Parsimonious Simulation of Daily Rainfall Fields
The spatial distribution of rainfall has a significant influence on catchment dynamics and streamflow generation. However, there are few stochastic models that can simulate long sequences of stochastic rainfall fields continuously in time and space. To address this issue, this paper presents a continuous daily spatial rainfall model which employs a latent variable approach. The use of a latent variable approach is parsimonious as it utilises a single distribution to model both the rainfall occurrence and the amount. A comprehensive evaluation approach was developed that uses a performance classification scheme to systematically evaluate model performance over a range of temporal and spatial scales. The evaluation of the model performance using the Onkaparinga catchment in South Australia showed that the model is able to reproduce the majority of statistics at individual sites within the region as well as realistic patterns of spatial rain fields. These statistics included rainfall occurrences/amounts, wet/dry spell distributions, monthly/annual volumes, extremes and spatial patterns, which are important from a hydrological standpoint. One of the few model weaknesses was that the total annual rainfall in drier years (lower 5%) was over-estimated by 15% on average over all sites. An advantage of the comprehensive evaluation was that it was able to attribute the source of this over-estimation to the poor representation of annual variability in rainfall occurrences. Given the strengths of this continuous daily rainfall field model it shows considerable promise for hydrological applications such as distributed catchment modelling.