Adapting Urban BMPs for Resilience to Long-term Environmental Changes

Abstract

Stormwater best management practices (BMPs) help mitigate the adverse effects of urban development on stream hydrology and water quality, and are widely specified in development requirements and watershed management plans. However, design of stormwater BMPs largely relies on experience with historic climate, which may not be a reliable guide to the future. To inform BMP design that is robust to future conditions, it is important to examine how potential changes in precipitation, temperature, and potential evapotranspiration will affect the performance of BMPs. We use continuous simulation modeling to examine BMP performance under current and potential future climatic conditions, and determine the changes needed in site configuration to address future impacts. We perform modeling for five development types in five different regions of the United States and explore both conventional (“gray”) and green infrastructure (GI) stormwater management approaches. If stormwater designs are adapted to address potential future climate conditions, this study suggests that the most cost-effective approaches may use both gray and green BMPs. If the magnitude of extreme weather events increases dramatically, then gray practices that provide detention storage may have better cost effectiveness. Incorporating risk of future climate impacts into stormwater design may help communities become more resilient.

Introduction

Stormwater management refers to the practice of controlling and conveying excess runoff from precipitation events occurring on the built environment to minimize negative impacts to communities and receiving waters. The specific approaches and infrastructure of stormwater management have evolved over time from the simple stone canals of pre-historical settlements to the more advanced active treatment systems of modern industrial stormwater systems (see Reese, 2001 for a comprehensive [and humorous] history of stormwater management). Historically, stormwater infrastructure has focused on drainage and flood control and has taken the form of routing stormwater runoff away from buildings, roadways, and other critical infrastructure as quickly and efficiently as possible. While effective at protecting infrastructure near the source of the excess runoff, this potentially leads to increases in downstream flooding and structural damage. In response, flood control approaches were adopted extensively throughout the 20th century that incorporated temporary runoff storage upstream of areas of flooding concern. Conventional methods often involve use of detention ponds to control peak runoff from large storms. Stormwater drainage systems that rely on curbs, gutters, culverts, and detention basins are examples of “gray” infrastructure.

Stormwater treatment evolved further as concerns about excess pollutant loading from urban areas led to practice designs that targeted capturing pollutants and reducing loads in outflow. Typical control in the early days of this stormwater treatment phase involved treating runoff volume from a small storm (e.g., 1-inch rainfall) to achieve a performance standard (e.g., 85% sediment reduction). A stormwater wet pond was the typical best management practice (BMP) used. This was followed by the emergence of enhanced water quality control using practices such as stormwater wetlands, and later various “green” infrastructure (GI) and low impact development practices. These practices were developed to address a wide variety of regulatory and community goals including pollutant load reduction, promoting infiltration, and maintaining runoff volume/water balance. The specific drivers, goals, and objectives dictate practice selection and design. Examples of GI include bioretention, permeable pavement, impervious surface disconnection, and stormwater capture/reuse. In some cases, a practice may be designed to meet multiple objectives (e.g., initial storage provides treatment of the 1-inch storm event for water quality goals, while additional storage is used to mitigate peak flows from flooding events).

Stream protection and hydromodification control is a third primary function of stormwater management and the third phase in its evolution. Increases in runoff volumes, peak discharges and duration of flow velocities can accompany urban development and result in failing culverts, scour, and stream bank erosion. Initial stream protection approaches involved control of small storm peak flow volume (e.g., 1 or 2-year peak control). More recent research (Geosyntec Consultants, 2004) has suggested that additional volume control is needed beyond peak management, which has led to more holistic hydromodification requirements that target matching the upper end of the site flow duration curve to a specific (often pre-developed) condition.

While much is known about stormwater management in general, little is known about how stormwater infrastructure performance is affected by changes in climate. We can make inferences based on knowledge of unit processes in BMPs; for instance, if precipitation volume and storm event intensity increase, the following performance changes are likely:

• For BMPs providing volume-based treatment, the target depth is typically based on a percentile of the distribution of rainfall depths (e.g., the 90^th percentile event is equivalent to the 1-inch storm). If precipitation volume and intensity increase, the target rainfall depth would increase as well. Treatment efficiency of the BMP is likely to decrease.

• Some BMPs are designed to achieve a certain minimum pollutant removal (e.g., 80% of inflow sediment). Pollutant load reduction may decrease as residence time decreases.

• BMPs providing flood control generally target one or more design storm depths (such as the 10-year event and the 25-year event) and use detention structures with outlet control to match pre-developed conditions. An increase in precipitation volume and intensity would change the hydrograph shape and increase the volume of the target storm events. Peak flows in excess of design standards are a likely result.

• Likewise, for BMPs providing for more complex hydromodification control, the flow duration curve is likely to change, and the BMP may not fully mitigate the risk of downstream channel erosion.

There is a need to evaluate how stormwater infrastructure performance in specific places is affected by changes in climate, and to understand how to adapt stormwater infrastructure to accomplish treatment objectives. This study attempts to address the following questions:

1. How might projected changes in meteorological forcing affect the performance of conventional stormwater infrastructure and GI (developed under existing design standards) compared to current conditions,

2. How can conventional designs and GI designs be adapted so that the stormwater infrastructure will provide current levels of performance under future conditions, and

3. What do the results suggest regarding the adaptation potential of gray and green infrastructure for increases in extreme precipitation events?

