Assessment of Long-Term Streamflow Response to Flash Drought in the São Francisco River Basin over the Last Three Decades (1991–2020)

Abstract

Flash droughts, characterized by a rapid onset and severe intensity, pose a serious threat to water resource management. Extensive research has indicated that drought has lagged impacts on streamflow. Nevertheless, the hydrometeorological conditions by which streamflow dynamics respond to drought within the São Francisco River Basin (SFRB) remain ambiguous. To bridge this gap, we conducted a study on long-term streamflow responses to flash drought in the SFRB from 1991 to 2020, combining the Standardized Antecedent Precipitation Evapotranspiration Index (SAPEI) and quantile streamflow with a trend analysis. This study employed the SAPEI, a daily drought index, to identify flash droughts and assess the response of streamflow to the identified events across the SFRB. Our findings reveal insights into the direct response of streamflow to flash drought conditions, stimulated by the application of the SAPEI for analysis of flash drought events. The interannual flash droughts fluctuated over the years, with the middle part of the SFRB experiencing frequent, longer flash droughts, while the south part encountered shorter but less frequent events. About 55% of the study area is trended towards drying conditions. A comparative analysis of the SAPEI and streamflow identified a synchronized response to the onset of flash drought events, but the recovery timescale for the SAPEI and streamflow varied among the events. This study enhances understanding of the flash-drought–streamflow relationship in the SFRB and provides theoretical support for addressing drought risks under climate change.

1. Introduction

Drought is characterized by its wide range and long duration, profoundly impacting human life and water resource management [1,2]. Conventionally, drought is considered a slow-developing climate phenomenon, taking months or years to reach peak extent and intensity [3]. There are five types of droughts including meteorological, agricultural, hydrological, socio-economic, and ecological droughts, taking months or years to reach peak extent and intensity [4]. Each type reflects different phenomena and influencing factors. For instance, meteorological droughts are caused by a deficit of moisture in the atmosphere, and agricultural droughts occur when there is no sufficient rainfall and soil moisture during the growing period, while hydrological droughts are caused by a shortage of water resources, and socio-economic and ecological droughts are related to the impacts of drought on human production, life, and natural ecosystems [5]. Recent studies have defined a new category of a rapidly developing drought, called flash drought (hereinafter FD) [6,7,8].

In 2012, the São Francisco River Basin (hereinafter SFRB) experienced a rapid onset of extreme drought conditions [9,10]. In this context, the SFRB has received unique interest since it is an important watershed of Brazil in terms of hydropower and agricultural production. Numerous studies have documented a trend of increasing drought severity within the SFRB and a strengthening of meteorological monitoring and early warning systems in the northeastern Brazil, e.g., [8,9]. However, in this watershed, there is a noticeable gap in the literature, as there has been a scarcity of studies that have investigated the relationship between FD and streamflow [10]. Europe and the United States were also impacted by FDs in 2012 [7]. In recent years, considerable efforts have been directed towards elucidating the occurrence patterns of FDs, e.g., [8].

The association between the onset of FD conditions and subsequent streamflow reductions emphasizes the interconnected dynamics of meteorological and hydrological aspects during these extreme events [9]. FDs occur when both agricultural and ecological drought conditions escalate quickly in response to below-average precipitation, often combined by other environmental anomalies such as elevated temperatures and high evapotranspiration rates [11,12]. The extensive impacts of FDs have led to the investigation of flash drought events with respect to a varied set of hydrometeorological variables [13]. These include precipitation deficits [14], soil moisture depletion [15], elevated evaporative demand [16], vegetation conditions [17], and heat waves [18,19]. Studies have suggested that increased temperatures and net radiation accelerated soil evaporation and plant transpiration, e.g., [20]. This reduces soil water content, which, in turn, inhibits soil evaporation and plant evapotranspiration. Increased soil–air latent heat interactions lead to decreased soil moisture [21]. However, decreased soil moisture leads to increased soil–air sensible heat interactions [22]. FDs have always existed and were first described in 2002 by Svoboda et al. [12].

