Green Infrastructure Implementation in Urban Parks for Stormwater Management A Thesis

Green Infrastructure Implementation in Urban Parks for Stormwater Management


Andrew Feldman
A rain garden was constructed in Shoelace Park in Bronx, New York with the goal of capturing stormwater runoff generated on part of 228 th Street. As a pilot study implemented to test the overall strategy of utilizing urban park space to manage street runoff, the rain garden was monitored over a ten month period from October 2014 to July 2015. Field observations and analysis of the stormwater runoff entering the site through an inlet on 228 th Street indicate that, due to the contributions from many adjacent tributary areas, the rain garden receives more water than anticipated. Results of the monitoring campaign show that the stormwater inlet on 228 th Street captures runoff from an effective catchment that is about 1.6 times the designed tributary area. The rain garden retained an average of 77% of all inflows for all storms with water typically leaving the system in storms over 1 cm. It retained 96% and 45% of all inflow for all storms less than 1 cm and greater than 1 cm in total precipitation, respectively. The rain garden managed all off-site runoff from an area of 228 th Street that was about 9 times its size during a 2.5 centimeter storm while receiving additional runoff from other adjacent areas. New York City is 72% impervious and 19.5% parkland. With as little as 5% of all New York City urban park space retrofit with green infrastructure performing similarly to the Shoelace Park rain garden, runoff from 35% of New York's impervious surfaces can be treated. This limited impact to the park ameliorates concerns of infringing on the park's current recreational purpose. Other cities in the United States with high impervious coverage and available park space are urged to consider this stormwater management opportunity.

Introduction and Literature Survey
Precipitation generates large volumes of runoff in urban environments due to the proliferation of impervious cover. Traditional approaches to stormwater management collect, convey, and treat or discharge runoff to surface water bodies (Miskewitz et al. 2013). In those cities that combine their stormwater sewer and sanitary flows in the same pipe, overflows are generated when conveyance capacities are exceeded (Miskewitz et al. 2013). Federal policy requires wastewater utilities to reduce the frequency and volume of combined sewer overflows (CSOs) due to the ecological and public health risks associated with untreated discharges. Green infrastructure (GI) is a distributed approach to reduce the volume of runoff generated within urban watersheds. Through infiltration, evapotranspiration, and detention, the goal of GI is to reduce, or prevent, occurrence of CSOs without having to expand the conveyance capacity of the existing sewer system. Now a billion dollar industry, GI programs in cities such as: New York (NY), Philadelphia (PA), and Portland (OR) are now well underway (Kurtz 2008;McLaughlin et al. 2014;Rajan et al. 2008;Gunther et al. 2010). New York City's plan calls for the capture of the first ~25 mm (e.g. 1 in) of street runoff over 10% of the portion of the city served by combined sewers with GI (NYC DEP 2012). In Philadelphia, the spatial coverage goal is even higher, and is eventually predicted to reduce CSO volumes by 70% in a typical year (Rajan et al. 2008). Due to the presence of underground infrastructure, inadequate soil infiltration capacity, and conflicts with surface features such as driveways, the space available for GI in the public right-of-way is, however, limited (Gunther et al. 2010), and the need to identify GI retrofit opportunities on other portions of the urban landscape is well understood. This paper investigates one of the earliest attempts to infiltrate street runoff in parklands in New York City. In contrast to the surrounding impervious land covers, urban parks are green spaces, often valued as a means of reducing the urban heat island effect (Feyisa et al. 2014;Skoulika et al. 2014). Recently, the potential value of urban green space for stormwater management has begun to receive some attention. There have been a series of recent pilot projects that use GI installed within existing urban parks to manage off-site runoff, including Glencoe Elementary School in Portland, OR, Clark Park in Philadelphia, PA, and Flowers Park in New Rochelle, NY (Perry 2003;PWD 2015;Sorge 2010). Other parks such as Doyle Hollis Park in Emeryville, CA, Gas Tank Park in Queens, NY, and Tanner Springs Park in Portland, OR were constructed specifically to manage stormwater (Youngerman 2009;Compton 2009;City of Portland 2015). Of concern to some park users and officials is the introduction of large bodies of standing water, sometimes perceived as causing flooding and/or encouraging insect growth. To avoid such issues, GI is often designed at a small enough scale so as not to significantly infringe on and/or reduce existing park uses, and can be designed for rapid infiltration and/or with subsurface water storage so as to avoid standing water.
No studies quantifying the amount of stormwater that can be managed in any urban park were found in published literature. The pilot parkland GI project described in this paper was retrofit with an engineered rain garden designed to receive and infiltrate runoff originating in the right-of-way of streets external to the park boundary. The United States Environmental Protection Agency (USEPA) defines rain gardens as "landscaping features adapted to provide on-site treatment of stormwater runoff" (USEPA 2014). Rain gardens are a common form of GI, typically installed in private yards, and can be designed to capture up to 80% of the design storm precipitation volume, in the case of a project built in Portland, OR (Kurtz 2008). However, installation of rain gardens in urban parks is not common.
Coarse spatial considerations suggest that the potential value of parkland rain gardens for reducing urban runoff in New York City appears worthy of focused study. The city's current GI plan calls for the capture of runoff from 10%, or about 57 km 2 , of the impervious surfaces served by combined sewers by 2030. Parklands make up 19.5 percent of the city's total land area, or approximately 154 km 2 . If 20% of each existing park were retrofit to manage runoff originating on double that area, the city could have exceeded its GI policy goals, at least in terms of citywide aerial coverage. As a precursor to such an investigation, the goal of this study was to quantify the ability of one urban park to manage off-site runoff using rain garden GI. Performance indices were developed to assess the efficiency of a new stormwater inlet introduced in the streetscape to intercept street runoff and direct it to a rain garden retrofit into the parkscape.

