The benefit of climatological and calibrated reforecast data for simulating hydrological droughts in Switzerland

Streamflow droughts are characterized by reduced runoff and can cause significant damage to nature and society. Forecasts of streamflow droughts could improve risk management and preparedness. How the quality of the meteorological forcing of a hydrological model affects the hydrological response during streamflow droughts was explored in case studies of two Swiss rivers: the Thur, which is mainly rain-fed and the Landquart, which is dominantly snow-fed. A fully distributed version of the PREcipitation-Runoff-EVApotranspiration HRU model (PREVAH) was used with input from: (1) meteorological observations, (2) reforecast from the probabilistic numerical weather prediction model Limited-area Ensemble Prediction System developed and run by the COnsortium for Small-scale MOdelling (COSMO-LEPS) and (3) COSMO-LEPS reforecasts that are calibrated with the bias correction method quantile mapping. The meteorological input variables such as precipitation, air temperature, wind speed, relative humidity, global radiation and sunshine duration are required to initialize PREVAH. Different combinations of the input variables, for example observed precipitation with the other variables from COSMO-LEPS, were used to assess the sensitivity of each input variable to the performance of the low-flow simulation (1981–2000). Streamflow droughts are defined by indices of duration, severity and magnitude, by applying a seasonally varying threshold. Simulations for low-flow and streamflow drought performed better when meteorological observations or calibrated COSMO-LEPS reforecasts were used as forcing instead of the COSMO-LEPS reforecast. It was sufficient to replace precipitation with observations or calibrated COSMO-LEPS values and keep the other variables from COSMO-LEPS to improve the low-flow simulations. Depending on the catchment characteristics, better quality of temperature and relative humidity may also be relevant.