Scholarlink

<p>In urban flood risk simulations, vegetation canopy interception and green space Manning coefficients are crucial for estimating stormwater runoff and flood risk, given that these parameters are modulated by the vegetation type and density. However, coarse vegetation classification and difficulties in measuring vegetation density lead to significant inaccuracies in urban flood risk assessments. This study offers a method that integrates diverse remote sensing data for vegetation classification and density evaluation. This approach refines vegetation distribution parameters to calculate canopy interception and determine Manning coefficients for green spaces. The study employs Sentinel-1, Sentinel-2, and GF-2 satellite imagery to extract multi-source feature information, such as spectral data, vegetation indices, texture features, and backscattering coefficients. An ensemble learning classifier, used for vegetation type classification, achieves higher efficacy than established methods, including Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Multi-layer Perceptron (MLP). The ensemble model demonstrated an Overall Accuracy (<i>OA</i>) of 92.70%, reflecting an average improvement of 12.6 percentage points. Additionally, the median accuracy obtained during cross-validation surpassed 0.85, indicating the model’s robust and consistent performance throughout diverse validation groups. Furthermore, a novel method for vegetation density assessment is introduced, which leverages a combined threshold of vegetation coverage and texture features. This method enables effective differentiation of vegetation density, as confirmed through parameter sensitivity analysis. The findings confirms that the introduced method markedly enhances the precision of vegetation classification and density assessment, offering a robust framework for calculating canopy interception and Manning coefficients in green spaces. Consequently, this methodology enhances the accuracy of urban flood risk simulations, contributing valuable insights for improving flood risk management strategies.</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

Fine-Scale Classification and Density Assessment of Urban Vegetation Based on Multi-source Remote Sensing Data

  • 期刊论文

  • 发表日期：

    2025年10月28日

查看全文
  • Qiuxu Yan,
  • Jingcheng Yuan,
  • Dong Wu,
  • Wang Cao,
  • Weidong Yan

摘要

In urban flood risk simulations, vegetation canopy interception and green space Manning coefficients are crucial for estimating stormwater runoff and flood risk, given that these parameters are modulated by the vegetation type and density. However, coarse vegetation classification and difficulties in measuring vegetation density lead to significant inaccuracies in urban flood risk assessments. This study offers a method that integrates diverse remote sensing data for vegetation classification and density evaluation. This approach refines vegetation distribution parameters to calculate canopy interception and determine Manning coefficients for green spaces. The study employs Sentinel-1, Sentinel-2, and GF-2 satellite imagery to extract multi-source feature information, such as spectral data, vegetation indices, texture features, and backscattering coefficients. An ensemble learning classifier, used for vegetation type classification, achieves higher efficacy than established methods, including Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Multi-layer Perceptron (MLP). The ensemble model demonstrated an Overall Accuracy (OA) of 92.70%, reflecting an average improvement of 12.6 percentage points. Additionally, the median accuracy obtained during cross-validation surpassed 0.85, indicating the model’s robust and consistent performance throughout diverse validation groups. Furthermore, a novel method for vegetation density assessment is introduced, which leverages a combined threshold of vegetation coverage and texture features. This method enables effective differentiation of vegetation density, as confirmed through parameter sensitivity analysis. The findings confirms that the introduced method markedly enhances the precision of vegetation classification and density assessment, offering a robust framework for calculating canopy interception and Manning coefficients in green spaces. Consequently, this methodology enhances the accuracy of urban flood risk simulations, contributing valuable insights for improving flood risk management strategies.