Scholarlink

<p>Accurate identification of growth stress factors is essential for the conservation and restoration of Suaeda salsa in coastal wetlands, where vegetation dynamics are jointly affected by chemical, hydrological, and biological disturbances. This study developed a multimodal machine learning framework by integrating remote sensing imagery, near-surface observations, and ground survey data to assess the relative contributions of key stress factors in the Liaohe Estuary wetland. The main novelty of this work lies in the explicit quantification of crab burrow density as a biological stressor and the identification of its interaction with soil salinity and moisture conditions. The results showed that soil salinity was the dominant stress factor affecting Suaeda salsa growth, followed by crab burrow density, mean NDVI, precipitation, and soil moisture. The proposed CNN-RF hybrid model improved prediction accuracy compared with conventional machine learning methods and showed better robustness under compound stress conditions. Interpretability analysis based on the SHAP-LIME framework further revealed a salinity–moisture–crab burrow interaction mechanism, indicating that biological disturbance can amplify the negative effects of salt and water stress. The GIS-integrated decision support system enabled spatial visualization and early warning of growth degradation risks, providing a practical tool for precision management and ecological restoration of coastal wetlands.</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

Intelligent assessment of Suaeda salsa growth stress factors based on multimodal fusion and transfer learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月16日

查看全文
  • Hanyu Wang,
  • Zengwen Gao

摘要

Accurate identification of growth stress factors is essential for the conservation and restoration of Suaeda salsa in coastal wetlands, where vegetation dynamics are jointly affected by chemical, hydrological, and biological disturbances. This study developed a multimodal machine learning framework by integrating remote sensing imagery, near-surface observations, and ground survey data to assess the relative contributions of key stress factors in the Liaohe Estuary wetland. The main novelty of this work lies in the explicit quantification of crab burrow density as a biological stressor and the identification of its interaction with soil salinity and moisture conditions. The results showed that soil salinity was the dominant stress factor affecting Suaeda salsa growth, followed by crab burrow density, mean NDVI, precipitation, and soil moisture. The proposed CNN-RF hybrid model improved prediction accuracy compared with conventional machine learning methods and showed better robustness under compound stress conditions. Interpretability analysis based on the SHAP-LIME framework further revealed a salinity–moisture–crab burrow interaction mechanism, indicating that biological disturbance can amplify the negative effects of salt and water stress. The GIS-integrated decision support system enabled spatial visualization and early warning of growth degradation risks, providing a practical tool for precision management and ecological restoration of coastal wetlands.