Doctoral Dissertation Research: Active and Passive Remote Sensing for Predicting Tropical Tree Species Richness across Spatial Scales

本博士论文研究项目探讨了如何利用遥感生物物理变量在一定空间尺度上预测和模拟热带树种丰富度和森林结构。生态学家长期以来一直寻求能够利用遥感扩展他们的分析的空间范围，超越传统上研究树种丰富度和森林结构的样地水平空间尺度。该博士生将测试生物物理环境中的环境变异性在决定不同种类的热带树木如何能够分配资源并在有限的森林空间中共存方面具有解释力的假设。他将使用一系列遥感解释变量，包括垂直植被结构、光照有效性、地形变化和光谱反射率，来解释从现有森林普查样地数据计算出的物种丰富度、多样性和森林结构（基底面积和茎密度）的差异。他将使用广义最小二乘回归模型从遥感变量中确定树种丰富度的解释方差比例。他期望加强对哪些遥感变量解释物种丰富度和森林结构变异的最大比例，以及这些预测在哪些空间尺度上具有统计意义和生态相关性的理解。根据初步分析，他确定预测模型的准确性随着空间分辨率的提高而提高（1公顷的地块大小），并且当包括普查中的所有茎时具有最高的预测能力。该学生将包括额外的树木地块，仅关注大树（+20厘米胸径），以将预测扩展到覆盖更大比例景观的大型树冠树木。拟议的研究将提供一个框架，该框架可应用于任何类似的森林环境，包括实地清查样地数据和遥感数据。拟议的研究是针对特定地点的，但所测试的生态和生物地理学原理具有更广泛的意义。由于广泛的地面森林普查和遥感数据的可用性，拟议的研究将在巴拿马的巴罗科罗拉多自然纪念碑进行，但项目期间开发的方法将在其他拥有现有森林普查数据的热带森林科学中心（CTFS）站点使用。在其他CTFS站点测试这些新方法将为全球森林研究人员提供更多的机会。通过在全球范围内测试这些假设，生态学家和生物地理学家可以研究高密度树种共存的潜在原因和远程检测丰富度的信号。该项目将向与森林保护有关的政策组织提供有关预测树木丰富度和森林结构的资料性数据，并将使科学家和决策者了解与这种遥感预测有关的局限性和错误。通过使用树种丰富度预测图，对话小组可以在大空间尺度上最大限度地利用可用资金来瞄准物种丰富的森林。这个建议的博士论文研究改进奖将提供支持，使有前途的学生建立一个独立的研究生涯。

This doctoral dissertation research project investigates how remotely sensed biophysical variables can be used to predict and model tropical tree species richness and forest structure across a range of spatial scales. Ecologists long have sought to be able to use remote sensing to extend the spatial extent of their analysis beyond the plot-level spatial scale at which tree species richness and forest structure traditionally have been studied. The doctoral student will test the hypothesis that environmental variability in the biophysical environment contains explanatory power in determining how different species of tropical trees are able to partition resources and co-exist in limited forested space. He will use a range of remotely sensed explanatory variables, including vertical vegetation structure, light availability, topographic variation, and spectral reflectance, to explain the variance in species richness, diversity, and forest structure (basal area and stem density) calculated from existing forest census plot data. He will use generalized least-squares regression modeling to determine the proportion of explained variance in tree species richness from the remotely sensed variables. He anticipates enhancing understanding of which remotely sensed variables explain the largest proportion of variance in species richness and forest structure and at which spatial scales these predictions are statistically significant and ecologically relevant. Based on preliminary analysis, he has determined that prediction model accuracy increases as spatial resolution gets coarser (1-ha plot size) and has the highest predictive power when all stems in the census are included. The student will include additional tree plots that focus solely on large trees (+20 cm DBH) to extend the predictions to large canopy trees that cover a larger proportion of the landscape. The proposed research will provide a framework that can be applied to any similar forested environment with field inventory plot data and remote sensing data. The proposed research is site specific, but the ecological and biogeographic principles being tested have much wider-ranging significance. The proposed research will occur in the Barro Colorado Nature Monument of Panama due to the availability of extensive ground forest census and remote sensing data, but the methods developed during the project will have utility at other Center for Tropical Forest Science (CTFS) sites that have existing forest census data. Testing these new methods at other CTFS sites will provide additional opportunities for forest researchers globally. By testing these hypotheses on a global scale, ecologists and biogeographers can investigate the potential causes of high-density tree species co-existence and signals to detect richness remotely. The project will provide informative data about predicted tree richness and forest structure to policy organizations concerned with forest conservation, and it will inform scientists and policy makers about the limitations and errors associated with such remote sensing derived predictions. By using tree species richness prediction maps, conversation groups can maximize available funding to target species rich forests over large spatial scales. This proposed Doctoral Dissertation Research Improvement award will provide support to enable a promising student to establish an independent research career.