Estimation of load capture ratio for evaluating LID facilities performance in Korea

As the population moves and increases, the expansion of urban areas is proceeding rapidly. Urbanization has a negative impact on the environment in urban areas, creating impervious areas and reducing the natural pervious area (Harbor 1994; Moscrip & Montgomery 1997; U.S.G.S. 1999; Ahiablame et al. 2012). The expansion of the impervious area distorts the natural hydrological cycle and increases the incidence of non-point source pollutants including nutrients and suspended solids (Lee et al. 2010; Kayhanian et al. 2012; Barbosa et al. 2012; Davis & McCuen 2015).

Various policies and techniques such as LID (Low Impact Development) in the United States, Distributed Urban Design (DUD) in Germany, and WSUD (Water Sensitive Urban Design) in Australia have been introduced to minimize these negative effects (Kim et al. 2013). LID is a technique that integrates urban development and environmental issues with a focus on water resources and water quality in urban areas (Davis 2005; Hunt et al. 2010). LID facilities aim to strengthen the control and purification function of stormwater by activating the infiltration of soil through expansion of the permeable area, and ultimately to establish a healthy urban hydrological cycle system (Coffman 2002; Han 2011).

In order to efficiently design LID facilities, accurate performance evaluation of LID facilities is required first. In principle, the performance of LID facilities to be installed through modelling at the design stage should be evaluated appropriately for the situation in the field (NCR 2006). However, similarly to other countries, most of the stormwater monitoring data for modelling a specific site is not available in Korea. In addition, due to practical problems such as the lack of expertise of individual engineering companies, the performance of LID facilities has been evaluated using the design manual (NIER 2014) provided by the Ministry of Environment of Korea.

The method proposed by NIER (2014) (hereinafter referred to as NIER 2014) is designed for non-point reduction facilities (NPRFs). However, since this method evaluates the performance of NPRFs using the standard design rainfall depth irrespective of the characteristics of the facilities, the performance of all NPRFs is calculated equally when the standard design rainfall depth is the same. In other words, regardless of the individual characteristics of the facility, only the storage capacity can influence the performance of the facility.

This design concept is appropriate for traditional NPRFs where the facility is composed of a relatively large reservoir space, which collects and processes large amounts of stormwater and non-point pollutant loads in large areas. However, although LID facilities are often classified as NPRFs of natural type (Elliott et al. 2009), LID facilities are structurally different from traditional NPRFs. LID facilities are distributed in many locations across small areas and control stormwater using voids present in engineered soil or gravel. Therefore, applying NIER (2014) to LID facilities is not appropriate. In addition, since LID facilities have a wide variety of structural characteristics in each facility, it is necessary to derive a new performance evaluation formula considering the characteristics of individual LID facilities.

This study is a follow-up study to that of Choi et al. (2019), which proposed new Stormwater Interception Ratio (SIR) formulae using Ratio of Facility area of LID facilities to Drainage area (RFD). Based on Choi et al. (2019), new Load Capture Ratio (LCR) formulae for three types of LID facilities including bio-retentions, infiltration trenches, and vegetative swales are proposed in this study. Long-term stormwater depths and TP loads are simulated by constructing the US Environmental Protection Agency's Storm Water Management Model (EPA SWMM) from the stormwater depth and total phosphorus (TP) monitoring data of the study area. The TP-related parameters are additionally calibrated, along with improvements in stormwater depth-related parameters of the EPA SWMM reported in Choi et al. (2019). Numerical experiments are then carried out under various conditions to calculate the LCR formula, and a new LCR formula suitable for each LID facility is proposed. The procedure performed in this study is shown in Figure 1.

Currently, the performance evaluation method for LID facilities applied in Korea adopts NIER (2014), which is a performance evaluation method for relatively large NPRFs. This method is performed by the procedure shown in Figure 2. First, the standard design rainfall depth (in other words, design capacity) is used to calculate the RIR. The LCR is then calculated using the estimated RIR. Then, LCR is multiplied by the pollution loads originated from the drainage area, calculated by using the pollutant unit method, and the Load Capture Amount (LCA) is calculated. Finally, by multiplying the LCA by the reduction efficiency, ⁠, of the LID facility, the Load Reduction Amount (LRA), which is the load that is treated by the LID facility, is calculated.

In this study, LCR suitable for each LID facility are estimated using EPA SWMM. Some areas of Noksan National Industrial Complex in Busan Metropolitan City in the south-eastern part of the Korean peninsula, where observational rainfall data, stormwater data, and TP loads data exist, were selected as the study area. The area is a typical urban impervious area of approximately 13,000 m² consisting of paved surfaces and steel structure buildings. When modeling the study area using EPA SWMM, the study area was divided into a building area and a non-building area for more accurate parameter estimation. Sub-watershed 1 (S1, 4,212 m²) and sub-watershed 3 (S3, 4,069 m²) are buildings and sub-watershed 2 (S2, 4,719 m²) is the non-building area. The location of the study area and the sub-watershed map of the EPA SWMM are shown in Figure 3.