Traditionally, droughts have been evaluated using indices such as the Palmer Drought Severity Index (PDSI), the Standardized Precipitation Index (SPI), or the Standardized Precipitation Evapotranspiration Index (SPEI), which have been crucial for understanding and responding to the duration and intensity of drought-induced impacts on various systems [23]. These commonly used indices are not suitable for identifying FDs because of the temporal dimension of FD events; there is no clear boundary among them [24]. However, several indices have been developed to support the establishment of early warning systems for FDs, each combining hydrometeorological variables to enhance the identification and assessment of FD conditions [13,17,25]. Among these is the Standardized Antecedent Precipitation Evapotranspiration Index (SAPEI) that was introduced to improve the accuracy of FD detection [26,27]. The SAPEI accounts for understanding the strength of meteorological–hydrological drought impacts. Additionally, it incorporates the influence of the antecedent water surplus/deficit of the current day for understanding the dynamics of FD events [26,28].

In this study, we explored the connection between flash drought events and streamflow in the SFRB, aiming to understand the rapid intensification of flash drought events and their impact on streamflow over the 1991–2020 period. The three goals of this paper comprise: (1) demonstrating the consistency and sensitivity of the SAPEI to effectively identify flash drought events for selected catchment in the SFRB; (2) evaluating the response of streamflow to the identified flash drought events, aiming to yield research results that contribute to the prevention and mitigation of drought disasters; and (3) examining the spatiotemporal trends in the SAPEI and the hydro-climatological conditions during the past three decades (1991–2020).

2. Materials and Methods

2.1. Study Area

The SFRB is fourth largest basin in South America, spans an area of 639,219 km², and extends from (7.2–21.1° S) to (36.3–47.6° W) in Brazil (Figure 1). The average annual discharge across the SFRB is 1961 m³/s. The topography (Figure 1a) ranges from high in the south to low in the north, subdivided into four regions: (1) the upper (10,227,504 km², covering about 16%); (2) middle (40,270,797 km², covering about 63%); (3) sub-middle (10,866,723 km², covering 17%); and (4) lower (2,556,876 km², covering about 4%) parts of the SFRB (Figure 1b).

The annual mean precipitation in the study area from 1991–2020 was 907 mm, ranging from around 574 mm to >1248 mm (Figure 1b). The climate of the SFRB is dominated by interactions between warm, dry air masses from northeastern Brazil and warm, humid air masses from southern Brazil. The impact of climate change on the SFRB is primarily characterized by changes in warming and precipitation patterns, which not only affect the water cycle process in the region but also precipitate a surge in both the recurrence and severity of extreme meteorological occurrences, including droughts and floods [29,30,31,32]. Hence, the SFRB, which is highly dependent on climatic conditions, faces more severe challenges in terms of ecosystem stability, economic development, and industrial development. High-temperature stress combined with water stress has caused severe FDs in the SFRB [8]. Barbosa [17] discovered a pronounced rise in the occurrence of FDs in semi-arid and arid areas within the SFRB since the late 1990s. SFRB is a typical ecologically fragile and climate-change-sensitive basin [9].

Vegetation within the SFRB was classified into four major land cover types (Figure 2), considering both vegetation degradation and irrigated agriculture (Figure 2a). These areas, characterized by intense human activities such as agricultural activities and land use alterations (Figure 2b), significantly affect vegetation ecosystems, resulting in considerable fluctuations in response to drought [8]. As an example, Figure 3 shows land condition and land use changes in the fluvial channel of the SFRB in 2018, mainly related to changes in soil degradation, irrigated areas, and deforested areas near Pilão Arcado, Bahia. This is causing an increase in the susceptibility to desertification [17].

