Stochastic disaggregation of spatial-temporal rainfall with limited data

Despite the wide coverage of radar data in many countries, rainfall data availability remains a severe limitation in modelling spatial-temporal rainfall in most parts of the world. In this case, a spatial-temporal stochastic rainfall model has to be fitted using raingauge data only. This problem is explored in a real-world case in the UK with results showing that this fit is feasible. If such a spatial-temporal stochastic rainfall model can be fitted in a reliable manner with limited raingauge data, this can then be utilised to enhance the available historical rainfall information for hydrological modelling purposes. For example, if daily raingauge data are available from several sites, but data at fine temporal resolution are much more limited (e.g. a single hourly time series), then a disaggregation modelling framework can be established to disaggregate the historical data of daily raingauges into hourly series. This framework integrates the detailed spatial-temporal model with simpler multivariate stochastic models and appropriate stochastic disaggregation techniques. It is tested in the same real-world case and results in consistent hourly time series and a satisfactory reproduction of the actual hyetographs.