Site Description

Cherokee Park stormwater pond is a 1.06-hectare stormwater retention pond in Madison, Wisconsin, with a permanent pool of approximately 1.3 meters near the center (fig. 1). The pond is underlined with an impermeable clay layer preventing exfiltration to underlying native soils. The pond was constructed in 2013 following guidelines outlined by the WDNR (2024) and was approximately 7 years old at the time of the study. Removal of accumulated sediment as part of normal maintenance was not done prior to the study. Runoff from an 89-hectare medium-density residential catchment drained to the pond through a 91-centimeter (cm) concrete pipe and a 107- by 213-cm rectangular, concrete culvert. Water levels in the pond were controlled at the effluent by a sharp-crested weir located just upstream from a 91-cm concrete pipe. During extended dry periods, water levels in the pond would decrease because of evaporation, thereby temporarily increasing volumetric storage capacity.

Estimates of overhead street tree canopy were made using a combination of aerial imagery coupled with geographic information system software and field surveys. Street trees within the catchment covered 30 percent of the right-of-way (area between the outer edge of sidewalks on opposing sides of the street) and were generally a mix of mature, deciduous hardwood species. The majority of tree species were characterized as Acer platanoides (Norway maple), Gleditsia triacanthos (honey locust), and Fraxinus pennsylvanica (green ash).

The landscape surrounding the pond consisted of mature forest around the perimeter, except where adjacent to the roadway, and well-established prairie vegetation along the banks. Submergent and emergent aquatic vegetation including Ceratophyllum demersum (coontail), Lemna minor (common duckweed), and a variety of Potamogeton spp. (pondweed), and Nymphaea spp. (water lily) species were present during the growing season (typically May through October).

Sample Collection and Measurement of Nutrient, Sediment, and Chloride Concentrations

USGS streamgages (Cherokee Park East Pond Influent 430854089223101 and Cherokee Park East Pond Effluent 430900089223201) measured stormwater discharge and collected water-quality samples during runoff events at the pipe and culvert influent to the pond and the pond effluent. A detailed description of equipment and methods used to collect a water sample are provided in Selbig and others (2020) and are briefly highlighted herein. Each streamgage was equipped with a volumetric water level and velocity sensor that was used to compute discharge. Precipitation data were collected using tipping-bucket rain gages calibrated to 0.25 millimeter. There was no dry weather flow in either the test or control storm drain network. Automated stormwater samplers collected flow-weighted discrete samples over the duration of an event hydrograph, which were later combined into a single, composite sample resulting in an event mean concentration (EMC). Sample collection was activated by a rise in water level in the pipe during a precipitation event. Water-quality samples were typically collected within 24 hours after runoff cessation. Samples were analyzed at the Wisconsin State Laboratory of Hygiene, in Madison, Wis. All samples were tested for TSS according to Standard Methods 2540D (Lipps and others, 2023), suspended sediment concentration (SSC) according to the American Society for Testing and Materials D3977–97 (ASTM International, 2019), TP and DP according to U.S. Environmental Protection Agency (EPA) Method 200.7 (EPA, 2001), TN and DN according to EPA Method 353.2 (EPA, 1993), and chloride according to Standard Methods 4500–Cl–E (Lipps and others, 2023). Concentration data are provided in Selbig (2025).

Measurement of Dissolved Oxygen and Temperature

A water-quality sonde was placed near the center of the pond at a depth approximately 1.2 meters below the pond surface and 15 cm above the bottom. Measurements of daily mean DO and temperature were recorded to determine whether pond mixing at the surface was sufficient to overcome stratification of the water column, thereby preventing anoxia near the bottom sediment. The sonde was deployed in the spring after ice no longer covered the pond and was removed before ice cover in the winter. Because of equipment failure, the measured timespan covered May 13–November 29, 2022; March 25–November 21, 2023; and March 13–October 21, 2024. Gaps in measured data occurred July 3–27, 2022; October 4–20, 2022; September 20–October 3, 2023; and June 24–July 12, 2024. Daily mean values of DO and temperature are available at Selbig (2025).