Figure 2c shows a notable difference in precipitation variability among the land use and land cover classification types in the northern SFRB. An extreme high deviation in precipitation was observed in dense caatinga vegetation, with an average (μ) of 434.44 mm and deviation (d) of ±131.12 mm, while the extreme low deviation in precipitation was closely related to both non-vegetated areas and non-forest vegetation (μ = 398.75 mm ± d = 136.22 mm and μ = 405.11 mm ± d = 128.37 mm), respectively. Changes in plant growth due to rainfall change have a direct impact on watershed processes potentially leading to an increased intensity in drought conditions [17].

An increase in population density has occurred within the SFRB, where irrigated agriculture has developed. Likewise, there has been a marked escalation in the number of individuals subjected to frequent drought events [9,10]. The onset of flash droughts, a consequence of the persistence of hydrometeorological anomalies over days or weeks, can lead to substantial agricultural losses and ecological damage [8]. The navigability of rivers may be hindered due to decreased water levels, impacting transportation. Low streamflow levels may also allow salt water to move upstream (especially in coastal watersheds), threatening water supplies in these areas. Given the importance of precipitation variability across the SFRB, it is necessary to understand the relationship between FDs and streamflow. This knowledge is essential for devising effective strategies to manage and mitigate the diverse impacts that FDs can have on both natural ecosystems and human activities [17].

2.2. Datasets

Meteorological variables from the Brazilian Daily Weather Gridded Data (BR-DWGD) were used to assess and identify periods of FDs across the SFRB from 1991 to 2020. The BR-DWGD do not directly provided estimates of potential evapotranspiration (PET), but provide daily weather parameters (e.g., precipitation, maximum temperature, minimum temperature, vapor pressure, and daylength) for calculating PET [33]. Although gridded-based estimations of 0.1° are susceptible to inaccuracies and indeterminacies, all datasets underwent rigorous quality assurance and consistency evaluation according to the standard protocols of the BR-DWGD [9]. BR-DWGD were downloaded from the website (https://github.com/AlexandreCandidoXavier/BR-DWGD/ accessed on 15 September 2023). Figure 1b illustrates the annual precipitation of the SFRB over the period of 1980–2022 from the BR-DWGD.

Daily streamflow data were obtained for gauges of the SFRB in the Brazilian National Water Agency (ANA) over the period of 1991–2020 from the website (https://www.snirh.gov.br/hidroweb/apresentacao/ accessed on 10 August 2023). Streamflow gauges with a record missing more than 15% or those experiencing gaps exceeding two weeks of consecutive data during the specified period were eliminated from consideration in this study [34]. The standardized daily anomalies and then quantile values of the daily anomalies of the streamflow data were calculated for four selected catchments in the SFRB. Figure 1b illustrates the geographic distribution of the streamflow gauges. Daily data and anomalies in the Quantile Streamflow (hereafter QS) data were divided into percentile classes (10, 20, 30, 40, 50, 60, 70, 80, 90, and 100%). For consistency, the QS data were interpolated, and the spatial resolution was fixed to 0.1° using resample methods [35].

To evaluate the relationship between flash droughts and streamflow, gauges were additionally filtered by disregarding catchments which had dams or reservoirs immediately upstream; this was done so that we could consider the streamflow response during flash drought periods. The four catchments were selected from the ANA dataset. The selected catchments had a streamflow gauge close to or at the outlet of the catchment (Figure 1a). The mean catchment areas comprised a total of 373 km². Further, we applied wavelet analysis to quantify and compare the interannual variability of the monthly SAPEI and streamflow during the 1991–2020 period [9].

2.3. Standardized Antecedent Precipitation Evapotranspiration Index (SAPEI)

The SAPEI is a daily-scale drought index proposed by Li et al. [26], which considers both precipitation (P) and potential evapotranspiration (PET). This index incorporates the influence of antecedent water balance conditions on the prevailing dry or wet conditions of the current day. Its effectiveness in identifying and monitoring drought events at daily and weekly scales has been demonstrated in recent flash drought studies, highlighting the use of the SAPEI as a valuable tool for pinpointing and understanding the dynamics of flash drought events [33,34]. Therefore, the D values calculated in the SAPEI represent a climatic water balance and incorporate the antecedent conditions of the previous days. The calculation formula is as follows:

$$D=\sum _{n=0}^N{\alpha }^n{(P-PET)}_n$$

(1)

