ESTIMATION OF CROP WATER STRESS INDEX USING MACHINE LEARNING TECHNIQUES AND IRRIGATION SCHEDULING

Abstract

Precision irrigation has emerged as a promising approach for enhancing agricultural sustainability and water conservation. In water-scarce areas, effective irrigation scheduling and accurate assessment of crop water stress are critical for improving irrigation water management. The crop water stress index (CWSI) plays a pivotal tool in agricultural water management, particularly in regions facing water scarcity due to its direct, reliable, and non-destructive nature and easy-to-use technique. CWSI provides insights into the water status of crops, helping farmers make informed decisions about when and how much to irrigate. The primary objective of this study is to ascertain the CWSI for rice (Oryza Sativa L.) and wheat (Triticum Aestivum L.) crops under varying irrigation conditions. Additionally, it aims to identify the optimal water usage efficiency to enhance crop productivity. Field experiments were conducted over two seasons at the Civil Engineering Department, Indian Institute of Technology Roorkee, India, where varying irrigation water levels were applied to rice and wheat plots. The experimental field was divided into seven (for rice) and six (for wheat) plots, each subjected to different irrigation treatments based on the depletion of total available soil water (ASW) within the crop's root zone. These irrigation treatments maintained varying levels of water depletion in the soil (WDS) of TASWas well as fully irrigated (non-stressed) and extremely dry (fully stressed) conditions. Crop parameters (crop height, root depth, and leaf area index) were measured from the experimental field, and meteorological parameters were obtained from the National Institute of Hydrology (NIH), Roorkee, India. These parameters were used for calculations of upper and lower baseline canopy temperature. Multiple regression analysis was conducted between meteorological and crop parameters to establish a baseline canopy temperature. The CWSI was subsequently calculated for various levels of WDS using an empirical method. Numerous methods and approaches exist for estimating CWSI, each with its own advantages and disadvantages depending on how they are utilized and the amount of data they require. Machine learning techniques have become increasingly popular in recent years for determining CWSI. In this study, the performance of three machine learning techniques, Adaptive Neuro-Fuzzy Inference System (ANFIS), self-organizing maps (SOM), and Feed Forward-Back Propagation Artificial Neural Networks (FF-BP-ANN), are compared while determining the CWSI of rice and wheat crops.