Multisite spatial-temporal rainfall analysis in the Onkaparinga catchment using a generalised linear model

Multisite daily rainfall is an important input in rainfall-runoff modelling. Spatial-temporal rainfall analysis is difficult due to its high variability and non-normal distribution characteristics. Moreover due to these characteristics of rainfall, catchment based daily rainfall analysis using stochastic methods are often used. In this study, we have applied a Generalized Linear Model (GLM) for multisite daily rainfall analysis using occurrence and amount models. Nine rainfall stations were selected within the Onkaparinga catchment in South Australia. NCEP/NCAR reanalysis data and SOI and DMI climate index data were used in addition to observed daily rainfall. The model was fitted and simulated for the calibration period (1991-2010) and validated for the period 1981-1990. Several summary statistics were used to check the performance of the model. The 2.5th and 97.5th percentiles of simulated rainfall were plotted against the statistics of observed rainfall. The GLM was able to adequately simulate daily rainfall for use in water resources studies. Moreover, it was concluded that the GLM can be used for downscaling of General Circulation Model (GCM) rainfall projections for the Onkaparinga catchment in South Australia