To answer these questions, continuous simulation modeling was used to represent conceptual development sites of various configurations with a variety of stormwater management practices, to test how performance changes under potential future climate conditions, and to optimize design modifications to achieve original performance objectives. A detailed discussion of the modeling and results is provided in USEPA (2017).

Methods

Modeling Approach

We use the Hydrologic Simulation Program – Fortran (HSPF; Bicknell et al, 2004) to simulate hourly unit area time series of runoff and pollutant loads from pervious and impervious land from conceptual development sites for thirty years of input meteorology. Current climate conditions were obtained from National Oceanic and Atmospheric Administration monitoring stations. We developed future scenarios of changes in precipitation, temperature, and potential evapotranspiration from the current time series using both percent changes and downscaled General Circulation Models (GCMs). The conceptual sites and associated stormwater management infrastructure were simulated using the SUSTAIN (System for Urban Stormwater Treatment and Analysis INtegration) model, a decision support system and modeling tool (USEPA, 2009). We developed each conceptual site in detail representing BMP footprint, volume, and configuration using local and/or state requirements and guidance. TR-55 and other tools were used to develop scoping-level designs for the practices, including details of outlet structures with orifices and weirs set to meet stormwater requirements. Soil properties of media in filtering BMPs such as bioretention and sand filters were represented using properties from design guidance as well as values from BMP research. SUSTAIN provided hourly output for each individual BMP and at the site outlet.

We modeled each stormwater management approach under current and a selected set of projected future climate conditions for the mid-21^st century. The water quantity and quality performance of the site practices were calculated from modeling results. In an additional model run, the site’s practices were modified for the potential future climate scenario to achieve the same performance as under the current climate conditions using SUSTAIN’s optimization function. Modifications targeted resizing the water quality treatment and peak flow control BMPs, which are the primary drivers controlling site performance. The performance metrics were defined as follows:

• Annual outflow volume to address stormwater volume treatment requirements,

• Flow duration curve to address channel erosion risk and flooding risk, and

• Pollutant mass export to address water quality performance. Pollutants include sediment as total suspended solids (TSS), total nitrogen (TN), and total phosphorus (TP).

We tabulated annual runoff volume, the flow duration curve, and pollutant loads for each site under current climatic conditions in an initial SUSTAIN model run. Under projected future climate scenarios, these metrics generally increased. When this occurred, an optimization was performed in SUSTAIN using numerous model runs to assess incremental changes in practice sizes. The goal of the optimization was to find a new “adapted” configuration that met all the current conditions metrics. For the flow duration curve, the optimization sought to minimize the area between the curves, thus mimicking the current condition in hydraulic response. During a given optimization model run, typically 500 to 1500 different incremental designs were simulated. The SUSTAIN optimization included scoping-level estimates of unit-area practice costs, so the optimal solution was selected as being the lowest overall cost that met all the performance metrics. Costs were based on present value of design-build and 20-year O&M, and did not include other triple bottom line benefits generally associated with GI. BMP lifecycle costs were estimated using literature sources and best professional judgment based on project experience. The primary sources were King and Hagan (2011) and the Green Values Calculator (Center for Neighborhood Technology, 2014). As an example, for one of the stormwater management approaches, the footprints of the bioretention cells and the dry detention basin were increased under projected future climate scenarios. This provided additional hydrologic control and pollutant removal so that the revised configuration performed as well as or better than the site under current climate conditions. Performance is evaluated at the site outlet, defined as the point to which all areas, BMPs, and conveyances ultimately drain.

Regions and Land Uses

The conceptual sites cover five types of developed land use in five geographic locations representing different hydroclimatic regimes throughout the US (Table 1). The developed land use types include residential, commercial, mixed use, ultra-urban, and transportation corridor. We selected five regions that leveraged existing EPA research on projected hydrological changes: each of these regions was modeled as part of an EPA project assessing the impacts of projected climate change on hydrology and water quality in 20 U.S. watersheds (referred to in this paper as the “20 Watersheds” study) (USEPA, 2013; Johnson et al., 2015). HSPF models developed for the EPA project were used as unit area input for the SUSTAIN simulation. Each land use type is paired with a geographic location, and a variety of stormwater BMPs are represented, ranging from conventional gray infrastructure to GI designs. For each land use type, one, two, or three stormwater management approaches are used to illustrate different ways to address site stormwater. Most of the land use types have one conventional stormwater management approach and one approach incorporating GI elements. In some cases, the GI approach also includes gray infrastructure practices to address peak flow control requirements for large storm events (i.e., 10-year through 100-year design storms). In other cases, a GI approach with no gray infrastructure meets local stormwater requirements. It is important to note that the sites are conceptual – in other words, they do not exist in the real world – but site layouts and stormwater infrastructure do reflect real world conditions.

Table 1.

