Gridded meteorological data products often fall short in accurately capturing the amount of precipitation and its patterns in regions characterized by high elevations and complex topography. However, realistic precipitation data is crucial for high-alpine hydrological modeling. To address these discrepancies, we analyze possible corrections for solid, liquid and total precipitation of the 1 km2 gridded meteorological INCA-product in the high-alpine catchment of the Kölnbrein hydropower reservoir operated by VERBUND Hydro Power GmbH in the Malta Valley in Austria. By leveraging information from a stereo-satellite-derived snow depth map with physically-based snowpack modeling with Alpine3D, we quantitatively adjust and spatially redistribute solid precipitation, complemented by a multiplicative, stepwise correction model for liquid precipitation. We compare and evaluate five approaches using the hydrological COSERO model to our a) baseline simulation with no corrections on INCA in contrast of correcting, b) the amount and distribution of solely solid precipitation, c) the amount of liquid and solid precipitation, d) the amount of liquid and solid precipitation and the spatial distribution of the latter, e) precipitation inversely by the inflow bias, and f) calibrating the precipitation correction factor. In evaluating these strategies to improve the accuracy of reservoir inflow predictions, we found that separately correcting solid and liquid precipitation yielded the best results (c &d), with a substantial increase of up to 65% over the study period (1.10.2015–30.9.2023), while the other correction variants ranged between 42 and 52%. The inflow predictions by COSERO showed an increase in Nash-Sutcliffe Efficiency (NSE) by 17% and in Kling-Gupta Efficiency by 57% and 59% for variants c and d, respectively, along with an almost complete elimination of model bias. The higher KGE values observed for variant d compared to c during spring, summer, and fall suggest that a more realistic snow distribution enhances the simulation of snowmelt-driven runoff dynamics. In contrast, using a global (i.e., spatially homogeneous) and uniform (i.e., not distinguishing between liquid and solid precipitation phase) correction factor, inversely derived from the inflow bias (e), or solely correcting solid precipitation (b), demonstrated less performance, with a KGE increase of 47% and 49%, respectively, compared to 59% for variant d. Conversely, the calibration of the global and uniform correction factor (f) resulted in significant performance metric improvements (17% NSE, 60% KGE and 90% pBias), similar to variant d, however also led to unrealistic simulations of evapotranspiration, sublimation and glacier net runoff. The simulated water balance components – evapotranspiration and sublimation, as well as glacier runoff – in variant d were deemed plausible based on our comparison with additional simulations using Alpine3D, as well as findings from other high-alpine catchments in Austria reported in the literature. Overall, our results underscore the importance of applying a dual correction strategy for both liquid and solid precipitation, particularly when significant deficiencies are present in meteorological datasets, and suggest that such corrections should be supplemented by a comprehensive analysis of the simulated high-alpine water balance components.
Read full abstract