Street Cleaning Practices

Two variations of street cleaning frequency were evaluated: a mechanical broom cleaner, like that described by Selbig and Bannerman (2007), operated at a frequency of once per month, and a mechanical broom and regenerative-air cleaner operated at a frequency of once per week (fig. 2). Water-quality samples collected during the weekly mechanical broom and weekly regenerative-air campaigns were combined into a single dataset representing a weekly street cleaning frequency. This was based on prior research concluding the performance of municipal street cleaning programs was more related to the frequency and not the form of technology (Selbig and Bannerman, 2007; Selbig and others, 2020). Street cleaners were typically deployed in late spring and continued through late November (fig. 2). No other maintenance practices (for example, sump cleaning) were done to augment delivery of pollutants to the pond during the study period.

Statistical Analyses

The performance of the stormwater pond was based on a calculated removal efficiency using comparisons of influent and effluent pollutant load from paired sampled events. Storm event loads at each monitoring location were computed by multiplying the EMC by event runoff volumes. Load data are provided in Selbig (2025).

Load data were first tested for normality by use of the Shapiro-Wilk test (Helsel and Hirsch, 2002). Generally, the data were not normally distributed in which case differences between paired influent and effluent loads between street cleaning phases were evaluated by use of the nonparametric Mann-Whitney U test on groups (Helsel and Hirsch, 2002). All statistical tests were first done using a two-tailed test in which the null hypothesis assumed that the central tendency of populations for each street cleaning phase was not different. If the null was rejected, tests were repeated using a one-tailed test to determine if the influent or effluent load for one street cleaning phase was statistically greater or less than another phase. Changes in pollutant load measured at the influent and effluent within each street cleaning phase were evaluated by use of the nonparametric Wilcoxon signed-rank test (Helsel and Hirsch, 2002). All tests used a 95-percent confidence level unless otherwise noted.

Comparison of Concentrations

EMCs measured at the effluent were compared to the flow-weighted average of the EMCs measured at the influent pipe and culvert among and between street cleaning phases. The pond was able to effectively reduce concentrations in monthly and weekly phases for all pollutants except chloride (table 1). Results of the Wilcoxon signed-rank test showed these reductions to be significant; however, concentrations of chloride did not follow this pattern with both monthly and weekly phases having influent concentrations that were significantly less than the effluent. Median concentrations of chloride measured at the effluent exceeded the influent by a factor of approximately 5 during the weekly phase and by a factor of 7 during the monthly phase.

When comparing between street cleaning phases, median and mean pond influent concentrations of TN, DN, TSS, and SSC were greater under a monthly frequency compared to weekly (table 1, fig. 3). This was confirmed by use of the nonparametric Mann-Whitney U test for all except TSS, which failed statistical significance (probability value [p-value]=0.12). Measurements of TSS in stormwater can be highly variable because of the bias incorporated into the analytical method (Selbig and Bannerman, 2011). Although TSS is commonly used by the regulated community, previous studies have documented TSS to be an inferior measurement of the amount of sediment suspended in stormwater, favoring SSC as a more suitable metric (Gray and others, 2000; Selbig and Bannerman, 2011). Higher influent concentrations of SSC during the monthly phase indicate the lower frequency of street cleaning may have allowed for greater buildup of sediment on street surfaces thereby increasing the amount of pollutants available for wash-off during subsequent rain events. Median influent concentrations of TP, DP, and chloride during monthly street cleaning were slightly lower than weekly (table 1) but were not significantly different when evaluated using the Mann-Whitney U test.

