Statistical correlations between rainfall and climate indices in Western Australia

Abstract

As the concept of 'sustainable development' is increasing day by day, rainfall has become the most significant and investigated hydro-climate variable in Australia. Rainfall variability is considered as a major economic factor in Australia. It was observed that Australian rainfall is affected by several climate patterns and long-term prediction remains a challenge for many years. However, forecasting rainfall can be beneficial for the design, maintenance and management of water resources infrastructures. Nevertheless, rainfall is the product of complex global atmospheric phenomena. Strong correlations between rainfall and several climate indices have already been observed throughout the world. Any such correlation with climate indices and rainfall afterwards can be used in forecasting long-term rainfall. For the prediction of rainfall in advance, statistical and dynamic systems can be used in practice. However, dynamic systems are too complex and expensive to use in a wide range of situations. This paper focused on the investigation of statistical correlations between rainfall and several climate indices as potential predictor of long-term Western Australian rainfall. Since Australian rainfall is highly variable both in time and space, this analysis was performed on regional scale. Several multiple regression models were investigated using the climate indices as potential predictors of rainfall. The models which satisfied the limits of statistical significance were used to forecast Western Australia rainfall in advance. Historical rainfall data were obtained from Australian Bureau of Meteorology. The rainfall station Roebourne in Western Australia was chosen as a case study. The station was selected based on their long term recorded data having fewer missing values. The major aim was deterministic forecasting of long-term rainfall in terms of climate indices in regional scale. The analysis showed that DMI-ENSO based combined multiple regression models could be used for long-term rainfall forecasting of Western Australia except extreme rainfall.