Scholarlink

Background and aims <p> &#xa0;Accurate assessment and mapping of soil salinity are essential for proper irrigation scheduling and sustainable land management. The research intended to develop a robust retrieval method using remote sensing that ensures reliable monitoring of soil salinity in arid irrigation districts with both high intra-annual accuracy and strong inter-annual generalizability.</p> Methods <p>This study proposed a novel strategy to improve soil salinity retrieval by integrating crop information hierarchically with different data preprocessing approaches. A series of cross-validation experiments, error analysis, and importance analysis was carried out to evaluate the performance of different incorporation approaches.</p> Results <p>The hybrid incorporation of three categories of crop information (i.e., multi-temporal spectral features, phenological metrics, and crop types) evidently improved the performance of soil salinity retrieval compared with only one or two categories. The match of data preprocessing approach and data types/characteristics was found important for enhancing the retrieval accuracy and generalizability.&#xa0;The combination of the One-Hot encoded crop type data, the normalized spectral features and the standardized of phenological metrics, achieved the best retrieval performance in the arid irrigation district.</p> Conclusion <p>The proposed incorporation strategy of crop information and data preprocessing can provide a theoretical basis for reliable estimation and mapping of soil salinization.</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

Enhancing the accuracy and generalizability of soil salinity monitoring in Arid Regions by incorporating crop information and data preprocessing

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Cunzhen Pan,
  • Jingwen Han,
  • Yuhang Sun,
  • Yunwu Xiong,
  • Guanhua Huang

摘要

Background and aims

 Accurate assessment and mapping of soil salinity are essential for proper irrigation scheduling and sustainable land management. The research intended to develop a robust retrieval method using remote sensing that ensures reliable monitoring of soil salinity in arid irrigation districts with both high intra-annual accuracy and strong inter-annual generalizability.

Methods

This study proposed a novel strategy to improve soil salinity retrieval by integrating crop information hierarchically with different data preprocessing approaches. A series of cross-validation experiments, error analysis, and importance analysis was carried out to evaluate the performance of different incorporation approaches.

Results

The hybrid incorporation of three categories of crop information (i.e., multi-temporal spectral features, phenological metrics, and crop types) evidently improved the performance of soil salinity retrieval compared with only one or two categories. The match of data preprocessing approach and data types/characteristics was found important for enhancing the retrieval accuracy and generalizability. The combination of the One-Hot encoded crop type data, the normalized spectral features and the standardized of phenological metrics, achieved the best retrieval performance in the arid irrigation district.

Conclusion

The proposed incorporation strategy of crop information and data preprocessing can provide a theoretical basis for reliable estimation and mapping of soil salinization.