Grantee Research Project Results
2000 Progress Report: A High Performance Analytic Element Model: GIS Interface, Calibration Tools, and Application to the Niagara Falls Region
EPA Grant Number: R827961Title: A High Performance Analytic Element Model: GIS Interface, Calibration Tools, and Application to the Niagara Falls Region
Investigators: Rabideau, Alan J.
Institution: The State University of New York at Buffalo
EPA Project Officer: Packard, Benjamin H
Project Period: March 1, 2000 through February 28, 2003 (Extended to April 1, 2004)
Project Period Covered by this Report: March 1, 2000 through February 28, 2001
Project Amount: $996,545
RFA: Computing Technology for Ecosystem Modeling (1999) RFA Text | Recipients Lists
Research Category: Environmental Statistics
Objective:
As outlined in the original project proposal, the primary objectives of the research project are to:- Develop an expanded parallel version of an existing groundwater flow model
based on the analytic element method (AEM), with an emphasis on message passing
interface (MPI)-based portability and scalable performance for very large-scale
problems using massively parallel computers. MPI was selected because it is
easy to use, widely accepted and supported throughout the scientific community,
available on all major platforms (including both shared and distributed memory
architectures), and functional on a heterogeneous cluster of workstations
or PCs.
- Develop and compare two calibration/optimization tools for use with the
newly expanded AEM software, based on two computational approachesnonlinear
regression and the genetic algorithm (GA), each optimized for parallel processing.
- Develop a GIS-based graphical user interface and database for the AEM model
and calibration tools.
- Implement and benchmark the AEM model and calibration tools on three high-performance
parallel computing platforms: the SGI Origin, the IBM SP, and a heterogeneous
Sun workstation cluster.
- Demonstrate the performance of the modeling system through a case study
of regional groundwater flow in the Niagara Falls area that was previously
analyzed using the finite difference approach (U.S. Geological Survey MODFLOW
Model).
- Examine the efficacy of a hybrid AEM/finite-difference approach by combining
a regional AEM model with local MODFLOW models at one or more hazardous waste
sites in the Niagara region.
- Coordinate the model development activities with concurrent efforts by other researchers and EPA programs, with particular attention to the multimedia integrated modeling system (MIMS).
Secondary objectives include developing and implementing a distance learning course on "Regional Groundwater Flow Modeling by the Analytic Element Method"; and developing a significant AEM initiative as part of a growing interdisciplinary research program in ecosystem modeling at the University at Buffalo (UB).
Progress Summary:
The project is proceeding well, with considerable progress made in the development of the primary project deliverable: a high-performance groundwater flow model based on AEM. Current versions of the model and associated tools are available from the UB Groundwater Research Group Web Site (www.groundwater.buffalo.edu). Because we have been able to recruit high-quality project staff, it is likely that the project tasks will be expanded to include significant new developments in AEM methodology. However, contract-related startup delays led to subsequent delays in project staffing, which will necessitate a no-cost extension to continue project activities into a fourth and (possibly) fifth calendar year without impact to the overall budget.Work Status
General. The project is proceeding well. By the end of Year 1 (February 28, 2001), the project team was assembled, preliminary simulation models were developed, and code development and model testing are well underway. The primary software product consists of a high-performance FORTRAN groundwater flow model (currently named SPLIT), which is distributed from the project Web site. SPLIT has been configured to operate with a GIS-based (ArcView) pre/postprocessor (currently named Arcflow), which also is available from the Web site. Although considerable development remains to meet the project objectives for high-performance computing, the basic building blocks are in place. The result of initial flow model testing are summarized in the SPLIT manual, which is available from the project Web site.
Model Development. The primary product of the project is a high-performance groundwater flow model based on AEM equipped with a GIS-based graphical user interface and optimized calibration tools. The original proposal explicitly stated that new AEM theoretical developments would not be conducted as part of this work. However, several factors have resulted in the need for new model development work:
- Although the STAR solicitation had indicated that approximately $2 M in
additional funding would be allocated to other AEM projects, ours was the
only AEM project funded. At the time of this writing, there is minimal research
funding within the AEM community to stimulate ongoing method development.
- The original proposal identified a program developed through previous EPA
funding (MODAEM) as the most likely vehicle for optimization for high-performance
applications. However, at the time of the proposal submittal, MODAEM had not
yet been officially released, and the source code was not available for review.
