Investigating watershed hydrology from a data-driven causal perspective is an attractive opportunity to characterize and understand relationships between water storages and fluxes. Here we assess integrally how the water balance components interact with themselves, aiming to find relevant time-lags or dependency patterns. Granger’s causality test and time-lagged mutual information were used in a pairwise approach to examine cause-effect relationships between precipitation, streamflow, groundwater levels under different land-covers, and evapotranspiration data (daily timescale) from 2009 to 2019 in a Brazilian watershed (52 km²), located in a recharge area of the Guarani Aquifer System. A verification assessment using synthetic datasets shows that the methods are effective to identify the underlying generating mechanisms. Statistically significant causal connections were confirmed in practically all pairs of observed data. Granger’s causality indicates that groundwater and streamflow responses are influenced by precipitation even with a lag of 1-day, while evapotranspiration can take more than 200 days to influence groundwater responses, depending on the water table depth and surrounding land-cover. Mutual information curves show dependency patterns between hydrological processes that are different from the ones obtained by cross-correlation functions. The causal analysis provides a complementary view of the hydrological system’s functioning and may lead us to develop predictive models that reproduce not only the target variables but also the diverse time-lagged dependencies observed in environmental data.