Matrix of Regions, Locations, Land Uses, Future Climate Scenarios, Stormwater Management Approaches, and Stormwater Practices

Region	Location	Land Uses and Characteristics	Climate Scenarios	Stormwater Management Approach	Stormwater Practices
Southeast	Atlanta, GA	Ultra-Urban 2 acres 90% impervious	•GCM High Intensity	Conventional (gray) infrastructure	Underground sand filter, underground dry detention basin
				GI with gray infrastructure	Green roof, permeable pavement, bioretention, and underground dry detention basin
Arid Southwest	Maricopa County, AZ	Commercial 10 acres 80% impervious	•GCM High Intensity	Conventional (gray) infrastructure	Detention/infiltration basin
				GI only	Permeable pavement, cistern, bioretention, and stormwater harvesting basin
Pacific Northwest	Portland, OR	Transportation Corridor 0.35 acres 89% impervious	•GCM High Intensity	GI only	Bioretention swales, permeable pavement
Mid-Atlantic	Harford County, MD	Mixed Use 20 acres 65% impervious	•GCM High Intensity •Minus 10 Percent •Plus 10 Percent •Plus 20 Percent	Conventional (gray) infrastructure	Surface sand filters, extended dry detention basin
				GI with gray infrastructure	Infiltration trenches, infiltration basins, permeable pavement, and dry detention basin
				Conventional (gray) infrastructure with distributed GI	Surface sand filters, extended dry detention basin, distributed infiltration trenches
Midwest	Scott County, MN	Residential 30 acres 48% impervious	•GCM Low Intensity •GCM Medium Intensity •GCM High Intensity •Minus 10 Percent •Plus 10 Percent •Plus 20 Percent	Conventional (gray) infrastructure	Wet pond
				GI with gray infrastructure	Distributed bioretention and dry detention basin
				GI only	Distributed bioretention, permeable pavement, and impervious surface disconnection
				Conventional (gray) infrastructure with distributed GI	Wet pond, distributed bioretention

For the Mid-Atlantic site and the Midwestern case studies, we modified the stormwater management designs using two different management strategies – increasing the size of the structural practices (as done for the other locations) and addressing the performance gap by incorporating additional distributed GI practices. In the latter approach, the current conventional infrastructure configuration was unchanged and distributed infiltration trenches (Mid-Atlantic site) and bioretention (Midwestern site) were added to provide treatment equivalent to current conditions.

Projected Future Climate Scenarios

For this analysis, we examined 10 climate change scenarios previously used in the “20 Watersheds” study. The study selected six future climate scenarios from the North American Regional Climate Change Assessment Program (NARCCAP) (Mearns et al., 2009, 2013) . The NARCCAP scenarios provided high resolution climate change data for the contiguous US and the full suite of meteorological variables needed to implement HSPF simulations that use an energy balance approach to estimate evapotranspiration (ET). The remaining four scenarios in the study used statistically downscaled output from the same set of GCMs used by NARCCAP from the bias correction and statistical downscaling (BCSD) methodology. BCSD is provided by the World Climate Research Programme’s Coupled Model Intercomparison Project phase 3 multi-model dataset (Maurer et al., 2007). All climate change scenarios in the study were implemented using a change factor approach (Anandhi et al., 2011) to modify 30 years of observed local weather data to ensure realistic patterns in time series. For further details, see USEPA (2013) and Johnson et al. (2015).

For this study, we selected the scenario (from the pool of ten) representing the largest projected increase in precipitation intensity to assess effects on stormwater infrastructure, thus characterizing an upper bound for potential impacts. For each of the ten climate scenarios, we calculated both the change in monthly precipitation depth and the fraction of precipitation contained within events greater than the 70^th percentile daily event by comparing runs of the same GCM for future and historic 30-year time periods. Projected future climate time series were created from observed historic time series by applying a multiplicative change factor approach to adjust total volume and redistributing the fraction of this volume to events above and below the 70^th percentile event. The scenario with the month that had the highest projected volume of events above the 70^th percentile was chosen as approximating the highest storm event volumes among the available scenarios (hereafter called the “high intensity scenario”). Note that the scenarios used in the “20 Watersheds” study do not represent the full range of projected future precipitation driven storm volume and intensity changes at these sites that would be derived from the CMIP3 or CMIP5 ensembles (USGCRP 2017).

In terms of the analysis, projected future meteorological inputs have been prepared that are representative of local conditions using local National Centers for Environmental Information (NCEI, formerly the National Climatic Data Center) weather stations. In addition, the redistribution of precipitation changes between smaller (<70^th percentile) versus larger (>70^th percentile) events helps account for not only changes in volume but also changes in intensity. (Potential changes in frequency of events independent of volume were not investigated.) The approach is believed to be appropriate for the goal of illustrating the potential impacts of projected changes in meteorological forcing on the performance of stormwater BMPs in urban watersheds distributed across the United States; however, the specific results for each geographic location are in part dependent on the characteristics of the individual weather station. Details regarding the change factor approach, final GCMs selected for each region, and the local NCEI weather stations used for each region are provided in the online supplemental material.

In addition, two types of climate sensitivity analyses were conducted as part of the study. One sensitivity analysis was performed by selecting two additional downscaled GCM scenarios from the pool of ten for the Midwestern site that represented the lowest change and a medium change in large storm event intensity. Another sensitivity analysis conducted at the Midwestern site and the Mid-Atlantic site evaluated the effects of set percent changes in all precipitation events. The current precipitation record was modified to represent potential future climate conditions by applying a graduated set of percent changes to the entire precipitation record (across the entire range of hourly precipitation values from a trace to the largest rainfall value). The percent change factors were −10 percent (a decrease in intensity), +10 percent, and +20 percent (both increases in intensity). Air temperature was adjusted as well using the Clausius-Clapeyron relationship (Clausius, 1850; Clapeyron, 1834), such that a 10 percent change in precipitation intensity was paired with a 2.6 °F change in air temperature. Potential ET was then recalculated from the modified air temperature and precipitation time series using the same method employed for the “20 Watersheds” study (USEPA 2013; Johnson et al., 2015).

