Green Infrastructure Benefits for Floodplain Management: A Case Study
Green infrastructure (GI), also known as low-impact development (LID), is a stormwater management approach that emphasizes minimization of runoff through a combination of sound site planning principles and concepts that encourage disconnection of impervious areas from waterways, infiltration, and evapotranspiration. A basic tenet of GI is to control small storms, which typically comprise upward of 80% of the average annual rainfall. Runoff from these storms is diverted to pervious areas or collected in micro-controls distributed throughout a site. It has been surmised that these controls have minimal impact on extreme events that result in flooding; therefore GI has not been deemed capable of reducing flood losses. This article illustrates that GI can indeed provide substantial benefits regarding flood protection.
Performance of GI Controls During Large Storms
Once site planning principles of GI such as runoff minimization and imperviousness disconnection have been implemented, the resulting runoff is stored in relatively small controls dispersed throughout the site and designed to capture the runoff as close as possible to its source. Bioretention filters, pervious pavement, green roofs, and cisterns are designed to store a given volume of water that corresponds to a threshold rainfall depth, such that a large fraction of the annual rainfall volume would be controlled if that threshold is captured and returned to natural hydrologic pathways, mostly evapotranspiration and shallow groundwater flow (interflow) that subsequently becomes baseflow in the receiving streams. The threshold varies depending on the region and the desired fraction of the annual rainfall to be captured. For an 80 to 85% capture goal, the threshold is around 1 inch in the mid-Atlantic and places in the midwestern United States; in southeastern locations it is around 1.2 inches; and in semi-arid climates, it is around half an inch.
Large storms that can produce flood conditions easily exceed these thresholds. For example, in Columbia, SC, the two-year, 24-hour storm has a depth of 3.6 inches, and the 100-year event is 8.4 inches, according to the National Oceanic and Atmospheric Administration’s Atlas 14. In Delta, UT, the two-year, 24-hour storm is 0.8 inch and the 100-year storm is 2 inches. Therefore, it is reasonable to assume that capturing a volume of 0.5 to 1.2 inches is not going to make much of a difference when a severe storm occurs.
The effects of GI are best evaluated using a case study–in this case a watershed in the southeastern United States, shown in Figure 1. The total area is 13.2 square miles, and the watershed drains to an 8-mile-long creek and its tributaries. The watershed is heavily urbanized, resulting in an imperviousness of 39%.
The distribution of soil types in the watershed according to their infiltration capacity is shown in Figure 2; A soils are the most pervious and D the least. B soils comprise 91% of the watershed; C soils account for 7% of the area, followed by 2% for D soils and 0.1% for A soils. Therefore, the watershed has good overall infiltration capacity, except that a large fraction is covered with pavement.
The annual rainfall depth in this area is about 50 inches. Table 1 shows the depths for several 24-hour storms of various frequencies.
The watershed was developed before requirements for stormwater management; therefore, most of the runoff reaches the stream network unabated. This exercise will compare the hydrologic response and flood elevations caused by the current conditions with simulated conditions in which GI has been implemented in the entire watershed to manage stormwater resulting from development. The assumption for this analysis is that 1.2 inches of runoff will be captured by various GI controls. Runoff in excess of this design capacity will bypass the controls. Captured runoff is evapotranspired, infiltrated into the native soil to replenish shallow aquifers, or, if the soil’s infiltration capacity is limited, released slowly to the existing storm drain system. In either case, the goal is to reestablish the effects of interflow to restore baseflow in the streams.
Modeling the Effects of GI
The effects of GI deployment were investigated using the Hydrologic Modeling System (HMS) software and the River Analysis System (RAS) hydraulic modeling software developed by the Hydrologic Engineering Center (HEC) of the US Army Corps of Engineers.
For hydrologic modeling, the Natural Resource Conservation Service’s curve number (CN) methodology was used to simulate rainfall-runoff processes. The watershed was divided into 75 subwatersheds, each of which was assigned a CN depending on land use and soil type. The CN concept was developed to predict peak flows during large storms. When present, stormwater management controls are simulated as individual storage facilities draining large tracts of land. The methodology was not designed to account for numerous small controls distributed throughout the landscape as is the case with GI; therefore, the procedure was modified to simulate this condition. Since the inception of GI in the early 1990s, several approaches have been proposed to adapt the CN concept to GI; this study uses a variant of the one proposed by Prince George’s County, MD (1999). This approach entails calculating the runoff volume generated with the current condition’s CN; subtracting the GI design capture volume, in this case 1.2 inches; and back-calculating the CN corresponding to this reduced volume of runoff. Because runoff volume is a function of rainfall depth, the modified CNs vary depending on storm frequency. Both sets of CNs were entered in the watershed’s HEC-HMS model to estimate the corresponding peak flows for the storms in Table 1.
