Summary
The Rocky Mountain Arsenal was a United States military facility that operated between 1942 and 1992. Its main purpose was to manufacture chemical weapons in support of the United States' involvement in World War II. Due to improper disposals, however, the site has been plagued by environmental issues (specifically groundwater contamination) ever since its inception. Cleanup efforts over the decades eventually led to the official establishment of the Rocky Mountain Arsenal National Wildlife Refuge in 2004. Cleanup is scheduled to be complete in 2011.
Atwood and Gorelick's article was written in the mid 1980s. By this time, the environmental impact of the Rocky Mountain Arsenal had been fully realized, and scientists were attempting to determine the most effective and efficient method of groundwater contaminant cleanup.
In their article, the authors document a hypothetical test of groundwater remediation at the Rocky Mountain Arsenal site. The method is broken into two distinct stages. In Stage I, the "[groundwater] contaminant transport is simulated through time using an assumed velocity field." This simulation yields an approximate contaminant boundary over time. In Stage II, the results of Stage I are integrated with linear programming to "select the wells and determine the pumping/recharge schedules that most effectively stop the migration of the plume by controlling the hydraulic gradient during cleanup."
Without first breaking the problem into two parts, this problem could not be analyzed in a linear fashion. Stage I yields an "expected geometry" of the contaminant boundary - a simplification which Stage II would be unable to function without. Given the expected geometry, the authors' linear program is capable of finding the optimal operation of wells around the site within a 16-year "management period". In the case of this particular problem, optimal operation is defined by the minimal amount of pumping/recharge required of wells which would sufficiently maintain the hydraulic gradient requirements outlined by the groundwater flow model from Stage I. In addition, the authors ran two iterations of the linear program: one "global", which accounted for the entire planning horizon, and another "sequential", in which the most optimal course of action was selected based on the current status of the system.
Discussion
I found this article to be both interesting and significant. Because the applicability of theory was so clear in this particular study, I found myself interested throughout the entire article. The authors did an excellent job of not delving too much into the background behind their linear program, but rather focusing on the approach to the problem and analysis of results. I found the authors' work significant simply by the nature of the issue they chose to tackle - any work involving the improvement of public health is inherently significant. In addition, shaving years of time off of the remediation effort is a major contribution which would have not happened without the authors' linear programming contribution.
Future work in this area will likely focus on combining the authors' two-stage process into one. Because Stage I of this process yields an approximate plume boundary location over time, the results of Stage II will only be as certain as the approximation of Stage I, regardless of how well the linear program is constructed. If these two stages could somehow be combined, a more accurate result would likely be obtained, and the process of modeling would likely be expedited. Nevertheless, I give the authors a lot of credit for the creativity of their approach, and I do not feel that the errors of these approximations take away from the significance of their findings.
The Rocky Mountain Arsenal was a United States military facility that operated between 1942 and 1992. Its main purpose was to manufacture chemical weapons in support of the United States' involvement in World War II. Due to improper disposals, however, the site has been plagued by environmental issues (specifically groundwater contamination) ever since its inception. Cleanup efforts over the decades eventually led to the official establishment of the Rocky Mountain Arsenal National Wildlife Refuge in 2004. Cleanup is scheduled to be complete in 2011.
Atwood and Gorelick's article was written in the mid 1980s. By this time, the environmental impact of the Rocky Mountain Arsenal had been fully realized, and scientists were attempting to determine the most effective and efficient method of groundwater contaminant cleanup.
In their article, the authors document a hypothetical test of groundwater remediation at the Rocky Mountain Arsenal site. The method is broken into two distinct stages. In Stage I, the "[groundwater] contaminant transport is simulated through time using an assumed velocity field." This simulation yields an approximate contaminant boundary over time. In Stage II, the results of Stage I are integrated with linear programming to "select the wells and determine the pumping/recharge schedules that most effectively stop the migration of the plume by controlling the hydraulic gradient during cleanup."
Without first breaking the problem into two parts, this problem could not be analyzed in a linear fashion. Stage I yields an "expected geometry" of the contaminant boundary - a simplification which Stage II would be unable to function without. Given the expected geometry, the authors' linear program is capable of finding the optimal operation of wells around the site within a 16-year "management period". In the case of this particular problem, optimal operation is defined by the minimal amount of pumping/recharge required of wells which would sufficiently maintain the hydraulic gradient requirements outlined by the groundwater flow model from Stage I. In addition, the authors ran two iterations of the linear program: one "global", which accounted for the entire planning horizon, and another "sequential", in which the most optimal course of action was selected based on the current status of the system.
Discussion
I found this article to be both interesting and significant. Because the applicability of theory was so clear in this particular study, I found myself interested throughout the entire article. The authors did an excellent job of not delving too much into the background behind their linear program, but rather focusing on the approach to the problem and analysis of results. I found the authors' work significant simply by the nature of the issue they chose to tackle - any work involving the improvement of public health is inherently significant. In addition, shaving years of time off of the remediation effort is a major contribution which would have not happened without the authors' linear programming contribution.
Future work in this area will likely focus on combining the authors' two-stage process into one. Because Stage I of this process yields an approximate plume boundary location over time, the results of Stage II will only be as certain as the approximation of Stage I, regardless of how well the linear program is constructed. If these two stages could somehow be combined, a more accurate result would likely be obtained, and the process of modeling would likely be expedited. Nevertheless, I give the authors a lot of credit for the creativity of their approach, and I do not feel that the errors of these approximations take away from the significance of their findings.
NOTE
Additional information for this post was obtained from the following source:
http://www.rma.army.mil/site/sitefrm.html
1 comment:
It is interesting that you comment on combining the two stages. My undergrad research had a problem like this, where we were optimizing the hydroelectric power generated, but generation was dependent on flow*head in the reservoir, and since head is dependent on the flow (decision variable) the program was nonlinear. We had to assume an initial head value, run the program, find the resulting head, then plug that head back in and run the program again. We thought that running the program multiple times would help us converge on the actual average head, but it turns out it would jump between two different head values each time we re-ran the program. I still haven't wrapped my mind around why that happened. Anyways that is one method of making a nonlinear program into a linear program.
Post a Comment