Results

A description of each site’s characteristics, current post-development stormwater requirements, and adaptation (stormwater management) scenarios is provided below. Next, a comparison of the current and projected future climate scenarios is explored. Percent change scenarios are not discussed, since they are simply multiples of the corresponding current precipitation series.

Southeast, Ultra-Urban Site

The ultra-urban conceptual site is assumed to be in a heavily developed downtown area of Atlanta, GA in the southeastern US, on a two-acre parcel with 90 percent impervious cover. A combination of local and state stormwater requirements apply, the most important of which are retention or treatment of runoff from the first inch of rainfall (the water quality volume or WQv), detention of runoff from the 1-year 24-hour storm over a 24-hour period (the channel protection volume, or CPv), and matching peak flows to pre-development (natural) conditions for the 2-year through 100-year 24-hour storm events. The two stormwater management approaches at the site (see Table 1) use a combination of practices to achieve both pollutant removal and temporary storage of runoff. As discussed in Methods, one of the simulation runs is used to optimize and resize the BMPs under potential future precipitation conditions and achieve the same performance as the site under current conditions. For the conventional stormwater management scenario, both the sand filter and the underground detention basin were allowed to increase in size. For the GI with gray infrastructure stormwater management scenario, all practices were allowed to increase in size except for permeable pavement, which was considered to already occupy the maximum available area.

For the Southeast future precipitation scenarios, the projected changes in average annual precipitation show an increase for all years with a low degree of variability. Projected average annual depth increases from 55.7 inches to 59.4 inches, or by 6.6 percent. Projected monthly precipitation increases in some months and decreases in other months. In addition, daily sums of precipitation depth were calculated and were used to determine percentiles of 24-hour depth. For the 85^th percentile, the projected depth increases from 1.04 to 1.14 inches, while for the 95^th percentile the depth increases from 1.71 to 1.87 inches. An hourly precipitation comparison for the very highest storm events (those with the potential to cause flooding) shows an increase of 1.2 – 1.5x between current and future conditions.

Arid Southwest, Commercial Site

Maricopa County (home of Phoenix) is in the arid Southwest, where annual precipitation volume is much lower than at the other locations in this study. The closest available weather station from the “20 Watersheds” study is a short distance northeast of Maricopa County in the Salt River watershed. Annual precipitation across Maricopa County is highly variable, ranging from less than 5 inches per year to over 20 inches per year. Annual average precipitation at the weather station used in this analysis is over 18 inches per year, which is at the high end of the range for the county. The conceptual commercial shopping center site is assumed to be 80 percent impervious covering 10 acres. Maricopa County requires (where infiltration rates permit) full capture and infiltration of the volume of runoff from the 100-year 2-hour storm event. For both stormwater management scenarios (Table 1) the selected practices are designed and represented in SUSTAIN to fully capture the target runoff volume. For the conventional stormwater management scenario, a single infiltration basin is used; note that while an infiltration basin is often considered a GI BMP, in this context it is conventional in terms of implementation. The optimization modeling runs allow all the practices to be increased in size for both stormwater management scenarios.

The annual precipitation comparison reveals that projected future conditions are highly variable, with increases seen in some years and decreases in other years. Average annual totals are 18.4 inches for current conditions and 19.6 inches under potential future conditions, reflecting a change of 6.5 percent. Monthly changes are somewhat variable, with the largest changes in January. For the percentiles of 24-hour precipitation depth, the current and future projected 85^th percentile is the same at 0.71 inches. However, for higher percentiles the future values exceed the existing; for instance, the 95^th percentile depth increases from 1.16 inches to 1.57 inches between current and projected future climate. The comparison of the highest hourly precipitation volumes indicates a steadily increasing gap between current and future intensity, ranging from 1.2x at the 1-year recurrence interval to as much as 1.8x at the highest recurrence interval.

Pacific Northwest, Transportation Corridor Site

The approach for the Portland, OR stormwater management scenario differs from the other locations; the stormwater management scenario reflects a specific style for which Portland has gained recognition – the Green Street. Green street projects use land adjacent to roads as an opportunity to retrofit practices into the urban landscape. GI elements placed in medians and along right-of-ways (ROWs) are used to address water quality treatment and stormwater volume reduction goals. Green street projects in Portland do not fall under the City’s requirements since they are implemented in road ROWs that are exempt from post-construction stormwater requirements. The City does attempt to meet the requirements, but this is not always possible due to site limitations. One stormwater management scenario was modeled for this location, based on a design typology reported in a master plan published by the City (City of Portland, 2008). The simulated site represents one side of a Green Street, an area of 0.7 acres with 89 percent impervious area. The site is designed to completely capture the 10-year 24-hour storm event. For the optimization modeling run, the bioretention cells were allowed to increase in size.

The Pacific Northwest potential future precipitation scenario differs from the previous locations, where the scenarios with the highest increase in intensity also showed an overall volume increase. Here, the precipitation volume is projected to decrease from 42.0 inches to 39.7 inches (−5.5 percent drop). Monthly precipitation volume decreases in most months from current to future conditions, though increases are seen during December and January. For the 85th through 95th percentiles of 24-hour precipitation depth, the change in 24-hour depth decreases from current to projected future climate conditions. However, there is an increase of 0.14 inches for the 99th percentile, indicating a modest increase in intensity for the very largest events. In terms of highest hourly precipitation volumes, projected future intensity tracks current intensity, with the highest increase about 1.16x.

