Document Type : Original Article
Authors
1
Department of Remote Sensing and GIS, University of Tehran, Tehran, Iran
2
هیات علمی دانشگاه تهران
3
Department of Remote sensing & GIS, Faculty of Geography, University of Tehran
4
Department of Irrigation and Drainage, Faculty of Water Sciences Eng., Shahid Chamran University of Ahvaz,
5
Department of Remote Sensing and GIS, Sugarcane Development Research and Training Institute, Ahvaz, Iran,
6
Lancaster Environment Center, Faculty of Science and Technology, Lancaster University, Lancaster, UK,
10.22034/iwrr.2024.443011.2742
Abstract
Soil moisture is one of the most important environmental parameters for water resources management and irrigation planning systems in agricultural areas. In agricultural areas, most soil moisture retrieval models are unstable in terms of their accuracy and performance during crop growth season. As a result, there is no consensus on which model performs optimally during the agricultural season. This is because of the uncertainties associated with model physics, input data, vegetation attenuation and soil characteristics. To deal with these practical concerns, in this research, a simple but effective soil moisture retrieval method for using combination of multiple models based on machine learning has been introduced. Firstly, the WCM with different vegetation descriptors were calibrated and validated in sugarcane fields for Sentinel-1 backscattering coefficients . For this purpose, soil moisture measurements of sugarcane fields (400 samples in total) during the plant growing season in 2020 were used. Then optimization of calculations was done using the GRNN. The results showed that WCMNDWI retrieves soil moisture more accurately than other models in the early stages of sugarcane growth, while WCMVWC and WCMLAI were more accurate in the late sugarcane growth. Time-series soil moisture retrieval accuracy using the combined method based on GRNN was higher than that of single WCM models. According to the results of the in situ validation for sugarcane fields, with the optimal combination of models, the minimum MAE) is less than 0.02 m3m-3, the RMSE is approximately 0.085 and the R it was equal to 0.7 for the growing season.
Keywords
Main Subjects
)