Application of a space-time stochastic model for downscaling future rainfall projections

Abstract

A space-time statistical downscaling model, based on current NWP techniques, is being developed at Monash University in collaboration with Bureau of Meteorology. This model, in combination with regional climate models, will provide high-resolution projections of the future rainfall over the major Australian cities together with reliable estimates of the uncertainty in these projections. These results will be used to design a storm water harvesting system over Australian cities. The brief description of the methodology is as follows.

City-scale rainfall patterns are classified into 5-6 clusters using a K-means algorithm and 22 years of AWAP data. A multiplicative cascade model will be used to generate a high resolution rainfall distribution scenario for the given area average precipitation from the coarser climate model output. Parameters such as predominant cloud motion, convective-stratiform fraction, mean and standard deviation of rainfall are required to generate a space-time distribution of rainfall. These are computed using 10 minute interval radar data and shown to be significantly different for these rainfall classes. The space-time cascade model will use the pre-calculated parameters to downscale the area averaged daily rainfall at finer space-time resolution.

Due to the computational efficiency of statistical model and the need for only coarser resolution parameters from dynamical model, space-time model can be run many (perhaps a hundred) times with slightly perturbed initial conditions, thereby generating an estimate of the uncertainty in the modeled time series.

Note: Journal articles and conference papers (and links where available) are available under open access arrangements where possible. Otherwise please contact your institution’s library, the authors, or publishers to organise full access.