Estimating the height of wetland vegetation using airborne discrete-return LiDAR data
Estimating the height of wetland vegetation using airborne discrete-return LiDAR data
复制标题
使用机载离散回波激光雷达数据估算湿地植被的高度
DOI：
10.1016/j.ijleo.2017.10.016
复制
发表时间：
2018
期刊：
影响因子：
3.1
通讯作者：
Tian Jianlin
中科院分区：
文献类型：
--
作者：
Nie Sheng;Wang Cheng;Xi Xiaohuan;Luo Shezhou;Li Shihua;Tian Jianlin
已经进行了几项研究，以估计湿地植被高度使用机载激光雷达数据。然而，没有研究已经进行了探讨植被覆盖对植被高度估计模型的影响。本研究的目的是在不同植被覆盖度的湿地中估算植被高度，以进一步分析植被覆盖度的影响。首先，我们进行了线性和对数回归分析之间的实地测量的高度和每个激光雷达派生的度量在湿地不同植被覆盖。然后通过多元回归分析将LiDAR导出的指标组合起来估计湿地植被高度。最后用留一法交叉验证法对高度估计值进行验证。结果表明，对数回归分析在估算湿地植被高度方面优于线性回归分析，无论植被覆盖率如何，几乎所有的LiDAR数据都可以用对数回归分析来估算。激光雷达返回的最大高度（Hmax）提供了最好的协议与实地测量的高度在湿地高植被覆盖。相比之下，最强的相关性观察到使用激光雷达高度（HSD）的标准差低植被覆盖。多元回归分析结果表明，在植被覆盖度较高的湿地，最佳模型是基于Hmax和激光雷达回波的90百分位高度（H90），而在植被覆盖度较低的湿地，最佳模型可以表示为HSD和H90的函数。因此，植被覆盖度是影响植被高度模型的重要因素。此外，我们的研究将提供有价值的指导实地测量数据收集和植被高度估计使用激光雷达数据。
Several studies have been conducted to estimate wetland vegetation height using airborne LiDAR data. However, no studies have been conducted to explore the influence of vegetation cover on vegetation height estimation models. The objective of this research is to estimate vegetation height in wetlands with varying vegetation cover for further analyzing the effects of vegetation cover. First, we performed both linear and logarithmic regression analyses between field-measured heights and each LiDAR-derived metric in wetlands with different vegetation cover. Then LiDAR-derived metrics were combined through multiple regression analysis to estimate wetland vegetation height. The height estimates were finally validated by leave-one-out cross-validation method. The results showed that the logarithmic regression analysis performed better than the linear regression analysis in estimating wetland vegetation height for almost all LiDAR-derived metrics regardless of vegetation cover. The max height of LiDAR returns (Hmax) provided the best agreement with field-measured heights in wetlands with high vegetation cover. In contrast, the strongest correlation was observed using the standard deviation of LiDAR heights (HSD) for low vegetation cover. Results of multiple regression analysis indicated that the best model was based on Hmaxand 90 percentile height of LiDAR returns (H90) in wetlands with high vegetation cover, while the best model can be expressed as a function of HSDand H90 for low vegetation cover. Therefore, our study concluded that vegetation cover is an important factor of vegetation height models. Additionally, our study will provide valuable guidance for field-measured data collection and vegetation height estimation using LiDAR data.