MCA: Leveraging Artificial Intelligence to Improve Understanding of Biogenic Volatile Organic Compound Emissions and Chemistry over Heterogeneous Forest Landscapes

这个中期职业发展（MCA）项目将利用成像技术、人工智能（AI）和机器学习（ML）技术的进步，推进采样策略和预测方法，以绘制异质景观中生物源性挥发性有机化合物（BVOC）的排放。BVOCs通过影响对流层的氧化能力和大气微量气体的相关化学循环，在大气化学中发挥重要作用。活性挥发性有机化合物的排放对臭氧产生、空气质量、健康影响和气候都有影响。该项目将对森林成像数据应用人工智能方法，在缅因州中部和巴西亚马逊州玛瑙斯附近两种不同的森林生态系统类型中规划BVOC现场采样点。由于许多目标区域难以进入，将使用无人驾驶飞行器（uav）对选定森林上方的bvoc进行采样。该项目将解决以下问题：(1)如何在有限的采样地点对异质景观中的BVOCs进行最佳采样？(2)不同类型森林的BVOC浓度差异有多大？(3)不同的林冠上BVOC浓度对下伏森林亚型BVOC排放率的差异意味着什么？(4) BVOC排放的空间变化对区域和全球排放估算的显著影响在何种尺度和程度上？这项研究的最终目标是将BVOC浓度图与MEGAN等标准排放模型预测的排放量进行比较。这个项目有可能开发出在环境传感方面有广泛应用的工具，并对大气化学和更广泛的环境科学界有很大的兴趣。该项目包括支持一名暑期本科生参与研究。此外，还计划开设一个人工智能驱动的传感和环境建模暑期学院，为本科生参与创新的STEM研究提供新的机会。该项目由地球科学理事会（Directorate for Geosciences）共同资助，以支持地球科学领域的人工智能/机器学习进步，并由既定计划（Established Program）共同资助，以刺激竞争性研究。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Mid-Career Advancement (MCA) project will leverage advances in imaging technology, artificial intelligence (AI), and machine learning (ML) techniques to advance sampling strategies and prediction methods for mapping biogenic volatile organic compound (BVOC) emissions in heterogeneous landscapes. BVOCs play an important role in the chemistry of the atmosphere by influencing the oxidative capacity of the troposphere and the associated chemical cycles of atmospheric trace gases. The emissions of reactive BVOCs have implications for ozone production, air quality, health effects, and climate. This project will apply AI methods to forest imaging data to plan BVOC field sampling locations in two disparate forest ecosystem types: Central Maine and near Manaus, Amazonas, Brazil. As many of the target areas are inaccessible, unmanned aerial vehicles (UAVs) will be used to sample BVOCs above the selected forests. This project will address the following questions: (1) How can BVOCs over a heterogeneous landscape be optimally sampled with a limited number of sampling locations? (2) To what extent do BVOC concentrations vary across heterogeneous forest features? (3) What do varying above-canopy BVOC concentrations suggest about differences in BVOC emission rates from the underlying forest subtypes? and (4) At what scale and to what extent do spatial variations in BVOC emissions significantly affect regional and global emissions estimates? The final goal of the study is to compare the maps of BVOC concentrations with emissions predicted by standard emission models such as MEGAN.This project has the potential to develop tools that could have a wide range of applications in environmental sensing and be of great interest to the atmospheric chemistry and broader environmental science community. The project includes support for a summer undergraduate student to participate in the research. There are also plans for a summer institute on AI-driven sensing and environmental modeling provide new opportunities for undergraduate students to become involved in innovative STEM research. This project is co-funded by the Directorate for Geosciences to support AI/ML advancement in the geosciences and by the Established Program to Stimulate Competitive Research.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.