Grantee Research Project Results
2015 Progress Report: Performance and Effectiveness of Urban Green Infrastructure: Maximizing Benefits at the Subwatershed Scale through Measurement, Modeling, and Community-Based Implementation
EPA Grant Number: R835555Title: Performance and Effectiveness of Urban Green Infrastructure: Maximizing Benefits at the Subwatershed Scale through Measurement, Modeling, and Community-Based Implementation
Investigators: McGarity, Arthur E , Hobbs, Benjamin F. , Rosan, Christina , Welty, Claire , Heckert, Megan , Szalay, Shandor
Current Investigators: McGarity, Arthur E , Hobbs, Benjamin F. , Rosan, Christina , Welty, Claire , Heckert, Megan
Institution: Swarthmore College , AKRF, Inc. , The Johns Hopkins University , Temple University , University of Maryland - Baltimore County
Current Institution: Swarthmore College , Temple University , The Johns Hopkins University , University of Maryland - Baltimore County
EPA Project Officer: Packard, Benjamin H
Project Period: October 1, 2013 through September 30, 2017 (Extended to September 30, 2018)
Project Period Covered by this Report: October 1, 2014 through September 30,2015
Project Amount: $1,000,000
RFA: Performance and Effectiveness of Green Infrastructure Stormwater Management Approaches in the Urban Context: A Philadelphia Case Study (2012) RFA Text | Recipients Lists
Research Category: Watersheds , Water
Objective:
(1) Evaluate selected Green Infrastructure (GI) demonstration projects in the Philadelphia CSO area; (2) develop methodology for creating zones of green infrastructure (ZGIs); (3) develop quantitative GI benefit-investment functions for each zone; (4) incorporate benefit functions into tools for use by municipal officials and community stakeholders; and (5) incorporate STEM learning at multiple levels.
Progress Summary:
Task 1: Subsurface modeling and monitoring; Co-PI Claire Welty, UMBC.
UMBC expenditures on this grant commenced on 7/1/14. Tasks undertaken in 2014-2015 included groundwater modeling and instrumentation installation. Based on site selection criteria discussed in last year’s annual report, the Wingohocking was chosen as our focus watershed, as well as three sites within this watershed where GI is installed. The Wingohocking lies in the Piedmont Physiographic Province and is about 21 sq km in area. There is no surface expression of drainage. All streams are buried in combined sewers. Final GI sites selected within the Wingohocking include: Wakefield Park (rain garden), Leeds School (tree trenches) and Waterview Recreation Center (permeable pavement).
- Groundwater Modeling. ParFlow.CLM is being used to model the coupled groundwater/surface water/land atmosphere system at watershed and site scales.
- Watershed-scale modeling. A rectangular domain encompassing the Wingohocking watershed was constructed. Final gridding chosen was 40 m x 40 m in the horizontal and 1 m in the vertical owing to the computational demands of the model. The model was populated with physical data including topography (slopes); pipe locations; hydraulic conductivity of soil, saprolite, and fractured bedrock; and land cover (impervious surfaces and vegetation). Precipitation and energy terms for the atmospheric modeling component were obtained from NLDAS database. Leakage from PWD water supply pipes was estimated and incorporated into the model as additional input. Production runs were carried out for calendar years 2004–2014. Output data evaluated were subsurface storage and pressure head across space and time. Figure 1 shows a snapshot in time of the spatial distribution of pressure head of the model surface layer during a dry period. Figure 2 shows integrated subsurface storage across model years. Annual cycles of storage are apparent from this figure. Figure 3 illustrates approximate comparison between depth to water predicted by the model and depth to water at one USGS well in the domain that is similar in depth to the ParFlow domain thickness.
Figure 1. Pressure head distribution at land surface of Wingohocking domain during a dry period.
The buried streams are saturated and the surrounding landscape exhibits spatial variability in negative pressure head.
Figure 2. Simulated subsurface storage in Wingohocking watershed, 2004-2015.
Figure 3. Comparison of simulated and actual depth to water at USGS Well PH 550 in the Wingohocking watershed. - Site-scale modeling. Site-scale modeling is in progress for Wakefield and Waterview. The gridding for both models is 1.5 m in the horizontal with a variable vertical “dz” so as to be able to capture fine-scale near-surface processes. The vertical dz is as small as 10 cm near the land surface. The Wakefield model has been spun up for 5 years followed by simulation of calendar years 2008 and 2009 to date. The model will be run to present time. We recently gridded the Waterview site and now are in the process of assimilating physical data and conducting runoff tests. This will be followed by spin up and production runs for 2008 to present. Once these models are completed, comparisons to field data collected at the sites will be carried out, and links to the watershed-scale model will be made.
