Predicting pest population using weather variables an ARIMAX time series framework

Abstract

Farmers are encountering several issues in endeavour to increase crop productivity. Despite several successful new agricultural technologies related with crop cultivation, India is unable to attain world average mark in productivity. One of the main reasons for this is climatic conditions and abundance of insects and pests. To mitigate the loss due to pest attacks and for better yield, forecasting of pest population based on historical data and pertinent external climatic information is considered. Autoregressive Integrated Moving Average with Exogenous variables (ARIMAX) time-series model is applied for modelling and forecasting the pest population after testing for stationarity. Primary weekly data (2008-2012) for three pests namely Jassids, Whitefly and Thrips in Guntur and Faridkot Districts along with weekly maximum temperature, minimum temperature, rainfall, maximum RH and minimum RH have been used for model development. Evaluation of forecasting is carried out with relative mean absolute prediction error (RMAPE). Diagnostic test were applied and results showed that maximum temperature and minimum temperature along with maximum relative humidity have a significant role for Whitefly and Thrips at Guntur district respectively. Rainfall was found to be significant at Faridkot district in case of Thrips. The fitted models along with the data points are also presented. A perusal of figures indicates that in both districts, the population of Whitefly is best predicted followed by Jassids and Thrips.