUND-SCHULTZE, M. et al. Paternalism or participatory governance? Efforts and obstacles in implementing the Brazilian water policy in a large watershed. Land Use Policy, [s. l.], v. 48, p. 120–130, 2015. Disponível em: SILVA, I. A. de S.; BARROS, J. R. Degradação ambiental, cobertura e uso das terras: uma análise geográfica do município de Gilbués-PI. Revista Equador (UFPI), [s. l.], v. 5, n. 2, p. 190–204, 2016. SILVA, P. F. Da et al. Seasonal patterns of carbon dioxide, water and energy fluxes over the Caatinga and grassland in the semi-arid region of Brazil. Journal of Arid Environments, [s. l.], v. 147, p. 71–82, 2017. SIMMONS, A. J. et al. Comparison of trends and low-frequency variability in CRU, ERA-40, and NCEP/NCAR analyses of surface air temperature. Journal of Geophysical Research: Atmospheres, [s. l.], v. 109, n. 24, p. 1–18, 2004. SIMPSON, L. D. The Brazilian Northeast region and the Rio São Francisco. International Journal of Water Resources Development, [s. l.], v. 14, n. 3, p. 399–404, 1998. SINGH, C. V. Pattern characteristics of Indian monsoon rainfall using principal component analysis (PCA). Atmospheric Research, [s. l.], v. 114, n. 3, p. 317–328, 2006. SINGH, V. P.; GUO, H.; YU, F. X. Parameter estimation for 3-parameter log-logistic distribution (LLD3) by Pome. Stochastic Hydrology and Hydraulics, [s. l.], v. 7, n. 3, p. 163–177, 1993. SMITH, C. A.; SARDESHMUKH, P. D. The effect of ENSO on the intraseasonal variance of surface temperatures in winter. International Journal of Climatology, [s. l.], v. 20, n. 13, p. 1543–1557, 2000. SOUTO, J.; BELTRÃO, N.; TEODORO, A. Performance of Remotely Sensed Soil Moisture for Temporal and Spatial Analysis of Rainfall over São Francisco River Basin, Brazil. Geosciences, [s. l.], v. 9, n. 3, p. 144, 2019. STAGGE, J. H. et al. Candidate Distributions for Climatological Drought Indices (SPI and SPEI). International Journal of Climatology, [s. l.], v. 35, n. 13, p. 4027–4040, 2015. STOLF, R. et al. Water transfer from São Francisco river to semiarid northeast of Brazil: technical data, environmental impacts, survey of opinion about the amount to be transferred. Engenharia Agrícola, [s. l.], v. 32, n. 6, p. 998–1010, 2012. Disponível em: STONE, R.; AULICIEMS, A. SOI phase relationships with rainfall in eastern Australia. 157 International Journal of Climatology, [s. l.], v. 12, n. 6, p. 625–636, 1992. SUBRAHMANYAM, V. P. Incidence and spread of continental drought. In: Reports on WMO/IHD projects. 2. ed. Geneva: Secretariat of the World Meteorological Organization, 1967. p. 52. SUN, T. et al. Water availability of São Francisco river basin based on a space-borne geodetic sensor. Water, [s. l.], v. 8, n. 213, 2016. SUN, W. et al. Spatiotemporal vegetation cover variations associated with climate change and ecological restoration in the Loess Plateau. Agricultural and Forest Meteorology, [s. l.], v. 209–210, p. 87–99, 2015. Disponível em: SVOBODA, M. et al. The drought monitor. Bulletin of the American Meteorological Society, [s. l.], n. April, p. 1181–1190, 2002. TANSER, F. C.; PALMER, A. R. The application of a remotely-sensed diversity index to monitor degradation patterns in a semi-arid, heterogeneous, South African landscape. Journal of Arid Environments, [s. l.], v. 43, p. 477–484, 1999. TEDESCHI, R. G.; GRIMM, M.; CAVALCANTI, I. F. A. Influence of Central and East ENSO on precipitation and its extreme events in South America during austral autumn and winter. International Journal of Climatology, [s. l.], 2016. TEIXEIRA, H. Determining regional actual evapotranspiration