α^N = c

(2)

where D is a measure of the dry or wet conditions of each date, N is the number of accumulated days, and α and c are the decay and the amount of the last day’s precipitation parameters, respectively. Here, N is 90 days and α is 0.95, within the range suggested in previous studies [36,37,38]. The time series of the D values was calculated using the gridded precipitation and PET obtained from the daily gridded data. A series of probability distributions were then used to fit the D time series, and the Kolmogorov–Smirnov (K–S) test with a 0.05 significance level was used to determine the optimal probability distribution. The distribution candidates considered were the generalized extreme value, lognormal, normal, generalized pareto, and log-logistic, as these are recommended in the literature for fitting drought indices, e.g., [32]. The generalized extreme value distribution was the best fit distribution for the highest percent of grids across the SFRB, covering 44% of the basin. Further, we applied trend and wavelet analyses to quantify and compare the interannual variability of the monthly SAPEI and hydro-climatological data of the SFRB from 1991 to 2020. Through these analyses, our goal was to gain an initial understanding of the response of streamflow to FDs and provide valuable insights for FD management in the SFRB.

2.4. Definition of Flash Drought Events

The delineation and computational approaches for FDs in this study followed the methodology set forth by Barbosa [8,17], using SAPEI values. The SAPEI threshold criteria for detecting the FD events were: (1) two or greater in a 30-day period; (2) less than −1.5; and (3) below −0.5 for at least 30 days. We used different threshold combinations to identify FD events in the SFRB as done in previous studies. The first, second, and third criteria help separate flash droughts and dry spells and identify flash drought events that may have potential environmental impacts over more extended periods.

FD events were identified with the five-day average SAPEI time series using a rolling window of six consecutive intervals (equivalent to a 30-day period) based on the threshold criteria. Here, our emphasis was on capturing the onset of flash drought events as this reflects the rapid fluctuations of FD events, and we were interested in understanding the response of streamflow to the period of rapid intensification. The choice to set the average at 5 days considered the potential for longer-term dependencies in the data, allowing for an examination of how past values of the SAPEI may have influenced current quantile streamflow.

3. Results

3.1. Mapping Flash Drought Events with the SAPEI

We classified five FD events from different lengths to explore the mapping capabilities of the SAPEI. This helped in studying how persistent and severe the FD was and how long the drought would last for a particular event.

Table 1. Summary of statistics for FD events in the SFRB from 1991 to 2020 identified using the threshold criteria for the SAPEI.

Event	Onset [Date]	End [Date]	Length [Months]	Average SAPEI [-] 1	Dry Area Peak [%]	Severity [-]
E1	March 1993	February 1994	12	−2.18	99.60	21.74
E2	April 1998	February 1999	11	−1.46	91.34	14.29
E3	March 2012	November 2013	21	−2.63	100.00	42.01
E4	November 2015	December 2015	2	−1.47	89.71	2.59
E5	January 2017	January 2018	13	−2.63	95.40	23.66

Note(s): The results indicated that the extreme dry events from 1991 to 2020 were identified using a threshold of the minimum value of ¹ SAPEI ≤ −1.5.

shows the results of statistics for FD events across the SFRB during the 1991–2020 period. For every event other than E2 and E4, a SAPEI intensity less than −1.5 was predominantly seen. But E3 showed a more extensive and severe drought range compared to the other four events. For E3, which occurred from 12 April 2012 to 12 November 2013, the drought severity for the whole of the SFRB was significantly high, with a value for the mean SAPEI less than −2.6; hence, this dry condition could be linked to its rapid onset and high severity, which have had detrimental effects on water resources, agriculture, and ecosystems [29]. As seen in Table 1, E3 indicated the worst extreme drought conditions in the climatic water balance in this region, representing severity (−42), length (126 pentads), and spatial coverage of about 92%. This is consistent with previous studies [9].

