Scholarlink

<p>Shallow water bathymetry mapping is essential for coastal management, navigation safety, and resource assessment. Traditional surveying methods are often costly and time-consuming, making satellite-derived bathymetry (SDB) a promising alternative. This research presents a novel framework based on the integration of ensemble machine learning and multispectral satellite imagery for enhancing SDB accuracy in shallow water zones. Four advanced machine learning models, including Extra Tree Regression (ETR), Extreme Gradient Boosting (XGB), Voting Regression (VR), and Random Forest (RF), were used to estimate water depth from satellite-derived reflectance values. The models were trained and validated using satellite imagery and ground data collected from the New Mansoura coastal area in Egypt. To investigate the accuracy of the proposed integrated techniques, PlanetScope and Sentinel-2 images were used and compared. The results demonstrate the effectiveness of the integrated ensemble approach and PlanetScope images, where the bathymetry estimation accuracy is improved compared to conventional techniques. The ETR model outperformed other proposed models in mapping bathymetry of coastal zones with a coefficient of determination (R) equal 0.99 and 0.97 along with root mean square error (RMSE) equal 15&#xa0;cm and 26&#xa0;cm for PlanetScope and Sentinel-2 images, respectively. Based on the International Hydrographic Organization’s (IHO) standards, the accuracy of the developed SDB-based ETR model when implemented with PlanetScope images was found to fulfil all hydrographic survey orders accuracy requirements. The practical implications of the research was realized through the adherence to the IHO standards that adds a practical relevance to hydrographic survey applications. The sensitivity analysis shows the significant impact of green and blue bands of PlanetScope images in bathymetric modeling. Therefore, the developed approach can offer a valuable tool for coastal monitoring, environmental assessment, and decision-making processes, for similar regions.</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

Integrated Ensemble Learning for Enhanced Accuracy in Satellite-Derived Bathymetry of Shallow Coastal Zones

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Mohamed T. Elnabwy,
  • Mosbeh R. Kaloop,
  • Furquan Ahmad,
  • Mohammed El-Diasty,
  • Jong Wan Hu,
  • Pijush Samui,
  • Tamer ElGharbawi,
  • Fawzi Zarzoura

摘要

Shallow water bathymetry mapping is essential for coastal management, navigation safety, and resource assessment. Traditional surveying methods are often costly and time-consuming, making satellite-derived bathymetry (SDB) a promising alternative. This research presents a novel framework based on the integration of ensemble machine learning and multispectral satellite imagery for enhancing SDB accuracy in shallow water zones. Four advanced machine learning models, including Extra Tree Regression (ETR), Extreme Gradient Boosting (XGB), Voting Regression (VR), and Random Forest (RF), were used to estimate water depth from satellite-derived reflectance values. The models were trained and validated using satellite imagery and ground data collected from the New Mansoura coastal area in Egypt. To investigate the accuracy of the proposed integrated techniques, PlanetScope and Sentinel-2 images were used and compared. The results demonstrate the effectiveness of the integrated ensemble approach and PlanetScope images, where the bathymetry estimation accuracy is improved compared to conventional techniques. The ETR model outperformed other proposed models in mapping bathymetry of coastal zones with a coefficient of determination (R) equal 0.99 and 0.97 along with root mean square error (RMSE) equal 15 cm and 26 cm for PlanetScope and Sentinel-2 images, respectively. Based on the International Hydrographic Organization’s (IHO) standards, the accuracy of the developed SDB-based ETR model when implemented with PlanetScope images was found to fulfil all hydrographic survey orders accuracy requirements. The practical implications of the research was realized through the adherence to the IHO standards that adds a practical relevance to hydrographic survey applications. The sensitivity analysis shows the significant impact of green and blue bands of PlanetScope images in bathymetric modeling. Therefore, the developed approach can offer a valuable tool for coastal monitoring, environmental assessment, and decision-making processes, for similar regions.