Subsequent evaluation has indicated that the algorithms and four features
of MODAEM are unsuitable for the high-performance applications that are the
focus of this work.
- Review of the available AEM groundwater models indicated that only a single modelthe commercial product MLAEMwas available that incorporates multiple aquifer layers. Because it is a proprietary software produce, MLAEM is unsuitable for use in this project.
A straightforward solution was identified for the above problems: the development of a new AEM model based on SPLIT, the FORTRAN program written by Dr. Jankovic. SPLIT is ideal for the project because it was designed using state-of-the-science computational features (iterative solver, high-order elements, superblocks). The single drawback associated with SPLIT is that it currently is configured for a single aquifer layer. Therefore, a significant component of the project will include extension of SPLIT to multiple layers. We currently are evaluating a new algorithm for this purpose, and the original results are promising. Because of the strong additions to the "software side" of the project staff, it is not yet clear whether the need to develop a multilayer model will require diversion of effort from other tasks. Modification of SPLIT for parallel processing is ongoing.
In conjunction with development of the model user interface, we have initiated a dialogue with the Minnesota Department of Health (MDOH), which utilizes the SPLIT program in conjunction with Arcflow, a GIS-based user interface. During the early stages of the project, we will utilize Arcflow as one of several interfaces for SPLIT. Ongoing collaboration with MDOH will lead to the development of an updated version of Arcflow that accommodates both the high-performance aspects of SPLIT and the user needs as identified by MDOH and others. Conversations with ESRI (the developers of ArcInfo/ArcView) indicated that upcoming Version 8 of the GIS software will facilitate the use of modules created using Microsoft's Visual Basic to a much greater degree. Until we develop a better understanding of ESRI future plans, we have expanded our short-term effort to include a second user interface programmed entirely using Visual Basic.
For the calibration tools, we have expanded the project scope to include an interface to the PEST program. Work is ongoing to link SPLIT to the calibration tools identified: the U.S. Geological Survey program UCODE, the commercial model PEST, and a custom module based on the GA. Both of the nonlinear regression models (UCODE and PEST) have been updated recently to include parallel versions, and public domain parallel libraries are available for GA-based programs.
Conversations held at the August 2000 EPA workshop indicated that OO programming is strongly encouraged for multimedia ecosystem modeling. We have therefore initiated development of an OO version of SPLIT programmed using C++. During the next year, we will evaluate the advantages and disadvantages of switching to a completely OO framework.
The current version of SPLIT is available for download from www.groundwater.buffalo.edu. This version includes improvements to the model made as a result of this project, as well as a current version of the Arcflow interface. Over the next 3 months, we expect to release a beta version of the OO (C++) version of SPLIT as well as the Visual Basic user interface.
Collaboration With Other Research Groups. Because the original proposal focused on computational issues and did not address developments in AEM, collaboration with other researchers was identified as critical. The original proposed tasks emphasized a workshop to be held at UB to bring together AEM researchers early in the project. However, in light of activities conducted by other organizations, we have elected to redirect the workshop effort to a series of targeted meetings with various "constituencies" as follows:
- All of the project investigators attended the 2nd International Conference
on Analytical Element Modeling held in Brainerd, MN, in March 2000.
- Project Director Alan Rabideau and Co-PI Igor Jankovic attended the 3-day
EPA-sponsored workshop on multimedia ecosystem modeling held in Research Triangle
Park, NC, August 2000. The primary purpose of the workshop was to coordinate
efforts of STAR grant recipients toward support of EPA's MIMS.
- A 1-day workshop was held at UB in November 2000, in which Dr. Otto Strack
(University of Minnesota) and Dr. W. de Lange (the Netherlands) presented
recent developments on AEM theory and applications. A focus of this workshop
was on methods for extending single-layer models to consider leakage across
multiple aquifer layers.
- Dr. Christ Peters-Lidard (Georgia Tech) visited UB in December 2000 to discuss
methods for coupling AEM groundwater models with models for unsaturated zone
flow and transport. Dr. Peters-Lidard has extensive involvement on developing
MIMS models for other media, and he provided valuable suggestions for future
developments.
- The UB Groundwater Research Group sponsored a 1-day workshop "Groundwater Modeling for Wellhead Protection," which included presentations by representatives from the MDOH on the use of SPLIT and Arcflow. Guests included representatives from New York state regulatory agencies who have expressed interest in using the UB models.