Evidence of street cleaning frequency preferentially influencing concentrations of dissolved or particulate nutrients was inconclusive. Concentrations of nitrogen measured at the pipe inlets were primarily in the dissolved phase with the median value of DN, as a percentage of TN, approximating 60 percent for monthly and weekly phases. Changes in season did not appear to alter this pattern with median values of DN remaining near 60 percent in all three seasons monitored (55 percent in fall, 61 percent in summer, 63 percent in spring). Contrary to nitrogen, the amount of phosphorus in the dissolved phase varied by season. Fall concentrations of DP, as a percentage of TP, had the highest median value at 68 percent, whereas spring and summer had only 27 percent and 37 percent, respectively. Comparison of the DP to TP ratio between the street cleaning phases may contain temporal bias, because there were no fall samples during the monthly phase. With fall samples included (n=9), the median concentration of DP, as a percentage of TP, was higher when street cleaning was done weekly compared to monthly at 51 percent and 31 percent, respectively. If fall samples were excluded, the proportions become nearly identical at 34 percent and 31 percent, respectively.

Median concentrations observed at the effluent showed little variation between street cleaning phases for most pollutants. Coefficients of variation were generally less than one, indicating relatively low variability (table 1). TSS and SSC showed considerable variability in effluent concentrations during the weekly phase with coefficients of variation exceeding two. Low variability in effluent concentrations may indicate consistent pollutant removal efficiency regardless of influent concentrations related to changes in street cleaning; however, Spearman’s rank test showed weak to moderately positive correlation (Spearman’s rho coefficient [ρ]=0.39 to 0.45, p-value<0.01) between influent and effluent concentrations for all pollutants except TSS and SSC, which showed no correlation. Although relations between influent and effluent concentrations were not linear, the pond appears to be limited in its ability to reduce pollutants other than sediment as influent concentrations increase.

Comparison of Loads

Similar to reductions observed in pollutant concentrations, the pond was able to reduce the cumulative pollutant load discharged from the effluent for both phases of street cleaning, except chloride (fig. 4, table 2). Overall, the pollutant removal capacity of the pond was greater during the monthly phase except for DP. Both phases showed the largest percentage reductions in sediment load (TSS and SSC), which is typical for stormwater management practices that use settling as the primary form of treatment (EPA, 1983; Gu and others, 2016). Dissolved forms of nitrogen, when compared cumulatively, had slightly higher percentage reductions than the total. Chloride was the only pollutant where more was observed leaving the pond than coming in. Export of chloride ranged between 109 percent in the weekly phase to 860 percent in the monthly. A closer inspection of individual paired samples showed all but six events exported chloride from the pond, three of which occurred in the fall when concentrations (influent and effluent) were generally at their lowest (Selbig, 2025).

Table 2 shows that more load was received by the pond during the monthly phase for all pollutants except DP and chloride. This was supported by use of the Mann-Whitney U test on individual paired events. Results showed TN, DN, TSS, and SSC loads measured at the influent during the monthly phase were significantly greater (p-value<0.05 for TN, DN, and SSC; p<0.10 for TSS) than the weekly phase, whereas TP, DP, and chloride showed no difference. This indicates the reduced frequency of street cleaning allowed for more accumulation of sediment and debris on the street surface, which became available for washoff during subsequent rain events. A similar conclusion was made by Selbig and Bannerman (2007) who observed significant reductions in street dirt for vacuum-assisted and mechanical broom street cleaners operated once per week compared to no street cleaning, whereas a mechanical broom street cleaner operated once per month showed no significant reductions in street dirt. Because the calculated percentage reduction in pollutant load for most treatment practices has been shown to be strongly affected by the influent concentration (Barrett, 2005), the contrast in performance between the monthly and weekly phase was not unexpected.

