International Stormwater BMP Database Enhancements, Updated Findings, and New Additions
The International Stormwater BMP Database (BMPDB) has grown significantly since the first analysis of available data was completed and presented at CASQA in 2005 and then again in 2008. New BMP performance data evaluations were completed in 2012 and the website has been updated and improved in 2013. In addition, in 2014 we will be migrating the National Stormwater Quality Database to the BMPDB website in coordination with its original developer Dr. Robert Pitt emeritus Professor at the University of Alabama. Eventually the databases may be completely merged with the BMP inflow datasets to create a larger database of urban stormwater quality. Finally, we have designed and began populating an Agricultural BMP Database schedule for initial public release in 2015. Urban users of this database will benefit from being able to assess potential trading options for pollutant trading as well as utilize the information in larger watershed planning efforts that encompass both land uses.
The BMPDB now has grown to over 530 BMP studies as compared to about 200 in 2004 and 300 in 2008. We will be entering data from another 40 to 60 studies this year. The project team has again completed performing a re-evaluation of the data contained in the BMPDB to assess the overall performance of BMPs as well as compare BMP design attributes to performance. The evaluations have included the assessment of various BMP types as categorized in the BMPDB with regards to their ability to reduce runoff volumes and pollutant concentrations, as well as the median effluent quality that is achieved. BMPs with wet pools or filtration media tend to achieve lower effluent concentrations than other BMPs such as dry detention basins or vegetated swales for most constituents. However, these latter two BMPs are also showing greater ability to reduce runoff volumes thereby expected to still provide significant load reductions depending on infiltration capacity of underlying soils.
These larger data sets are also now producing tighter confidence intervals about the median predicted effluent concentrations and statistically significant influent/effluent correlations are now being found for some BMPs and constituents. For these analyses we have developed standardized statistical approaches to better characterize the distributions of inflow and outflow water quality. Specifically, the use a robust regression-on-order statistics (ROS) algorithm provides a strong handling of censored (non-detect) results when computed aggregate statistics. Bootstrapping algorithms are then employed to provide refined estimates of the confidence limits around those statistics. Non-parametric hypothesis tests are also regularly employed to compare and correlate influent and effluent data sets.
Leveraging the rapidly growing scientific computing tools built around the Python programming language (http://numfocus.org/projects.html), Geosyntec has developed a set of publically available tools to rapidly select, and apply the ROS and bootstrapping algorithms, conduct hypothesis testing, and visualize data stored in the BMPDB. The source code is available on Github (www.github.com), a popular web platform for sharing and contributing to open source code.
This presentation will demonstrate how the database, website, and analysis tools can be used by a broad range of entities involved in stormwater management, standards development, or research such as public works administrators and engineers, state transportation departments, consulting engineers, regulators, and university researchers. The BMPDB is a scientifically-based resource that offers relevant information and tools to all of these user types. This presentation will provide a demonstration of the reports and interactive resources now available from the www.bmpdatabase.org updated website and identify how each user type can best use the data. A brief overview of the data analysis and visualization methods will also be provided.