Statistical Testing of Dynamically Downscaled Rainfall Data for the East Coast of Australia

Tuesday, 16 December 2014
Parana Manage Nadeeka1, Natalie Lockart1, Garry R Willgoose2, George A. Kuczera2 and A.F.M. Kamal Chowdhury1, (1)University of Newcastle, Callaghan, NSW, Australia, (2)University of Newcastle, Callaghan, Australia
This study performs a validation of statistical properties of downscaled climate data, concentrating on the rainfall which is required for hydrology predictions used in reservoir simulations. The data sets used in this study have been produced by the NARCliM (NSW/ACT Regional Climate Modelling) project which provides a dynamically downscaled climate dataset for South-East Australia at 10km resolution. NARCliM has used three configurations of the Weather Research Forecasting regional climate model and four different GCMs (MIROC-medres 3.2, ECHAM5, CGCM 3.1 and CSIRO mk3.0) from CMIP3 to perform twelve ensembles of simulations for current and future climates.

The validation has been performed in the Upper Hunter region of Australia which is a semi-arid to arid region 200 kilometres North-West of Sydney. The analysis used the time series of downscaled rainfall data and ground based measurements for selected climate stations within the study area. The initial testing of the gridded rainfall was focused on the Auto Regressive characteristics of time series. The focus on correlations was because the reservoir performance depends on long-term average runoffs. In order to compare the datasets, a correlation analysis was performed at daily, fortnightly, monthly, three monthly and annual time scales. Moreover, the spatial variability of statistics of gridded rainfall series were calculated and plotted at the catchment scale. The NARCLiM data were able to successfully reproduce the autocorrelations of observed rainfall at each time scale. The correlation analysis shows a good agreement between NARCliM data and ground truth at three months and annual time scales. The spatial variability plots show a possible link between the statistics and orography at the catchment scale. Preliminary results for an orographic analysis will be shown.