- Watershed-scale modeling. A rectangular domain encompassing the Wingohocking watershed was constructed. Final gridding chosen was 40 m x 40 m in the horizontal and 1 m in the vertical owing to the computational demands of the model. The model was populated with physical data including topography (slopes); pipe locations; hydraulic conductivity of soil, saprolite, and fractured bedrock; and land cover (impervious surfaces and vegetation). Precipitation and energy terms for the atmospheric modeling component were obtained from NLDAS database. Leakage from PWD water supply pipes was estimated and incorporated into the model as additional input. Production runs were carried out for calendar years 2004–2014. Output data evaluated were subsurface storage and pressure head across space and time. Figure 1 shows a snapshot in time of the spatial distribution of pressure head of the model surface layer during a dry period. Figure 2 shows integrated subsurface storage across model years. Annual cycles of storage are apparent from this figure. Figure 3 illustrates approximate comparison between depth to water predicted by the model and depth to water at one USGS well in the domain that is similar in depth to the ParFlow domain thickness.
- Tensiometer Deployment. Three TS1 tensiometers purchased in 2014 were lab- and field-tested at UMBC from March 2015-August 2015 before deployment in Philadelphia at the GI sites. Wakefield, Waterview, and Leeds GI sites were outfitted with TS1 tensiometers deployed at multiple depths during summer/fall 2015. Given that the cost of these instruments was higher than originally quoted, the revised deployment strategy was to place one nest in or as close to each GI treatment as possible, with another nest in a nearby control urban soil/grass test patch. We currently are collecting data from all three sites. An example data set from Wakefield is shown in Figure 4. Each site deployment consists of a number of peripherals including data logger, padlocked utility box, solar controller, solar panel, 12V battery, cables, conduit, and grounding rod. CR1000 data loggers are being rented from USGS for the duration of the project. All other peripherals were purchased.
Figure 4. Wakefield tensiometer data, from TS1s installed under the center of the rain garden. - Well drilling and piezometer installation. We have contracted with Eichelbergers, Inc., to drill boreholes and install piezometers at the GI sites. Wells are scheduled to be installed at two sites in November 2015. As soon as access issues are resolved with Leeds, well installation will be carried out at that site.
- STEM Education. UMBC Masters student Elvis Andion-Nolasco is writing his thesis on modeling and monitoring GI in the Philadelphia CSO area. UMBC Ph.D. student Matthew Schley is learning to model green infrastructure with the ParFlow model while applying it to the Waterview Recreation Center site in the Philadelphia CSO area.
Task 2: Philadelphia Community Engagement through Research Partnerships; Co-PI Christina Rosan, Temple University, Co-PI Ben Hobs, Johns Hopkins University, and Co-PI Megan Heckert, Swarthmore College.
- GreenPhilly Advisory Research Board Partners. One of our primary goals is to incorporate “community benefits” into the decision-making about green infrastructure investment. In year 2, Co-PI Christina Rosan further focused our work with the project team’s partners in the GreenPhilly Community Advisory Research Board by hosting several workshops at Temple University. One outgrowth of these workshops is a new partnership with the Village of Arts and Humanities, a Registered Community Organization in North Philadelphia recognized by the City of Philadelphia (see http://www.phila.gov/CityPlanning/projectreviews/Pages/RegisteredCommunityOrganizations.aspx Exit ). Our goal is to work with community experts in refining the StormWISE multiobjective decision-making model to prioritize green infrastructure investments in Philadelphia and to develop tools that are transferable to other cities where CSOs have major impacts on water quality. We are actively developing a similar partnership with the Overbrook Arts and Environmental Education Center in West Philadelphia. These two partnerships, as well as others under active development, will provide our models with neighborhood-scale inputs and will enable further development of our approach to delineating “zones of green infrastructure.”
- JHU Team Participation. Co-PI Ben Hobbs and Johns Hopkins Ph.D. student Fengwei Hung have developed innovative analytical tools for eliciting and quantifying preferences from community stakeholders relating to the benefits of GI. Hobbs and Hung planned and implemented sessions at each of the advisory board workshops that have provided insightful commentary from our community partners. Progress has been made in developing a subjective scale for quantifying GI aesthetic and amenity values using visual tools showing before and after images of different kinds of properties (e.g., vacant lots, rights of way, rooftops, etc.) transformed by installation of various GI practices. These developments are being incorporated into a web-based toolbox that will be used to generate the parameters necessary to incorporate community benefits into the StormWISE model’s multiobjective optimization framework.
