Simulation of soil water content through the combination of meteorological and satellite data

Recent investigations have supported the possibility of predicting soil water content (SWC) through a simplified soil water balance (WB) model fed with remotely sensed actual evapotranspiration (ETa) estimates. This approach, however, requires information on main soil features (i.e. depth, wilting point, field capacity) which are generally difficult to retrieve over large regions. The current paper proposes an alternative model which directly predicts SWC relying on the same logic of the recently proposed ETa estimation method, i.e. the combination of meteorological and normalized difference vegetation index (NDVI) datasets. The theoretical bases of the old and new SWC estimation methods are first described. Both methods are then applied in Central Italy using ancillary and Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI data having a spatial resolution of 250 m; the outputs obtained are assessed versus SWC measurements taken both continuously and instantaneously. In the former case three ecosystems are analyzed (i.e. the grassland of Amplero and the forests of Barbialla and Amiata), where SWC was measured by probes during three different years (2003, 2012 and 2018, respectively). The latter experiment concerns 362 sample sites where SWC gravimetric measurements were taken during a similar time span (2000-2018). In both cases SWC is first normalized into relative SWC (RSWC), which is more directly influential on vegetation conditions. The results of both experiments indicate that the two simulation methods perform similarly when the former is driven by adequate information on soil features. The main limitations of the two simulation approaches are due to the spatial resolution mismatch between SWC measurements and estimates, which has a relatively minor impact on the first three homogeneous areas but is decisive for the other sampled sites. In general, the new simulation method is capable of predicting RSWC with relatively high spatial and temporal resolution without the use of specific soil information.