Mid-Atlantic, Mixed Use Site

The conceptual mixed-use site, assumed to be in Harford County MD north of Baltimore, is 20 acres and 65 percent impervious. Local and state requirements are similar to those for the Southeastern Site – there is a WQv requirement for water quality treatment and a CPv requiring 24-hour detention of the 1-year 24-hour storm event. In addition, pre-development peak matching is required for the 10-year 24-hour storm event, and a groundwater recharge volume that must be completely infiltrated into site soils. The resulting suite of practices for the two stormwater management scenarios are designed and configured to meet the multiple applicable requirements. For the optimization runs, the practices were allowed to increase in size (except for the permeable pavement, which was implemented to the maximum extent practicable). In an additional optimization run, distributed infiltration trenches were added to the conventional design rather than resizing the practices for adapting to future climate conditions.

For the high intensity precipitation scenario, annual precipitation volume is projected to increase for each year of the simulation period when compared to current conditions; however, the increase varies from a few inches per year to over ten inches per year. The overall annual averages are 44.3 inches for current conditions, and 50.0 inches for future conditions, reflecting a 12.8 percent increase in volume. For average monthly depth, precipitation increases for some months and decreases in other months. The largest change is seen in September, where projected average depth increases from about four inches to about seven inches. For percentiles of 24-hour precipitation depth, the change in depth between current and future ranges from 0.09 inches for the 85th percentile to 0.43 inches for the 99th percentile. For the largest hourly precipitation depths, there is an approximately 1.5x increase in hourly precipitation depth across the recurrence range. The depth corresponding to 10-year recurrence under current conditions is projected to take place at a 2-year recurrence under future climatic conditions.

Midwest, Residential Site

The conceptual residential site covers 30 acres of single family and townhome development, with 48 percent impervious area. The assumed site location, Scott County MN, is south of Minneapolis beyond the built-out portions of the region. A combination of local and state stormwater regulations requires infiltration or treatment of the WQv, equivalent to one inch of runoff from impervious surfaces. Peak matching is required for the 2-year, 10-year, and 100-year 24-hour storm events. The site is assumed to have soils with poor infiltration rates, so infiltration of the WQv is not feasible. Instead, practices were selected that achieve an 80 percent reduction in sediment load, thus meeting the treatment requirement. Three stormwater management scenarios were simulated for the site (reflecting conventional, GI with gray, and purely GI management approaches). The optimizations runs under projected future climate conditions allow all the practices to increase in size for the conventional and GI with gray infrastructure stormwater management scenarios, while bioretention alone is resized for the GI only management approach. In an additional optimization run, distributed bioretention cells were added to the conventional design rather than resizing the practices for adapting to future climatic conditions.

As discussed in Methods, three future precipitation scenarios were selected for simulation reflecting a range in changes in intensity – the lowest intensity change, a medium intensity change, and the highest intensity change. Average annual precipitation is 30.1 inches for current conditions and is projected to be 28.7 inches for the future low intensity precipitation change scenario, 33.3 inches for future medium intensity change scenario, and 33.6 inches for future high intensity change scenario, corresponding to changes of −4.4 percent, 10.7 percent, and 11.6 percent respectively. Simulated changes in monthly average precipitation depth are highly variable by month. Percentiles of daily precipitation depths were also calculated for each scenario. For the future low intensity scenario, there is a decrease for all percentiles from the 85^th through the 99^th. The future medium intensity and future high intensity scenarios have comparable depth increases across the percentiles, with the future high intensity scenario showing a larger increase for the 99th percentile from 1.94 inches to 2.33 inches. For the comparison of the highest hourly precipitation volumes, the low intensity scenario projects a decrease in intensity compared to current climate for all hours except the single largest precipitation depth in the 30-year time series. The medium intensity scenario is only slightly higher than current conditions, while the high intensity scenario has a projected increase of about 1.25x to 1.4x that of current conditions.

Discussion

The introduction presented three central study questions:

1. How might projected changes in meteorological forcing affect the performance of conventional stormwater infrastructure and GI (developed under existing design standards) compared to current conditions?

Projected changes in meteorological forcing suggest a risk of increased flow and pollutant loading that BMPs designed under current regulations cannot fully mitigate. We first focus on changes in runoff volume, maximum peak flow, and pollutant mass loading prior to any reductions due to BMPs. Results are shown for the high intensity precipitation scenarios only since those were modeled for all five sites (Figure 1). Note that presenting a numeric measure of change in the flow duration curve between current and projected future climate conditions can be difficult to interpret, so the highest hourly peak flow during the simulation is provided as a proxy for large storm event response. Annual average runoff decreases for the Pacific Northwest site due to lower precipitation and higher summer ET projected under the high intensity scenario. However, the percent change in maximum hourly outflow increases; this suggests that while overall precipitation volume decreased, the intensity of the largest storm events increased. Percent changes in sediment and nutrient loads range from 1.5 percent to 26.7 percent in Southeast, Mid-Atlantic and Midwest sites. Load decreases in the Pacific Northwest correlate to reduced overall runoff. For the Arid Southwest, sediment more than doubles and TP more than triples. These larger increases are the result of one storm event with wet antecedent conditions where precipitation doubles, and a large increase in pervious runoff depth results. Surface and rill erosion have a non-linear increase with runoff depth, so a large mass of sediment and bound phosphorus are exported. It is possible that the model simulation represents an extremely rare occurrence. If the storm is omitted from the analysis, the percent increase drops to 58.4 percent for sediment and 32.0 percent for TP.

