Spatial variability of stochastically generated rainfall

Abstract

The spatial variability of rainfall is a significant source of error when estimating runoff as part of any rainfall based flood design method. While there are numerous stochastic rainfall models that simulate point rainfall, there is still some way to go with reliable simulation of spatial variability. To highlight this point, the spatial properties of a clustered point process model are investigated. The model is compared to an existing daily coarse-scale (0.05° ~5km) grid covering the region of Sydney and significant differences between the two are shown. The proposed remedy for these limitations is to separate the stochastic model's role of reproducing multisite statistics from the role of spatial infilling. This observation will enable a more flexible approach to generating accurate spatial rainfall distributions that match empirical data.