A Theoretical and Computational Framework for Linking Tree form and Function to Forest Diversity and Productivity

怀俄明大学获得了一笔赠款，用于开发一个规模框架，以了解森林多样性和生产力。本研究解决了开发该框架的三个最重要的问题。(1)与树形相关的性状（如异速生长、形态）和功能（如生理、生长、分配、生存）在不同物种之间如何变化，进化和环境驱动因素如何影响性状变异？(2)准确描述群落和生态系统特性（如多样性、演替、生产力、碳循环）是否需要特定物种的形式和功能表征？(3)如何建立一个包括物种特异性性状变异和关键生理机制在内的大尺度森林动态预测的通用尺度框架？为了解决这些问题，本研究开发并应用了数据模型集成方法，包括：(1)将树的形态和功能联系起来，结合植物的关键功能性状，并适用于广泛的时空尺度的动态过程模型；（ii）新的元分析方法，用于分析包含系统发育关系的物种特异性特征的大量文献信息，并克服了经典元分析方法的共同局限性；（iii）采用严格的统计和计算方法，为过程模型提供大量不同的数据源（即文献、森林清查和树木年轮宽度数据库）。这种高度集成的方法将为构建和测试通用扩展框架迈出重要的一步。这项工作的广泛影响包括本科生到研究生科学家在数据模型集成方法方面的多个培训机会。本研究中开发的方法将通过美国生态学会年会的为期一天的生态学贝叶斯分析年度研讨会部分传播。许多大学课程缺乏数据模型集成，特别是贝叶斯方法的培训，因此将进一步开发和整合贝叶斯数据分析和高级/计算贝叶斯两门新的研究生课程，为怀俄明大学（UW）提供应用统计建模和计算的现代课程。现代统计建模的培训将为那些在华盛顿大学新的和充满活力的生态学研究生课程的博士生提供一个独特的教育机会。该研究还将为UW本科生创造独立研究的机会，并将为一名博士后和两名博士提供独特的跨学科教学，指导和研究培训，包括生态学，统计学，数学和计算科学。

The University of Wyoming is awarded a grant to develop a scaling framework for understanding forest diversity and productivity. This study addresses three questions that are paramount to developing this framework. (1) How do traits related to tree form (e.g., allometries, morphology) and function (e.g., physiology, growth, allocation, survival) vary between species, and how do evolutionary versus environmental drivers affect trait variability?(2) Is a species-specific representation of form and function necessary to accurately describe community and ecosystem properties (e.g., diversity, succession, productivity, carbon cycling)? (3) How do we develop a general scaling framework for predicting large-scale forest dynamics that includes species-specific trait variability and key physiological mechanisms? Towards addressing these questions, this study develops and applies data-model integration methodologies, including: (i) dynamic process models that link tree form and function, incorporate key plant functional traits, and are applicable to broad spatial and temporal scales; (ii) new meta-analysis methods for analyzing vast amounts of literature information on species-specific traits that incorporate phylogenetic relationships and overcome limitations common to ³classical² meta-analytic approaches; and(iii) rigorous statistical and computational methods for informing the process model with large and disparate data sources (i.e., literature, forest inventory, and tree-ring width databases). This highly integrative approach will provide a major step towards building and testing a general scaling framework. The broader impacts of this work include multiple training opportunities in data-model integration methods for undergraduates through post-graduate scientists. Methods developed in this study will be partly disseminated through an annual, daylong workshop on Bayesian analysis in ecology for the Ecological Society of America annual meetings. Training in data-model integration, and specifically Bayesian methods, is lacking in many university curriculums and two new graduate-level courses in Bayesian data analysis and advanced/computational Bayesian will be further developed and integrated, providing a modern curriculum in applied statistical modeling and computing at the University of Wyoming (UW). Training in modern statistical modeling will offer a unique educational opportunity for those PhD students in UW¹s new and vibrant graduate Program in Ecology. This study will also create independent research opportunities for UW undergraduates, and it will provide one post-doctoral and two PhD students with unique interdisciplinary teaching, mentoring, and research training in ecology, statistics, mathematics, and computational science.