Figure 1. Percent Change in Site Measures (no BMPs) between Current and Projected Future Climate Conditions (GCM High).

Figure 2 shows unit area annual average post-treatment site export of runoff and pollutant mass for current and projected future climate conditions including the impacts of BMPs. Export rates of runoff and pollutant mass for the adaptation scenarios are not shown because performance with adapted BMPs is the same or better than current performance, and results do not provide insight into how changes in precipitation and temperature affect BMP function. The results were normalized to site area to facilitate comparison across sites and regions. Even though site practices generally remove more mass under projected future conditions, the overall site export rates of volume/mass increase under future conditions. The highest peak flows recorded during the simulation increase as well. The percent increase in export load is very high for some climate scenarios (e.g., sediment mass export for Mid-Atlantic GI plus Gray nearly triples), but the absolute increase (future minus current) is fairly stable. Note that by design, the Pacific Northwest and Southwest stormwater management requirements stipulate zero discharge for the 10-yr and 100-yr events respectively, so average annual export rates are very low for these locations.

Figure 2. Normalized Site-Export under Current and Projected Future Climate Conditions (GCM High) with Existing BMPs.

Stormwater management approaches notation: (a) Conventional, (b) GI with Gray, (c) GI Only

• 2.How can conventional designs and GI designs be adapted so that the stormwater infrastructure will provide current levels of performance under future conditions?

Increasing BMP size can mitigate the potential increase in flow, runoff volume, and pollutant loads, but result in increased costs. As an alternative, additional BMPs can be added to a site to address projected climate change impacts. Costs associated with current and future adapted stormwater management approaches are summarized in Table 2 for all combinations of sites, stormwater management approaches, and climate scenarios. Note that we do not present climate scenarios resulting in a reduction in all performance metrics, because they already meet the objectives (e.g., the minus 10 percent volume precipitation scenarios). Current cost of stormwater infrastructure reflects the 20-year present value of the capital cost and O&M for new development. Following the adaptation simulations under projected future climate conditions, the 20-year present value was recalculated for resized/adapted BMPs. Note that the adapted cost reflects new development cost (i.e., the cost for new construction of the adapted BMPs) rather than retrofit costs of changing BMP configurations on an already developed site. The reason for using the same basis for calculating costs is threefold; first, it allows the results to be more comparable; second, retrofit cost data tend to be highly variable and difficult to generalize from literature values; and third, when BMPs reach the end of their life, it is likely the replacement BMPs would conform to revised design standards that reflect adaptations to the future climate. The current cost was subtracted from the adapted total cost to obtain the adaptation cost increase. Both metrics were normalized to the contributing impervious area. Current costs are highly variable even when normalized, and range from $0.212 million to $1.28 million; the costs per impervious acre tend to increase with increasing percentage of impervious area. Adaptation cost increases are also highly variable, but do not follow the same pattern.

Table 2.

Projected Costs for Current and Future Adapted Stormwater Management Approaches

Location	Climate Scenario	Stormwater Management Approach	Current Cost (106 $/impervious acre)	Adaptation Cost Increase (106 $/ impervious acre)	Percent Increase in Cost
Southeast	GCM High	Conventional	0.768	0.495	64%
	GCM High	GI with Gray	1.282	0.162	13%
Arid Southwest	GCM High	Conventional	0.599	0.255	43%
	GCM High	GI Only	0.498	0.293	59%
Pacific Northwest	GCM High	GI Only	0.624	0.624	100%
Mid-Atlantic	GCM High	Conventional	0.408	0.497	122%
	Plus 10%			0.207	51%
	Plus 20%			0.374	92%
	GCM High	GI with Gray	0.396	0.538	136%
	Plus 10%			0.084	21%
	Plus 20%			0.164	41%
	GCM High	Conventional with Distributed GI	0.408	0.812	199%
	Plus 10%			0.125	30%
	Plus 20%			0.273	67%
Midwest	GCM Medium	Conventional	0.212	0.413	195%
	GCM High			0.524	248%
	Plus 10%			0.411	194%
	Plus 20%			0.766	362%
	GCM Medium	GI with Gray	0.341	0.171	50%
	GCM High			0.688	201%
	Plus 10%			0.222	65%
	Plus 20%			0.444	130%
	GCM Medium	GI Only	0.591	0.228	39%
	GCM High			0.551	93%
	Plus 10%			0.295	50%
	Plus 20%			0.598	101%
	GCM Medium	Conventional with Distributed GI	0.212	0.114	54%
	GCM High			0.355	168%
	Plus 10%			0.109	52%
	Plus 20%			0.191	90%

When the SUSTAIN future adaptation scenarios were simulated, the algorithm attempted to find a new set of BMP configurations (via resizing) that met all five of the current conditions performance objectives simultaneously. The result is that the performance improvements “overshot” some of the metrics while seeking to meet all metrics. When we reviewed the optimizations, it became clear that certain metrics were the most limiting (i.e., costliest to achieve). In many cases, only one metric was the limiting factor, while in other cases, multiple metrics were limiting factors. Meeting the flow duration curve metric was the limiting or co-limiting factor in over 80 percent of the optimization runs. This indicates that control of flood event runoff volume is generally the most difficult objective to meet when adapting site BMPs to projected future climate conditions. Practices that can address flood event volume control are therefore a critical component of adaptation to climate change, assuming there is a substantial increase in large storm event intensity. Matching the annual runoff volume was the limiting or co-limiting factor 40 percent of the time, indicating the challenging nature of either infiltrating or evaporating the projected increases in runoff volume. Of the three pollutants, sediment load was the most common limiting factor (20 percent of the scenarios).