Methods
In this section, the rain garden, surrounding site, and ancillary hydraulic systems are described in detail. Next, the monitoring equipment and their use in the analysis are discussed.
Finally, the water balance is defined with subsections describing the method to calculate each respective inflow and outflow.

Site Description
In January 2012, an agreement was signed between Drexel University and the National Fish and Wildlife Foundation (NFWF) to retrofit a rain garden into Shoelace Park (Bronx, NY).
The project team lead by the university included an engineering consulting firm, eDesign Dynamics LLC responsible for site assessment and design development, a landscape contractor, Olson's Creative Landscaping responsible for construction, and a local quasi-governmental entity, the Bronx River Alliance to assist in maintenance activities. Drexel was responsible for installation and upkeep of the monitoring equipment for a period lasting from October 2014 to June 2015. The total cost of the GI system including construction, maintenance, and monitoring was $318,371. The construction, maintenance, and monitoring costs were $237,728, $22,000, $58,643, respectively, over the project duration.
The rain garden was constructed inside the park adjacent to the intersection of East 228 th Street and Bronx Boulevard. The site plan is shown in Figure 1. The full set of design drawings are in Appendix A. A narrative description of the main features of the design, as depicted in Figure 1, is presented here. Stormwater originating on 228 th is intercepted by a new stormwater inlet installed in the streetscape just east of a catch basin (Combined Sewer Inlet A as shown in Figure 1) located on the southeast corner of the intersection of 228 th Street with Bronx Boulevard.
Before the project, all runoff from 228 th Street entered Combined Sewer Inlet A and entered the city's combined sewer system. The new stormwater inlet was connected via a 30.5 cm diameter pipe to a new Type A-3 modified shallow manhole, positioned in the sidewalk. A 26.8 m long, 30.5 cm diameter ductile iron inflow pipe directs water from the shallow manhole into a surcharge pit located on the opposite side of Bronx Boulevard in the rain garden. The inflow is distributed across the rain garden surface at a depth of 0.5 m vis-à-vis a 10.2 cm diameter perforated PVC distribution pipe. An overflow riser positioned inside the rain garden conveys water ponded more than 8.9 cm over the surface, which is still below grade, to Combined Sewer Inlet B (as shown in Figure 1), located 3 m to the west, by means of a 30.5 cm diameter PVC pipe with 26 perforations in its cap.
The oblong rain garden is 37 m 2 in area and was constructed by excavating the in-situ soils and replacing them with 45.7 cm of engineered soil on top of 30.5 cm of crushed stone though it was designed to receive runoff from an approximately 600 m 2 tributary area including the streets and sidewalk of the south side of 228 th Street between Bronx Boulevard and Carpenter Avenue (catchment C1 as presented in Table 1 and depicted in Figure 2) as well as a sloped lawn inside the park (catchment C2). Post-construction field observations during a 8.89 cm storm event on April 20 th , 2015 revealed that the actual tributary area of the rain garden is approximately 4,850 m 2 area and includes a larger than anticipated portion of the sloped lawn (catchment C3), Bronx Boulevard extending from 226 th to 229 th Street (catchment C4), as well as a segment of the pedestrian trail inside the park (catchment C5). The aforementioned site visit also revealed several roof drains contributing to the 228 th Street runoff entering the stormwater inlet. However, these tributary estimates were not evaluated as it was not possible to gain roof access and determine the tributary area contributing to the stormwater inlet. These design versus observed tributary areas are presented in Table 1, and shown graphically in Figure 2.

