Author(s): Likith Muni Narakala; Manoj Yadav; Ghanshyam Giri; Hitesh Upreti; Gopal Das Singhal

Linked Author(s):

Keywords: Canopy temperature; Artificial neural networks; Irrigation scheduling

Abstract: Efficient irrigation management is crucial for sustaining crop yields, particularly in the water-scarce regions. In this regard, plant-based indicators, such as canopy temperature (Tc), offer promising solutions for the assessment of water stress. This study aims to enhance the Tc predictions using artificial neural networks (ANNs), utilizing meteorological and soil moisture data collected for wheat crops during the two winter crop seasons: 2021-22, and 2022-23 at the Water Management Field Laboratory (WMFL), Shiv Nadar Institution of Eminence (SNIoE), Greater Noida, India. Five input configurations (C1 to C5) were tested, with varying soil moisture depth layers, to identify the optimal combination of predictors. The best-performing model, combination C4, which included vapor pressure deficit (VPD), air temperature (Ta), net radiation (Rn), wind speed (U), and soil moisture at 0-20 cm, resulted in the lowest mean absolute error (MAE) of 1.06 °C during training and 1.27 °C while testing. Removing soil moisture from the input parameters affected the prediction performance adversely where the combination (C5) resulted in higher MAE value. Also, including soil moisture at 0-40 cm along with the meteorological variables increased the testing MAE to 1.46 °C. These results underscore the importance of integrating shallow soil moisture data with meteorological parameters to improve the prediction accuracy of Tc. This ANN-based approach demonstrates a potential tool for crop water stress management, paving the way for further research in diverse agro-climatic conditions.

DOI: https://doi.org/10.64697/978-90-835589-7-4_41WC-P1911-cd

Year: 2025