Neural computing modelling of the crop water stress index

In this study, two artificial neural network models viz. supervised Feed-Forward Back Propagation (FF-BP) and unsupervised Kohonen Self-Organizing Map (K-SOM) have been developed to predict the Crop Water Stress Index (CWSI) using air temperature, relative humidity, and canopy temperature. Field experiments were conducted on Indian mustard to observe the crop canopy temperature under different levels of irrigation treatment during the 2017 and 2018 cropping seasons. The empirical CWSI was computed using well-watered and non-transpiring baseline canopy temperatures. The K-SOM and FF-BP CWSI predictions were compared with the empirical CWSI estimates and both performed satisfactorily. Of the two, however, the K-SOM was better with R2 (coefficient of determination) of 0.97 and 0.96 for model development and validation, respectively; corresponding values for FF-BP were 0.86 and 0.75. The results of the study suggest that neural network modelling offers significant potential for reliable prediction of the CWSI, which can be utilized in irrigation scheduling and crop stress management.