The years 1998 (E2) and 2012 (E3) were the extreme high FD proportions, respectively. The year 1998 had a strong El Niño and the average SAPEI value was less than 1.5 for more than three consecutive pentads, indicating extreme drought conditions. In 2012, the entire SFRB was severely impacted, where the experienced persistent FD proportions were higher than those observed during the El Niño drought events. This year was a La Niña year; hence, dry condition could not be linked to the effect of the El Niño. In the comparison of these two FD events, the SAPEI demonstrated the ability to reflect droughts from different causes. However, the causes for the severity in all five FD events call for a separate study.

In terms of spatial distribution, FD persistency (percentage) across the SFRB was identified using the threshold of a minimum SAPEI ≤ −1.5 from 1991 to 2020 as shown in Figure 4. At each grid cell, the corresponding value of 100% means that the FD conditions occurred throughout the entire length (event duration). It is observed that the areal extent of the SAPEI ≤ −1.5 indicated a clear FD persistency signal in this region for each event. However, it included areas without FDs. The SAPEI results not only encompassed the drought-free areas identified by persistency values ≤ 40% but also provided a clearer and more detailed delineation of the FD persistency between areas experiencing an extreme high proportion (80%) and an extreme low (40%) proportion. In E3, the extreme high proportion of the FD persistency occurred in the whole of the SFRB. In this event, the SAPEI showed similar FD ranges, while the FD persistency in E2, E4, and E5 was relatively weaker. This may suggest that, in this specific region and time, the SAPEI had a stronger capability to capture severe FD signals. In E2, E4, and E5, the extreme high values of FD persistency proportion were in the central and northern areas of the SFRB. This finding concurs with previous research indicating that episodes of FD have grown more frequent since the year 2010 [8,17].

The spatial characteristics of extreme high FD persistency events from the 1991 to 2020 period showed distinct differences between the center–southern (upper streamflow) and center–northern (lower streamflow) regions of the SFRB. Apart from E1 and E3, E2, E4, and E5 had the highest persistency of FD events but lower durations and severities. Notably, the center–southern areas of E1 and E3 experienced slightly higher extreme FDs than other events. However, these areas exhibited significantly higher FD durations, intensities, and severities compared to E2, E4, and E5. The center–northern region of the basin is commonly in a state of drought, with a high intensity, wide range, and profound impacts [9]. The primary reasons include scarce rivers in the northern areas of the SFRB, which are influenced by atmospheric circulation anomalies. The intensity, weakness, and fluctuations of the tropical high-pressure system restrain low-latitude climate [39,40]. The placement and strength of the tropical high-pressure system can affect precipitation patterns, whereas the displacement of the Inter-Tropical Convergence Zone (ITCZ) is related to El Niño Southern Oscillation (ENSO) events [41,42]. For example, the emergence of an EI Niño results in deviations in global climate, serving as a significant driver behind droughts and floods in the SFRB [9]. In addition, excessive irrigation, urban and industrial water use, and human activities have exacerbated water shortages and vegetation degradation (desertification), further worsening the drought conditions [17]. Although the number of FD events in the center–southern area is less frequent, their intensity and duration are higher than in other areas. Paredes-Trejo et al. [9] indicated that the duration of droughts in the upper and center parts of the SFRB is likely to increase.

In terms of spatial distribution, the minimum SAPEI values across the SFRB ranged from −1 to −3.5. According to Figure 4, different spatial patterns were observed in these extreme events. Spatial differences in the dry extreme events among E1, E2, E3, E4, and E5 in the SFRB were prominent, particularly in the northern and southern SFRB. However, the mean minimum SAPEI value was −1.8, which likely explains the decreased precipitation and increased temperatures [26,28]. Areas with shorter minimum lengths may experience more frequent but shorter-lived droughts, whereas those with longer maximum lengths may face prolonged periods of drought stress, potentially leading to more severe impacts on water availability and ecosystem health [39,40].