In order to estimate LCR using EPA SWMM, it is necessary to accurately reproduce the hydrological and water quality characteristics of the study area. In this study, the main EPA SWMM parameters to be calibrated were selected by referring to US EPA (2015, 2016a, 2016b). The parameters selected for the calibration of the stormwater depth are the sub-watershed width (W), depression storage depth (d_s), and %Zero, which represents the ratio of the area where d_s is zero in the sub-watershed. The parameters selected for calibration of the water quality are the maximum accumulation amount (B_max) and build-up rate constant (K_B) for the exponential growth curve associated with the TP build-up and Event Mean Concentration (EMC) associated with TP wash-off.

Parameters related to stormwater depth were estimated first, and parameters associated with TP loads were estimated with the estimated parameters associated with stormwater depth fixed. For 20 rainfall events from April 2009 to July 2012, stormwater depth and TP loads monitoring data, which were directly observed at the end of the study area, were used for parameter estimation.

The EPA SWMM provides LID modules to simulate urban catchment stormwater and water quality responses to the installation of LID facilities. Among the LID facilities available for simulation of the EPA SWMM, bio-retention cells, infiltration trenches, and vegetative swales were used in this study. It was assumed that the standard LID facility was installed at the exit of S2 (see Figure 3) among the sub-watersheds in the study area. When one LID facility was installed, the stormwater and TP loads were simulated and the SIR and LCR of each facility were estimated. The standard parameters for simulating the performance of the LID facility are given in Tables 1–3, as used in Choi et al. (2019).

In Choi et al. (2019), a number of documents (US EPA 1999; KECO 2009; Palhegyi 2010; CVC 2012; DOEE 2013; VWRRC 2013; US EPA 2015; DER 2002) were examined to determine the parameters of the EPA SWMM to construct a standard bio-retention facility. In general, bio-retention systems have a drainage area in the range of 500 –4,000 ⁠, with 4% of the drainage area as the facility area, with a surface layer depth of 300 mm, a soil layer depth of 600 mm, and a storage layer depth of 300 mm. Therefore, it was assumed that a standard bio-retention LID facility with a drainage area of 2,359.5 m² would be installed at the exit of the S2 (see Figure 3) sub-watersheds in the study area.

Also, since LID facilities treat stormwater and non-point pollutant sources mainly by using the infiltration process, the drainage capacity of the original native soil below the facility plays an important role in the performance of the facility. Therefore, bio-retention is generally placed on soil of group-A or group-B with high drainage capacity among the hydrologic soil groups proposed by the US Natural Resources Conservation Service (ADEM 2007). The purpose of this study is to present a performance evaluation formula as a design guideline, so it is necessary to estimate the performance of the facility conservatively. Therefore, the most conservative drainage capacity (i.e., 4 mm/hour) in the soil of group-B was applied.

Parameters for the design of standard infiltration trenches were determined (see Table 2) by reference to various documents (US EPA 1999, 2015; WMS 2008; Palhegyi 2010; MC 2011; RC 2011; VWRRC 2013; Tao et al. 2017). Infiltration trenches are typically installed in a drainage area in the range of 200–8,000 ⁠, with 3% of the drainage area being set as the facility area. In this study, it was assumed that 2,000 of drainage area is set in the sub-watershed S2 and an infiltration trench of 60 m² is installed at the sub-watershed exit. The drainage capacity of the original native soil was set to have a value of 4 mm/hour, the same as the bio-retention.

The parameters for the standard vegetative swale of the EPA SWMM were set using several references (ASCE 2001; DEP 2006; KECO 2009; CVC 2012; SOWP 2016; TDEC 2015; US EPA 2015, 2016a, 2016b; VWRRC 2013). The facility area of the vegetative swale is recommended to be installed at 10% of the drainage area, and drainage area is recommended to be in the range of 4,000 m²–8,000 m². Therefore, the drainage area of the vegetative swale was set to have a value of 4,719 ⁠, which is the total area of the sub-watershed S2, and the vegetative swale was assumed to be installed at a facility area of 471.9 ⁠.

Unlike bio-retentions and infiltration trenches, vegetative swales need to be installed on original native soil that has a higher drainage capacity than other types of LID facilities since there is no storage layer in vegetative swales. This means that the vegetative swale needs to be installed on soil in group-A of the NRCS. Therefore, drainage capacity of the original native soil of the vegetative swales was set to 10 mm/hr.

