当前位置: X-MOL 学术Remote Sens. Ecol. Conserv. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Detecting overmature forests with airborne laser scanning (ALS)
Remote Sensing in Ecology and Conservation ( IF 3.9 ) Pub Date : 2022-07-15 , DOI: 10.1002/rse2.274
Marc Fuhr 1 , Etienne Lalechère 2 , Jean‐Matthieu Monnet 1 , Laurent Bergès 1
Affiliation  

Building a network of interconnected overmature forests is crucial for the conservation of biodiversity. Indeed, a multitude of plant and animal species depend on forest structural maturity attributes such as very large living trees and deadwood. LiDAR technology has proved to be powerful when assessing forest structural parameters, and it may be a promising way to identify existing overmature forest patches over large areas. We first built an index (IMAT) combining several forest structural maturity attributes in order to characterize the structural maturity of 660 field plots in the French northern Pre-Alps. We then selected or developed LiDAR metrics and applied them in a random forest model designed to predict the IMAT. Model performance was evaluated with the root mean square error of prediction obtained from a bootstrap cross-validation and a Spearman correlation coefficient calculated between observed and predicted IMAT. Predictors were ranked by importance based on the average increase in the squared out-of-bag error when the variable was randomly permuted. Despite a non-negligible RMSEP (0.85 for calibration and validation data combined and 1.26 for validation data alone), we obtained a high correlation (0.89) between the observed and predicted IMAT values, indicating an accurate ranking of the field plots. LiDAR metrics for height (maximum height and height heterogeneity) were among the most important metrics for predicting forest maturity, together with elevation, slope and, to a lesser extent, with metrics describing the distribution of echoes' intensities. Our framework makes it possible to reconstruct a forest maturity gradient and isolate maturity hot spots. Nevertheless, our approach could be considerably strengthened by taking into consideration site fertility, collecting other maturity attributes in the field or developing adapted LiDAR metrics. Including additional spectral or textural metrics from optical imagery might also improve the predictive capacity of the model.

中文翻译:

使用机载激光扫描 (ALS) 检测过熟森林

建立相互关联的过熟森林网络对于保护生物多样性至关重要。事实上,许多动植物物种都依赖于森林结构成熟度属性,例如非常大的活树和枯木。在评估森林结构参数时,激光雷达技术已被证明是强大的,它可能是识别大面积现有过熟森林斑块的有前途的方法。我们首先建立了一个指数(IMAT),结合了几个森林结构成熟度属性,以表征法国北部前阿尔卑斯山 660 个田地的结构成熟度。然后,我们选择或开发了 LiDAR 指标,并将它们应用于旨在预测 IMAT 的随机森林模型中。模型性能通过自举交叉验证获得的预测均方根误差和在观察到的和预测的 IMAT 之间计算的 Spearman 相关系数进行评估。当变量被随机排列时,预测变量的重要性是根据袋外误差平方的平均增加来排序的。尽管 RMSEP 不可忽略(校准和验证数据组合为 0.85,单独验证数据为 1.26),但我们在观察到的 IMAT 值和预测的 IMAT 值之间获得了高度相关性(0.89),表明场图的准确排名。LiDAR 高度指标(最大高度和高度异质性)是预测森林成熟度的最重要指标之一,与海拔、坡度以及在较小程度上描述回波分布的指标一起 强度。我们的框架使得重建森林成熟度梯度和隔离成熟度热点成为可能。尽管如此,我们的方法可以通过考虑场地肥力、收集该领域的其他成熟度属性或开发适应的 LiDAR 指标来大大加强。包括来自光学图像的额外光谱或纹理指标也可能会提高模型的预测能力。
更新日期:2022-07-15
down
wechat
bug