Weather generators are used for spatio-temporal downscaling of climate model outputs (e.g., precipitation and temperature) to investigate the impact of climate change on the hydrological cycle. In this study, a multiplicative random cascade model is proposed for the stochastic temporal disaggregation of monthly to daily precipitation fields, which is designed to be applicable to grids of any spatial resolution and extent. The proposed method uses stationary distribution functions that describe the partitioning of precipitation throughout multiple temporal scales (e.g., weekly and bi-weekly scale). Moreover, it explicitly considers the intensity and spatial covariance of precipitation in the disaggregation procedure, but requires no assumption about the temporal relationship and spatial isotropy of precipitation fields. A split sampling test is conducted on a high-resolution (i.e., 4×4 km 2 grid) daily precipitation data set over Germany (≈ 357 000 km 2 ) to assess the performance of the proposed method during future periods. The proposed method has proven to consistently reproduce distinctive location dependent precipitation distribution functions with biases less than 5% during both a calibration and evaluation period. Furthermore, extreme precipitation amounts and the spatial and temporal covariance of the generated fields are comparable to those of the observations. Consequently, the proposed temporal disaggregation approach satisfies the minimum conditions for a precipitation generator aiming at the assessment of hydrological response to climate change at regional and continental scales or for generating seamless predictions of hydrological variables.
Architecture, Civil Engineering, Surveying