Prediction of crop yield using climate variables in the south-western province of India: a functional artificial neural network modeling (FLANN) approach

To meet the demand of the growing population, there exists pressure on food production. In this context, appropriate prediction of crop yield helps in agricultural production planning. Given the inability of the traditional linear models to provide satisfactory prediction performance, there is a need to develop a crop yield prediction model that is simple in complexity, accurate in prediction, and less time-consuming during training and validation phases. Keeping these objectives in view, the present paper focuses on building an adaptive, low complexity, and accurate nonlinear model for the prediction of crop yield. A time series dataset for the period 1991–2012 of Karnataka, a southwestern state of India, is used for yield prediction. An empirical nonlinear relation between crop yield and the four independent attributes has been obtained from the proposed ANN model. The independent attributes employed are total rainfall, the cumulative distribution of temperature, the proportion of irrigated land, and the average amount of fertilizer used. It is demonstrated that the developed model exhibits better prediction accuracy, less root mean square error in the range of 0.07–0.14, less mean square error in the range of 0.01–0.04, and mean absolute error in the range of 0.07–0.15 compared to its corresponding linear regression model. It is recommended that the proposed ANN model can also be applied to predict other agricultural products of the same or other geographical regions of the globe.