Statistical Modelling and Projection of Future Rainfall using SARIMA and Hybrid SARIMA-GARCH Models in Various Zones of Kerala

Abstract

Water is an important natural resource considered as basic need for all living things around the world. The volume of pure water present in the Earth is regulated by the amount of rainfall received over the years. Sudden climatic changes are observed in throughout the world which led to flood, drought and uneven rainfall over the years. In this study, SARIMA and SARIMA-GARCH models are applied for forecasting rainfall in different zones of Kerala. The presence of heteroscedasticity in residuals obtained from SARIMA model was identified using ARCH-LM test and it was eliminated by applying SARIMA-GARCH model to the same. The ARCH-LM test results confirmed the presence of heteroscedasticity in residuals. The comparison of models used for predicting rainfall revealed that hybrid SARIMA-GARCH model is more efficient in projecting future values of rainfall in the northern and southern zones of Kerala whereas SARIMA model is showing more accuracy in the central zone of Kerala even in the presence of heteroscedasticity of residuals. The comparison of rainfall forecasted in different zones of Kerala clearly indicated that rainfall is higher in the northern zone whereas lower in the southern zone. In the northern and central zones, the rainfall showed a peak from June to September and almost negligible rainfall from December to February. The outperformed model in each zones of Kerala was applied for projection of future rainfall for next 5 years (2021-2025). Compare to previous years, the rainfall in the northern and central zones is expected to decrease whereas in southern zone of Kerala, rainfall will be almost same.