Geological and climatic controls of low flows in Brazil

Low flows are generated by the interplay of climatic variability and basin water storage dynamics, which depend on basin attributes such as geology and soil properties. Even though low flows have been predicted worldwide, their controls and generation mechanisms remain elusive in many regions, such as South America. Here, we investigate the relative importance of climate and basin attributes in the spatial variability of minimum annual 7-day streamflow magnitudes (Q_min) of 1412 river basins in Brazil. We analyze time series of observed daily streamflow, precipitation and evaporation from 1980-2020; geology such as rock type and hydraulic conductivity; topography; soil properties such as sand content and class; and land cover. We estimate Q_min with a simple conceptual model that separates the roles of climate and basin attributes in regulating low flows. For each river basin, we identify the longest annual dry spells and estimate Q_min using an exponential decay model with three components: (i) initial flow, which indicates the basin’s water storage at the dry spell onset; (ii) dry spell length (T_dry), representing climate seasonality, computed from a 31-day precipitation minus evaporation series; and (iii) flow recession rate (Q_rec), indicating how quickly the basin releases the water stored. We estimate the initial flow component with the fraction (β) of mean annual precipitation minus evaporation (PE_m) that recharges the aquifer. This fraction is estimated from basin attributes using model-based recursive partitioning, a method similar to regression trees, in which we found soil properties as the predominant attributes. The flow recession component is estimated likewise, in which we found rock type and composition as the predominant attributes. We found that the model explains 56% of the variance in observed Q_min. The large-scale patterns show a close match. Results show that the relative importance of climate and basin attributes depends on the spatial scale of analysis. Climate and basin attributes are similarly important at the national scale, in which changing PE_m, T_dry, Q_rec, and β by one spatial standard deviation change estimated Q_min on average by 41%, 57%, 66%, and 31% respectively. On the other hand, basin attributes control low flow variability on subnational scales. Analyzing blocks sized 300 by 300 km, changing PE_m, T_dry, Q_rec, and β by one spatial standard deviation in each block change estimated Q_min on average by 19%, 11%, 36%, and 19%. Our interpretation is that the spatial variability of low flows is regulated mainly by the basin’s water storage capacity, here driven by rock type and composition, which even compensates for the highly seasonal climate from the South American monsoons. For example, most of the highest low flows (i.e., Q_min above 1 mm/d) are located in high-storage sandstone aquifers, a common aquifer type in Brazil. These highest low flows rarely occur in low-storage aquifers, which require a combination of high annual precipitation (i.e., above 3000 mm/yr) and the absence of a dry season such as in northwestern Amazonia. These findings can contribute to water security by estimating the impacts of climate change and variability on droughts.