Trend Analysis and Forecasting for Paddy Production in Sri Lanka
Munasingha, M.A.P. and Napagoda, N.A.D.N.
ABSTRACT
Sri Lanka is mainly an agricultural country and about 40 per cent of its working population is engaged in agriculture island-wide. Rice is cultivated during two seasons; Maha season (October-March) usually accounts for about 65% of annual production with the remaining 35% coming from the Yala season (April-September). The objectives of this study are to investigate the present trend of paddy production and to develop the most appropriate time series models for paddy production in Yala and Maha seasons separately. The paddy production data were obtained from the Department of Census and Statistics in Sri Lanka from 1952 to 2020. Shapiro-Wilks test was applied to check the normality of the dataset. According to the results, the Mann-Kendall trend test, and Cox – Stuart trend test, were used to detect the presence of trends in the data. It was confirmed that there was an increasing tendency in both seasonal models but the slope of paddy production in the Yala season was less compared to the paddy production in the Maha season. In this study, Auto-Regressive Integrated Moving Average (ARIMA) method was applied to forecast based on the historical data. The well-fitted ARIMA model for the paddy production of Yala season was ARIMA (2,1,1) and paddy production of Maha season was ARIMA (2,1,0). The performances of these models were mainly validated with the Akaike Information Criterion (AIC), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) values. Finally, the best model for the Yala and Maha seasons was applied separately to predict the values of the variable over the next three years. As the result of forecasting of paddy production is a requirement for planning purposes and the import policy of rice, should be based on this kind of research.
KEYWORDS: ARIMA model, Cox-Stuart trend test, Mann-Kendall trend test