• 3.What do the results suggest regarding the adaptation potential of gray and green infrastructure for increases in extreme precipitation events?

How do current stormwater infrastructure costs and additional costs for adaption compare between Conventional versus GI-based stormwater management scenarios? What does this say about various approaches to stormwater management? In Figure 3, we show stormwater infrastructure costs for meeting current regulatory requirements, along with the cost of adapting site BMPs to meet current performance metrics. The two costs (current cost and additional cost of adaptation) are indicated in different colors and stacked to provide total infrastructure cost. All stormwater management scenarios where current practices were resized to adapt to projected future conditions are shown, including results for the climate sensitivity analyses. In general, the original (current) cost of stormwater infrastructure using GI practices is more expensive on a per impervious acre basis than the equivalent scenario using only conventional practices. The cost for the GI Only scenario for the Arid Southwest site is less, but that may be due in part to limited cost data for representing an infiltration basin (used in the Conventional scenario) in an arid environment leading to an overestimation of the current cost. However, the cost of adaptation is frequently less for approaches using GI compared to the Conventional-only approaches. For the Arid Southwest, the GI Only adaptation cost is higher than for Conventional, but the net cost (current plus adaptation) is less for the GI-Only scenario than for the Conventional scenario. For the Southeast, the adaptation cost is much less for the GI plus Gray approach, though the combined cost is somewhat higher than for the Conventional cost. For the Mid-Atlantic, the adaptation and combined costs of GI plus Gray are lower than Conventional for the percent change future precipitation scenarios, but slightly higher for the downscaled GCM scenario using high storm event intensity change. For the Midwest, the GI plus Gray scenario has both the lower adaptation and combined cost (compared to Conventional and GI Only) for the percent change and medium intensity downscaled GCM scenarios, but the trend is not held for the high intensity downscaled GCM scenario. Combined costs are highest for the GI-only approach. These results lead to two generalizations:

Figure 3. Current Cost and BMP Cost Increase for Adaptation Scenarios Where BMP Footprints are Increased.

1. Approaches that use a combination of conventional and GI components tend to have greater cost effectiveness compared to approaches relying on only conventional or only GI practices – in other words, the increase in cost of maintaining current performance under projected future climate conditions is less than for conventional-only or GI-only approaches, which indicates combined conventional/GI approaches are better equipped for adaptation. This likely reflects the combined advantages of having practices that better address large flooding events (such as wet ponds and detention basins) with GI practices that provide most holistic treatment of volume and pollutants.

2. However, GI practices appear to be at a disadvantage in some cases when there is a large projected increase in the most extreme precipitation events. The adaptation optimization forced GI components (which tend to be more expensive on a unit basis) to be larger to provide sufficient volume control for the highest runoff events to meet the flow duration curve metric. This leads to more expensive approaches when GI is relied on than when conventional infrastructure counterparts are used.

We evaluated two different approaches to adapting a site under projected future climate conditions to address increase in performance measures for the Mid-Atlantic and the Midwest Conventional stormwater management scenarios – resizing currently implemented practices versus adding distributed GI to the site. Figure 4 shows the results using the same stacked-cost format as described above for Figure 3. The trends shown in the results are consistent with those seen in Figure 3. Adding distributed GI to a site to adapt to climate change is generally less expensive than resizing conventional practices – again, the approach that combines conventional and GI practices has the better cost effectiveness. However, this is not the case for the Mid-Atlantic site for the high intensity downscaled GCM climate scenario. In this case, so much additional volume control is needed that the higher cost of GI outstrips a simple resizing of the less expensive extended detention basin. For the Midwestern site, the distributed GI approach remains less expensive for the high intensity downscaled GCM climate scenario, but this is driven in part by the large footprint needed by the wet pond to provide sufficient evaporation to control the runoff volume increase. In addition, the GI adaptation cost is highest for the high intensity downscaled GCM.

Figure 4. Current Cost and BMP Adaptation Cost Increase Comparing Conventional with Resized BMPs to Conventional with Added Distributed BMPs.

To adapt stormwater BMPs to address potential future climate conditions, practices were either resized, or distributed GI was added to the sites. BMPs take up physical space, and the SUSTAIN optimizations provided future practice dimensions. This raises two questions: How do changes in BMP footprints (relative to the entire site) compare across scenarios and locations? And; Are the increases realistic? BMP footprints as a percent of overall site area (as shown in Table 1) are shown in Figure 5 for current and projected future adapted stormwater management scenarios. The importance of the change in BMP physical footprint depends largely on whether space is readily available for BMP expansion. While the figures show total practice area as a percent of the total site area, it is important to note where the increase in practice footprint is occupying pervious versus impervious areas. This is important because practices that increase coverage of impervious surfaces are simply replacing existing impervious surfaces (i.e., expansion of green roof area and increased use of permeable pavement) versus taking up current site pervious area, which was previously used as landscaping, open space, or even private yard area. Trends are most evident for the Mid-Atlantic and the Midwest where multiple climate scenarios were modeled. For the Mid-Atlantic, the biggest increase in footprint across the stormwater management approaches is always associated with the high intensity downscaled GCM climate scenario. This corresponds to the previous conclusion that large increases in the most intense rainfall leads to larger practice footprints to provide sufficient storm event volume control. This trend is less evident for the Midwest, where both the high intensity downscaled GCM and the plus 20 percent change climate scenarios show the largest footprint increase.

