Multi-site doubly stochastic Poisson process models for fine-scale rainfall

Abstract

We consider a class of doubly stochastic Poisson process models in the modelling of fine-scale rainfall at multiple gauges in a dense network. Multi-site stochastic point process models are constructed and their likelihood functions are derived. The application of this class of multi-site models, a useful alternative to the widely-known Poisson cluster models, is explored to make inferences about the properties of fine time-scale rainfall. The proposed models, which incorporate covariate information about the catchment area, are used to analyse tipping-bucket raingauge data from multiple sites. The results show the potential of this class of models to reproduce temporal and spatial variability of fine time-scale rainfall characteristics.