Comparative Evaluation of Machine Learning and CROPWAT for Wheat Water Requirement Estimation in Peshawar District

Abstract

Agriculture plays a pivotal role in Pakistan’s economy by providing nearly 90% of the country’s food and fiber requirements. Efficient management of irrigation water is therefore essential for ensuring food security, particularly under increasing water scarcity conditions. Wheat, the major staple crop of Pakistan, requires precise irrigation scheduling to achieve optimal productivity and water use efficiency. Accurate estimation of crop evapotranspiration (ETc) is a fundamental component of irrigation planning and water resource management. This study evaluated the effectiveness of remote sensing technology for estimating wheat water requirements through a comparative assessment of the Surface Energy Balance Algorithm for Land (SEBAL) and the FAO CROPWAT model during the 2016–2017 wheat growing season in Peshawar, Pakistan. The CROPWAT model relies on extensive climatic data, including temperature, humidity, wind speed, sunshine duration, and rainfall, whereas SEBAL utilizes satellite imagery and surface energy balance principles to estimate evapotranspiration. Results revealed that the seasonal ETc estimated by SEBAL was 299.7 mm, while CROPWAT estimated 322.39 mm. Although CROPWAT produced slightly higher values, the difference between the two methods was relatively small, indicating a strong agreement and validating the reliability of remote sensing-based estimations. Furthermore, remotely sensed monthly ETc maps successfully captured temporal and spatial variations in crop water consumption throughout different growth stages of wheat. The findings demonstrate that SEBAL provides a rapid, practical, and cost-effective alternative to conventional methods for estimating crop water requirements. Consequently, remote sensing-based ETc estimation can significantly support irrigation management, enhance water use efficiency, and promote sustainable utilization of limited water resources in agricultural regions.