Combining remotely sensed actual evapotranspiration and GIS analysis for groundwater level modeling

This paper describes a combination of large-scale spatially remote-sensed actual evapotranspiration and geographical information systems of surface runoff to estimate groundwater recharge potential, by water balance, to model groundwater level. We selected an area nearby the city of Águas de Santa Bárbara, central-western part of São Paulo State, Brazil to be analyzed between 2014 and 2018, during recent El Niño-Southern Oscillation most active period (2016/2017) and verify its posteriori effects on vegetation (until early 2018). The Simple Algorithm for Evapotranspiration Retrieving (SAFER), with biweekly MODIS and seasonal Sentinel-2 imagery, and the rational method for runoff modeling were applied, which was applied in a daily-basis hydrological model, which is the main novelty of this report. The average annual groundwater recharge potential for each of the land uses (pasture, sugarcane crop, silviculture, and forest) varied between 15 and 50% of the rainfall. Silviculture showed higher evapotranspirations rates than forest and sugarcane crops. Groundwater levels measured at 46 monitoring wells were analyzed to obtain enough data to create the hydrographs required for the validation. 36 shallow wells (which reached depths smaller than 3 m) had the best results (R² > 0.92), where the root mean squared absolute error (RMSE) term appeared to be less than 20% of the mean groundwater level, indicating that it has a faster response to remote-sensed actual evapotranspiration and that the water balance is sufficiently understood for policy and decision making. The results of this study lead to the conclusion that including spatiotemporal thermal and soil moisture conditions by remote-sensing tools in the evapotranspiration account improved the modeling of groundwater levels for shallow wells at daily basis.