Scholarlink

<p>The expansion of intensive agriculture has led to increasing soil salinity worldwide. highlighting the critical need for accurate soil salinity measurements is essential to address this situation. In this context, digital soil salinity mapping becomes necessary to properly manage soil resources in limited data regions. Therefore, this study aimed to map soil salinity using a Random Forest (RF) model that incorporated several spectral indices and physicochemical properties in the Tadla Plain. 149 samples were used to investigate the physical and chemical characteristics of soils in the study area. the dataset was divided into 70% of the ground data for model training and 30% for validation. The results show that 81.1% of the studied soil is non-saline, 15,5% is slightly saline, and 3,4% is moderately saline. However, statistical metrics showed that the RF model performed well with a salinity index (SI6), achieving a correlation coefficient (R<sup>2</sup>) of 0.80 and Root Mean Square Error (RMSE) of 0.084. Salinity indices SI1, SI2, SI3, SI4, SI5, and SI7 yielded results with low precision, with R<sup>2</sup> values below − 0.2. These findings provide valuable insights for developing strategies to mitigate soil salinity in semi-arid 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

Spatial mapping of soil salinity in a semiarid region using a machine learning model based on spectral indices and ground data

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Khalid El Bahjaouy,
  • Ahmed Barakat,
  • Amine Oussilkane,
  • Abdessamad Hilali,
  • Nawal Amrani,
  • Hassan Mosaid

摘要

The expansion of intensive agriculture has led to increasing soil salinity worldwide. highlighting the critical need for accurate soil salinity measurements is essential to address this situation. In this context, digital soil salinity mapping becomes necessary to properly manage soil resources in limited data regions. Therefore, this study aimed to map soil salinity using a Random Forest (RF) model that incorporated several spectral indices and physicochemical properties in the Tadla Plain. 149 samples were used to investigate the physical and chemical characteristics of soils in the study area. the dataset was divided into 70% of the ground data for model training and 30% for validation. The results show that 81.1% of the studied soil is non-saline, 15,5% is slightly saline, and 3,4% is moderately saline. However, statistical metrics showed that the RF model performed well with a salinity index (SI6), achieving a correlation coefficient (R2) of 0.80 and Root Mean Square Error (RMSE) of 0.084. Salinity indices SI1, SI2, SI3, SI4, SI5, and SI7 yielded results with low precision, with R2 values below − 0.2. These findings provide valuable insights for developing strategies to mitigate soil salinity in semi-arid areas.