Next generation forest dynamics modelling using remote sensing data

全球生态系统受到气候变化和生物多样性丧失的威胁。许多国家计划利用森林减缓气候变化，但许多森林生态系统受到气候变化本身的威胁。例如，到2100年，气候引起的物种迁移可能使欧洲森林的价值减半。未来森林如何吸收和储存碳，关键取决于各个物种对气候变化的反应，因此预测气候变化对森林的影响是一个优先事项。现有的模型是不合适的，因为它们是围绕小规模的地块数据构建的，所以不能预测国家，大陆或全球范围内的森林的未来。准确预测欧洲森林的未来需要大型监测网络，最佳利用现有数据，尖端的测量技术和预测性，数据驱动的模型。该研究金将开发全新的方法，通过利用机器学习和人工智能技术，将现有信息与新的尖端遥感技术的数据相结合，测量和模拟森林动态，包括多样性和生态细节的森林模型捕捉长期的演替动态和多样性变化，并可以预测碳储存的变化，但是由于它们需要不能广泛获得的详细地面数据，因此在空间上是有限的。国家森林资源清查包含来自空间广阔的地块网络的关于多样性和人口比率的信息。然而，这些调查需要大量劳动力，因此通常每十年才进行一次，并且只包含简单的地面结构测量（如树干直径和高度）。地球观测卫星数据可以在大的空间和长的时间尺度上提供，并提供关于森林功能的信息。然而，由于解释方面的挑战、低空间分辨率以及从空间可测量的内容之间的不匹配，这些数据未被生态学家充分利用。（主要是树冠特性）以及基于地面测量的模型中所代表的内容地面激光扫描（Terrestrial Laser Scanning）和无人机遥感（Drone Remote Sensing）等新技术可以以前所未有的细节捕捉单个树木和整个森林树冠的3D信息，提供地面和地球观测数据之间的联系。它们可以测量树木和树冠形状、叶面积和排列，以及树冠中的树冠镶嵌，这些都是已知的生产力和动态驱动因素，为森林建模开辟了新的方法。此外，新的地球观测卫星，如欧洲航天局的哨兵卫星，提供高空间分辨率的全球数据，为监测创造新的机会，该研究金将建立一个新的概念框架，用于模拟森林动态，用欧洲各地的森林数据进行参数化和测试。该模型将把地面激光扫描和无人机数据联系起来，以绘制关于多样性和动态的信息，并将通过吸收地球观测数据，有力地预测森林对气候变化的反应。这将提高模型的空间覆盖范围和准确性，并有可能在每月而不是十年的时间步长进行校准和验证。新的3D测量将为冠层结构如何影响动态提供新的见解。机器学习和人工智能方法将用于无人机数据的自动物种检测，从而实现大空间区域的生态监测。该奖学金将通过一个完全数据驱动的模型，结合尖端监测，创造关于欧洲森林如何运作以及如何应对气候变化的新知识。该方法将能够对气候变化对森林多样性和动态的影响进行可靠和可更新的预测，并灵活地纳入未来的数据流，从而为整个非洲大陆的气候变化减缓政策提供信息。

Ecosystems are threatened globally by climate change and biodiversity loss. Many countries plan to use forests for climate change mitigation, but many forest ecosystems are threatened by climate change itself. For example, climate-induced species' shifts may halve the value of European forests by 2100. How forests will absorb and store carbon in the future depends critically on individual species' responses to climate change, so predicting the impact of climate change on forests is a priority. Existing models are not suitable because they are constructed around small-scale plot data, so cannot predict the future of forests at national, continental or global scales.Accurate predictions of the future of European forests require large monitoring networks, optimal use of existing data, cutting-edge measurement techniques, and predictive, data-driven models. This fellowship will develop wholly novel approaches to measuring and modelling forest dynamics by integrating existing information with data from new, cutting-edge remote sensing technologies using techniques drawn from machine learning and artificial intelligence.Forest models that include diversity and ecological detail capture long-term succession dynamics and diversity shifts, and can predict changes in carbon storage, but are spatially limited because they require detailed ground data not widely available. National forest inventories contain information on diversity and demographic rates from spatially extensive plot networks. However, they are labour intensive to survey so are often only carried out once per decade, and contain only simple ground measurements of structure (such as trunk diameter and height). Earth Observation satellite data are available at large spatial and over long temporal scales, and provide information on forest function. However, these data are underused by ecologists due to challenges in interpretation, low spatial resolution, and mismatches between what is measurable from space (primarily canopy properties) and what is represented in models based on ground measurements (primarily of individual trunks).New technologies such as Terrestrial Laser Scanning and drone remote sensing can capture 3D information on individual trees and whole-forest canopies in unprecedented detail, offering a link between ground and Earth Observation data. They can measure tree and crown shape, leaf area and arrangement, and crown tessellation in canopies, which are known drivers of productivity and dynamics, opening up the possibility of new approaches to forest modelling. Additionally, new Earth Observation satellites such as the European Space Agency's Sentinels provide global data at high spatial resolution, creating new opportunities for monitoring.This Fellowship will create a new conceptual framework for modelling forest dynamics, parameterised and tested with forest data from across Europe. The model will link Terrestrial Laser Scanning and drone data to plot information on diversity and dynamics, and will predict forest responses to climate change robustly by additionally assimilating Earth Observation data. This will improve model spatial coverage as well as accuracy, with calibration and validation possible at monthly rather than decadal time-steps. New 3D measurements will give novel insights into how canopy structure influences dynamics. Machine learning and artificial intelligence approaches will be used to automate species detection from drone data, allowing ecological monitoring across large spatial areas.The Fellowship will create new knowledge of how European forests function and how they will respond to climate change, with a fully data-driven model that incorporates cutting-edge monitoring. The approach will enable robust and updatable predictions of climate change impacts on forest diversity and dynamics, with flexibility to incorporate future data streams, that could inform climate change mitigation policy across the continent.