AbstractStreamflow regionalization is a technique used in areas where hydrological data are scarce or nonexistent. Although numerous studies have been conducted with the aim of improving this technique, unsatisfactory results are still evident. This study, therefore, aims to propose and evaluate the performance of a new explanatory variable for regionalization, which represents the streamflow formation process and considers the actual evapotranspiration (ETR) obtained from remote sensing products. For this purpose, the regional regression method was used to estimate long-term mean streamflow and minimum flow with 90% of permanence in time. The explanatory variables were the precipitated volume (Peq), precipitated volume minus the empirical value of 750 mm (Peq750), water balance for each segment of the water course (WBeq), and water balance for each hydrologically homogenous region (WBeqM). These variables were obtained using a combination of drainage area, precipitation, and ETR. The ETR was estimated using two remote sensing products: MOD16 and Global Land Evaporation Amsterdam Model. The models of streamflow regionalization were evaluated by statistical, physical, and risk analyses. The study was applied for Grande River basin. All the variables, with the exception of Peq, presented good statistical performance, good representativeness of regionalized streamflows, and safe estimates for the planning and management of water resources. The WBeqG was the most recommended variable for streamflow regionalization in Grande River basin, because it considers variations in edaphoclimatic and vegetative conditions along the basin. This work contributes to improving the predictive capacity of the streamflow through a method that is potentially applicable to other areas of study, allows easy physical interpretation, utilizes easily obtained variables, and represents the streamflow formation process.