Monitoring Equipment Analysis
The main components of the monitoring system included a Thel-Mar weir, two Global Water pressure transducers, and a rain gauge. The Thel-Mar weir was fitted into the upstream end of the 30.5 cm pipe leading to the rain garden from the shallow manhole. Pressure transducer A (PT-A) was placed within the shallow manhole, and was used to provide continuous measurements at five minute time steps of the water depth upgradient of the weir. Depths Monitoring was performed between October 2014 and June 2015. Due to a prolonged period of sub-zero temperatures, data from PT-A and PT-B was not deemed reliable during February 2015. Also, the onsite rain gauge did not function properly between April 17 th and June 9 th , and storms during this time period were not considered in the analysis.

Pressure Transducer Calibration
In order to validate PT-A and PT-B readings, each sensor was placed in a bucket of water. The bucket was filled with water at 1.5 cm intervals. At each water level, the manually measured water depth was compared to the pressure transducer reading. The calibration data for PT-A and PT-B are shown in Table 2 and in Figure 3. In both cases, the pressure transducer readings correlated to the manual measurements with the coefficient of determination at nearly unity. The regression line equation generated from the calibration exercise was used to convert the raw pressure transducer reading to actual water levels in all further calculations.
To further validate the pressure transducer readings, in-situ manual measurements were made during various site visits throughout the monitoring period (Table 3 and Table 4). The PT-A data behaved as if outflow was occurring 10 cm below its expected outflow invert of 34.14 cm through the Thel-Mar weir during the Fall 2014, Winter 2015, and Summer 2015 months.
However, the manual in-situ measurements refuted this assertion by validating that PT-A readings were correct. This suggests slight water losses from the manhole during periods when the water level was below the invert of the Thel-Mar weir. Because such losses cannot be explained by evaporation, a slight leak perhaps by a crack in the bottom of the manhole is suspected. A crack analysis is discussed in further detail in Appendix C. Overall, the manual measurements revealed a small root mean square error (RMSE) of 1.9 between the calibrated PT-A readings and manhole manual measurements between January 2015 and June 2015. The RSME between the manual and logged water levels at PT-B was 1.2. However, since the in-situ measurements were performed only sporadically, they were used as a validation of the bucket calibration. This is discussed further in Appendix D.

Weir Invert Analysis
Once the water level in the manhole reaches the V-notch of the Thel-Mar weir located in the outlet pipe of the manhole, water begins to flow out of the manhole towards the rain garden.
PT-A readings were used to estimate the water depth inside the manhole. Since the water level at the conclusion of various storms did not align with the 34.14 cm measured invert, the invert height was originally estimated as the calibrated PT-A reading corresponding to the end of flow through the Thel-Mar Weir. However, the 5.14 average percent difference between in-situ measurements and PT-A readings validates the PT-A readings and the 34.14 cm constant invert was used to calculate flow through the Thel-Mar weir. The behavior of the water level in the manhole is discussed in further detail in the discussion.

Storm Discretization
The continuous rainfall record was discretized into individual storms based on the inflow into the rain garden through the Thel-Mar weir. The beginning of each storm was defined as the onset of recorded precipitation, and the end of each storm was defined as the time corresponding to when the post-storm water level in the manhole reached the Thel-Mar weir invert elevation.
The sum of all precipitation that was logged in this period was used to compute the total storm depth. A retrospective analysis of the particular storm discretization methodologies indicated that the same events would have been identified if a four hour interevent dry period was used to separate storms. The same approach was used in the analysis of the PT-B readings since a very short time-lag was observed between the two measurement locations. Rain garden outflow to Combined Sewer Inlet B included all flow through the riser that was recorded between the onset of recorded precipitation and the first time when the ponded water level subsided to the elevation corresponding to the invert of the riser pipe.