3.2. Relationship between the SAPEI and Streamflow during Flash Drought Events

It can be observed from Figure 5 that the correlation of the SAPEI and QS revealed a linear relation with a statistically significant value. This relationship between the two variables was found to be 0.32, significant at the 0.01 level. According to Goodman–Kruskal’s gamma coefficient, the wet–dry conditions induced the strongest response in SAPEI values during a period of six pentads (i.e., 30 days) if a time lag equal to zero is used; the intensity of the SAPEI–QS linkage was higher when using a maximum lag of six pentads (gamma = 0.24; p < 0.05). The gamma coefficient computes the variance rate of two time series where one is lagged [43]. The chosen maximum lag provides basic insights for understanding the correlation between two variables. However, the decision to set the maximum lag at six pentads considered the potential for longer-term dependencies in the data, allowing for an examination of how past values of the SAPEI may influence the current streamflow. This may be because, although the SAPEI represents the accumulation of meteorological drought, it does not accurately depict the specifics of hydrological drought as effectively as QS [26,28].

Figure 5a shows the pentad moving average of QS and the SAPEI from 1 January 1991 to 31 December 2020. The QS values began to decline around the same time that the SAPEI crossed the −0.5 drought threshold, with the QS from the Boqueirão streamflow gauge exhibiting a decrease below the 40th quantile threshold. The minimum QS values were experienced on 22 July 2017, reaching less than the 10th percentile. The QS increased back to the 30th percentile while the SAPEI experienced a peak on 29 December 2020.

Figure 6 shows the changing trend of SAPEI values from 1991 to 2020, representing statistically significant results at the 0.95 confidence levels. Here, dry SAPEI values (the red color) are commonly observed in the central SFRB, whereas wet SAPEI (the blue color) values predominantly cluster in the southwest areas, aligning with the drying and wetting trends in the SFRB. In contrast, the rising trend of extreme dry events was greater than the wetter trend. Nevertheless, the SFRB has undergone a remarkable climate transition from wet to dry [17], particularly in the central areas. Overall, the trend distribution of increasing dry conditions was the highest percentage of grids across the SFRB, covering about 55% of the basin. However, the driving factors of the drying trend in this basin may be more related to extreme climatic events or specific surface water conditions rather than solely temperature or vegetation conditions [8]. The central SFRB, being arid and semi-arid, has low precipitation and vegetation cover, including soil degradation, water resource depletion, and ecological disruptions [17,29]. Agricultural production in this area is mostly irrigation-based, leading to a non-linear relationship between soil moisture, vegetation, and temperature changes, which might explain the wetness condition [44]. Hence, the SAPEI can depict the wet–dry conditions in the study region.

Further, we analyzed the local- to sub-regional-scale hydro-climatological conditions in the central and northern areas of the SFRB. It is important to note that streamflow gauges across the region are very limited. On the catchment scale, the Boqueirão streamflow gauge, located in the Mansidão municipality in the state of Bahia (11°21′ S, 43°50′ W; in the center part of the SFRB; 401 m above sea level), was used daily in the validation process. The Boqueirão station is labeled with an ID# 1143010, following the nomenclature of ANA. Due to this lack of data, it is often understood that the streamflow over the entire center region of the SFRB varies in the phases of interannual and longer time scales, including over the center–northern part where little to no information is available [9]. This understanding is supported by comparing the trends of streamflow from the Boqueirão gauge and the climatology of both precipitation and temperature over the center part of basin during the 1991–2020 period. The trends obtained from the annual estimates of streamflow, precipitation, and temperature are shown in Figure 7. From the results, the streamflow decreased 950 m³/s per year during the last 30 years, i.e., from 1991 to 2020. Annual rainfall on the central SFRB scale showed a decreasing trend (−1.20 mm/ year) with statistically significant values (Figure 7a). The mean annual rainfall for the period of 1991 to 2020 was 450 mm (Figure 7c). On the contrary, the increasing trend of air temperature (0,021 °C per year) was identified in the study period (Figure 7b). These results indicate that the decreased trend in streamflow was attributed to a decline in precipitation and due to an increase in air temperature observed during the study period.

