Scholarlink

<p>Soil erosion poses a serious threat to agricultural productivity and ecosystem stability, especially in drought-prone regions. This study estimated soil loss in Purulia, West Bengal, using the&#xa0;Revised Universal Soil Loss Equation model enhanced by machine learning (ML) and geospatial technology. A key focus was the K-factor, a critical soil erodibility parameter, estimated for the first time using a Quantile Regression Forest model (QRF). The QRF model achieved 63% accuracy, highlighting soil texture and terrain ruggedness as major influences on soil erodibility. Among variables, satellite-derived band-B4 (red spectral band) was the most impactful in predicting the K-factor (over 20%), followed by normalized difference vegetation index (19%), channel network depth (18%), and topographical features like profile curvature and slope (16%). Climate factors, including mean annual precipitation, mean annual temperature, and minimum temperature, showed lower influence (10–14%). Erosion-susceptibility analysis revealed high spatial variability, with western regions facing severe erosion due to steep slopes and inadequate vegetation, while eastern lowlands showed lower rates. Approximately 30% of the area is at high erosion risk, emphasizing the need for targeted conservation efforts. This study’s integration of ML offers a powerful tool for assessing erosion risks, supporting sustainable land management in Purulia and similar vulnerable areas.</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

Potential Soil Loss Estimation Using Machine Learning and Geospatial Technology in Part of the Eastern Plateau and Hill Region of West Bengal

  • 期刊论文

  • 发表日期：

    2025年4月12日

查看全文
  • Arindam Chattaraj,
  • Amrita Daripa,
  • Sudipta Chattaraj,
  • Sah Kausar Reza,
  • Sudip Dey,
  • Swapan Paul,
  • Rajkumar Porel,
  • Feroze Hasan Rahman

摘要

Soil erosion poses a serious threat to agricultural productivity and ecosystem stability, especially in drought-prone regions. This study estimated soil loss in Purulia, West Bengal, using the Revised Universal Soil Loss Equation model enhanced by machine learning (ML) and geospatial technology. A key focus was the K-factor, a critical soil erodibility parameter, estimated for the first time using a Quantile Regression Forest model (QRF). The QRF model achieved 63% accuracy, highlighting soil texture and terrain ruggedness as major influences on soil erodibility. Among variables, satellite-derived band-B4 (red spectral band) was the most impactful in predicting the K-factor (over 20%), followed by normalized difference vegetation index (19%), channel network depth (18%), and topographical features like profile curvature and slope (16%). Climate factors, including mean annual precipitation, mean annual temperature, and minimum temperature, showed lower influence (10–14%). Erosion-susceptibility analysis revealed high spatial variability, with western regions facing severe erosion due to steep slopes and inadequate vegetation, while eastern lowlands showed lower rates. Approximately 30% of the area is at high erosion risk, emphasizing the need for targeted conservation efforts. This study’s integration of ML offers a powerful tool for assessing erosion risks, supporting sustainable land management in Purulia and similar vulnerable areas.