Early detection of GPP-related regime shifts after plant invasion by integrating imaging spectroscopy with airborne LiDAR
Early detection of GPP-related regime shifts after plant invasion by integrating imaging spectroscopy with airborne LiDAR
复制标题
DOI：
10.1016/j.rse.2018.02.038
复制
发表时间：
2018-05-01
影响因子：
13.5
通讯作者：
Oldeland, J.
中科院分区：
文献类型：
--
作者：
Grosse-Stoltenberg, A.;Hellmann, C.;Oldeland, J.
入侵植物物种可对生态系统的结构和功能产生高度的、自我强化的影响，从而导致生态系统特性的永久性变化。因此，需要早期发现和及时管理，以减轻入侵的生态系统后果。将机载高光谱图像与激光雷达数据相结合，即使在入侵的早期阶段，也可以提供有关入侵者发生和生态系统变化的空间明确信息。然而，相关的“模型入侵者”和特征良好的生态系统需要被识别，以增加入侵理论的预测能力和优先管理。此外，仍然有一个知识差距，以传感器为基础的方法，是有效的空间和时间来评估入侵工程师对生态系统功能的影响，以及潜在的诱导政权转移。在这项研究中，发生和时空影响的入侵N-2固定灌木，长叶金合欢，进行了评估，在一个异质的，地中海沙丘生态系统。入侵者映射使用植被指数来自机载高光谱图像以及机载激光雷达数据使用随机森林分类与灵敏度为0.79，正预测值（PPV）为0.81，科恩的Kappa值为0.77。入侵的网站之间变化的早期阶段与低覆盖，其中孤立的补丁被检测到，严重侵扰A。长叶灌丛。对历史图像的分析表明，入侵者可以在开放的沙丘平原的恶劣条件下建立，可能是由人类干扰引发的。最近开发的近红外植被指数（NIRv），这是相关的总初级生产力（GPP），增加了线性和显着的入侵覆盖。这表明入侵者引起的GPP相关的制度转变，改变生态系统的生产力代表开放的灌木林的森林。这种转变甚至可以在入侵的早期阶段发现。因此，NIRv指数可以提供适当的基于传感器的“模型度量”来评估侵入性工程师的影响。这就提供了预测和预测政权转变的机会，作为及时管理的基础。
Invasive plant species can have high, self-reinforcing impacts on ecosystem structure and functioning that induce permanent changes of ecosystem properties. Therefore, early detection and timely management is required to alleviate ecosystem consequences of invasion. Integrating airborne hyperspectral imagery with LiDAR data can deliver spatially explicit information on invader occurrence and ecosystem transformations even at early stages of invasion. However, relevant "model invaders" and well-characterized ecosystems need to be identified to both increase predictive power of invasion theory and prioritize management. In addition, there is still a knowledge gap regarding sensor-based approaches that are valid in space and time to assess the impact of invasive engineers on ecosystem functioning as well as the potential to induce regime shifts. In this study, occurrence and spatio-temporal impact of the invasive N-2-fixing shrub, Acacia longifolia, was assessed in a heterogeneous, Mediterranean dune ecosystem. The invader was mapped using vegetation indices derived from airborne hyperspectral images as well as airborne LiDAR data using Random Forest classification with a Sensitivity of 0.79, a Positive Predicted Value (PPV) of 0.81, and Cohen's Kappa of 0.77. Invaded sites varied between early stages with low cover, where isolated patches were detected, to heavily infested A. longifolia thickets. Analysis of historical images showed that the invader could establish under the harsh conditions of open dune plains, possibly triggered by human interference. The recently developed Near-Infrared Vegetation Index (NIRv), which is related to Gross Primary Production (GPP), increased linearly and significantly with increasing invader cover. This indicated a GPP-related regime shift induced by the invader, changing ecosystem productivity representative of open shrublands to that of forests. Such a shift could even be identified at early stages of invasion. Thus, the NIRv index may provide an appropriate sensor-based "model metric" to assess impacts of invasive engineers. This offers the opportunity to predict and anticipate regime shifts as a basis for timely management.