To assess the relationship between QS values and the SAPEI, the discharge at the Boqueirão streamflow gauge was used as a benchmark time series; this was done so that we could consider the streamflow response to the extreme FD events in the central SFRB. This context may be used to understand the relationship between QS values and the SAPEI. The SAPEI provides insights into drought meteorological conditions, while QS data offer a direct assessment of hydrological responses. In this case, the QS values were the standardized anomalies of daily streamflow values assigned to ten quantiles, and the SAPEI values were spatially averaged for the middle part of the basin. For catchment-specific analyses (69,995 km²), the Boqueirão streamflow gauge within the SFRB (Figure 1b) was selected for the comparison.

As shown in Figure 5, the behavior of the QS and SAPEI was nearly similar, though the minimum SAPEI value was only −0.6. While the decline and recovery were nearly similar, the peak of the decline in the QS values happened before the peak in the SAPEI values, this time with minimum QS values below the 30th percentile from 2004 to 2011. Notably, the 5-year moving average SAPEI values decreased significantly more than the QS values. Between 2012 and 2018, the year with the most FD events in the SFRB was 2012. During this year, the SFRB experienced multiple high-temperature heat waves. The year 2012 was an extreme dry year, with the precipitation in the SFRB less than 30–60% of the annual average [9]. In Figure 7, regardless of the spatial scale, the trend of the Boqueirão streamflow was similar to the rainfall change. In areas where FD events increased, the temperature also increased, especially in the northern areas of the SFRB. The precipitation and streamflow in the northern SFRB had significant decreasing trends, at 3.43 mm per year and 31.40 mm³/s per year, respectively.

Prolonged periods of elevated temperatures and minimal rainfall intensify the loss of soil moisture through evaporation. Land–atmosphere feedbacks connect all these variables and, qualitatively, consider all principal processes in drought evolution (e.g., [8,17]). We even extended this interpretation to directly relate to higher surface albedo to higher radiance values of MSG-SEVIRI images covering Brazil. The visualization of the evolution of radiances in the MSG-SEVIRI images with the longest flash droughts (Figure S1) can be found in the supplementary material.

The squared wavelet coherence between the monthly SAPEI and streamflow series plots in Figure 7 exhibit an anti-phase relationship, passing the 99% confidence test. Periodicity with 4–12 months was predominantly high from 1991 to 2020, suggesting the complex interplay between meteorological and hydrological processes during FD events. Given that the SFRB is large and heavily managed, drought conditions in one region may not necessarily reflect conditions in another part of the basin. Therefore, even if one region is not experiencing meteorological drought conditions, drought conditions in another part of the basin could still have a significant effect on downstream locations, underscoring the interconnected nature of water resource management within the SFRB.

4. Discussion

Several studies have confirmed that climate change plays a crucial role in the generation and development of FD events [7,8,11,12,13,15,17,18]. This study aims to gain an understanding of the relationship between FDs and streamflow within the SFRB. The overarching goals were multifaceted, encompassing the demonstration of the efficacy of the daily-scale drought index, the SAPEI, in capturing FD events in the SFRB. Our findings revealed that this index can accurately assess drought in the SFRB from both temporal and spatial perspectives. Table 1 shows a summary of five extreme FD event statistics. Across all five events, the period of rapid intensification was marked by a transition in the SAPEI from above 0 to below −0.5, aligning with a simultaneous reduction in streamflow to below the 20th quantile. This indicates that there may be a significant synchronized response between atmospheric and hydrological conditions during the onset of FD events. In the 2012–2017 FD events, large precipitation deficits and high temperatures led to rapid drought conditions across the SFRB [45]. Additionally, it is noteworthy that at the basin scale, five FD events may occur, persisting into long-term droughts or expanding to larger areas. Subsequently, the D values computed for the SAPEI represented a climatic water balance and incorporated the antecedent conditions of the previous 90 days [26,33,46]. This was in line with what was observed in Figure 5, Figure 6 and Figure 7, where it is shown that these events mostly impacted the basin. Under climate change, the probability of FD events in the SFRB has substantially increased because of the influence of atmospheric circulation anomalies and land–air coupling processes [8,9,10,17]. Barbosa [8,17] confirmed the mechanism of FDs, and the decreased soil water content leads to the increased probability of rapidly developing drought. Increased temperatures accelerate soil evaporation and plant transpiration. This reduces soil water content, which, in turn, inhibits soil evaporation and plant evapotranspiration. However, decreased soil moisture leads to increased soil–air sensible heat interactions. This land–air interaction feedback mechanism in rapidly developing drought is a significant driver of FD events in the SFRB (Figure S1).

