Climate change in the Amazon includes the intensification of the hydrological cycle in association with the increase of extreme events. This study aims to identify the impacts on groundwater (GW) within a pristine micro-scale catchment in the Central Amazon related to the observed alterations on hydrological cycle. Precipitation and GW levels from seven piezometers distributed along a hydrological transect covering three zones (lowland, slope and plateau) were collected for the period 2001-2021. External (evapotranspiration, GW storage and climate patterns) and computed (annual recharge) data were used to strengthen the analyses. We identified a generalized growth trend in all compartments (rainfall, evaporation and water table). There was a heterogeneity in water table fluctuations and annual recharge, depending on distance from creek channel and water depth. Surface processes influenced shallows piezometers associated with rapid overflow and low recharge. Conversely, deep piezometers involved slower processes related to water movement, and were more representative of regional GW. Those presented larger seasonal and inter-annual fluctuations, annual recharge and positive trend than shallow piezometers. Besides, all piezometers showed large inter-annual variations in recharge. The El Niño-Southern Oscillation (ENSO) influenced GW level, recharge and storage: positive phase contributed to lower all variables, with recovery in during neutral and negative phase. The larger positive trends in the deeper piezometers are a sign of the resilience of the Alter do Chão aquifer, subject to anthropogenic pressure. As water table play a key-role in shaping the structure and productivity of Amazon forests, further hydro-ecological studies should be conducted to gather information about the fate of GW-dependent Amazonian ecosystems.

The spatio-temporal land cover dynamics of a medium‐size floodplain system along the Amazon/Solimões River (Janauacá Lake, 786 km 2) and their hydrological impacts are studied through remote sensing and modeling. Hence, the analysis of 5 satellite-derived land cover maps (1972-2016 period) reveals a decrease in natural environments (from 65% to 35%) to the benefit of anthropic classes (from 17% to 51%) through deforestation vectors (two highways and lake banks). Deforestation is a non-stationary process with significant increase over specific subperiods (1972-1986, and 2005-2016). It occurs in stages with conversions into secondary vegetation then into non-natural environments. 7 land cover scenarios (5 satellite-derived, 1 deforested and 1 forest, used as reference) are used as inputs to run simulations with the same meteorology over the 2006-2018 period. Beside high ( ≥ 24%) and low ( ≤ 7%) interannual variability of runoff-rainfall ratio (RRR) and evapotranspiration (ET), the numerical experiments evidence, on an annual scale, the RRR decreases and the ET increases with deforestation increases. Deforested scenario suggests a convergence: for the RRR, around 0.34 (-87%) and for the ET, around 1146 mm.yr -1 (+6%). At the seasonal scale, the landuse/landcover changes (LUCC) induce positive wet season ET anomaly (<9%) and large negative dry season RRR anomaly (-87%). The highest LUCC-induced disturbances (from -15% to 18%) in the FP mixture are recorded at seasonal scale, during LW and RW and, at interannual scale, during dry and normal HY. The LUCC-induced disturbances patterns of FP mixture mainly concern river and runoff. They are different regarding the hydrological period or HY type. Our experiments suggest the existence of a tipping point between present land cover (2016) and fully deforested cover associated with reversal phenomena and enhancing of seasonal and interannual LUCC-induced disturbance. At last, the model shows the LUCC augment the vulnerability associated with drought periods.