Scholarlink

<p>Soil erosion induced by sand dune movement significantly threatens desert ecosystems, requiring precise prediction methods for sustainable land management. This study employs Object-Based Image Analysis (OBIA), Convolutional Neural Networks (CNN), and the Fuzzy Analytic Hierarchy Process (FAHP) to accurately quantify morphometric characteristics of sand dunes across multiple scales in the Taklamakan Desert, China, and the Lut Desert, Iran. The research also examines the relationship between spectral reflectance values, such as albedo, and morphometric features concerning soil erosion and sand dune dynamics. Results demonstrate that the Self-Attention Generative Adversarial Network (SAGAN) model effectively enhances satellite image quality, enabling precise sand dune extraction. Furthermore, pairwise comparison of topographic parameters weighted by the FAHP method, combined with high-resolution imagery, facilitates accurate prediction of active desert processes and the generation of detailed landform maps. Findings indicate that the optimal scale for sand dune delineation and morphometric feature identification for soil erosion prediction depends on the spectral composition and spatial resolution of the imagery. Areas with finer particle accumulation and higher albedo values exhibit the highest soil erosion rates. Therefore, integrating neural network models with remote sensing techniques and selecting an appropriate scale are recommended for enhanced soil erosion prediction accuracy. The novelty of this research lies in the integration of spectral reflectance analysis with advanced deep neural network algorithms and remote sensing techniques for predicting sand dune movement and soil erosion.</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

Automatic prediction of soil erosion in arid regions using multi-sensor remote sensing integrated with object-based image analysis and deep neural networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Huayu Lu,
  • Marzieh Mokarram

摘要

Soil erosion induced by sand dune movement significantly threatens desert ecosystems, requiring precise prediction methods for sustainable land management. This study employs Object-Based Image Analysis (OBIA), Convolutional Neural Networks (CNN), and the Fuzzy Analytic Hierarchy Process (FAHP) to accurately quantify morphometric characteristics of sand dunes across multiple scales in the Taklamakan Desert, China, and the Lut Desert, Iran. The research also examines the relationship between spectral reflectance values, such as albedo, and morphometric features concerning soil erosion and sand dune dynamics. Results demonstrate that the Self-Attention Generative Adversarial Network (SAGAN) model effectively enhances satellite image quality, enabling precise sand dune extraction. Furthermore, pairwise comparison of topographic parameters weighted by the FAHP method, combined with high-resolution imagery, facilitates accurate prediction of active desert processes and the generation of detailed landform maps. Findings indicate that the optimal scale for sand dune delineation and morphometric feature identification for soil erosion prediction depends on the spectral composition and spatial resolution of the imagery. Areas with finer particle accumulation and higher albedo values exhibit the highest soil erosion rates. Therefore, integrating neural network models with remote sensing techniques and selecting an appropriate scale are recommended for enhanced soil erosion prediction accuracy. The novelty of this research lies in the integration of spectral reflectance analysis with advanced deep neural network algorithms and remote sensing techniques for predicting sand dune movement and soil erosion.