Scholarlink

<p>Estimation of soil moisture is essential for crop-water investigations on agricultural fields. Remote sensing provides significantly better spatial soil moisture measurements than field installation sensors. This study synergizes Sentinel radar and optical products to remotely extract topsoil moisture (~ 5&#xa0;cm). Sentinel-1 radar images were collected and analyzed for two distinct field sites in Quebec, Canada, for the growing seasons 2017 (field site A, clayey soil) and 2019 (field site B, sandy loam soil). Several model combinations of backscatter intensities, their mathematical band combinations (MBCs) and vegetation indices (DpRVI, NDVI, NDMI, SAVI) were formulated and statistically analyzed for their significance in soil moisture retrieval. Regression modeling was employed for the model training and testing, in which the models were cross-validated using a k-fold cross-validation technique. In site A, the maximum correlation between the predicted and in-situ soil moisture was 0.52, and 0.6 in site B. Several models from site B were below the set RMSE threshold of 0.05 m<sup>3</sup>&#xa0;m<sup>−3</sup>, whereas none of the models from site A met this criterion. Most of the best-performing retrieval models from each field site comprised the MBCs and one of the vegetation indices as their additional predictors. Therefore, this study’s demonstrated results infer that including MBCs and the vegetation indices would enhance soil moisture extraction, particularly during limited ground truth data. Moreover, these cross-validated retrieval models would further emphasize their future applicability in agricultural management and decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Investigating the Role of Intensity Band Combinations and Vegetation Indices in Topsoil Moisture Retrieval

  • 期刊论文

  • 发表日期：

    2025年5月30日

查看全文
  • Naresh Arumuga,
  • Chandra Madramootoo

摘要

Estimation of soil moisture is essential for crop-water investigations on agricultural fields. Remote sensing provides significantly better spatial soil moisture measurements than field installation sensors. This study synergizes Sentinel radar and optical products to remotely extract topsoil moisture (~ 5 cm). Sentinel-1 radar images were collected and analyzed for two distinct field sites in Quebec, Canada, for the growing seasons 2017 (field site A, clayey soil) and 2019 (field site B, sandy loam soil). Several model combinations of backscatter intensities, their mathematical band combinations (MBCs) and vegetation indices (DpRVI, NDVI, NDMI, SAVI) were formulated and statistically analyzed for their significance in soil moisture retrieval. Regression modeling was employed for the model training and testing, in which the models were cross-validated using a k-fold cross-validation technique. In site A, the maximum correlation between the predicted and in-situ soil moisture was 0.52, and 0.6 in site B. Several models from site B were below the set RMSE threshold of 0.05 m3 m−3, whereas none of the models from site A met this criterion. Most of the best-performing retrieval models from each field site comprised the MBCs and one of the vegetation indices as their additional predictors. Therefore, this study’s demonstrated results infer that including MBCs and the vegetation indices would enhance soil moisture extraction, particularly during limited ground truth data. Moreover, these cross-validated retrieval models would further emphasize their future applicability in agricultural management and decision-making.