Rain Gauge Data Analysis
Since the onsite rain gauge did not function properly between April 17 th and June 9 th , a rain gauge located at the Fieldston Ethical Cultural School 3.6 kilometers west of the rain garden was assessed to use the precipitation data during this time. To validate whether the Shoelace Park and Fieldston Ethical Culture School rain gauges were similar, the total precipitation values collected during each storm by each rain gauge were compared in the time periods that both rain gauges were functioning properly. Despite the rain gauges' close proximity and approximately equal elevations, the average percent difference between the rain gauge total precipitation values was 28.9%. This is significantly different and, therefore, the Fieldston Ethical Cultural School was deemed unreliable in estimating precipitation in Shoelace Park. Since no other rain gauges obtained comparative readings to the rain gauge in Shoelace Park, the four storms between April 17 th and June 9 th were not included in the analysis.

Rain Garden Water Balance
Inflows to the rain garden include direct precipitation onto the rain garden surface ( The ability of the rain garden to retain stormwater will vary from storm to storm, in part due to the relationship of its available storage capacity to the antecedent dry weather conditions. The methodology used to compute each of the inflows and outflows is described below.

Inflow from 228 th Street
Inflow from 228 th Street (I228)  In order to obtain a better approximation of performance efficiency (PE) values, the volume of 228 th Street runoff captured by the stormwater inlet that enters the manhole when the water level is below the invert of 34.14 cm is computed. This was computed solely for a better approximation of performance efficiency values and is not included in the calculation of rain garden inflow. These values significantly affected the performance efficiency values. This value is added to the flow through the Thel-Mar weir when determining the total water volume captured by the stormwater inlet. The peak water depth below the weir (HPeak) is 34.14 cm for storms that produce flow through the Thel-Mar weir. For storms that do not produce a water depth that reaches the weir, the HPeak value is the peak water depth below the Thel-Mar weir measured by PT-A over the duration of the storm.

Additional Inflow to Rain Garden
The Soil Conservation Service (SCS) Curve Number Method (Viessman et al. 2003) was used to estimate additional observed inflow into the rain garden from tributary areas C2-C5. Note that this equation was empirically developed using English units and is not valid using metric units. Therefore, inputs were in imperial units for calculations using this equation and outputs were converted back to the metric system. The method is summarized below: = ( 1000 ) − 10 (4) Valid for: Where:  With a higher curve number, more runoff is generated from a given precipitation amount.
228 th Street (C1), Bronx Boulevard (C4), and the Shoelace Park walkway (C5) all have a curve number of 98 (Viessman et al. 2003). This is due to their asphaltic surfaces. The sloped lawn adjacent to the rain garden has an estimated curve number of 61 (Viessman et al. 2003). This is because the surface is covered by more than 75% of grass and the underlying soil is best represented by Soil Group B. This soil group has slightly higher rates of permeability and inhibits runoff.
Runoff values were also estimated for 228 th Street, C1, in order to validate the Thel-Mar weir inflows. This is later discussed in the performance efficiency methods section.

Direct Precipitation Volume
In addition to the runoff, the direct precipitation on the rain garden area was calculated for each storm. This is simply the precipitation depth generated by the rain gauge multiplied by the rain garden area as shown in the equation.

Rain Garden Outflow
The outflow through the domed riser was calculated by the flow rate through the individual holes within the perforated cap. There are two separate rows of 1.27 cm diameter holes on the cap. The first row includes 15 holes at a 32.4 cm elevation above the elevation of PT-B.
The second row includes 11 holes at a 3.8 cm height above the first row. PT-B measured the head as the distance from the observation well sump to the water table in the rain garden. These dimensions are shown in elevation view on Figure 6. Note that the figure is not to scale. Also refer to Appendix F for photographs of the domed riser and observation well. The orifice flow equation was used to calculate the flow rate through these perforations. This is shown as Equation

7
. The coefficient of discharge includes the friction losses from the texture of the edges of the holes. Since the holes are sharp-edged, the coefficient of discharge is 0.61 (Street et al. 1996).

Rain Garden Infiltration
Evapotranspiration, infiltration, and rain garden storage during the storm are assumed to be the difference between all inflows and the outflow to the riser, as shown in Equation 1.
Equation 11 was developed to estimate the amount of rain garden infiltration over the duration of each storm. This estimation requires an approximation of rain garden soil storage during a storm. (%) = 228 + _228 × 100 (13) Where: When the PE is 100%, the site inlet has successfully diverted all of the 228 th Street runoff to the rain garden. When the PE is below 100%, it is likely that there was bypass at the stormwater inlet leading to the rain garden. In the case where the PE is greater than 100%, the effective 228 th Street catchment was likely larger than assumed, for example due to bypass of other combined sewer inlets located further upslope beyond Carpenter Avenue or lot level runoff discharged to this segment of 228 th Street.
The second indicator of the rain garden stormwater capture performance is the percent retained (PR) considering all inflows and outflows to the rain garden during each storm. Where:

Results
This section displays the results of the water balance calculations and performance indicators. Tables 6, 7, 8, and 9 display the results of the water balance, performance efficiency, percentage retained, and storm data relationship calculations respectively. A comparison between the inflow and outflow hydrographs and hyetographs will follow. Finally, the correlations between independent variables and PE and PR results from the tables are discussed.

Hydrograph and Hyetograph Comparison
The comparison between the hydrograph and hyetograph for a storms on November 17 th , 2014 (Figures 7 and 8), March 31 st , 2015 (Figures 9 and 10), and June 27 th and 28 th , 2015 ( Figures 11 and 12) depict the hydraulic systems' response to precipitation. There is a slight delay between the beginning of the storm and Thel-Mar weir inflow as shown in Figures 7, 9, and 11.
Water accumulates within the shallow manhole on 228 th Street and, generally, after about an hour of rainfall, stormwater begins to enter the rain garden via the Thel-Mar weir. In the case of the November 17 th Storm as shown in Figure 7, there were almost 12 hours of light rainfall before flow began through the Thel-Mar weir. This is because the initial water height in the manhole was about 23 cm and required a water level rise of more than 11 cm before reaching the Thel-Mar weir invert. There is an even longer delay between rainfall accumulations and rain garden outflows as shown in Figures 8, 10, and 12. Since the rain garden is unable to fully infiltrate the water during the event, ponding and subsequent outflows will tend to occur later in the event. The Thel-Mar weir flow and runoff from the surrounding tributary areas will continue to increase rain garden ponding and subsequent outflow a few hours after the conclusion of rainfall. Many other storms exhibited the same behavior.

Performance Efficiency Results
Overall, the average performance efficiency was 175% which has a 75% variation from the expected PE of 100%. Assuming normally distributed data, the 95% confidence intervals of PE are 65% to 284%. Performance efficiency was compared to various metrics in Tables 6, 7, 8, and 9 to identify whether any of these assumed independent variables appear to have played a role in the inlet hydraulic performance.  Figure 13 shows this relationship. Similarly, lower PE values correlate with longer storm durations as described by a correlation coefficient of -0.685. This is shown in Figure 14. However, Thel-Mar weir inflows (I228) and storm intensities did not correlate with PE and had low levels of significance.
Therefore, these metrics were not shown. The relationship between the PE and date of the event were graphed in Figure      June 2015 to July 2015 340%

Percentage Retained Results
The rain garden was able to retain 11.9 m 3 or 77% of inflows on average. which is indicative of a loss in rain garden performance with increasing precipitation depths. Figure 16 shows this relationship. Similarly, lower PR values correlate with larger rain garden outflows as described by a correlation coefficient of -0.951. This is shown in Figure 17. However, storm duration and storm intensities did not correlate with PR. Therefore, scatterplots of these metrics were not shown. Relationships with other variables were statistically insignificant and were not considered.   in PT-A for a few storms as explained in Appendix G. Higher PT-A readings consequently cause a higher Thel-Mar inflow volume and up to 50% higher PE. As this is a possible explanation for these higher PE, it is not an all-encompassing explanation as 11 other storms with PE over 100% did not exhibit abnormally high PT-A reading spikes during rain events and consequent Thel-Mar weir flows. The second scenario is that the volume of water captured by the stormwater inlet is in fact larger than expected. Sources of this extra water may include upstream combined sewer inlet blockages and unexpected water contributions from adjacent surfaces, such as nearby roofs or buildings. This scenario is likely as observations showed roof drains from adjacent structures contributing runoff to the stormwater inlet. However, while this may account for PE three times the expected 100% value, it does not account for PE 11 times larger than expected in the July 7 th storm. The third scenario is that flow through the Thel-Mar weir is hindered by objects in the manhole nullifying the developed hydraulic equations. There is no evidence of this phenomena occurring, though flow through the Combined Sewer Inlet B was observed to be hindered causing equivalent PT-B reading behavior in the spring and PT-A reading behavior in the fall. Since the site visit revealed that the design catchment area was fully contributing to the stormwater inlet with little or no stormwater inlet bypass to Combined Sewer Inlet A, bypass may be occurring in another fashion. Lower PE were characteristic of higher precipitation depths and storm durations as shown Figures 13 and 14, respectively. Therefore, higher precipitation and larger runoff values may cause inlet bypass. However, the April 20 th , 2015 field visit during a large precipitation event did not confirm this hypothesis. None of the scenarios were fully proven or refuted and the possibility of stormwater inlet bypass and a larger catchment area on 228 th Street cannot be denied.

PR and Rain Garden Outflow Relationship
In general, PR decreases as storm depth and storm duration increase. Additionally, overflow increases as precipitation depth increases, indicating that at about 1 cm of precipitation, the volume available within the soil matrix of the rain garden fills up completely at least in its superficial layers. This also indicates that the saturated infiltration rate of the rain garden is smaller than the water inflow rate. However, the majority of overflows are very small compared to their corresponding total inflow volumes. The five instances that the outflows were larger than inflows were analyzed in detail. This was attributed to the Combined Sewer Inlet B clogging scenario discovered during the field visit on April 20 th , 2015. The clogged inlet indicates that the hydraulic equations are invalidated and performance of the rain garden may consequently be underestimated. On the other hand, such findings draw attention to the issue of the maintenance of the hydraulic structures ancillary to the rain garden system. The storms on December 8 th , January 18 th , March 14 th , March 26 th , and April 9 th produced negative infiltration volumes and consequently negative PR. Three of these storms occurred about a month before the clogged Combined Sewer Inlet B was observed. The average precipitation depth from these five storms was 3.1 cm. Since the rain garden releases stormwater to Combined Sewer Inlet B during storms with 1 cm or larger of precipitation depth, Combined Sewer B would have likely clogged during these storms. This in effect causes an overestimation of outflow and underestimation of PR in storms with larger precipitation accumulations. This evidence supports the possibility of overestimation of outflow since the calculated outflow was erroneously greater than the total inflow. Therefore, these five storms were removed from the data analysis. Despite the occurrence of larger outflows, the rain garden performed quite well given the circumstances of a drainage area eight times larger than the design drainage area. As previously mentioned, a rainfall event greater than 1 cm is likely to induce ponding on the rain garden and consequent overflow. A precipitation depth of 1 cm is a small storm with a high reoccurrence interval, but if it only receives runoff from the design drainage area, the rain garden would be able to retain all inflows from more significant events such as 2.5 cm storms. Still, the analysis proves the rain garden's ability to retain an average of 11.9 m 3 of water per event from a drainage area 131 times its size.
The 37.1 m 2 rain garden successfully retains approximately 24.16 m 3 from a 2.5 centimeter storm from all tributary areas. On average, 10% of the inflow into the rain garden was from off-site 228 th runoff flowing via the Thel-Mar weir and 87% of the inflow was attributed to on-site runoff in Shoelace Park with the other 3% from direct rainfall. If the rain garden was designed to capture only runoff from off-site street locations, it could capture all runoff from a 950 m 2 area, or about 26 times the size of the rain garden, for a 2.5 centimeter storm.  Urban park GI allocation percentages up to 15% can theoretically manage all of the city's impervious cover. This includes managing on-site runoff as well. In the case of the Shoelace Park rain garden, about 87% of the stormwater managed is from parkland runoff which, in this case, still reduces the amount of stormwater entering the combined sewer. Regardless of runoff type, the Shoelace Park rain garden was able to manage large quantities of stormwater from both on and off site runoff. Retrofitting less than 15% of parkland space with GI has a large stormwater retention impact with possibly no influence on the park's current use.