It is noteworthy that at the basin scale, the length of FD events indicated that this basin experiences longer periods of FD and is expanding to larger areas. While this pattern may initially imply a lower immediate impact, it also raises concerns about the basin’s preparedness to manage the potential consequences of a severe flash drought event, given their increased historical intensity [29]. Furthermore, the central and southern areas of the SFRB stood out for experiencing the lowest minimum SAPEI values, indicating the most severe FD conditions. Ultimately, these aspects pose certain challenges for hydropower production and energy security, such as those observed between 2012 and 2017 [45,47,48]. With the heightened intensity of FDs in the upstream basin, this may be exacerbated by water management practices in downstream areas.

The observational data showed that, compared with the average from 1991 to 2020, the precipitation in the northern SFRB generally showed a significant decreasing trend, at 3.43 mm/yr, with high precipitation variability and extremely uneven spatial distribution (Figure 1), making it among the driest areas situated in northeastern Brazil. It was observed that dry conditions prevailed during El Niño events in the SFRB rainy season period. El Niño events had an impact on the amount of precipitation received by this basin. In addition, long-term rainfall conditions had a significant impact on the effect of drought on the NDVI. This high precipitation variability reflected different vegetation responses to drought (Figure 2).

According to Figure 6 and Figure 7b, in areas where extreme dry events increased, the temperature also increased. Multiple studies have confirmed trends of increasing temperature and decreased precipitation in the SFRB; however, our research indicated that the increased temperature was more significant than the decreased precipitation, which was confirmed by the changing trend of specific streamflow (as shown in Figure 7a). Further analysis between flash drought events and streamflow is needed to explore the probabilistic characterization of spatial characteristics of drought/flash droughts [49,50,51]. The results of the squared wavelet coherence between the SAPEI and streamflow values during all the periods of flash drought identified for four catchments in the SFRB (Figure 8) revealed the complex and localized nature of the interaction between meteorological conditions (Figure S1), as captured by the SAPEI, and the subsequent hydrological responses in terms of streamflow. Such variability highlights the need for catchment-specific analyses and the consideration of local factors influencing flash-drought–streamflow dynamics. In future research, it would be beneficial to expand these analyses by considering basin characteristics as potential predictors. In addition, by establishing more streamflow gauges, refining observation methodologies, and harnessing sophisticated algorithm approaches [48], the precision of future forecasts for FD events can be increased.

5. Conclusions

The assessment of SAPEI’s capability to effectively capture flash drought events revealed its proficiency in delineating the onset, development, and recovery phases. Overall, this paper shows that the middle part of the SFRB experiences frequent, longer droughts, while the south part encounters shorter but less frequent events. Furthermore, it underscores the variances in lag effects of flash droughts indicated by the SAPEI on subsequent reductions in streamflow, enriching our comprehension of the onset of flash drought responses to streamflow reductions. However, further analysis between the SAPEI and streamflow is needed to explore lagged impacts. This work furnishes the scientific direction for informing decision making and enhanced resilience against the impacts of flash drought events across the São Francisco River Basin (SFRB) in northeastern Brazil.