Grantee Research Project Results
1998 Progress Report: Advancement of Environmental Decision Support Systems Through High Performance Computing Communication
EPA Grant Number: R825196Title: Advancement of Environmental Decision Support Systems Through High Performance Computing Communication
Investigators: Brill, E. Downey , Ranjithan, S. Ranji , Fine, Steven S. , Baugh, John W. , Loughlin, Daniel
Institution: North Carolina State University
EPA Project Officer: Aja, Hayley
Project Period: October 1, 1996 through September 30, 1999 (Extended to September 30, 2000)
Project Period Covered by this Report: October 1, 1997 through September 30, 1998
Project Amount: $599,932
RFA: High Performance Computing (1996) RFA Text | Recipients Lists
Research Category: Human Health , Aquatic Ecosystems , Environmental Statistics
Objective:
We envision an environmental decision support system (DSS) as facilitating an iterative decision-making process in which a decision-maker/analyst incrementally learns more about a problem during the design process. As knowledge is gained, previous assumptions may be challenged. Through a number of iterations, one or more designs can be incrementally developed and refined.In addition to providing a flexible design framework, a DSS simplifies the design process by insulating the user from tedious tasks, such as accessing data, setting up modeling studies, quality assurance checks, etc. This allows the user to concentrate on developing, testing, and comparing alternative control strategies.
A potentially important component of such a system is optimization. Optimization allows the DSS to take a more proactive role in the design process by generating benchmark low cost alternatives that meet user-defined goals and objectives. These alternatives may be good starting places for analysis and may provide the user with valuable insights about how to efficiently solve the problem.
While this vision of an environmental management DSS is readily articulated, implementations that fit this description are rare. One reason is the complexity of many problems faced in environmental management. For example, developing air quality management strategies for tropospheric ozone includes the following difficulties: large numbers of sources and potential control options; uncertainties in emissions, control, and meteorological data; complex and potentially counter-intuitive pollutant behavior; and time-consuming air quality simulation models.
DSS features can be designed to facilitate consideration of these issues; however, the long runtime of simulation models makes identifying feasible control strategies difficult. This may make the ultimate goal of generating and comparing management alternatives impractical. Further, the complex behavior of many environmental problems cannot be easily incorporated into traditional optimization approaches. Nontraditional approaches such as genetic algorithms (GAs) and simulated annealing (SA) can be used instead, but the computational intensity of these approaches may limit their practicality.
The goal of our research has been to explore how HPCC tools and capabilities can be applied to make the use of DSSs more practical for complex environmental problems. Our primary focus has been in the areas of: 1) solving complex environmental problems using optimization in a distributed computing environment, 2) improving optimization algorithms for multiobjective analysis, chance-constrained optimization, and generation of very different alternative strategies, and 3) identifying features and prototyping a potential architecture for environmental decision support.
The progress summary that follows is divided into three sections that outline our work in each of these areas.
Progress Summary:
1. Solving Complex Environmental Problems Using Optimization in a
Distributed Computing Environment
An advantage of GAs and simulated
annealing SA is their ability to accommodate complex environmental process
models into the optimization search process. While this is an important
characteristic of GA and SA search, the process model may be executed from
roughly 5000 to 20000 times during either procedure. If the process model
requires only about a second to run, the optimization run would require several
hours. However, if each run of the process model requires 30 minutes, the
optimization runtime would be from 100 to 400 days. This duration is not
practical in most realistic decision-making scenarios.
In this project, we are exploring how distributed computing can make optimization of this type of problem practical. Approaches we are taking include:
- Using parallel versions of GAs and SA
- Conjunctively using simple and complex process models in the evaluation function.
- Conjunctively using global and local optimization search techniques
In the previous two years of the project we have explored the first two approaches listed above. During the third project year we will implement the third and test the combination of all three on a case study problem. The case study involves the development of tropospheric ozone control strategies. We plan to use a parallel GA with the Urban Airshed Model as the complex environmental process model. The GA will be implemented on a network of 7 to 30 computers. We anticipate that these approaches may reduce clock time by as much as a factor of 200.
2. Improving Optimization Algorithms for Multiobjective Analysis,
Chance-Constrained Optimization, and Generation of Very Different Alternative
Strategies
In this part of the project, we have explored modifications to
GAs to allow them to perform multiobjective optimization, chance-constrained
optimization, and generation of very different alternative solutions.
In multiobjective optimization, the GA population is directed to converge on a set of noninferior (or Pareto optimal) solutions corresponding to the tradeoffs among competing objectives. A new approach, called the Neighborhood Constraint Method (NCM), was developed for this purpose. Results of a comparison with other GA multiobjective approaches suggest that NCM performs better than those approaches for the type of environmental management problems that we are solving.
We also investigated how GAs could be used for chance-constrained optimization. The approach we implemented uses Monte Carlo simulations in the GA evaluation function. Results of trials provided an interesting conclusion: if a sampling technique like Latin Hypercube Sampling (LHS) is used to generate a new set of realizations each generation, and if all individuals in a generation are evaluated with the same set of realizations, then reasonable results can be obtained using sample sizes as low as 20 to 50.
Using approaches called niching techniques, GAs can be made to converge on a number of different solutions simultaneously. We developed a new niching technique that encourages these solutions to be very different from each other, yet all good within a small incremental fitness from that of the best solution. Preliminary results suggest this approach performs better than sharing inat achieving this goal.
3. Identifying features and Prototyping a Potential Architecture for
Environmental Decision Support
We performed two major tasks in this area:
1) continuing development of a prototype DSS for air quality management, and 2)
designing and beginning development of a new prototype. We began developing the
initial prototype, called the Strategy Development Tool (SDT), under a previous
project. It includes features for developing tropospheric ozone strategies
through trial and error or by using GA-based optimization. Other features allow
the outcomes of emissions charges and emissions trading programs to be
predicted.
Over the past several years we have demonstrated the SDT for several groups of potential users, including scientists from MCNC, the State of North Carolina, and the USEPA. Their inputs provided valuable insights and ideas for many new features. We used their suggestions, along with our insights from developing the SDT, to generate the architecture for a new prototype DSS. Architectural decision were made to facilitate capabilities such as optimization of pollutant strategies considering cross-media impacts and the use of distributed optimization. Development of the latest prototype will continue through the last year of the project.
Journal Articles:
No journal articles submitted with this report: View all 21 publications for this projectSupplemental Keywords:
Air quality management, ozone management, joint-cognitive system., RFA, Scientific Discipline, Ecosystem Protection/Environmental Exposure & Risk, computing technology, Ecology and Ecosystems, ecosystem modeling, decision support systems, environmental decision making, HPCC, computer science, data management, ecosystem simulation, data analysis, information technology, parallel computingProgress 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.