Development of short-term evapotranspiration forecasting model using time series method for supporting the precision agriculture management in tropics

Abstract

To support daily farm management through the environmental challenges, it is necessary to have short-term evapotranspiration forecasting to predict an n-hour step. Evapotranspiration (ET) is the sum of evaporation and transpiration from the soil surface and plant tissue that can be used to assess the water loss behaviour in open-field cultivation. The objective of this study was to develop a short-term evapotranspiration forecasting model using the time series method. The model is based on the Seasonal Autoregressive Integrated Moving Average (SARIMA). The environmental data of air temperature, relative humidity, and solar radiation were observed at Rejeki Tani Yogyakarta from January to August 2014. The ET was estimated using the FAO56 Penman-Monteith. A suitable parameter of non-seasonal autoregressive order (p), the degree of differencing (d), moving average order (q), and their seasonal parameter (P, D, Q)s were investigated to predict 12-hour ahead of ET. As the result, the suitable parameter was SARIMA(2, 1, 2)(1, 1, 1)24. From the six months model validation with the different monsoons, the MAE and RMSE ranged from 0.035636 to 0.063419, and 0.045893 to 0.079961 respectively. The R2 was between 0.8045 and 0.85902. The developed forecasting model can be used to predict the hourly evapotranspiration with acceptable error and accuracy.