The modified CNs are lower than the current condition’s CN, which translates into reduced peak flows. An additional scenario was evaluated using a forested condition to compare with an undeveloped watershed. Table 2 shows comparisons of the results using original CNs and the modified values after deployment of GI for three sample subwatersheds. It should be noted that the impact of the GI capture volume in reducing the CN value is relatively less for the more severe storms.
A HEC-RAS model of the main stream was developed using topographic data and the geometry of bridge and culvert crossings. Peak flows resulting from the hydrologic model for the various scenarios were entered to calculate the water surface elevations for the storm events in Table 1. These elevations were then intersected with the terrain to estimate inundation areas for each storm. Figure 3 compares the floodplains for the two-year and 100-year events. Figure 4 shows the variation of the inundated area for each scenario.
Figures 3 and 4 show that the effect on GI is negligible on the 100-year event but quite noticeable for the two-year storm. Figure 3 shows that the two-year condition with GI comes close to mimicking the forested watershed but is far from replicating it. Moreover, Figure 3 reveals that the extent of flooding increases dramatically with the presence of urbanization, regardless of whether, or how, stormwater is managed. In other words, GI can improve the situation but cannot completely mitigate it.
Imperviousness Effects
As noted earlier, this watershed possesses good infiltration capacity because of the predominant B-type soils. To evaluate the impact of GI on a less pervious watershed, the CNs were modified by changing B and C soils to D soils, and A soils to C soils. The analysis described above was replicated with the new soil types, and the resulting floodplains are shown in Figure 5. Figure 6 summarizes the areal extent of flooding for these cases.
Figures 5 and 6 show that GI has less of an effect on both the two-year and 100-year events compared to the analysis involving more pervious soils. In general, GI controls are less able to infiltrate water in tighter soils, but in this example it is assumed that GI controls are designed to remove the same 1.2 inches of runoff as in the case of the more pervious soils. For this reason, the GI controls must be larger than in the case of more pervious soils to maintain the same performance. Therefore, the reason there is less of a reduction in the extent of the floodplains is not due to reduced performance of the GI controls but because the watershed simply generates more runoff, and the fraction handled by the GI controls is proportionately less in the impervious watershed than in it was in the original case with more pervious soils. On the other hand, impervious soils are closer to pavement than pervious ones; therefore, development causes a smaller relative impact in runoff generation in this case. The proximity of the curves in Figure 6 illustrates this effect. In this case, GI has proportionately better success in closing the gap between the developed condition and the forest condition.
Impacts on Flooding Risk
The results of the analyses indicate that GI does not appear to have a significant impact in reducing the extent of the 100-year floodplain, which is the benchmark used by the Federal Emergency Management Agency (FEMA) for flood insurance purposes. However, the reduction in the extent of flooding associated with GI implementation for less severe but nonetheless flood-inducing events has significant impacts on overall flood risk exposure. This outcome is evaluated in this article through the application of FEMA’s Hazus software.
Built on a geographic information system (GIS) platform, Hazus is a public domain application used by FEMA and other emergency management organizations to estimate potential losses associated with natural disasters. The methodology in Hazus enables estimates of physical, economic, and social impacts of earthquakes, hurricanes, and floods. Physical damages include the cost to replace or repair residential and commercial buildings, critical facilities, and infrastructure; economic losses include lost jobs and business interruptions; and examples of social impacts include shelters needed, displaced households, and debris amounts. This information is used as the basis for hazard mitigation, disaster response and recovery, and community preparedness.
In the case of flooding, given the flood depths in a floodplain associated with a specified return period, Hazus uses a series of damage curves and built-in or user-supplied databases of assets at risk categorized by census blocks to quantify these impacts as economic damages. For example, the Federal Insurance Administration has developed damage curves that relate the depth of flooding at a house with the percent damage to the structure and contents. For a given return period, Hazus uses the location of the house to estimate the flood depth and the corresponding damage curve and replacement value of the house to estimate the damages. Hazus adds similarly calculated damages for all assets in the floodplain to arrive at the total economic losses caused by the flood. Additional information on Hazus can be found in www.fema.gov.
The scenarios described earlier were analyzed with Hazus to estimate the losses associated with each flooding event. The exposed assets and their value were taken from the built-in databases in Hazus. Figure 7 summarizes the resulting building damages for each of the storms in Table 1. The horizontal axis is the annual probability of exceedence, which is the reciprocal of the return period in Table 1. For example, under existing conditions, the five-year event–which has a probability of exceedence of 1/5, or 20%–causes $27 million in damages. In developing this curve, it was assumed that the one-year flood is fully contained in the stream channel in all scenarios and causes no damages. Only the main stem of the creek was analyzed; therefore, the damage estimate is conservative.
