Brief report on activities underway supported by NSF Award DEB-02-19269 "ITR: Parallel and grid computing for ecological multimodeling" to The University of Tennessee Louis J. Gross Departments of Ecology and Evolutionary Biology and Mathematics Director, The Institute for Environmental Modeling gross@tiem.utk.edu Michael W. Berry Department of Computer Science berry@cs.utk.edu March 12, 2003 This project focuses on the development of new computational methods, particularly using parallel and grid computing, to address environmental problems that require interactions between multiple types of computational models. These models include physical ones, hydrology and weather being examples, and biological ones, which may be for particular species populations or for communities of interacting populations. These models may be operating at different spatial and temporal resolutions, and may utilize quite different mathematical approaches, such as matrix models, differential equation models and individual-based models. An overall objective is to develop new techniques that allow for efficient solution of linked models for which parallel methods are appropriate. An additional objective is to provide a framework for potential use of these parallel methods by resource managers to compare and contrast the effects of alternative management of natural systems, and eventually lead to the capability to carry out optimal dynamic and spatial control of natural systems. As a case study, the project is utilizing the variety of ecological and physical models developed for Everglades restoration by the ATLSS (Across Trophic Level System Simulation) project of the US Geological Survey. See http://atlss.org/ for details on these. These models utilize a variety of different approaches for different groups of species including freshwater fish and wading birds, as well as particular species of interest such as alligators and the Florida panther. The initial phase of this project is focusing on two major models: (i) freshwater fish, using a spatially-gridded, discrete dynamical system approach for size structure in two main groups of fish - planktivores and piscivores; and (ii) wading birds using both a spatially-explicit index model that operates at a coarse, yearly time-step, and an individual-based model that operates within-days and is event-based rather than operating on fixed time-steps. Both models are linked to data provided by an hydrology model that operates over periods of several decades with daily time-steps. This hydrology model modifies the output of the major model used to analyze alternative hydrologic plans for South Florida (developed by the South Florida Water Management District), increasing its spatial resolution to a scale more appropriate for the biotic components of the system. An initial parallelization of the fish model was developed to analyze a key question regarding the impacts of human control systems on the Everglades ecosystem: how much does the compartmentalization of the system into subregions due to the variety of canals and roads which interupt the natural hydrologic flows affect the dynamics of the biota? This first parallel model was done by running simulations for different subregions, unlinked, on different processors. This parallel version is now being extended to incorporate realistic linkages between different subregions. At the same time, we have begun the process of connecting this parallel fish model to a wading bird model that will provide the ability to investigate the effects of flocks of birds in reducing local fish densities in a spatially-explicit manner, as well as compare the impacts of different hydrologic plans on the combined fish - wading bird system. In a step towards the second objective of providing a means for resource managers and interested stakeholders to analyze impacts of different hydrologic plans on biotic components and spatial regions of particular interest to them, we have initiated development of a version of ATLSS that would run across a computational grid. Since few of the resource agencies involved in planning for South Florida (e.g. the National Park Service, the Fish and Wildlife Service) have access to the computational facilities needed to carry out ATLSS modeling, we are utilizing the facilities of the Scalable Intracampus Research Grid (SInRG) Project at the University of Tennessee (supported by the National Science Foundation CISE Research Infrastructure Award EIA-99-72889 - see www.cs.utk.edu/sinrg/) as a test bed. The objective is to provide a means for users in the resource management community to transparently carry out ATLSS modeling without the need to know details such as which model is running on which grid componenet or where the data are stored. The initial user community for this test bed are the USGS managers of the ATLSS project. Financial support for the parallelization efforts described above is from NSF DEB-02-19269, making use of equipment supplied in part by NSF EIA-99-72889 and in part by matching funds from the University of Tennessee. Basic support for the development of the underlying serial ATLSS models comes from the USGS, with this development relying heavily on the expertise of field biologists from numerous agencies (including NPS, USGS, FWS) and universities (including Florida International University and University of Miami). USGS also provides, through funding from the Critical Ecosystems Study Initiative, funds for a variety of field studies which provide key data for both model development and evaluation. ATLSS staff collaborate with the hydrological modeling staff of the SFWMD to utilize and appropriately enhance the SFWMD hydrologic models for use in the biotic modeling.