In order to estimate the SIR before estimating the LCR, the method proposed by Choi et al. (2019) was applied in this study. Choi et al. (2019) proposed the SIR estimation formula using RFD to improve the disadvantages of the RIR estimation formula based on the standard design rainfall depth. Therefore, this study simulated SIR and LCR by applying various RFDs to the standard LID facility. At this time, considering the standard RFD, the range of applied RFDs was set to 0.004–1.0 for bio-retention and infiltration trench and to 0.03–1.0 for vegetative swale.

In this study, parameters were automatically optimized using the EPA SWMM-MATLAB linkage module. The parameters used for calibration of the stormwater depth were W, d_s, and %Zero for each sub-watershed. The watershed width W was estimated in proportion to the area of each sub-watershed, and d_s and %Zero were estimated to have the same value in all sub-watersheds. Table 4 shows the final parameters estimated through the parameter optimization process.

The simulated stormwater depth (mm) and the observed stormwater depth (mm) are shown in Figure 5. The coefficient of determination and the Nash-Sutcliffe model Efficiency coefficient were 0.98 and 0.94, respectively.

The parameters for implementing TP loads in the study area were parameters B_max and K_B related to the build-up, and parameter EMC related to the wash-off. The parameters of sub-watersheds S1 and S3 in the building area were set to the same value, and those of sub-watershed S2 were set to estimate a separate value. The parameters related to the TP loads estimated through parameter optimization are shown in Table 5.

The simulated TP loads (kg) and the observed TP loads (mm) are shown in Figure 6. The coefficient of determination and the Nash-Sutcliffe model Efficiency coefficient were 0.81 and 0.75, respectively.

Since the model had been reported to reflect the observed values when the coefficient of determination is over 0.5 and the Nash-Sutcliffe model Efficiency coefficient is above 0.4 (Chung et al. 1999; Green et al. 2006), optimized parameters using EPA SWMM and MATLAB linkage module were considered to reproduce the stormwater depth and TP loads in the study area relatively well.

SIRs and LCRs were estimated through long-term stormwater depth and TP loads simulation, considering regional rainfall characteristics using the hourly rainfall data from 2006 to 2016 of 61 rainfall observation sites operated by KMA. As a result, the regional rainfall characteristics had a great influence on the SIR estimation as shown in Choi et al. (2019) (see Figure 7).

As the regional rainfall characteristics affected the SIR of the LID facility, the LCR was also influenced by the local rainfall characteristics (see Figure 8). In particular, the LCR of the standard vegetative swales with RFD of 0.10 ranged from 0.371 (at Geoje site) to 0.469 (at Daegwallyeong site), depending on where it was installed, with a deviation of about 24%. Standard bio-retentions (RFD = 0.04) and infiltration trenches (RFD = 0.03) also showed variations of 15% and 16% based on the average LCR of each facility. Therefore, local rainfall characteristics need to be considered in order to more accurately evaluate the performance of the LID facility.

However, the final goal of this study is to present a design formula for guideline use that can be used uniformly throughout the country. Therefore, to minimize the performance error of the LID facility due to regional differences in rainfall characteristics, the national average SIR and LCR were calculated using simulated SIRs and LCRs reflecting regional rainfall characteristics. For reference, in Figures 7 and 8, the red dotted line represents the error range of the SIR or LCR due to the rainfall characteristics, and the dotted black line represents the national average SIR and LCR.

So far, the LCR of each facility was estimated based on the standard LID facility. However, if the design variables of the LID facility are different from the standard values, it will be a practical matter to see how the LCR responds. As shown in Table 6, the LCR was calculated by changing the depth of each layer by ±50 mm and ±100 mm based on the standard values of design variables. Rainfall data from nine major cities including Busan, Daegu, Daejeon, Gangneung, Gwangju, Incheon, Jeju, Seoul, and Ulsan were applied. The RFD of each facility was fixed to the standard value.

As can be seen in Figure 9, the sensitivity of the design variables to LCR was found to be insignificant. For reference, each solid line in Figure 9 represents the LCRs calculated by applying rainfall data in each region, and the red dotted line represents the national average LCR calculated by averaging them. The depth of the surface layer of bio-retention was the most sensitive, and when the depth of the surface layer increased by 100 mm, the LCR showed a rate of change of 7%. In other words, when the LID facility is designed with an increase of 200 mm from the standard value of the surface layer (i.e. the surface layer depth is 500 mm), the LCR is increased by about 14%. This is similar to the LCR sensitivity due to regional rainfall characteristics, but it is impractical to design the bio-retention to have a surface layer depth of 500 mm. Therefore, considering practical coverage, the effect of changes in the design variables of bio-retention on LCR will be much smaller than that shown in Figure 9.

