Answers are grounded in this paper and cite their sources.
Fractional vegetation cover (FVC) is a key parameter in assessing the condition of surface vegetation, playing a significant role in maintaining ecosystem health and preventing natural disasters. Recently, optical remote sensing data have been widely utilized for FVC estimation. However, its sensitivity to weather conditions limits the spatial and temporal continuity of the data. Therefore, a method for estimating FVC based on synthetic aperture radar (SAR) images is proposed, which incorporates SAR feature optimization and ensemble learning models to improve the accuracy of FVC estimation. The proposed approach extracts twenty-five SAR features and assesses their sensitivity to vegetation through Pearson, Spearman, and Kendall correlation analyses, as well as time series trend analysis of the SAR features. Ultimately, sixteen representative SAR features are selected for FVC estimation. To enhance the accuracy and robustness of FVC estimation, the strong ensemble learning models Ada-Xgboost and Ada-Catboost are constructed using majority voting. The Dongting Lake area, with its complex terrain, is selected as the study area, and a layover removal step is introduced to optimize the quality of SAR images. The validation results show that the Ada-Xgboost model outperforms other ensemble learning models (Ada-Xgboost: coefficient of determination (R2) = 0.8950, root mean square error = 0.1187, mean square error = 0.0141, mean absolute error = 0.0952, number of features = 9, standard deviation = 0.0352, interval width = 0.1380, coverage probability = 29.73%). This study provides a reliable method for time series FVC estimation in topographically complex areas and also has the potential to accurately estimate FVC in other terrains.
No takes yet. Share an insight, caveat, or question.
Wu et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: