Scholarlink

Purpose <p>Accurate mapping of chromium (Cr) concentration in farmland soils is essential for risk assessment and targeted remediation. However, conventional soil sampling is time-consuming and provides limited spatial coverage. This study aims to develop a synergistic inversion framework that integrates hyperspectral remote sensing imagery with environmental variables to estimate soil Cr concentration at the regional scale in an agricultural area in North China.</p> Methods <p>A total of 291 topsoil samples were collected across the study area. ZY1-02D satellite hyperspectral data were preprocessed using Savitzky-Golay (SG) smoothing and first-order derivative (FD) transformation. Fourteen environmental variables were incorporated to characterize topographic conditions and anthropogenic influences. Feature selection was conducted using least absolute shrinkage and selection operator (LASSO), recursive feature elimination (RFE), and model-importance filtering. Partial least squares regression (PLSR), random forest (RF), and extreme gradient boosting (XGBoost) models were applied. Model interpretability was evaluated using SHAP analysis.</p> Results <p>Soil Cr concentrations ranged from 30.3 to 81.1 mg/kg, with a mean of 52.3 mg/kg, indicating weak spatial variability and overall clean conditions. Integrating preprocessed hyperspectral data with environmental variables significantly improved model performance. The RF model combined with RFE achieved the best results (test R² = 0.666, RMSE = 4.607 mg/kg, MAE = 3.602 mg/kg, RPD = 1.761). SHAP analysis identified distance to mining areas (33.1%) and elevation (15.8%) as the dominant factors controlling the spatial distribution of soil Cr.</p> Conclusion <p>The proposed multi-source fusion approach provides a reliable framework for regional-scale mapping of soil Cr in agricultural areas with low spatial variability and shows potential for soil heavy metal monitoring and management.</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

Inversion of Cr concentration in farmland soils by coupling hyperspectral remote sensing images with environmental variables: a case study in an agricultural Region of North China

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Haitao Wei,
  • Yaning Zhang,
  • Dujuan Zhang,
  • Hengliang Guo,
  • Shan Zhao,
  • Tian He,
  • Gang Wu,
  • Mengjie Hou,
  • Baowei Zhang,
  • Guangsheng Zhou

摘要

Purpose

Accurate mapping of chromium (Cr) concentration in farmland soils is essential for risk assessment and targeted remediation. However, conventional soil sampling is time-consuming and provides limited spatial coverage. This study aims to develop a synergistic inversion framework that integrates hyperspectral remote sensing imagery with environmental variables to estimate soil Cr concentration at the regional scale in an agricultural area in North China.

Methods

A total of 291 topsoil samples were collected across the study area. ZY1-02D satellite hyperspectral data were preprocessed using Savitzky-Golay (SG) smoothing and first-order derivative (FD) transformation. Fourteen environmental variables were incorporated to characterize topographic conditions and anthropogenic influences. Feature selection was conducted using least absolute shrinkage and selection operator (LASSO), recursive feature elimination (RFE), and model-importance filtering. Partial least squares regression (PLSR), random forest (RF), and extreme gradient boosting (XGBoost) models were applied. Model interpretability was evaluated using SHAP analysis.

Results

Soil Cr concentrations ranged from 30.3 to 81.1 mg/kg, with a mean of 52.3 mg/kg, indicating weak spatial variability and overall clean conditions. Integrating preprocessed hyperspectral data with environmental variables significantly improved model performance. The RF model combined with RFE achieved the best results (test R² = 0.666, RMSE = 4.607 mg/kg, MAE = 3.602 mg/kg, RPD = 1.761). SHAP analysis identified distance to mining areas (33.1%) and elevation (15.8%) as the dominant factors controlling the spatial distribution of soil Cr.

Conclusion

The proposed multi-source fusion approach provides a reliable framework for regional-scale mapping of soil Cr in agricultural areas with low spatial variability and shows potential for soil heavy metal monitoring and management.