- STEM Education
- Undergraduate training. 1 credit for Temple undergraduate summer research course.
- Regular coursework. In Fall 2014, Prof. Rosan taught Temple University’s Urban Environment course (ES/GUS 2051). She designed the class to connect with the Swarthmore EPA STAR grant on Green Infrastructure. She brought in numerous guest speakers on green infrastructure, took students on field trips, and asked them to work on a group policy paper about green infrastructure planning in Philadelphia. Many students said they liked the course because they learned about possible green careers. Partially, as a result of taking this course, one undergraduate became very interested in green infrastructure and was offered a job after he graduated at Avazea, Inc., working on a GIS project for the Philadelphia Water Department. One reason that he was given the job was because he was able to share with them the video he made about green infrastructure for the class https://www.youtube.com/watch?v=kq_pqKCZoHM Exit . Another student who took the course interned over the summer for the U.S. Forest Service’s Philadelphia Field Station working on urban greening. Another student, who was an exchange student from Brazil, Iporã Brito Possantti, returned to Brazil and in October 2015 presented his research at the ISRRU meeting (Revitalizing Urban Rivers Symposium).
Task 3: Develop Benefit Functions for Direct Benefits and Co-Benefits of Green Stormwater Infrastructure; PI Arthur McGarity and Co-PI Megan Heckert, Swarthmore College.
- Simulation with Multiobjective Evolutionary Optimization for Runoff Reduction Benefits. Further progress has been made in driving EPA’s SWMM simulation model with the multiobjective evolutionary optimization engine (MOEA) Borg, under development at Penn State and Cornell Universities. The latest developments in both SWMM and Borg have been incorporated to create an efficient interface between the two programs. The SWMM Engine input file format now directly supports specification of parameters for GI practices such that they can be treated as variables in an optimization driven by Borg. We have adapted Borg’s recently developed Python code wrapper to interface with the Python code we developed to drive the SWMM Engine. Experimentation with a prototype watershed developed from an example in the SWMM Applications Guide reveals that runoff reduction benefits exhibit a saturation behavior as the land area treated by each type of GI practice increases (see McGarity, et al.). Our next step is to create a SWMM model of the Wingohocking sewershed and drive it with Borg to test our methodology in one of the major sewersheds in the Philadelphia CSO area.
- Cost Model Development. Our project consultant, AKRF, Inc., has been tasked with developing a cost model for use in development of GI benefit functions. Progress has been made on a statistical regression model that determines, for various categories of GI practices, which parameters significantly influence costs. Cost data from publicly available sources are being combined with the design and monitoring experience accumulated by AKRF engineers in their work for clients within the Philadelphia CSO area.
- Geospatial Statistics to Estimate Co-Benefits. Megan Heckert continues to make progress on modeling the impact of GI on property values. Because such a large proportion of existing installations was completed in 2013 and 2014, we are seeking additional real estate sales data from 2014 and 2015 to have a larger “after” sample. We are actively developing a need-based index to be used to measure equity as an outcome and to operationalize it in a way that can be built into the StormWISE model. A pilot version is complete and will be presented to the GreenPhilly Advisory Research Board for review, commentary, and revision during Year 3.
- STEM Education. Three Swarthmore College undergraduate engineering students in PI McGarity’s Fall 2014 engineering elective course “Operations Research” chose topics related to Philadelphia stormwater GI for their practicum case study projects. One of these students has chosen a topic related to Tasks 3 and 4 of this project for his senior engineering design project to be conducted during 2015-16. Four Swarthmore College undergraduate engineering students received research stipends for 10 weeks during Summer 2015 to support research supervised by PI McGarity. Three of these students are extending their summer research during the Fall of 2015 by selecting case study projects related to Philadelphia stormwater GI.
Task 4. Develop a Stochastic and Multi-Stage StormWISE Model; Co-PI Benjamin Hobbs, Johns Hopkins University and PI Arthur McGarity, Swarthmore College.
- GI Benefits Quantification and Prioritization. During Year 2, Co-PI Benjamin Hobbs and Johns Hopkins Ph.D. candidate Fengwei Hung focused on development of methodologies for quantifying and prioritizing community benefits associated with stormwater GI. These methods are being designed to generate the parameters necessary for the StormWISE model to generate trade-off analyses for multiple objectives, including maximization of physical water quality benefits in the city’s waterways related to CSO reductions as well as localized community benefits realized in the neighborhoods where the GI practices are deployed. These efforts integrate tightly with Task 2.
