Collaborative Research: MSA: Upscaling soil organic carbon measurements at the continental scale: Evaluating emergent ecosystem properties using multivariate quantitative methods

大气中二氧化碳（CO2）的增加是全球气候变化的主要原因。应对这一挑战的最有效的自然解决方案之一就在我们的脚下——土壤。在全球范围内，土壤中的碳含量比地球大气和植被中的碳含量加起来还要多。为了应对气候变化，各国和国际上都采取了增加土壤有机碳（SOC）含量和储存能力的举措。有机碳储存的多方面好处还可以确保地球人口的粮食和营养安全，并有助于实现许多联合国可持续发展目标。然而，目前尚不清楚土壤能为我们的全球社区提供这些生态系统服务多久。部分原因是来自各种来源的SOC数据和基于计算机模型的预测并不一致。该项目旨在为美国相邻地区（CONUS）提供可靠的SOC估算，这有助于确定不同模型之间不一致的潜在原因，并最终促进决策者做出有关气候变化的明智决策。它还将为学生提供研究培训机会，并为教师提供讲习班和培训课程。对于美国来说，利用空间聚类方法减少碳有机碳动态的不确定性，并通过扩大国家生态观测站网络（NEON）的站点测量来限制大陆尺度上的模型，这是一个独特的机会。新兴生态系统的特性将通过使用多元定量方法进行外推或内插，从NEON陆地站点的空间星座到CONUS进行点尺度SOC测量。NEON地面站点收集的数据将与一系列多元地理聚类算法（k-means聚类，集成聚类）和机器学习（卷积神经网络，人工神经网络）方法相结合。这些定量分析还将使有机碳空间代表性的不确定性量化成为可能，并有助于确定潜在的未来可重新定位（或移动）地点，以便对与陆地碳循环过程相关的变量进行额外的地面真值测量。现有的NEON生物地球化学、微生物、水文学、传感器和遥感数据产品将利用气候、生态、环境、地球化学和微生物变量的类似组合，为CONUS生成定量的SOC区域地图。利用NEON数据开发的算法将在SoDaH（土壤数据协调数据库）和ISNC（国际土壤碳网络）等其他点尺度数据上进行验证。利用现有的网格化数据库：SoilGrids、世界土壤协调数据库（HWSD）、北环极土壤碳数据库（NCSCD）和网格化美国土壤调查地理数据库（gSSURGO），对CONUS基于代表性的SOC区域图的空间失配进行评估。CONUS基于eon的SOC区域地图也将与耦合模型比较项目第6阶段（CMIP6）参与模型的缩减历史SOC预测相结合。CONUS对SOC的鲁棒（和可扩展）估计将使使用历史CMIP6模型运行来诊断陆地碳循环过程成为可能。更广泛的影响将包括本科生和研究生的培训机会，以及讲授数据分析工作流程方法的车间和培训课程。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。

Increase in atmospheric carbon dioxide (CO2) is a major cause of global climate change. One of the most effective nature-based solutions to this challenge lies right under our feet - the soil. Globally, soil contains more carbon than in the Earth's atmosphere and vegetation combined. National and international initiatives are in place to increase soil organic carbon (SOC) content and storage capacity to combat climate change. The multifaceted benefits of SOC storage can also ensure food and nutritional security for the Earth's human population and help meet many of the United Nations Sustainable Development goals. However, it is not clear how long soil can provide these ecosystem services to our global community. This is partly because SOC data available from various sources and predictions based on computer models don’t agree with each other. This project aims to provide a robust estimate of SOC for the conterminous United States (CONUS), which can help identify potential reasons for inconsistency across different models and ultimately facilitate policy-makers in making informed decisions about climate change. It will also offer research training opportunities for students as well as workshops and training courses for teachers. For the U.S., there is a unique opportunity to use spatial clustering approaches to reduce uncertainties in SOC dynamics and constrain models at the continental scale by upscaling site-based measurements across the National Ecological Observatory Network (NEON). Emergent ecosystem properties will be evaluated by using multivariate quantitative methods to extrapolate or interpolate point-scale SOC measurements from a spatial constellation of NEON terrestrial sites to CONUS. Data collected across NEON terrestrial sites will be coupled with an array of multivariate geographic clustering algorithms (k-means clustering, ensemble clustering) and machine-learning (convolutional neural network, artificial neural network) approaches. These quantitative analyses will also enable uncertainty quantification of spatial representativeness of SOC and help identify potential future relocatable (or mobile) sites for additional ground-truth measurements of variables related to terrestrial C cycle processes. Existing NEON biogeochemistry, microbial, hydrology, sensor, and remote sensing data products will be leveraged to produce quantitative SOC regional maps for CONUS using similar combinations of climatic, ecological, environmental, geochemical, and microbial variables. The algorithms developed with NEON data will be validated with other point-scale data like SoDaH (SOils DAta Harmonization database) and ISNC (International Soil Carbon Network). The spatial mismatch of derived representativeness-based SOC regional maps for CONUS will be evaluated with existing gridded databases: SoilGrids, Harmonized World Soil Database (HWSD), Northern Circumpolar Soil Carbon Database (NCSCD), and gridded U.S. Soil Survey Geographic Database (gSSURGO).EON-based SOC regional maps for CONUS will also be integrated with downscaled historical SOC predictions from participating models of the Coupled Model Intercomparison Project Phase 6 (CMIP6). The robust (and scalable) estimate of SOC for CONUS will enable the diagnosis of terrestrial C cycle processes using historical CMIP6 model runs. Broader impacts will involve training opportunities at the undergraduate and graduate levels, and workhops and training courses to teach data analysis workflow methods.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.