MSB-FRA: Testing abiotic drivers of activity, abundance, and diversity of ground-dwelling arthropod communities at a continental scale

昆虫是生物圈中数量最多、生态最重要的动物之一。它们包括严重的农作物害虫和入侵物种，造成数百万人的损失。这个项目的一个关键目标是了解昆虫的数量和活动如何以及何时随着一个人在美国各地的移动而变化，以及为什么这些数字每年都在波动。这样的理解可以帮助预测虫害的爆发、入侵物种的传播，以及生态系统提供食物和纤维以及保存土壤养分能力的变化。该项目将使用美国国家生态观测站网络（NEON）在美国47个地点收集的昆虫样本该项目将使用环境条形码（从样品的防腐剂中分析昆虫DNA）和图像分析技术来训练计算机算法，以计数和识别保存在样品中的昆虫。总之，这些技术将自动化和简化NEON的监测网络，提供第一个关于美国丰富度、活动和多样性的全国性数据集土壤昆虫。在此过程中，他们将服务于各种利益相关者：生态学家测试和完善预测未来昆虫群落的模型；想知道害虫爆发可能性的土地管理者；自然保护生物学家和城市规划者希望预测入侵蚂蚁和甲虫的传播。该项目的一个关键目标是量化和预测地球上巨大的非生物驱动因素——温度、降水和生物地球化学——如何控制土壤昆虫的生态群落在不同地方的变化。大陆范围内的群落数据远远不能代表陆生节肢动物，部分原因是需要付出巨大的努力来计数、大小和识别从螨虫到蚂蚁、甲虫到蜘蛛的分类群。然而，现有的为数不多的节肢动物数据表明，当一个人从沙漠到雨林旅行时，陆生节肢动物群落在丰度（个体数量）、大小（个体质量）、活动（个体对系统的作用速率）和多样性（物种/形式的数量）方面会发生数量级的变化。结合起来，这四个变量有助于预测节肢动物如何调节生态系统过程，如分解、食草和种子传播。这一知识缺口将由NEON陷阱网络（埋藏在土壤中的陷阱阵列，每两周捕获并储存乙醇中的节肢动物样本）的样本分析来填补。它将为此制定两种互补的方法。环境条形码样品和识别陷阱分类从乙醇提取物。图像分析使用机器学习对样本中的节肢动物进行计数、大小和分类。包含地球节肢动物关键目的陷阱样本将从NEON的47个地点进行分析，环境条形码和图像分析管道将从这些样本中进行计数，大小和识别分类群的开发，测试，磨砺和验证，然后用于分析两年的NEON样本。

Insects are among the most abundant and ecologically important animals in the biosphere. They include serious crop pests and invasive species that cause millions in damage. A key goal of this project is to understand how and when the number, and activity of insects change as one moves from place to place across the U.S., and why those numbers fluctuate from year to year. Such an understanding can help predict insect pest outbreaks, the spread of invasive species, and changes in an ecosystems ability to provide food and fiber and conserve soil nutrients. This project will use insect samples collected by the National Ecological Observatory Network (NEON) at 47 locations spanning the U.S.'s major ecosystems to determine how abundance, activity, and diversity of soil insects vary across the U.S. The project will use environmental barcoding (analyzing insect DNA from the samples' preservative) and image analysis techniques to train computer algorithms to count and identify insects preserved in the samples. Together, these technologies will automate and streamline NEON's monitoring network, providing the first such nationwide dataset on abundance, activity, and diversity of the U.S.'s soil insects. In doing so they will serve a variety of stakeholders: ecologists testing and refining models that predict future insect communities; land managers who want to know the likelihood of a pest eruption; conservation biologists and urban planners hoping to anticipate spread of invasive ants and beetles. A key aim of this project is to quantify and predict how Earth's great abiotic drivers--temperature, precipitation, and biogeochemistry--govern how ecological communities of soil insects vary from place to place. Community data at continental extents vastly underrepresent the terrestrial arthropods in part due to the immense effort required to count, size, and identify taxa ranging from mites to ants to beetles to spiders. Yet the few existing arthropod datasets suggest that as one travels from deserts to rainforests, terrestrial arthropod communities vary by orders of magnitudes in abundance (the number of individuals), size (mass per individual), activity (the rate at which individuals do work on the system), and diversity (the number of species/forms). Combined, these four variables help predict how arthropods regulate ecosystem processes like decomposition, herbivory, and seed dispersal. This knowledge gap will be filled by the analysis of samples from the NEON pitfall network (arrays of traps, sunk in the soil, that capture and store biweekly samples of arthropods in ethanol). It will develop two complementary methods to do so. Environmental Bar Coding samples and identifies pitfall taxa from extracts of ethanol. Image Analysis uses machine learning to count, size, and classify arthropods in a sample. Pitfall samples containing key orders of Earth's arthropods will be analyzed from NEON's 47 sites, and Environmental Bar Coding and Image Analysis pipelines that count, size, and identify taxa from these samples will be developed, tested, honed, and validated, and then used to analyze two years of NEON samples.