of irrigated crops and natural vegetation in the São Francisco river basin (Brazil) using remote sensing and Penman- Monteith equation. Remote Sensing, [s. l.], v. 2, n. 5, p. 1287–1319, 2010. TEIXEIRA, M. da S.; SATYAMURTY, P. Trends in the frequency of intense precipitation events in southern and southeastern Brazil during 1960-2004. Journal of Climate, [s. l.], v. 24, n. 7, p. 1913–1921, 2011. THORNE, P. W. et al. Reassessing changes in diurnal temperature range: Intercomparison and evaluation of existing global data set estimates. Journal of Geophysical Research: Atmospheres, [s. l.], v. 121, p. 5138–5158, 2016. THORNTHWAITE, C. W. An approach toward a rational classification of climate. Geographical Review, [s. l.], v. 38, n. 1, p. 55–94, 1948. Disponível em: THORNTHWAITE, C. W.; MATHER, J. R. The Water Balance. Publications in Climatology, [s. l.], v. 8, p. 1–86, 1955. TIAN, M.; WANG, P.; KHAN, J. Drought forecasting with vegetation temperature condition index using ARIMA models in the Guanzhong Plain. Remote Sensing, [s. l.], v. 8, n. 9, 2016. TIMMERMANN, A. et al. El Niño – Southern Oscillation complexity. Nature, [s. l.], 2018. Disponível em: TINÔCO, I. C. M. et al. Characterization of Rainfall Patterns in the Semiarid Brazil. Anuário 158 do Instituto de Geociências - UFRJ, [s. l.], v. 41, n. 2, p. 397–409, 2018. TOMASELLA, J. et al. Desertification trends in the Northeast of Brazil over the period 2000– 2016. International Journal of Applied Earth Observation and Geoinformation, [s. l.], v. 73, n. August 2017, p. 197–206, 2018. Disponível em: TRATAR, L. F.; STRMCNIK, E. The comparison of Holt e Winters method and Multiple regression method : A case study. Energy, [s. l.], v. 109, p. 266–276, 2016. TUCKER, C. J. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, [s. l.], v. 8, n. 2, p. 127–150, 1979. UNCCD. United Nations convention to combat desertification in countries experiencing serious drought and/or desertification, particularly in Africa. Paris, France: A/AC, 241-27, United Nations Convention to Combat Desertification, 1994. UTIDA, G. et al. Tropical South Atlantic influence on Northeastern Brazil precipitation and ITCZ displacement during the past 2300 years. Scientific Reports, [s. l.], v. 9, n. 1, p. 1– 8, 2019. UVO, C. B. et al. The Relationships between Tropical Pacific and Atlantic SST and Northeast Brazil Monthly Precipitation. Journal of Climate, [s. l.], v. 11, p. 551–562, 1998. VAN LOON, A. F. et al. Drought in a human-modified world: Reframing drought definitions, understanding, and analysis approaches. Hydrology and Earth System Sciences, [s. l.], v. 20, n. 9, p. 3631–3650, 2016. a. VAN LOON, A. F. et al. Drought in the Anthropocene. Nature Geoscience, [s. l.], v. 9, n. 2, p. 89–91, 2016. b. Disponível em: VAN ROOY, M. P. A Rainfall Anomaly Index Independent of Time and Space. Notos, [s. l.], v. 14, p. 43, 1965. VASCONCELLOS, F. C.; CAVALCANTI, I. F. A. Extreme precipitation over Southeastern Brazil in the austral summer and relations with the Southern Hemisphere annular mode. Atmospheric Science Letters, [s. l.], v. 11, n. 1, p. 21–26, 2010. VERBESSELT, J. et al. Detecting trend and seasonal changes in satellite image time series. Remote Sensing of Environment, [s. l.], v. 114, n. 1, p. 106–115, 2010. Disponível em: VICENTE-SERRANO, S. M. et al. Global assessment of the standardized evapotranspiration deficit index (SEDI) for drought analysis and monitoring. Journal of Climate, [s. l.], v. 31, n. 14, p. 5371–5393, 2018. VICENTE-SERRANO, S. M.; BEGUERÍA, S.; LÓPEZ-MORENO, J. I. A multiscalar drought index sensitive to global warming: The standardized precipitation evapotranspiration index. Journal of Climate, [s. l.], v. 23, n. 7, p. 1696–1718, 2010. VIEIRA, R. M. S. P. et al. Identifying areas susceptible to desertification in the Brazilian northeast. Solid Earth, [s. l.], v. 6, n. 1, p. 347–360, 2015. 