In this study, recurrent networks to downscale meteorological fields of the ERA-40 re-analysis dataset with focus on the meso-scale water balance were investigated. Therefore two types of recurrent neural networks were used. The first approach is a coupling between a recurrent neural network and a distributed watershed model and the second a nonlinear autoregressive with exogenous inputs (NARX) network, which directly predicted the component of the water balance. The approaches were deployed for a meso-scale catchment area in the Free State of Saxony, Germany. The results show that the coupled approach did not perform as well as the NARX network. But the meteorological output of the coupled approach already reaches an adequate quality. However the coupled model generates as input for the watershed model insufficient daily precipitation sums and not enough wet days were predicted. Hence the long-term annual cycle of the water balance could not be preserved with acceptable quality in contrary to the NARX approach. The residual storage change term indicates physical restrictions of the plausibility of the neural networks, whereas the physically based correlations among the components of the water balance were preserved more accurately by the coupled approach.

This study presents an extended version of a single site daily weather generator after Richardson. The model is driven by daily precipitation series derived by a first-order two-state Markov chain and considers the annual cycle of each meteorological variable. The evaluation of its performance was done by deploying its synthetic time series into the physical based hydrological model BROOK90. The weather generator was applied and tested for data from the Anchor Station at the Tharandt Forest, Germany. Additionally its results were compared to the output of another weather generator with spell-length approach for the precipitation series (LARS-WG). The comparison was distinguished into a meteorological and a hydrological part in terms of extremes, monthly and annual sums and averages. Extreme events could be preserved adequately by both models. Nevertheless a general underestimation of rare events was observed. Natural correlations between vapour pressure and minimum temperature could be conserved as well as annual cycles of the hydrological and meteorological regime. But the simulated spectrums of extremes, especially, of precipitation and temperature, are more limited than the observed spectrums. While LARS-WG already finds application in practice, the results show that the data derived from the presented weather generator is as useful and reliable as those from the established model for the simulation of the water balance.