Comparison of influent and effluent loads are further detailed in table 3 along with the calculated percentage change. Negative percentages indicate reduction, whereas positive percentages indicate export. Contrary to other pollutants, chloride was the only pollutant that consistently had more leaving the pond than entering with median percentage differences for weekly and monthly phases of 258 percent and 639 percent, respectively (Selbig, 2025). Median percentage reductions of TN, DN, and TP were similar for both phases ranging from −39 percent to −51 percent. These values met or exceeded the expected removal performance for ponds based on prior studies (Weiss and others, 2007; Koch and others, 2014; Janke and others, 2022). Sediment (TSS and SSC) had the highest percentage reduction of all pollutants ranging from −78 percent for TSS in the weekly phase to −92 percent for SSC in the monthly phase, meeting target reduction goals established by the WDNR (2024). Reduction of DP showed the greatest disparity between phases. Although both street cleaning phases showed reductions of DP (median values for monthly and weekly at −39 percent and −67 percent, respectively), statistical comparison of individual percentage reductions across all paired events using the Mann-Whitney U test showed the weekly phase to be significantly higher than the monthly phase (p-value=0.01). Similar statistical comparisons of percentage difference for all other pollutants showed no difference between the two street cleaning phases.

It is unclear why reduction of DP during the weekly phase was significantly greater than monthly, whereas all other pollutants were not. One explanation could be related to the lack of samples in the fall during the monthly phase, a time when organic detritus on streets is at its highest. Previous studies have documented the disproportionate amount of leachable phosphorus stemming from leaf litter in the fall compared to spring and summer (Selbig, 2016; Janke and others, 2017; Wang and others, 2020). The seasonal influx of DP during the weekly phase accounted for 26 percent of the total cumulative load observed at the influent. This proportion was not similarly represented in the monthly phase. Repeating the Mann-Whitney U test with fall events removed showed percentage reductions in the weekly phase were still statistically higher than the monthly phase, but with less confidence (p-value=0.11); therefore, differences in the pond’s ability to remove DP can only be partially explained by the discrepancy in seasonal loading between street cleaning phases. Another explanation could be related to internal phosphorus loading within the pond. The release of phosphorus from sediment has been well documented in small ponds under select conditions (Frost and others, 2019; Taguchi and others, 2020). Closer inspection of all paired samples revealed 10 events having DP load at the effluent exceeding what was observed at the influent. Of those 10 events, eight occurred during the monthly phase representing 28 percent of the total cumulative DP load observed at the effluent (Selbig, 2025). Export of DP in pond effluent as a result of internal loading would have falsely misrepresented the reduction capabilities of the pond during this phase.

Although concentrations and rates of DP from pond sediment were not directly measured, evidence of internal loading was apparent. Field measurements of DO revealed anoxic conditions near the pond bottom throughout much of the sampling period from May through October (Selbig, 2025). Stormwater ponds with hypolimnetic anoxia have been shown to be highly susceptible to internal release of sediment phosphorus, which can potentially contribute to downstream eutrophication (Taguchi and others, 2020). Despite persistent anoxia, Janke and others (2022) suggest even large release rates of DP from internal loading may not strongly influence effluent loads. This is largely due to high levels of autotrophic biomass and organic matter (floating vegetation) that maintain high rates of biological production. Despite evidence of anoxic conditions near the pond bottom, it is difficult to quantify the effect, if any, on phosphorus export. More information is needed to accurately assess the influence of internal phosphorus loading on pond performance.

Other factors may also have contributed to pond performance that are unrelated to street cleaning. Janke and others (2022) found storage volume to be a major factor for maintaining high rates of nutrient capture of three stormwater ponds in Minnesota. Increased storage volume from water losses would effectively increase water residence time, which promotes settling and uptake of sediment and nutrients. Because the pond was underlined with an impermeable clay layer, water losses through exfiltration to underlying native soils was unlikely; therefore, increased volumetric storage would come primarily from evaporative losses between storm events. Interevent pond levels were not measured so it is difficult to determine the volumetric storage prior to each storm event. Instead, differences between paired effluent and influent volume, represented as a percentage reduction, was used as a surrogate for storage prior to a storm. Increasing percentage reductions would indicate more storage was available prior to the storm event. These values were statistically compared to corresponding percentage differences in pollutant load by use of the Spearman’s correlation test. Results indicate changes in volumetric storage did not influence changes in pollutant load reductions (table 4).