Large-Scale Evidence of Non-stationarity in Brazilian Streamflows Across Hydrographic Regions

We investigate non-stationarity in streamflow regimes across Brazil through a large-scale assessment of trends in annual maximum, mean, and minimum discharges under climate variability and change. The analysis is based on daily streamflow records from 515 Brazilian catchments with at least 30 years of continuous observations (1980–2010), obtained from the Catchment Attributes for Brazil (CABra) dataset. For each catchment, annual maximum discharge (Qmax), annual mean discharge (Qmean), and annual minimum 7-day average discharge (Q7,min) were derived using the hydrological year. The primary objective was to identify coherent spatial patterns of hydrological change across major hydrographic regions and to determine which components of the flow regime are most sensitive to non-stationary signals. With this design, we aim to address two research questions: (i) Do non-stationary signals exhibit distinct spatial patterns across Brazil’s diverse hydroclimatic regions? (ii) Which streamflow metrics (high, mean, or low flows) are most sensitive to long-term changes? We adopt a regionalized assessment approach, applying the non-parametric Mann–Kendall test and Sen’s slope estimator to quantify the significance and magnitude of trends. The findings reveal a marked spatial dichotomy and strong metric-dependent sensitivity to non-stationarity. A pervasive decline in minimum flows (Q7,min) is observed across central and northeastern Brazil, indicating a systematic loss of catchment buffering capacity and baseflow resilience. Specifically, the São Francisco basin emerges as the most critically affected region, where 86.3% of catchments exhibit significant reductions in Q7,min  and 45.2% show decreasing trends in  Qmean. Similarly pronounced declines in low flows were identified in the Parnaíba (70%), East Atlantic (>60%), and Tocantins–Araguaia (54%) basins. Conversely, the Amazon basin displays an intensification of the regional hydrological cycle, with approximately 20–27% of catchments showing increasing trends across all flow metrics ( Qmax, Qmean, and Q7,min). Outside the Amazon, trends in  Qmax remain largely stable, suggesting that changes in extreme high flows are less widespread than those affecting low-flow conditions. Overall, our results demonstrate that minimum streamflows are the most sensitive indicators of non-stationarity in Brazilian hydrology. The severe depletion of baseflows—particularly in the São Francisco basin—poses significant risks to water security, hydropower generation, and the viability of large-scale interbasin water transfer projects. The study underscores the limitations of the stationarity assumption in traditional water management and emphasizes the urgent need for region-specific adaptation strategies to manage the increasing vulnerability to hydrological droughts.