Conclusion
The analyses conducted throughout the monitoring period indicate that the rain garden built within Shoelace Park is extremely effective in managing the stormwater generated in its design drainage area on 228 th Street. The rain garden appears to receive much more water than expected during its design since the design hydraulic loading ratio (HLR) of 16 is an order of magnitude lower than the observed HLR of 131. The additional inflow is likely associated with water captured by the stormwater inlet in excess of that expected by the inlet drainage area and water conveyed to the rain garden from adjacent areas, namely the Shoelace Walkway, Bronx Boulevard, and Sloped Lawn. The former contribution is partly due to rooftop stormwater discharge from buildings located on 228 th Street, although other tributary areas upstream may have further increased the water load. When all these additional contributions are factored in, it was determined that the rain garden was receiving over 8 times the design stormwater load.
The rain garden captured a large area of stormwater in addition to the 228 th Street diverted water. Ponding and consequent outflow from the rain garden into Combined Sewer Inlet B was detected in some instances throughout the monitoring seasons. For all monitored storms, the rain garden was able to retain an average of 77% of the stormwater load. The rain garden outflow into Combined Sewer B began with events that were greater than 1 cm in depth. The recurrence interval of a storm with this precipitation depth is fairly frequent. The 600 m 2 catchment area that the rain garden was originally designed for is much smaller than the 4,850 m 2 observed catchment area. Combined Sewer Inlet B was designed to capture the stormwater within Shoelace Park which has an approximate catchment area of 4,400 m 2 . However, the rain garden was observed to capture this runoff instead. Even though overflows occur at precipitation depths of 1 cm, the rain garden is capturing an area 131 times its size. Therefore, the rain garden is performing very well considering the excess runoff inflows from Shoelace Park.
With limited right-of-way space in urban environments, the analysis determined that urban parks can efficiently manage runoff from off-site, impervious cover. The rain garden at Shoelace Park successfully manages all runoff from an off-site tributary area about 26 times its size in 2.5 centimeter and below storms. This means that only a small percentage of New York City's 154 km 2 of parklands can manage a great deal of runoff from its 72% impervious cover.
This small amount of implemented GI also allays urban park users' concerns about detriments to existing recreational uses. Other cities with high impervious coverages and existing parkland space such as Philadelphia, Pennsylvania, Washington, D.C., and Los Angeles, California can follow suit. All being considered, the analysis of the system installed at Shoelace Park suggests that the effective hydrological footprint of parklands extend beyond the physical confines of parks themselves and that parklands can be effectively used to support urban stormwater runoff management. This study assumes that all impervious coverage areas are within the combined sewer areas of NYC and that urban parks are located adjacent to a large portion of impervious coverage. A thorough study of how much adjacent impervious coverage runoff each urban park can realistically manage using GI is in order. It is also uncertain how well the Shoelace Park rain garden would perform if all inflows were from off-site runoff. For these reasons, the development of a comprehensive monitoring effort is encouraged to further assess the potential of parks within a large scale stormwater management strategy.

Appendix B: Thel-Mar Flow Rate Validation
Thel-Mar, LLC provided their experimental flow rate data as shown in Tables 15 and 16. The table converts a head value above the weir V-notch in millimeters to a corresponding volumetric flow rate in a variety of units. This was plotted in Figure 19 as a conversion from head in millimeters to the outflow in cubic centimeters per second. Since Thel-Mar collected the data through experimentation, a curve fit was the best way to develop an equation to convert the calibrated head data from PT-A and PT-B to a flow rate. After observing the complex curve in Figure 19, it was determined that a single curve fit equation would not provide accurate flow rate values. Therefore, splitting the curve into three different sections would reduce error in the curve fit.
A polynomial curve fit was determined to be the best fit to the complex data. However, the order of the polynomial curve fit was uncertain and elicited a statistical comparison technique.
Akaike Information Criterion (AIC) was utilized to statistically compare models with different numbers of parameters to determine the model that fits the data the best. This is shown in  Table 17 below. Based on these results, the fifth order polynomial curve fit has the lowest AIC value and is therefore the best predictor of all of the models. The data are plotted on Figure 19 below. Due to the complexity of the data, the data were divided into three sections. The fifth order polynomial curve was fit to all three sections as shown in Figures 20, 21, and 22. The equations generated in the chart were used to compute Thel-Mar weir flow rates with respect to each head in millimeters.   The experimental flow rates obtained from Thel-Mar, LLC were validated using theoretical weir flow formulas (Street et al. 1996). Since the shape of the weir as shown in Figure   5 is slightly more complex due to a combination of triangle and rectangular cross sections, two different formulas were required. The triangular weir flow rate equation is shown below.
The flow rate through the rectangular section of the weir is shown below.