Figure 7 shows that watershed-wide implementation of GI clearly reduces flood damages for both the original case and the case involving less pervious soils. This effect can be summarized by computing the average annualized losses (AAL), which are the summation of damages for each individual event multiplied by the probability of occurrence, which mathematically is equal to the area under each of the curves in Figure 7. Computation of these areas results in damages of $13 million for existing conditions and $8 million if GI had been implemented. Therefore, in the long run, GI reduces flood damages by an average of $5 million a year, almost a 40% savings. Because these avoided damages are for the test watershed only, the analysis does not include the avoided losses associated with other watersheds downstream that will benefit from the reduced volume of incoming water. Figure 7 also shows the results for the watershed with low-infiltration soils. In this case, the AAL is $17 million without GI and $10 million with GI; therefore, the average avoided damages are $7 million a year. As expected, both values are greater than in the case of more pervious soils, but the difference of $7 million between the with-GI and without-GI scenarios is still about 40% of the damages without GI. The AAL results are summarized in Figure 8.
Another useful way to look at these benefits is to calculate the dollars of avoided damages per unit area of floodplain, as shown in Figure 9. This estimation assumes that the damages are homogeneously distributed across the floodplain, which is consistent with how the Census data are published. In reality, damages tend to be more severe close to the source of flooding depending on the value of the assets at risk and their location in the watershed. The values found for the 100-year event are comparable to those derived by Johnston, Braden, and Price (2006) for a watershed in Illinois, where practices that encourage onsite storage of stormwater were shown to reduce the losses by an amount between $6,700 and $9,700 per acre. As expected, Figure 9 shows that GI is more effective at avoiding losses for the more frequent flooding events. An interesting observation is that there appears to be a maximum for the unit avoided losses. The losses are assumed to be zero for the one-year event, after which the losses avoided per unit area increase and then decrease for the more severe events. The figure also shows that around the maximum the impervious soils are associated with greater values than for the pervious soils, but the opposite relationship occurs for the more severe events.
Impact on Cost-Effectiveness of GI Implementation
In a watershed already built out, GI is implemented by retrofitting existing impervious areas in an effort to improve water quality, reduce stream erosion, and protect aquatic habitat. For example, a bioretention facility can be installed in a parking lot to capture runoff from paved areas, or a green roof can be installed on an existing building to capture rain that falls on the roof. Planning-level cost estimates for these types of retrofits in the mid-Atlantic are around $100,000 per impervious acre. In new development, the cost is substantially less because it can be folded into a site’s landscaping plan. However, utility relocation increases the cost in urban redevelopment projects.
The 13.2-square-mile watershed in this case study is 39% impervious, which means that it comprises about 3,300 impervious acres. Retrofitting this entire area using GI would cost $330 million. Determining the cost-effectiveness of this type of watershed-wide GI implementation, if it is done, typically involves assessing such factors as the value of environmental and societal assets, property values, savings in stream restoration efforts, reduced stormwater infrastructure costs, and avoided penalties for noncompliance with regulations. Avoided flood losses are seldom included. However, assuming avoided damages between $5 million and $7 million a year, and an annual discount rate of 4% over a 30-year period, the present value of damages avoided can add up to about 20% of the implementation cost, assuming that annual operation and maintenance expenditures for GI controls amount to 3% of the construction cost. It is essential to recall here that the modeling and AAL analyses for this study accounted only for the main stem of the creek; there are additional losses avoided associated with the eight tributaries shown in Figure 1 that were not included. Therefore, the avoided damages are a conservative estimate. Overall, GI can have a measurable effect on flood-loss reduction that needs to be considered in an economic analysis.
Conclusion
In addition to its known benefits pertaining to water quality and channel and habitat protection, GI has a positive impact in reducing exposure to flood hazards. The analysis for the case study above suggests that GI has a small impact in reducing the extent of the 100-year floodplain, which is the regulatory instrument for flood insurance purposes. However, when all of the flood-inducing events are considered, GI can reduce average economic losses in the long run. To be effective, implementation of GI needs to take place on a watershed-wide basis. This positive effect is more apparent in watersheds with pervious soils.
GI must be deployed as part of a holistic approach to watershed management that affords benefits in water quality, channel stability, reduced flood risk, ecosystem integrity, natural resource protection, and recreation and aesthetic benefits to the public. Based on this analysis, floodplain management agencies would appear to benefit from encouraging the application of GI techniques for new development and redevelopment. Given the economic benefits measured as damages avoided, actuaries should consider GI implementation as a factor that could potentially reduce flood insurance premiums.
Acknowledgements
The authors wish to thank Amelia Bergbreiter, P.E., CFM of Atkins North America Inc. and Lisa Hair, P.E., from EPA, Office of Water, for their constructive commentary on this article.
References
Johnston, D. M., J. B. Braden, and T. H. Price. 2006. “Downstream Economic Benefi ts of Conservation Development.” ASCE Journal of Water Resources Planning and Management 132(1).
Prince George’s County. 1999. Low-Impact Development Design Strategies An Integrated Design Approach. Largo, MD: Prince George’s County, Department of Environmental Resource Programs.