Perez-Pedini C, Limbrunner JF, Vogel RM (2005) “Optimal location of infiltration-based best management practices for storm water management,” JOURNAL OF WATER RESOURCES PLANNING AND MANAGEMENT, 131(6) pp. 441-448
Summary
This week's article had a very similar theme to our previous reading assignment. As indicated by the title, this study focuses on infiltration-based best management practices (BMPs), as opposed to storage-based BMPs, which we explored in last week's discussion. By infiltration-based BMPs, the authors refer to techniques "ranging from infiltration basins, rain gardens, and pervious pavements, to the use of swales and curb cuts to direct runoff from impervious surfaces to nearby pervious surfaces and depressions." The study ultimately aims to determine the optimal number and location of infiltration-based BMPs which would most effectively reduce peak runoff at a watershed outlet.
The authors selected the highly-urbanized Aberjona River watershed (near Boston) as the basis of this study. The watershed was modeled as a network of 4,533 hydrologic response units (HRUs). Each HRU was modeled as a 120m x 120m square cell and considered as a potential candidate for containing an infiltration-based BMP. Hydrologic fluxes of precipitation, infiltration, groundwater flow, and runoff were considered in each HRU. To model runoff, the NRCS curve number method was employed. Once all was said and done, the authors' model was validated and calibrated using known rainfall data from storm events in 2002 and 2003. A genetic algorithm was then used to determine the most effective locations of infiltration-based BMPs.
Because of the sheer size of the problem, the authors had to make further simplifications during the modeling process. Infiltration-based BMPs were treated as a binary decision. If this decision was a "1", the curve number on that particular HRU would be reduced by five, regardless of whether this reduction may have been less or more. Furthermore, the authors had to reduce the decision space for the genetic algorithm by narrowing down candidate HRUs. With these simplifications, the results indicated that a 20% reduction in peak flow could be achieved by implementing less than 200 BMPs. Furthermore, the authors examined the relationship between peak flow reduction and number of BMPs. Once beyond 600 BMPs, this relationship exhibited minimal reductions in peak flow.
Discussion
This article aided in my understanding of a situation which would require a genetic algorithm. Prior to reading this article, genetic algorithms seemed to be an alternative method, but not a necessary one. In this particular situation, however, I can't imagine a different technique which would yield a solution to such a highly complex problem. Even with simplifying assumptions, I personally can't comprehend a decision space of 7.0 E 168 possibilities.
I found certain aspects of this study more valuable than others. In terms of BMP distribution, I thought the authors' findings were an extension of common sense; it is obvious that industrial and commercial areas are going to be highly impermeable, so it is no surprise that the model targeted these areas as critical locations. What I did find useful, however, was the authors' examination of diminishing returns. Once 600 BMPs are in place in this particular watershed, the installation of any more is almost useless. Without this study, it would seem reasonable to assume that the installation of BMPs would have a continually beneficial effect, but this is clearly not so. This finding could save municipalities huge sums of money in the future as they move toward infiltration-based BMPs.
Overall, I feel that the authors' method could be improved by being made more general. It appears that a lot of work went into defining the Aberjona watershed with an array of specific parameters. If this process could be made more general and efficient, I believe the method would gain more recognition.
Summary
This week's article had a very similar theme to our previous reading assignment. As indicated by the title, this study focuses on infiltration-based best management practices (BMPs), as opposed to storage-based BMPs, which we explored in last week's discussion. By infiltration-based BMPs, the authors refer to techniques "ranging from infiltration basins, rain gardens, and pervious pavements, to the use of swales and curb cuts to direct runoff from impervious surfaces to nearby pervious surfaces and depressions." The study ultimately aims to determine the optimal number and location of infiltration-based BMPs which would most effectively reduce peak runoff at a watershed outlet.
The authors selected the highly-urbanized Aberjona River watershed (near Boston) as the basis of this study. The watershed was modeled as a network of 4,533 hydrologic response units (HRUs). Each HRU was modeled as a 120m x 120m square cell and considered as a potential candidate for containing an infiltration-based BMP. Hydrologic fluxes of precipitation, infiltration, groundwater flow, and runoff were considered in each HRU. To model runoff, the NRCS curve number method was employed. Once all was said and done, the authors' model was validated and calibrated using known rainfall data from storm events in 2002 and 2003. A genetic algorithm was then used to determine the most effective locations of infiltration-based BMPs.
Because of the sheer size of the problem, the authors had to make further simplifications during the modeling process. Infiltration-based BMPs were treated as a binary decision. If this decision was a "1", the curve number on that particular HRU would be reduced by five, regardless of whether this reduction may have been less or more. Furthermore, the authors had to reduce the decision space for the genetic algorithm by narrowing down candidate HRUs. With these simplifications, the results indicated that a 20% reduction in peak flow could be achieved by implementing less than 200 BMPs. Furthermore, the authors examined the relationship between peak flow reduction and number of BMPs. Once beyond 600 BMPs, this relationship exhibited minimal reductions in peak flow.
Discussion
This article aided in my understanding of a situation which would require a genetic algorithm. Prior to reading this article, genetic algorithms seemed to be an alternative method, but not a necessary one. In this particular situation, however, I can't imagine a different technique which would yield a solution to such a highly complex problem. Even with simplifying assumptions, I personally can't comprehend a decision space of 7.0 E 168 possibilities.
I found certain aspects of this study more valuable than others. In terms of BMP distribution, I thought the authors' findings were an extension of common sense; it is obvious that industrial and commercial areas are going to be highly impermeable, so it is no surprise that the model targeted these areas as critical locations. What I did find useful, however, was the authors' examination of diminishing returns. Once 600 BMPs are in place in this particular watershed, the installation of any more is almost useless. Without this study, it would seem reasonable to assume that the installation of BMPs would have a continually beneficial effect, but this is clearly not so. This finding could save municipalities huge sums of money in the future as they move toward infiltration-based BMPs.
Overall, I feel that the authors' method could be improved by being made more general. It appears that a lot of work went into defining the Aberjona watershed with an array of specific parameters. If this process could be made more general and efficient, I believe the method would gain more recognition.
1 comment:
Once again, another great post. You never cease to amaze me.
Post a Comment