- We have identified 23 community benefit categories. Within the scope of this project, it will not be possible to quantify all of these benefits for the entire range of GI practices and also assess their uncertainties. Some of these benefits are subjective, such as aesthetics and amenities, and the links between them and GI practices are not clear. In response to these challenges, we have developed three tools to prioritize benefits and to understand the links between benefits and various GI practices: (1) GI investment exercise, (2) GI investment goal list exercise, and (3) plus and minus exercise. We have tested these tools during advisory board meetings. GI investment and GI goal list exercises show that aesthetics, amenities, green jobs, heat stress reduction, and maintenance costs are the most important factors and the plus and minus exercise helps us understand what benefits each GI practice might or might not provide. We also have tested three weighting methods: (1) direct weighting (DW), (2) simple swing weighting (SSW), (3) and pair-wise swing weighting (PW). Figure 5 shows the average weights calculated from the three methods. The results of conducting these exercises with our advisory board partners suggest that amenities and green jobs should have higher weights and maintenance costs, and stormwater fee savings and green canopy should have lower weights.
Figure 5. Average weightings calculated from the test run in our Oct. 1st Advisory Board meeting - Web Version of StormWISE. PI McGarity and his undergraduate research assistants have continued development of the web version of the StormWISE model. Based on our evaluation of alternative platforms during summer 2015, open source web app (Django) and GIS (GeoDjango) platforms have been selected for creation of a user interface for StormWISE that facilitates exploration of multiobjective GI benefit and decision spaces.
- Stochastic and Multi-stage Extensions of the StormWISE Model. The model theory and development subgroup consisting of investigators Hobbs, Hung, and McGarity continued work on the theoretical foundations of extensions to StormWISE to explicitly account for uncertainties in GSI performance and cost parameters and to incorporate multiple stages enabling modeling of adaptive decision making to reduce risks and improve GI implementation through learning over time. We have developed four models that can be used within an adaptive management framework. The models emphasize the concept of learning, which is realized by updating stochastic coefficients to indicate ways that GI implementation strategies can be improved over time. Figure 6 shows a decision tree that demonstrates the learning mechanism. If learning happens, the uncertainty is reduced, the model incorporates the learning opportunity, and it makes future decisions accordingly.
Figure 6. Stochastic Multi-Stage StormWISE Model Development: a schematic decision tree with learning that reduces uncertaintyAll four models are built as two-stage stochastic programming models but they differ in the mechanism used to model learning. The first model is the most idealized. It learns, during stage 1, perfect information, without effort (extra cost), that is used in stage 2 to make changes in GI implementation strategies so as to improve outcomes. The second model can choose either to learn nothing, without effort, or to learn perfect information with some degree of effort. The third and fourth models can choose to expend some effort to learn, but they can only obtain more realistic partial information, which has some chance of being incorrect. The fourth model additionally assumes that learning will trigger technology (efficiency) improvements, while the other three models assume no technology improvement over time. These models can be used to specify an optimal investment strategy to pursue in the present while accounting for the potential of adaptive management to adjust strategies so as to achieve improved outcomes in the future by incorporating knowledge gained through early deployments of GI.
- STEM Education. Johns Hopkins Ph.D. candidate Fengwei Hung’s dissertation research relates directly to Tasks 2 and 4 of this project. Several groups of graduate students in the Department of Geography and Environmental Engineering at Johns Hopkins have participated in mock exercises to help evaluate tools for quantification of community benefits of GSI.
Future Activities:
- Completion of sensor installation and continuation of data collection from field sites in Philadelphia along with continued hydrologic modeling and comparisons with monitored data and improvements in StormWISE model capabilities to model subwatershed runoff reduction resulting from GI.
- Community engagement activities associated with Task 2 will proceed by creating partnerships in additional Philadelphia neighborhoods and further development of tools to quantify and prioritize community benefits and work on an equity index for urban GI will continue.
- Statistical, simulation, and evolutionary optimization modeling will be combined to create benefit functions incorporating community input and further developments in our equity index, moving from prototype models to models implemented in the Philadelphia CSO area.
- Building on tools developed during Years 1 and 2 for quantifying and prioritizing GI benefits and costs, StormWISE extensions will be implemented in the Philadelphia CSO area including an interactive user interface to improve its ability to inform stakeholders about GI benefits, costs, and necessary tradeoffs, and inclusion of theoretical developments related to the two-stage stochastic model for integrating adaptive management into the modeling framework.
Journal Articles:
No journal articles submitted with this report: View all 31 publications for this projectSupplemental Keywords:
green infrastructure community benefitsRelevant Websites:
GreenPhilly Research Group ExitProgress 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.