Figure 5. Percent Change in BMP Footprint from Current Conditions to Adapted for Projected Future Climate Conditions.

Some of the adaptation results are likely financially infeasible. For instance, the wet pond in the Midwest Conventional scenario must be tripled or even quadrupled in size to provide sufficient control of runoff volume. While there is technically sufficient pervious surface available for expansion, in practice this would be difficult if not impossible to achieve given the site layout. This suggests that distributed solutions are a better option when the alternative is to tear out roads and properties. However, if there is a large increase in high intensity storm events, as noted previously GI alone may not be suitable for peak flow volume control. In the Mid-Atlantic adaptation scenario using distributed GI, the infiltration trenches occupy nearly half of the available remaining pervious area on the site. Another factor to consider is how site design and stormwater management are typically conducted. Development often maximizes the site footprint to meet the goals of the project – to make a profit, or at least to utilize the available space. Stormwater management is generally minimized to just comply with regulation, and little if any thought is given to setting aside space for increasing resilience to climate change. As a result, if a site is to be adapted, other site uses (e.g., parking, amenities) may need to be converted to stormwater management. This approach is often impractical.

Conclusions

This research demonstrates a risk that projected changes in meteorological forcing will negatively affect BMP performance for both gray and green stormwater management approaches. If stormwater designs are adapted in the future, the most cost-effective approaches (in terms of BMP lifecycle costs) use both gray and green BMPs. However, if the magnitude of extreme precipitation events increases dramatically, then gray practices that provide detention storage have better cost effectiveness. Incorporating risk of future impacts into stormwater design may help communities become more resilient.

Lessons Learned

Control of flooding events is a requirement used throughout the United States. The most difficult performance measure to mitigate projected future climate conditions was usually control of large flooding event outflows as represented by the flow duration curve metric. This study suggests that currently built practices will need greater temporary volume storage and/or reconfiguration of outlet structures to mitigate flooding and channel erosion risk in locations where the magnitude of extreme events is projected to increase. Stormwater requirements will likely need to be adapted in the future to address projected increases in precipitation. GI practices that rely on treatment without volume storage will be at a disadvantage for climate change adaptation while approaches that rely on adaptation of conventional practices only may not have the flexibility to address multiple performance objectives. For instance, the conventional practices for the ultra-urban Southeast site could not be adapted to address runoff volume increase or fully mitigate the increase in TN load. Likewise, the stormwater wet pond for the Midwest site provided poor annual volume reduction and was resized excessively in the adaptation scenarios to address annual runoff increases.

Conventional stormwater management approaches tended to be more cost-effective than their GI counterparts under current climate conditions. However, when climate scenarios with smaller increases in large storm event intensities are considered, the additional cost of adapting sites using GI approaches tended to be less than adapting conventional only approaches. Overall, approaches to stormwater management that combined both conventional and GI elements tended to have the best combined cost effectiveness. This was further reflected in the stormwater management scenarios that added distributed GI to a conventional approach site versus resizing the conventional practices. Again, the combination of conventional and GI practices had better cost effectiveness; however, the trend did not hold up for many future climate scenarios with the high projected changes in intensities for large storm events. In these cases, GI was at a disadvantage for providing temporary detention storage needed to mitigate flooding risk.

Recommendations

Projections of future seasonal average increases in air temperature are relatively consistent between various climate models, but changes in precipitation regime are much more uncertain and often location-specific. There would be a cost regret if practices were dramatically up-sized in anticipation of extreme precipitation events that did not actually occur. GI may have an advantage in flexibility, because it typically has a shorter design life before rehabilitation is required – so it would be possible to commit less in the present and use a more incremental approach as projected climate conditions evolve.

An important issue to consider is the flexibility of different types of practices, for both gray and green infrastructure. On an already-developed site, it will likely be difficult to increase the area or types of practices, especially if all the developable area is occupied. Adding dispersed GI may be considerably easier at a later date than resizing hard structures. However, it may be possible to use the existing footprint of BMPs and excavate them to provide more storage and treatment – an option that is not explored in this study. This is less likely to be a viable alternative on sites with a low elevation gradient. Another option is to build flexibility into site design, setting aside space for potential future BMP addition and/or expansion. Regardless of how it is addressed, flexibility is a key factor to consider in adaptation and resilience planning, especially when changing climate conditions may result in hydrologic changes downstream of development sites. Analysis of this complexity could then lead to changes in policy and management decisions.

Practitioner Points.

• There is a risk that projected changes in meteorological forcing will negatively affect stormwater BMP performance.

• Under projected future climate conditions, this study suggests the most cost-effective approaches may use both gray and green BMPs.

• If the magnitude of extreme weather events increases dramatically, gray practices that provide detention storage may have better cost effectiveness.

• Flexibility is beneficial in adaptation and resilience planning due to uncertainty in projected precipitation volume and intensity changes.