Academic Course. One of the secondary objectives of the proposal was the development of a distance learning graduate course on analytic element modeling. A trial version of this course will be offered to UB graduate students in fall 2001 under the title "Advanced Topics in Groundwater Engineering." This course will serve as a prototype for the distance learning version, which will be offered in fall 2002.
Dissemination of Results. The project has not advanced to the point of developing research journal articles. Current "beta" versions of the project simulation models are available on the UB Groundwater Research Group Web Site. Drs. Rabideau and Jankovic participated in EPA's Cross-Discipline Ecosystem Modeling and Analysis Workshop (August 1517, 2000, Research Triangle Park, NC). As part of this workshop, they presented a talk entitled "A high-performance analytic element model: GIS interface, calibration tools, and application to the Niagara Falls region."
Equipment. Because parallel processing is a central component of the project model development tacks, the first 2 years of the project included targeted equipment expenditures: $42,500 in Year 1 for individual multiprocessor workstations; and $62,400 in Year 2 for a set of project-dedicated processors for the SGI 2000 supercomputer maintained by UB's Center for Computational Research (CCR). After consultation with CCR and other researchers working in this area, we have reconfigured the equipment budget to construct a dedicated 32-processor-distributed memory workstation cluster for use in the project.
Project staff will access this shared resource using low-cost desktop computers, eliminating the need for individual multiprocessor workstations. The cost for the reconfigured system is very close to the original equipment budget (~ $100,000). Several low-end desktop workstations were purchased in Year 1; the 32-processor system will be purchased and assembled during summer 2001. In conjunction with this system, project staff will maintain access to CCR's other massively parallel systems (SGI Origin, IBM SP).
The reconfigured equipment will facilitate the testing of project software using all three of the parallel processing memory management systems (shared, distributed, and distributed/shared). In particular, the new Linux-based workstation cluster will enhance project capability in distributed MPI-based processing, which consistently was identified as a high priority by our research collaborators.
Budget. Due to the delays in project startup and staffing, actual expenditures in Year 1 were considerably less than budgeted. These budget savings will be needed to complete project tasks through a no-cost extension that will be requested to continue staffing for a fourth and (if needed) fifth year.
Publications/Presentations:
Rabideau AJ, Jankovic I. A high-performance analytic element model: GIS interface, calibration tools, and application to the Niagara Falls region. Presented at the EPA Cross-Discipline Ecosystem Modeling and Analysis Workshop, Research Triangle Park, NC, August 15-17, 2000.
Future Activities:
Planned future activities include: (1) sponsorship of a special session in analytic element modeling at the fall 2001 meeting of the American Geophysical Union; and (2) participation by Dr. Rabideau in a July 2001 training session on the use of the PEST software for model calibration. Also, the UB Groundwater Research Group has submitted a research proposal to several other sponsors (NSF, USDA, NASA) to develop "spin-off" projects based on the groundwater models under development in the work. Co-PI Matt Becker recently was awarded a NASA Young Investigator Award to explore the use of remote sensing data in groundwater modeling. In conjunction with this effort, we will configure the user interfaces developed in this work to include options for accessing and utilizing remote sensing information. These efforts will lend visibility to the STAR project and provide opportunities to leverage the EPA investment.Journal Articles on this Report : 3 Displayed | Download in RIS Format
Other project views: | All 46 publications | 15 publications in selected types | All 15 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Becker MW, Georgian T, Ambrose H, Siniscalchi J, Fredrick K. Estimating flow and flux of ground-water discharge using water temperature and velocity. Journal of Hydrology 2004;296(1-4):221-233. |
R827961 (2000) R827961 (2001) R827961 (2002) R827961 (Final) |
Exit Exit |
|
Craig JR, Jankovic, I, Barnes R. The nested superblock approach for regional-scale analytic element models. Groundwater 2006;44(1):76-80. |
R827961 (2000) R827961 (2001) R827961 (2002) R827961 (Final) |
Exit Exit Exit |
|
Fredrick KC, Becker MW, Flewelling DM, Silavisesrith W, Hart ER. Enhancement of aquifer vulnerability indexing using the analytic-element method. Environmental Geology 2004;45(8):1054-1061. |
R827961 (2000) R827961 (2001) R827961 (2002) R827961 (Final) |
Exit Exit |
Supplemental Keywords:
groundwater, ecosystem, analytic element method, AEM, modeling, calibration tools.,Relevant Websites:
http://www.groundwater.buffalo.edu Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.