Application of artificial intelligence-based modelling for the prediction of crop water stress index

Abstract

Abstract The study evaluates the performance of three artificial intelligence (AI) techniques viz. support vector regression (SVR), artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) for predicting the crop water stress index (CWSI) using relative humidity, air temperature, and canopy temperature. Field crop experiments were conducted on Wheat (during 2018, 2019) and Indian mustard (during 2017, 2018) to observe the canopy temperature in different irrigation levels. The experimentally obtained empirical CWSI was considered as the reference CWSI. Different configurations of ANN, SVR and ANFIS models were developed and validated with the empirical CWSI. The most optimal model structures for predicting CWSI were ANN5 (ANN with 5 hidden neurons), SVRQ (SVR with Quadratic kernel) and ANFIS2 (ANFIS with 2 membership functions) in Wheat; and ANN3 (ANN with 3 hidden neurons), SVRQ and ANFIS2 in Indian mustard. Based on the values of error statistics during validation, all three models presented a satisfactory performance, however, the efficacy of the models was relatively better in the case of Wheat. The model predictions at low CWSI values indicate deviations in the case of both crops. Overall, the study results indicate that data-driven-based AI techniques can be used as potential and reliable alternatives for predicting CWSI since the performance of the models is reliable for CWSI values commonly encountered in irrigation scheduling.