159 WAGNER, P. D. et al. Comparison and evaluation of spatial interpolation schemes for daily rainfall in data scarce regions. Journal of Hydrology, [s. l.], v. 464–465, p. 388–400, 2012. Disponível em: WANG, J. et al. Optimization of the time series NDVI-rainfall relationship using linear mixed- effects modeling for the anti-desertification area in the Beijing and Tianjin sandstorm source region. Theoretical and Applied Climatology, [s. l.], v. 132, p. 1291–1301, 2018. WANG, W. et al. Propagation of drought: From meteorological drought to agricultural and hydrological drought. Advances in Meteorology, [s. l.], v. 2016, n. 4, 2016. WASHINGTON-ALLEN, R. A. et al. Quantification of the ecological resilience of drylands using digital remote sensing. Ecology and Society, [s. l.], v. 13, n. 1, 2008. WILHITE, D. A. Drought as a natural hazard: Concepts and definitions. In: WILHITE, D. A. (Ed.). Drought: A Global Assessment. London, UK. p. 3–18. WILHITE, D. A.; GLANTZ, M. H. Understanding : the Drought Phenomenon : The Role of Definitions. Water International, [s. l.], v. 10, n. 3, p. 111–120, 1985. WILKS, D. S. Statistical Methods in the atmospheric sciences. 2. ed. San Diego, USA: Elsevier, 2006. v. 14 WILM, H. G. et al. Reports of the committee on transpiration and evaporation, 1943-44. Transactions, American Geophysical Union, [s. l.], v. 25, p. 683–693, 1944. WINTERS, P. R. Forecasting Sales by Exponentially Weighted Moving Averages. Management Science, [s. l.], v. 6, n. 3, p. 324–342, 1960. Disponível em: . Acesso em: 23 out. 2018. WU, H. et al. Appropriate application of the Standardized Precipitation Index in arid locations and dry seasons. International Journal of Climatology, [s. l.], v. 27, n. June 2006, p. 65– 79, 2007. WU, H.; QIAN, H. Innovative trend analysis of annual and seasonal rainfall and extreme values in Shaanxi , China , since the 1950s. International Journal of Climatology, [s. l.], v. 37, n. August 2016, p. 2582–2592, 2017. XAVIER, A. C.; KING, C. W.; SCANLON, B. R. Daily gridded meteorological variables in Brazil (1980-2013). International Journal of Climatology, [s. l.], v. 2659, n. October 2015, p. 2644–2659, 2016. XU, D.; YOU, X.; XIA, C. Assessing the spatial-temporal pattern and evolution of areas sensitive to land desertification in North China. Ecological Indicators, [s. l.], v. 97, n. October 2018, p. 150–158, 2019. Disponível em: ZARGAR, A. et al. A review of drought indices. Environmental Reviews, [s. l.], v. 19, n. NA, p. 333–349, 2011. Disponível em: ZHANG, B.; HE, C. A modified water demand estimation method for drought identification 160 over arid and semiarid regions. Agricultural and Forest Meteorology, [s. l.], v. 230–231, p. 58–66, 2016. Disponível em: ZHANG, X. et al. Monitoring vegetation phenology using MODIS. Remote Sensing of Environment, [s. l.], v. 84, n. 3, p. 471–475, 2003. ZILLI, M. T. et al. A comprehensive analysis of trends in extreme precipitation over southeastern coast of Brazil. International Journal of Climatology, [s. l.], v. 37, n. August 2016, p. 2269–2279, 2017. ZILLI, M. T.; CARVALHO, L. M. V.; LINTNER, B. R. The poleward shift of South Atlantic Convergence Zone in recent decades. Climate Dynamics, [s. l.], v. 52, n. 5–6, p. 2545– 2563, 2019. Disponível em: