广西-东盟多源遥感数据的森林高度反演及应用

The inversion and application of forest height of multi-source remote sensing data in Guangxi-ASEAN region

森林生态系统在调节生态气候与碳循环方面发挥着重要作用,森林高度是衡量森林生态系统功能的重要参数。利用单一遥感数据获取森林冠层高度会受到多种制约。因此,本文使用星载激光雷达ICESat-2提供的高质量离散森林冠层高度点,结合Sentinel-1、Landsat 8及地形数据,采用随机森林方法建立不同影像特征组合森林冠层高度的回归模型,并分析各特征对森林高度反演的影响,最后将模型应用于广西森林冠层高度制图。试验结果表明,多源遥感数据可有效提高森林冠层高度反演精度,在所利用遥感数据中,特征重要性从大到小依次为光学特征、地形特征、SAR特征,“L8+SRTM+Sentinel-1+邻域均值”特征组合的反演精度最高,加入邻域均值特征进行森林冠层高度反演效果最佳,随机森林模型能精确绘制森林冠层高度。

Forest ecosystems play a critical role in regulating the ecological climate and carbon cycling,and forest height is a fundamental parameter for assessing their functional capacity.However,the acquisition of forest canopy height using single remote sensing data is subject to various constraints.Therefore,in this paper,we use high-quality discrete forest canopy height points from the spaceborne laser altimeter ICESat-2,combined with Sentinel-1,Landsat-8,and terrain data,to establish regression models of forest canopy height with different combinations of image features by using the random forest method,analyzing the impact of each feature on forest height inversion,and applying the model to forest canopy height mapping in Guangxi.The experimental results indicate that multisource remote sensing data effectively enhance the accuracy of forest canopy height estimation.Among the utilized remote sensing data,the feature importance descends in the order:optical features,terrain features,SAR features.The combination of“L8+SRTM+Sentinel-1+neighborhood mean”features achieves the highest accuracy in canopy height estimation,with the inclusion of neighborhood mean feature yielding the best results.The random forest model demonstrates precision in mapping forest canopy height.