Collaborative Research: Harvesting Long-term Survey Data to Develop Zooplankton Distribution Models for the Antarctic Peninsula

该项目由地球科学理事会和高级网络基础设施办公室共同资助，旨在支持地球科学领域的人工智能/机器学习和开放科学活动。该项目将使用机器学习模型，根据海洋学特性预测南极半岛西部（wAP）五种浮游动物的分布。该项目将利用帕尔默长期生态研究（LTER）计划收集的长期系列数据，该计划收集有关物理，化学，浮游植物（或食物），浮游动物和捕食者（海鸟，鲸鱼和海豹）的年度数据。通过分析该数据集并将其与国家和国际计划收集的其他数据相结合，该项目将提供对wAP中浮游动物分布和丰度的理解和预测。机器学习模型将基于从遥感图像中提取的环境特性，从而提供生态系统知识，因为它减少了人类在南极洲的足迹。物种分布和生境之间的关系是区分自然变异与气候对浮游动物及其捕食者影响的关键。这项研究有利于NSF的使命，扩大南极系统，生物区系和过程的基础知识，以及与极地研究中的数据和样本再利用策略保持一致。该项目将通过支持两名女性早期职业科学家、一名博士后研究员和一名研究生来造福社会。极地扫盲将通过与校外活动的现有伙伴关系来促进，这些活动针对科学，技术，工程和数学（STEM）教育，预计每年将有120，000名来自STEM中代表性不足的少数民族的学生。该项目还将有助于评估拟议中的海洋保护区的生态系统，以磷虾渔业为主题。结果将通过一个互动式网络应用程序公布。主要研究人员提出解决三个主要问题：1）地貌特征，冬季预处理和夏季海洋条件可以用来预测浮游动物物种沿着wAP南部夏季分布？2)模拟的浮游动物物种分布沿着wAP的空间和时间模式是什么？浮游动物和捕食者物种的重叠模式是什么？该模型将生成浮游动物分布与环境变量之间的函数关系，并提供沿着南极半岛的浮游动物分布模型（ZDM）。帕尔默LTER数据库将与NOAA AMLR的北方wAP数据和KRILLBASE数据相结合，由英国南极调查局极地数据中心公布。该项目将生成1）研究区域内Palmer LTER分辨率网格上的年度环境空间层，2）来自不同调查的年度特定物种标准化浮游动物净数据，3）标准化网格上的年度特定物种捕食者目击，以及4）生态模型输出。生态模型输出将包括浮游动物物种分布的年度预测，包括5个主要浮游动物群体的三维场（x，y，t），包括南极磷虾，海鞘和翼足类动物。预测结果将通过将现场调查数据与现场或遥感收集的环境数据相结合得出。该奖项反映了NSF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 project is co-funded by a collaboration between the Directorate for Geosciences and Office of Advanced Cyberinfrastructure to support Artificial Intelligence/Machine Learning and open science activities in the geosciences.Machine learning model will be used in this project to predict the distributions of five zooplankton species in the western Antarctic Peninsula (wAP) based on oceanographic properties. The project will take advantage of a long-term series collected by the Palmer Long-Term Ecological Research (LTER) program that collects annual data on physics, chemistry, phytoplankton (or food), zooplankton and predators (seabirds, whales and seals). By analyzing this dataset and combining it with other data collected by national and international programs, this project will provide understanding and prediction of zooplankton distribution and abundance in the wAP. The machine learning models will be based on environmental properties extracted from remote sensing images thus providing ecosystem knowledge as it decreases human footprint in Antarctica. The relationship between species distribution and habitat are key for distinguishing natural variability from climate impacts on zooplankton and their predators. This research benefits NSF mission by expanding fundamental knowledge of Antarctic systems, biota, and processes as well as aligning with data and sample reuse strategies in Polar Research. The project will benefit society by supporting two female early-career scientists, a post-doctoral fellow and a graduate student. Polar literacy will be promoted through an existing partnership with Out Of School activities that target Science, Technology, Engineering and Mathematics (STEM) education, expected to reach 120,000 students from under-represented minorities in STEM annually. The project will also contribute to evaluate the ecosystem in the proposed Marine Protected Area in the wAP, subject to krill fishery. Results will be made available publicly through an interactive web application. The Principal Investigators propose to address three main questions: 1) Can geomorphic features, winter preconditioning and summer ocean conditions be used to predict the austral summer distribution of zooplankton species along the wAP? 2) What are the spatial and temporal patterns in modeled zooplankton species distribution along the wAP? And 3) What are the patterns of overlap in zooplankton and predator species? The model will generate functional relationships between zooplankton distribution and environmental variables and provide Zooplankton Distribution Models (ZDMs) along the Antarctic Peninsula. The Palmer LTER database will be combined with the NOAA AMLR data for the northern wAP, and KRILLBASE, made public by the British Antarctic Survey’s Polar Data Center. This project will generate 1) annual environmental spatial layers on the Palmer LTER resolution grid within the study region, 2) annual species-specific standardized zooplankton net data from different surveys, 3) annual species-specific predator sightings on a standardized grid, and 4) ecological model output. Ecological model output will include annual predictions of zooplankton species distributions, consisting of 3-dimensional fields (x,y,t) for the 5 main zooplankton groups, including Antarctic krill, salps and pteropods. Predictions will be derived from merging in situ survey data with environmental data, collected in situ or by remote sensing.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.