Abstract
The predictive uncertainty of hydrological forecasts, that is, predictions of river runoff or gauge level, is often estimated by means of hydrological ensemble forecasts. Biases and dispersion errors of the meteorological ensembles that drive the hydrological models typically cascade down to the hydrological ensemble. Deficiencies in the hydrological part of the rainfall-runoff modeling chain lead to additional systematic errors. Hence, ensemble calibration is just as important for hydrological ensembles. Many of the meteorological postprocessing techniques can also be applied to hydrological ensembles. However, hydrological forecasts have some special characteristics of particular importance. Postprocessing approaches tailored to hydrological forecasts must take account of skewness, and depending on the needs of the end-user, autocorrelations and cross-correlations between different gauges need to be modeled. This chapter reviews univariate and multivariate postprocessing methods that have been applied in hydrological studies. It focuses on univariate methods that have been developed, particularly for hydrological ensembles, such as copula Bayesian model averaging, as well as on the multivariate modeling of temporal, spatial, and spatio-temporal dependencies.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call