A multisite daily rainfall data generation model for climate change conditions

Concerns over climate change caused by increasing concentration of CO2 and other trace gases in the atmosphere have increased in recent years. A major effect of climate change may be alterations in regional hydrologic cycles and changes in regional water availability. Predictions produced from General Circulation Models (GCMs) are the main source at present to get information about the future climate. However, the resolution of the GCMs is too coarse to use in hydrologic modelling to assess the impact of climate change. As a result, a number of statistical downscaling techniques (Hughes et al., 1999; Chandler, 2002; Charles et al., 2004; Mehrotra and Sharma, 2005) have been developed to obtain rainfall data at catchment scale. While the techniques produce reasonable results (see Frost et al., 2009) for a comparison of models applied within Australia), they can be complicated and time consuming to apply. As a result, simple techniques, such as constant or daily scaling, are still used to obtain daily rainfall data under future climate for either a single site or catchment average rainfall. There are difficulties in applying these methods to multiple sites and maintaining spatially consistent rainfall values. At present, there is no simple technique available to obtain rainfall data for future climate for multiple stations apart from the constant scaling. The eWater CRC has already developed a multi-site two-part model for the generation of spatially consistent daily rainfall data for a number of sites for the current climate. In this paper, this model is modified by adjusting the parameters based on the GCM output to take into account of the effect of climate change. The parameters modified are the means and standard deviations at the monthly and annual time scales and the transition probabilities and gamma distribution parameters at the daily time scale. The cross correlations were not modified. The nested multisite two-part model was applied to 30 stations in the Murrumbidgee River catchment for the current climate (1981-2000). One hundred replicates, each of length 20 years, were generated to evaluate the performance of the model. A number statistics at the daily, monthly and annual time scale were calculated and the results indicated a satisfactory performance of the model. For the future climate (2046 2065) with A2 emission case, daily rainfall data were obtained from the Max Planck Institute GCM simulation. Ten grid points lying in and around the catchment were used to adjust the parameters of the model. The ratios of the parameters for the future and current climate GCM daily rainfall were obtained first at the ten grid points and then interpolated to 30 sites using inverse distance weighting. The ratios calculated monthly appeared to vary considerably between successive months due to small sample size (20 years). To overcome this, the ratios were averaged over the four seasons and the same value was used for the three months within a season. Again, one hundred replicates each of length 20 years were generated and the above statistics were calculated. The statistics calculated from the generated data were compared with the adjusted values of the parameters used in the model. The comparison showed that the model satisfactorily preserved the input model parameters for the future climate.