Application of vector autoregressive model for rainfall and groundwater level analysis

Abstract

Groundwater is a crucial water supply for industrial, agricultural and residential use, hence it is important to understand groundwater system. Groundwater is a dynamic natural resource and can be recharged. The amount of recharge depends on the rate and duration of rainfall, as rainfall comprises an important component of the water cycle and is the prime source of groundwater recharge. This study applies Vector Autoregressive (VAR) model in the analysis of rainfall and groundwater level. The study area that is focused in the study is along the East-West Highway, Gerik-Jeli, Malaysia. The VAR model with optimum lag length 8, VAR(8) is selected to model the rainfall and groundwater level in the study area. Result of Granger causality test shows significant influence of rainfall to groundwater level. Impulse Response Function reveals that changes in rainfall significantly affect changes in groundwater level after some time lags. Moreover, Variance Decomposition reported that rainfall contributed to the forecast of the groundwater level. The VAR(8) model is validated by comparing the actual value with the in-sample forecasted value and the result is satisfied with all forecasted groundwater level values lies inside the confidence interval which indicate that the model is reliable. Furthermore, the closeness of both actual and forecasted groundwater level time series plots implies the high degree of accurateness of the estimated model.