Since the fundamental purpose of this study is to derive the LCR formula for the design guideline, the performance of the LID facility needs to be treated most conservatively. Therefore, the drainage capacity of the original native soil where the LID facility was located was set to the most conservative value. However, since the drainage capacity of the original native soils may have a significant impact on the performance of the LID facility, this section analysed the sensitivity of drainage capacity of the original native soil to LCR. Numerous numerical experiments were performed after the infiltration capacity of the original native soils was set within the range of the minimum infiltration capacity (4 mm/hour) of the NRCS group-B soil to the maximum infiltration capacity (12 mm/hour) of the group-A soil. For the numerical experiments, the rainfall data of nine major regions including Busan, Daegu, Daejeon, Gangneung, Gwangju, Incheon, Jeju, Seoul, and Ulsan were applied, and all design variables except the drainage capacity of the original native soils were fixed to standard values.

As can be seen in Figure 10, the higher the drainage capacity of the original native soils, the higher the LCR of the LID facility, but the smaller the LCR difference due to regional rainfall characteristics. The difference between LCR by conservative standard values (4 mm/hour for bio-retentions and infiltration trenches, 10 mm/hour for vegetative swales) and LCR by applying the largest drainage capacity was less than 10%. However, in case of vegetative swales, if the facility was installed on group-B soil, a performance loss of more than 30% could be expected. Therefore, vegetative swales should be only installed on group-A soil.

Looking at the bio-retentions, it could be seen that NIER (2014) overestimated the performance of a facility in a relatively small facility, and underestimated the performance of a facility when a large capacity facility was installed. The LCR of NIER (2014) for the standard design rainfall depth of 27.4 mm corresponding to the standard RFD of bio-retention was 0.716, and the LCR proposed in this study was not as large as 0.736. Looking at the infiltration trenches, it could be seen that the LCR of NIER (2014) underestimated the performance of the facility in the range of all standard design rainfall depths. The LCR of this study with standard RFD was 0.672, while the NIER (2014) had only 0.531 LCR. Conversely, vegetative swales showed that the NIER (2014) LCR overestimates the performance of the facility over the range of all standard design rainfall depths. The LCR of this study with standard RFD was 0.416, while the LCR of NIER (2014) was estimated to be 0.683.

These results indicate that the performance of the LID facility may be distorted when the method presented by NIER (2014) is used. In addition, as shown in Figure 11, the LCR estimation method proposed in this study calculates various LCR values depending on the type of facility, even though the same design capacity (i.e. standard design rainfall depth) is applied. Therefore, it can be seen that the performance evaluation formula considering the characteristics of the facility is needed rather than applying the LCR regardless of the type of facility.

The current NIER (2014) LCR formula used to evaluate the performance of LID facilities is not adequate to evaluate the performance of LID facilities. Therefore, in this study, the regression coefficient required for the LCR empirical formula is newly proposed considering the characteristics of the LID facility. A comparison of the proposed LCR empirical formula with the LCR formula for NPRFs used in NIER (2014) suggests that the NIER (2014) method may distort the interception performance of non-point pollutant sources loads in the LID facility. Therefore, if the LCR empirical formula proposed in this study is used, it is expected that the performance evaluation of the LID facility consistent with the actual phenomenon will be possible.

This study aimed to derive LCR empirical formulas for design guideline use. Therefore, a single formula that can be applied to all countries has been proposed. To minimize the error due to the characteristics of regional rainfall during the performance evaluation of the LID facility, the national average LCR empirical formula was obtained by averaging the simulated results from the hourly rainfall data of 61 rainfall observation sites across the country. However, since the regional rainfall characteristics play a very important role in the performance of the LID facility, it is necessary to derive the performance evaluation formula of the LID facility considering the area or event rainfall depth.

In addition, the effects of the LID facility design variables on the performance evaluation of the LID facility were analyzed, but it was found that there was no significant error using the proposed LCR empirical formula in practical application. The effect of the original native soil drainage capacity of the LID facility on the performance of the LID facility was also analyzed, but this also showed no significant problem in applying the proposed LCR empirical formula. However, in the case of vegetative swales, more stringent site conditions were required than for bio-retentions and infiltration trenches.

It is still not possible to verify the proposed method using observed values before and after the installation of LID facilities because there is not enough systematic monitoring data on the performance of the LID facility. Therefore, it will be necessary to verify the validity of the proposed method by ensuring long-term monitoring data.

This work was supported by Korea Environment Industry & Technology Institute (KEITI) through Public Technology Program based on Environmental Policy Project, funded by Korea Ministry of Environment (MOE) (2016000200002).

All relevant data are included in the paper or its Supplementary Information.