Assessing the potential of water cloud model and machine learning algorithms in estimating leaf area index of wheat at different growth stages using Sentinel-1 SAR data

Saptarshi Dey, Abishek Murugesan, Rucha Dave, Koushik Saha,
Assessing the potential of water cloud model and machine learning algorithms in estimating leaf area index of wheat at different growth stages using Sentinel-1 SAR data,
Advances in Space Research,
Volume 76, Issue 10,
2025,
Pages 6130-6140,
ISSN 0273-1177,
https://doi.org/10.1016/j.asr.2025.09.003.
(https://www.sciencedirect.com/science/article/pii/S027311772500969X)
Abstract: Leaf Area Index (LAI) is a key biophysical parameter closely linked to plant health and growth. Though LAI is usually estimated from optical remote sensing data, SAR data offers significant advantages over optical data due to its all-weather capability and ability to collect data both day and night. This study evaluates the use of Sentinel-1 SAR-derived vegetation descriptors in the Water Cloud Model (WCM) and the machine learning algorithms for estimation of LAI during different growth stages of wheat, with a comparative analysis of the performance of both models. The WCM performed best when VH backscatter coefficient and LAIn (n was computed by the model) were used as vegetation descriptors, compared to other descriptor combinations. The analysis was carried out for different growth stages of wheat, revealing that both models estimate LAI most accurately for the early growth stage, with WCM achieving an R value of 0.7, while Support Vector Regression (SVR) and Gaussian Process Regression (GPR) machine learning algorithms achieved R values nearing 0.86 and 0.84, respectively. The accuracy for both WCM and machine learning algorithms declined during the intermediate growth stage but improved during the final growth stage. This study demonstrates the reliability and applicability of various vegetation descriptors in the WCM and capability of various machine learning algorithms for estimation of LAI. Further, it is observed that irrespective of using a semi-empirical model like WCM or machine learning algorithms, the ability for these models for estimation of LAI varies with crop growth stage. The key contribution of this work is the stage-wise comparative assessment of the WCM and machine learning models for wheat, an aspect that has been limited in previous studies.
Keywords: Leaf Area Index; Sentinel-1; Water Cloud Model; Support Vector Regression (SVR); Gaussian Process Regression (GPR); Vegetation descriptors