ITR: Parallel and Grid Computing for Ecological Multimodeling

跨越区域地理范围的环境问题需要多种方法来模拟不同空间、时间和生物尺度上的自然系统反应。分析自然系统对不同人类活动的反应需要几个物理和营养尺度的数学和计算机模型。计算机建模工作需要大量的时空数据集，这些数据集通常来自遥感，并与降雨和水文等非生物因素的复杂模型相关联。由于与这些模型相关的大空间数据集和繁重的计算需求，工作站平台上的串行计算方法几乎不适合分析。研究人员将开发算法和软件，用于连接生态和物理模型，适用于多处理器计算机和跨计算网格的机器集群。包括在各种平台上对生态多模型并行化的替代方法和算法的性能进行比较。该研究对自然系统管理者有效利用目前收集和存档的大量遥感数据、将这些数据与自然系统的现实生态和物理模型联系起来以及提供管理方案有效性评估的能力具有潜在的广泛影响。考虑到替代采收、水资源调度和土地利用设计的成本和潜在的长期影响，优化决策和监测所采取行动的有效性非常重要。然而，实现有效利用还需要培养一批能够利用新技术的管理骨干。调查人员将着手编写教育材料，环境科学家可以通过这些材料了解如何使用多模型来解决区域环境问题。目标是制定一个明确的教育计划，使计算网格方法被接受为用于自然系统管理的区域规划的工具包的一部分。

Environmental problems that span regional geographic extent require multiple approaches to the modelling of natural system responses at various spatial, temporal and organismal scales. Analyzing responses of natural systems to alternative human actions requires mathematical and computer models for several physical and trophic scales. Computer modeling efforts require large spatio-temporal data sets, often derived from remote sensing, linked to complex models for abiotic factors such as rainfall and hydrology. Due to the large spatial data sets and heavy computational demands associated with such models, serial computing methods on workstation platforms are barely appropriate for analysis. The investigators will develop algorithms and software for linked ecological and physical models appropriate for both multiple-processor computers and clusters of machines across a computational grid. Included will be a comparison of the performance of alternative methods and algorithms for parallelization of ecological multimodels on a variety of platforms.The proposed research has potentially very broad impacts on the ability of managers of natural systems to effectively utilize the extensive remote sensing data currently being collected and archived, link these data with realistic ecological and physical models of natural systems, and provide assessments of the effectiveness of management scenarios. Given the costs and potential long-term impacts of alternative harvesting, water scheduling, and land-use designs, it is important to optimize decision making and monitor the effectiveness of the actions taken. Achieving effective use, however, also requires the development of a cadre of managers able to make use of new technologies. The investigators will initiate the development of educational materials through which environmental scientists can learn about the use of multimodels to address regional environmental problems. An objective is to formulate an explicit educational program that will lead to computational grid methods being accepted as part of the toolkit applied in regional planning for management of natural systems.