Articles | Volume 20, issue 4
https://doi.org/10.5194/hess-20-1387-2016
https://doi.org/10.5194/hess-20-1387-2016
Research article
 | 
08 Apr 2016
Research article |  | 08 Apr 2016

Downscaling future precipitation extremes to urban hydrology scales using a spatio-temporal Neyman–Scott weather generator

Hjalte Jomo Danielsen Sørup, Ole Bøssing Christensen, Karsten Arnbjerg-Nielsen, and Peter Steen Mikkelsen

Abstract. Spatio-temporal precipitation is modelled for urban application at 1 h temporal resolution on a 2 km grid using a spatio-temporal Neyman–Scott rectangular pulses weather generator (WG). Precipitation time series used as input to the WG are obtained from a network of 60 tipping-bucket rain gauges irregularly placed in a 40 km  ×  60 km model domain. The WG simulates precipitation time series that are comparable to the observations with respect to extreme precipitation statistics. The WG is used for downscaling climate change signals from regional climate models (RCMs) with spatial resolutions of 25 and 8 km, respectively. Six different RCM simulation pairs are used to perturb the WG with climate change signals resulting in six very different perturbation schemes. All perturbed WGs result in more extreme precipitation at the sub-daily to multi-daily level and these extremes exhibit a much more realistic spatial pattern than what is observed in RCM precipitation output. The WG seems to correlate increased extreme intensities with an increased spatial extent of the extremes meaning that the climate-change-perturbed extremes have a larger spatial extent than those of the present climate. Overall, the WG produces robust results and is seen as a reliable procedure for downscaling RCM precipitation output for use in urban hydrology.

Download
Short summary
Fine-resolution spatio-temporal precipitation data are important as input to urban hydrological models to assess performance issues under all possible conditions. In the present study synthetic data at very fine spatial and temporal resolution are generated using a stochastic model. Data are generated for both present and future climate conditions. The results show that it is possible to generate spatially distributed data at resolutions relevant for urban hydrology.