Reducing Hydrological Modelling Uncertainty by Using MODIS Snow Cover Data and a Topography-Based Distribution Function Snowmelt Model

This work introduces a general multi-objective parameter estimation framework to exploit MODIS-based snow cover maps to reduce predictive streamflow uncertainty in snow-dominated catchments. The well-known GLUE methodology is applied with a multi-objective approach, combining streamflow observations recorded at the outlet section and satellite-derived snow cover maps, aggregated to fractional values of the catchment area. The hydrological model used in this study includes a snowpack routine which exploits a statistical representation of the distribution of clear sky potential solar radiation - a significant advantage when parameter sensitivity and uncertainty estimation procedures are carried out. The study provides an assessment of this approach based on operational quality data from two medium-size mountainous basins (a nested one included in a larger parent basin) located in the eastern Italian Alps. The nested basin is considered as ungauged, thus allowing a spatial assessment of the multi-objective approach. Results show a positive feedback between streamflow and snow cover area likelihoods, highlighted by means of the Pareto plot. Moreover, a better identifiability of the parameters driving snowmelt rate is found and consequently a shrink of the predictive streamflow uncertainty is observed. A containing ratio of 0.54 and a mean sharpness of 0.11 are found at the outlet of the parent basin, while a containing ratio equal to 0.65 and a mean sharpness equal to 0.17 are estimated at the nested basin, used as a validation test. These results confirm the potential of MODIS snow cover maps as additional data to inform hydrological models leading to more reliable and sharper streamflow simulations. This approach might be also appealing when streamflow simulations are required for ungauged basins.