Continuous simulation for flood estimation in ungauged mesoscale catchments of Switzerland – Part II: parameter regionalisation and flood estimation results

Abstract

Flood estimations for ungauged mesoscale catchments are as important as they are difficult. So far, empirical and stochastic methods have mainly been used for this purpose. Experience shows, however, that these procedures entail major errors. In order to make further progress in flood estimation, a continuous precipitation–runoff-modelling approach has been developed for practical application in Switzerland using the process-oriented hydrological modelling system PREVAH (Precipitation– Runoff–EVApotranspiration-HRU related model). The main goal of this approach is to achieve discharge hydrographs for any Swiss mesoscale catchment without measurement of discharge. Subsequently, the relevant flood estimations are to be derived from these hydrographs. On the basis of 140 calibrated catchments (Viviroli et al., 2009b), a parameter regionalisation scheme has been developed to estimate PREVAH’s tuneable parameters where calibration is not possible. The scheme is based on three individual parameter estimation approaches, namely Nearest Neighbours (parameter transfer from catchments similar in attribute space), Kriging (parameter interpolation in physical space) and Regression (parameter estimation from relations to catchment attributes). The most favourable results were achieved when the simulations using these three individual regionalisations were combined by computing their median. It will be demonstrated that the framework introduced here yields plausible flood estimations for ungauged Swiss catchments. Comparing a flood with a return period of 100 years to the reference value derived from the observed record, the median error from 49 representative catchments is only -7%, while the error for half of these catchments ranges between -30% and +8%. Additionally, our estimate lies within the statistical 90% confidence interval of the reference value in more than half of these catchments. The average quality of these flood estimations compares well with present empirical standard procedures, while the range of deviations is noticeably smaller. Additionally, the availability of complete hydrographs and the process-oriented background bear potential for analyses that go beyond the mere estimation of peak flows.