Appendix C: Catch Basin Crack Analysis
Upon proof of the weir invert at 34.14 centimeters, PT-A behavior suggests there is a crack in the manhole about 11 centimeters below the Thel-Mar weir invert. In order to calculate the crack geometry, the crack was assumed to be perfectly circular and exit into open air on the outside of the manhole. The crack diameter was computed using a derivation of the orifice equation shown in Equation 28 and the coefficient of discharge was assumed to be a short tube equal to 0.8 (Street et al. 1996). The crack location was defined as the calibrated PT-A reading where the water level became static after the duration of the storm. The head above the crack was defined as a head location below the Thel-Mar weir where water was still exiting through a lower orifice. The volumetric loss is the instantaneous volume change at the defined head above the  The average computed crack diameter was 0.27 centimeters for all storms. This crack size is possible, but only for this idealized situation. A more comprehensive crack flow analysis will most likely compute a larger crack geometry as the flowing water exits into the surrounding soil matrix more slowly than the idealized open air situation. The crack location seems to be located around 23 centimeters above the manhole sump. However, flow does not exit through this orifice in storms during March and April 2015. An explanation could be a dynamic groundwater table where the adjacent soil matrix may be saturated and inhibiting flow during March and April 2015 and dry at all of the times of the monitoring campaign. It is also unlikely that the crack geometry is a perfect circular conduit in the wall of the manhole, but the possibility cannot be ruled out. There is always a chance that multiple smaller cracks or orifices cause this same water loss. This gives credence to the constant Thel-Mar weir invert and a possible crack about 11 cm below it.

Appendix D: Pressure Transducer In-Situ Validation
The PT-A and PT-B in-situ measurements were not used as a calibration. It was difficult to obtain an accurate PT-A and PT-B measurement in the manhole and rain garden well, respectively. Due to the difficulty in taking accurate field measurements, less significant digits were used in the calculation of percent error. This is shown by coefficient of determination values less than the bucket calibration coefficient of determination values in Figure 3. A calibration with the slope equations was attempted, but calculated flow values were not consistent with performance indicator metrics later defined in this section. Therefore, these values were solely used as a validation.    Despite the PT difficulties, the three other tests were conducted to collect qualitative data on the PT's sensitivity to turbulence. Swirling water was tested by pumping water through a hose into the side hole of the bucket and simultaneously allowing water to drain out through the bottom orifice at the same rate. This produced a constant water depth as shown in Figure 33. This caused a fluctuation of 0.6 centimeters which can be up to 48 centimeters based off of the calibration values as shown in Table 20. The bottom orifice in the bucket was then sealed and the pressure transducer's response to waves and splashes was tested as shown in Figure 34. The PT readings fluctuated 0.2 centimeters which can translate to a 16 centimeter change. Despite the PT errors, the test suggests that water disturbances may cause inaccurate PT readings which calls for future studies.  Fluctuations in PT-A readings during Thel-Mar weir inflow were observed in eight storms between October 2014 and July 2015. Assuming that these abnormally high PT-A readings were due to turbulence, the high PT-A readings were replaced with averaged PT-A readings fitting the normal trend of PT-A readings in close proximity. The purpose was to test the affects that these fluctuations on PT-A readings had on the total Thel-Mar weir flow and performance efficiency values.
The adjustments made to irregularly high PT-A readings during Thel-Mar weir inflow for eight storms are displayed in Table 21. Replacing the high PT-A readings with lower averaged ones decreases the Thel-Mar inflow volume as shown in the "turbulence inflow volume" column in the table. This decreased the average PE of these eight storms from 203% to 67% with an average PE percent change of -50%.