Assessing the ability GEOframe modeling system for water budget analysis of a challenging karst basin in the Apennines chains, Central Italy.

The analysis of the water budget in the Upper Nera River basin, typical karst and fissured rocks catchment located in the Apennines chains in central Italy, has been performed to simulate snow, Evapotranspiration (ET), and runoff through different components of the GEOframe system. During this study, we showed that using an unsupervised approach for extracting the basin boundary could provide significant issues in the correct estimation of water budget components. To overcome this problem, both hydrogeological and hydrological information -obtained through a new type of time-series analyses and recent geological surveys- have been considered to estimate the contribution area and time response of the karst discharge. According to the mentioned information and benefiting the flexibility of the GEOframe-NewAge modeling system, a conceptual reservoir with a 30-day time response, derived from the time series analysis, has been added to estimate the karst river discharge contributed to Nera. The model, evaluated by different signatures (including mean daily discharge, high flow, low flow, low flow duration frequency, and flow duration curve slope and a new empirical probability function) has been shown to reproduce the water fluxes of the hydrological cycle in the basin relatively well (KGE values equal to 0.61, 0.80, and 0.71 in different sections, respectively). The karst discharge flux has a significant effect on the water budget of the basin especially in the upstream part (Castelsantangelo section) and this effect decreased through the river downstream to the outlet of Visso. We showed that 85% of the total discharge at Castelsantangelo station comes from outside of the geomorphological boundary of the basin. According to the water balance analysis, the maximum karst flux that happened in 2014 could be mainly caused by the maximum precipitation that happened in 2013 over the basin.  

To further cross-validation of the model performance, MODIS ET and Sentinel-1 snow depth products were used. The comparison of remote-sensed MODIS ET and GEOframe ET shows a systematic difference, with higher values of MODIS ET than our model estimations. As well, the spatial correlation of snow cover retrieved from Sentinel-1 snow depth and GEOframe Snow Water Equivalent has been examined and a good correlation has been reported especially for Castelsantangelo. The values of Sentinel-1 were also verified through some in-situ snow depth data.