Evaluation of Noah-MP Simulations and Development of a Functionally Equivalent Surrogate Snow Module
Sadaf Moghisi1, Ali Behrangi1, Guo-Yue Niu1
1Department of Hydrology and Atmospheric Sciences, University of Arizona, Tucson, AZ, USA
Accurate prediction of snow accumulation, snowmelt infiltration and runoff, and thus the inflow to reservoirs over headwater regions is critical for forecast-informed reservoir operations (FIRO). We use the Noah-MP land surface model with a dynamic vegetation model and the RAPID (David et al., 2011) for streamflow routing model to improve the predictions of the interannual variability of snow and streamflow over the Salt and Verde rivers, which provides reliable, affordable water and power to more than 2 million people living in central Arizona. We evaluated the model’s performance with the SMAP-derived soil moisture and freeze-thaw states, UA Snow Water Equivalent (SWE), USGS streamflow data, and MODIS-based climatological Leaf Area Index (LAI). using the Kling-Gupta Efficiency (KGE). To further improve the model’s prediction accuracy, we plan to develop a differentiable, learnable, simple energy balance-based snow model and learnable soil hydrology and thus use the differentiable parameter learning (dPL) algorithm to optimize the model with spatially-varying optimum parameters.