Holistic analysis of the carbon and water cycles to quantify the human footprint in basin-wide hydrological processes in the Amazon

While land-cover clearing (LCC) immediately reduces evapotranspiration (ET), its effects on other water fluxes, such as river discharge and terrestrial water storage, exhibit contrasting responses depending on location and scale. One explanation for this is that LCC triggers a series of asynchronous disruptions in the equilibrium of hydrological processes that was established upon the long-term balance with regional climatological, edaphic, and geological characteristics. Water fluxes under these circumstances are not well represented by hydrological models that have Budyko-like approaches or rely on the stationarity of the hydrological responses. The complexity of such analysis is incremented once LCC is followed by the conversion to pastures and crops established over random spatial and temporal patterns throughout river basins. Here, we propose an analysis of river discharge and root zone storage capacity (RZSC) to unveil underlying relationships between stream dynamics and water consumption by plants. We use a time-series segmentation and residual trend analysis on streamflow and precipitation of high-order tributaries of the Tapajós River in the Amazon whose catchments underwent an intense land-use change over the past decades. We estimate the RZSC using the mass-curve balance method by considering the annual land-cover changes over a >30-year period. Despite the common belief that increases in river discharge are primarily caused by reduced ET when precipitation trends are not significant, we show that this might not be the main trigger of streamflow change in these major Amazon catchments. Instead, the reduction in the RZSC caused by changes in the water consumption by plants over the dry season is tightly associated with the increased baseflow contribution to rivers. Finally, we analysed gross primary productivity (GPP) and ET estimates generated by a model based on eco-evolutionary optimality that integrates the water and carbon cycles at the canopy level. We found that trends in ET from croplands are not as pronounced as trends in GPP. Although RZSC is quantified using the water deficit driven by ET, changes in RZSC are more correlated to changes in GPP. Our results highlight the importance of considering the carbon cycle in hydrological assessment studies.