Floods and droughts are the costliest natural hazards globally, and their risks are exacerbated by climate change and human activities. While increased rainfall often leads to more floods and fewer droughts, and vice versa, some models suggest a simultaneous increase in both flood and drought severity—an 'acceleration' of the water cycle. This acceleration could severely impact global food production, ecosystem health, and infrastructure. Several processes can cause water cycle acceleration, including increased atmospheric moisture capacity in a warming climate leading to extreme rainfall and changes in atmospheric and oceanic circulations. Land management practices like agriculture, river engineering, urbanization, and groundwater pumping also contribute. This study focuses on Brazil, a region with some of the world's largest river basins and mounting concerns about changing flood and drought patterns, to analyze the combined effects of climate variability and land management on streamflow extremes over the past four decades.

Previous research in South America has examined the impact of climate change and land management on streamflow, but often separately. Studies have shown various trends in streamflow changes, including shifts in the South American Convergence Zone (SACZ), changes in the Intertropical Convergence Zone (ITCZ), and impacts of El Niño-Southern Oscillation. Research on deforestation's effect on streamflow has yielded mixed results, with some studies suggesting increased low flows due to reduced transpiration and others showing increases in both high and low flows. The limited availability of comprehensive streamflow data and the complex interaction between human activities and the water cycle have hindered a comprehensive continental-scale assessment of these intertwined factors. This paper addresses this gap by combining climate, land cover, and human water use data sets for a thorough analysis of the changes in Brazilian streamflow.

The study uses daily streamflow observations from 886 hydrometric stations in Brazil from 1980 to 2015. Annual time series of annual minimum 7-day streamflow (drought flows), mean daily streamflow, and annual maximum daily streamflow (flood flows) were computed for each station. Trends were quantified using the Theil-Sen slope estimator and Mann-Kendall test. Regional trends were obtained through spatial interpolation with ordinary kriging. Three climate drivers were considered: mean daily atmospheric water balance (P-E), annual minimum 90-day P-E, and annual maximum 14-day P-E. Two non-climatic drivers were also included: water use for irrigation and other purposes, and native vegetation cover. Trends in streamflow and drivers were expressed as % per decade. Panel regressions with fixed effects for location and log-transformed variables were used to analyze the links between streamflow changes and drivers. Bivariate frequency distributions were used to analyze water cycle acceleration by examining the relationship between flood and drought flow trends. Four hotspots representing distinct streamflow regimes, land management, and locations in major basins (Amazon, São Francisco, Paraná, Uruguay, and Iguaçu) were identified for a more detailed analysis. Time series were standardized to facilitate comparison between hotspots. The return periods of droughts and floods were analyzed using generalized extreme value (GEV) distributions. Data sources included the CAMELS-BR dataset for streamflow, CHIRPS v2.0 for precipitation, GLEAM v3.3a for evaporation, ESA/CCI Land Cover v2.0 for land cover, and the ANA's Manual of Consumptive Water Use in Brazil for water use.

The analysis revealed widespread streamflow changes in Brazil. Decreasing low flows (increasing drought severity) were observed in southern Amazonia and central-eastern Brazil, while increasing flood flows occurred in Amazonia and the southeast. Regression analysis showed that drought trends were primarily driven by changes in mean daily P-E, with substantial effects of water use and minimum P-E. Water use impacts were particularly noticeable in central-eastern Brazil. Flood changes were linked to maximum P-E and mean daily P-E, indicating that floods respond to modified extreme precipitation and antecedent soil moisture. Analysis of four hotspots revealed distinct patterns. In southern Brazil and northern Amazonia, drought flows aligned with increasing mean and minimum P-E with minimal land management effects. Floods in southern Brazil increased with increasing maximum and mean P-E. In the Brazilian Highlands (intensive agriculture), reduced drought flows aligned with decreasing mean P-E and increasing water use. Southern Amazonia showed substantially decreased drought flows despite minimal climate variable changes, suggesting deforestation effects. Quadrant classification of streamflow trends revealed that 29% of the study area exhibited an accelerating water cycle (increased floods and droughts), 25% showed wetting (increased floods and droughts), and 42% showed drying. Analysis of drivers within these quadrants showed that positive mean P-E trends were associated with wetting, negative trends with drying, and increased water use amplified the decreasing trends in the drying quadrant. The accelerating quadrant was dominated by increasing maximum P-E trends and decreasing native vegetation cover. Changes in return periods showed that a 10-year return period drought in the Brazilian Highlands became a 1-year drought, and a 100-year flood in southern Amazonia became a 25-year flood.

The findings demonstrate a clear, spatially coherent signal of streamflow changes attributable to the combined effects of climate variability and land management. Drying trends in central and northeastern Brazil are likely linked to southward shifts in the SACZ and northward shifts in the ITCZ, along with increasing agricultural water use. Wetting trends in northern Amazonia may be related to northward ITCZ shifts, while those in southern Brazil might be associated with El Niño-Southern Oscillation and SACZ changes. Acceleration in southern Amazonia is linked to ITCZ shifts, increased evaporation, intensified extreme wet-season precipitation, and deforestation. Deforestation reduces soil infiltration, increasing surface runoff and reducing groundwater recharge, resulting in both increased floods and decreased drought flows. It can also increase extreme precipitation through warmer land surface temperatures and altered moisture recycling. These changes have substantial regional and global impacts, such as threats to agricultural productivity and food security, increased flood hazards, and potential shifts in the Amazon rainforest from a carbon sink to a source. The observed acceleration in Brazil aligns with global climate model projections, suggesting similar changes may occur in other regions.

This study demonstrates a significant acceleration of the terrestrial water cycle in Brazil due to the combined effects of climate change and land management, particularly deforestation and water use for irrigation. These changes pose considerable risks to agricultural production, infrastructure, and the Amazonian rainforest ecosystem. The findings highlight the urgent need for both climate mitigation efforts and adaptation measures, including sustainable land management practices, to mitigate the compound risks of floods and droughts and maintain food security and infrastructure safety.

The study's conclusions are based on data from 1980-2015. Changes in data collection methods or future trends could alter the results. The analysis relies on several data sources with inherent uncertainties, including those related to precipitation, evaporation, and water use estimates. The spatial resolution of some datasets might limit the precision of the analysis, especially in areas with limited gauge density. The study focuses on Brazil and its findings may not be directly generalizable to other regions with different climatic conditions and land management practices.