CIF21 DIBBs: PD: Cyberinfrastructure Tools for Precision Agriculture in the 21st Century

这个跨学科的项目将计算机科学方法和计算资源应用于大型多维环境数据集，并将这些信息合成成更精细的分辨率、空间上明确的产品，可以用其他变量进行系统分析。主要的重点是生态信息学，这是信息学的一个分支，分析生态和环境科学变量，如关于景观、土壤、气候、生物和生态系统的信息。该项目侧重于生产高分辨率土壤水分数据集的合成/计算方法，试点应用是精准农业。这项工作结合了分析地理空间方法、机器学习方法和高性能计算(HPC)技术，以建立能够改变生态信息学数据分析方式的网络基础设施工具。研究人员基于公开可用的数据集(土壤水分数据集、土壤属性数据集和地形数据集)来开发：(1)基于机器学习技术的工具，用于将粗粒度数据缩减为土壤水分信息的细粒度数据集；(2)基于HPC技术的工具，用于估计与土壤水分基础变化、趋势和模式发生的时间间隔相关的置信度和概率；(3)数据和用户界面，集成了数据预处理以处理数据的异质性和不准确性，集装箱化的环境以确保可移植性，以及建模技术以表示土壤水分动态的时间和空间模式。这些工具将通过为毗邻的美国生成和使用独特的土壤水分信息来为精确农业提供信息。通过轻量级虚拟化、移动设备和Web应用程序，使田间从业人员能够访问(例如，当地土壤湿度信息)。这项由高级网络基础设施办公室颁发的奖项由美国国家科学基金会地球科学局地球科学部联合支持。

This interdisciplinary project applies computer science approaches and computational resources to large multidimensional environmental datasets, and synthesizes this information into finer resolution, spatially explicit products that can be systematically analyzed with other variables. The main emphasis is ecoinformatics, a branch of informatics that analyzes ecological and environmental science variables such as information on landscapes, soils, climate, organisms, and ecosystems. The project focuses on synthesis/computational approaches for producing high-resolution soil moisture datasets, and the pilot application is precision agriculture. The effort combines analytical geospatial approaches, machine learning methods, and high performance computing (HPC) techniques to build cyberinfrastructure tools that can transform how ecoinformatics data is analyzed.The investigators build upon publicly available data collections (soil moisture datasets, soil properties datasets, and topography datasets) to develop: (1) tools based on machine-learning techniques to downscale coarse-grained data to fine-grained datasets of soil moisture information; (2) tools based on HPC techniques to estimate the degree of confidence and the probabilities associated with the temporal intervals within which soil-moisture-base changes, trends, and patterns occur; and (3) data- and user- interfaces integrating data preprocessing to deal with data heterogeneity and inaccuracy, containerized environments to assure portability, and modeling techniques to represent temporal and spatial patterns of soil moisture dynamics. The tools will inform precision agriculture through the generation and use of unique information on soil moisture for the coterminous United States. Accessibility for field practitioners (e.g., local soil moisture information) is made possible through lightweight virtualization, mobile devices, and web applications. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Earth Sciences within the NSF Directorate for Geosciences.