Improving precision of biomass estimation in field and remote sensing-based forest monitoring - considering continuous horizontal biomass distributions

森林生物量是现代森林监测的核心变量之一。估算精度是森林监测的一个重要特征。森林生物量评估的研究主要集中在优化采样设计和增强基于实地测量树木数据（异速生长模型）和遥感数据（用于区划和基于模型的推断）的统计生物量模型。我们将研究通过改善每块地（固定区域）地上生物量（AGB）预测来提高估计精度的方法：我们引入了一种新方法，将森林生物量视为具有连续水平分布的变量，因此每块地AGB预测是严格在地块周界范围内的树木生物量。我们将这种方法称为地块生物量预测的连续方法（CA），与传统的离散方法（DA）相反。在传统的离散方法（DA）中，将树内离散数量的总生物量作为地块生物量，而不管该生物量有多少真正高于地块面积。在CA中，排除了在地块边界上倾斜的部分内树，而需要考虑延伸到地块的外树的生物量。为了使这种方法可行，我们引入了单个树的水平生物量分布（THBD），并从经验数据和理论考虑中得出其形状。我们使用完全映射的林分来比较传统的DA和我们新提出的CA的统计性能，我们希望更好地理解林分特征、样地大小、样地形状和（不可避免的）假设之间的相互作用。在初步研究的基础上，我们期望在估计精度上有所提高；我们也会评估成本问题，因为CA要求包括在地块范围内具有树冠分数的额外外树。本提案的基础研究有两个主要的新元素，以前没有进行过：(1)THBD的推导，(2)树木生物量水平分布的推导。利用CA进行野外样地生物量预测，我们假设可以通过(1)森林清查现场采样和(2)将野外观测生物量与遥感数据相结合的监测方法来提高估算精度。对两者的分析是本课题研究计划的核心内容。THBD的进一步应用——这里没有研究——可以在模拟林分发展和森林火灾蔓延的背景下看到。

Forest biomass is among the core variables in modern forest monitoring. Precision of estimation is an important feature in forest monitoring. Research on the assessment of forest biomass focusses largely on optimizing sampling designs and enhancing statistical biomass models from field measured tree data (allometric models) and from remote sensing data (for regionalization and model-based inference). We will look at options to increase precision of estimation by improving the per plot (fixed area plots) prediction of above ground biomass (AGB): we introduce a new approach that looks at the forest biomass as a variable with a continuous horizontal distribution so that the per-plot AGB prediction is the tree biomass which is strictly within the confinements of the plot perimeter. We call this the continuous approach (CA) to plot-biomass prediction - as opposed to the conventional discrete approach (DA), in which the total biomass of the discrete number of in-trees is taken as plot biomass - regardless how much of this biomass is truly above the plot area. In the CA, parts of in-trees that slop over the plot boundary are excluded while biomass fractions of out-trees that extend into the plot need to be considered. To make this approach operational, we introduce the individual tree horizontal biomass distribution (THBD) and derive its shape from empirical data and theoretical considerations. We use fully mapped stands to compare the statistical performance of the conventional DA and our newly proposed CA, where we wish to better understand the interactions between stand characteristics, plot size, plot shape and (the unavoidable) assumptions in modelling the THBD. Based on a pilot study we expect a gain in precision of estimation; we do also evaluate the cost issue as the CA requires including additional out-trees that have crown fractions within the confinements of the plot. There are two major novel elements of basic research in this proposal that had not been worked on before: (1) the derivation of the THBD, (2) the derivation of the stand-wise horizontal distribution of tree biomass. Using the CA to field plot biomass prediction, we hypothesize to increase precision of estimation from both (1) forest inventory field sampling and (2) from monitoring approaches that integrate field observed biomass with remotely sensed data. Analyses on both are a core element of the research plan of this proposal. Further applications of the THBD - not researched here - are seen in the context of modelling the development of forest stands and the spread of forest fires.