Site Description

This study characterized concentrations and loads of selected nutrients in stormwater runoff from two residential catchments in Madison, Wisconsin (fig. 1). The catchments were selected to represent typical medium-density residential land use with moderate (greater than 25 percent) overhead tree canopy in the right-of-way. Surface runoff in each catchment was routed to a network of underground storm drainage pipes that flowed into a stormwater treatment pond. Two pipes were located at the terminus of the drainage network in each catchment: (1) a pipe that was 0.9 meter (m) in diameter and a box culvert that was 2.1 m long by 0.9 m wide at the outfall of the test catchment (USGS station 430854089223101) and (2) two pipes that were 0.9 m in diameter at the outfall of the control catchment (USGS station 430855089224701).

Estimates of overhead street tree canopy were made using a combination of aerial imagery coupled with geographic information system software and field surveys (City of Madison, oral commun., 2023). The percentage of tree canopy covering streets was slightly higher in the control catchment (36 percent) than in the test catchment (30 percent). Trees were generally a mix of mature, deciduous hard and softwood species and most were characterized as Acer platanoides L. (Norway maple), Gleditsia triacanthos L. (honey locust), and Fraxinus pennsylvanica Marsh. (green ash). Because the amount of leachable phosphorus can vary appreciably amongst different tree species (Dorney, 1986; Wang and others, 2020), the range of nutrient concentrations reported herein may not be like other areas with different tree species.

Sample Collection and Measurement of Nutrient Concentrations

A monitoring station was used to measure flow and collect water samples at the storm sewer outfall of the control and test catchments. Storm event loads were computed by multiplying the event mean concentration by stormwater runoff volumes. A detailed description of equipment and methods used to collect a water sample is provided in Selbig and others (2020) and is briefly highlighted herein.

Each monitoring station was equipped with water level and velocity sensors in each pipe that were used to compute discharge. Precipitation data were collected using tipping-bucket rain gages calibrated to 0.25 millimeter (mm). No dry weather flow was detected in either the test or control storm drain network. Automated stormwater-quality samplers collected flow-weighted discrete samples over the duration of an event hydrograph and were later combined into a single composite sample resulting in an event mean concentration. 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., for total phosphorus (TP) and dissolved phosphorus (DP) according to U.S. Environmental Protection Agency (EPA) Method 200.7 (U.S. Environmental Protection Agency, 2001) and total nitrogen (TN) and dissolved nitrogen (DN) according to EPA Method 353.2 (U.S. Environmental Protection Agency, 1993).

Leaf Collection and Street Cleaning Practices

Municipal leaf collection in the State of Wisconsin typically begins in late September to early October each year and continues through November, or until snowfall precludes further collection. Leaf collection vehicles are deployed to collect and remove bulk leaf litter and other organic detritus from residential areas. Residents are asked to pile their leaves adjacent to the street to limit excess debris in the street gutter. After leaf piles are removed, street cleaners pass through the service area to remove leftover organic residue from the street and gutter (Selbig and others, 2020). The form of leaf collection for this study adopted a “transfer” method, which uses a vehicle equipped with a modified bristled plow to rapidly transfer piles of leaves placed near the curb into the street. The leaves are then pushed into a garbage collection vehicle for removal. Leaf piles were collected three times in the fall of each study year in the control and test catchments, approximating a frequency of once every other week.

The frequency and form of street cleaning and leaf collection used in this study were designed to supplement the matrix of municipal practices previously described in Selbig and others (2020; table 1). Although the methods described in table 1 are not inclusive of all municipal leaf collection and street cleaning options, they represent practices that are more commonly used by municipalities throughout the Midwest and Northeast (Selbig and others, 2020). Starting in October 2020, mechanical-broom street cleaners operated at a frequency of once per week in the test catchment, whereas no cleaning was done in the control. During this time, leaf piles placed between the street curb and sidewalk were collected about every other week using the transfer method. This schedule represented the “treatment” phase. Because of the small number of storm events measured in 2020, the treatment phase was repeated in 2021. Starting in October 2022, the method and frequency of leaf pile collection remained the same as in 2020 and 2021; however, no street cleaning was done in either the control or test catchments. This schedule represented the “calibration” phase of the study.

Although leaf litter has been determined to be the primary source of nutrients in runoff during the fall (for example, Selbig, 2016; Janke and others, 2017; Selbig and others, 2020), contributions of phosphorus from leaf piles placed in the street terrace (that is, the pervious area between the street curb and sidewalk) were assumed to be minimal. This assumption was based on evidence reported by Selbig and others (2020), which indicated catchments having a rigorous leaf collection program, but no street cleaning resulted in no change in the amount of phosphorus measured in runoff. It was concluded that the mass of leaf litter and other organic detritus on a street has a much greater effect on phosphorus in runoff than leaves piled adjacent to the street. Therefore, results of the study described herein are more a reflection of the method and frequency of street cleaning than collection of leaf piles.

Statistical Analyses

A paired-catchment design was used to help evaluate the effectiveness of leaf collection based on differences in loads measured in the control and test catchments between calibration and treatment phases of the study. The basis behind the paired-catchment approach is discussed in detail in Selbig and others (2020) and is briefly described herein. The paired-catchment statistical design assumes that a quantifiable relation exists between paired water-quality data and that this relation is valid until a major change (that is, treatment) is made in one of the catchments (Clausen and Spooner, 1993). At that time, a new relation will develop. The strength of this approach is that it does not require the assumption that the control and test catchments are statistically the same; however, it does require that the two catchments respond in a predictable manner together and that their relation remains the same over time except for the effect of leaf collection and street cleaning.

Following procedures outlined in Clausen and Spooner (1993), the significance of the effect of leaf collection and street cleaning was determined using analysis of covariance (ANCOVA). The analysis is a series of steps determining the significance of the treatment regression; the significance of the overall regression, which combines the calibration and treatment phase data; the difference between the slopes of the calibration and treatment regressions; and the difference between the intercepts of the calibration and treatment regressions. A change in intercepts but not slopes between the calibration and treatment phase indicates an overall parallel shift in the regression equation. If the treatment regression shifted below that of the calibration phase, the form of treatment can be considered effective at reducing nutrient loads. Results were reported if they met the 80-percent significance level (probability less than or equal to 0.20). If the results of the ANCOVA test for slope and (or) intercept reveal a significant difference between the calibration and treatment regressions, the regression equation representing the calibration period can be used to predict what the mean runoff event loads should have been in the test catchment during the treatment period. The overall reduction because of leaf collection and street cleaning can then be expressed as a percentage change based on the mean predicted and observed values during the treatment phase (Clausen and Spooner, 1993).

Nutrient Concentrations in Stormwater With and Without Treatment

Mean concentrations of phosphorus during the calibration phase ranged from 1.75 to 2.45 milligrams per liter (mg/L) in the test and control catchments for TP, respectively, and from 1.47 to 2.04 mg/L for DP (table 2). Selbig and others (2020) determined medium-density residential catchments having less than 30-percent street tree canopy generally resulted in mean TP concentrations of less than 1 mg/L in the fall when no leaf collection or street cleaning occurred, whereas catchments having more than 40-percent street tree canopy had mean concentrations of more than twice that value. With street tree canopy in the control and test catchments at less than 40 percent (36 percent in the control and 30 percent in the test), mean concentrations of TP reported in table 2 are generally higher than reported by Selbig and others (2020) for catchments without leaf collection or street cleaning. This difference could be due, in part, to the small number of samples representing the calibration phase (number of samples=5). A larger dataset would reduce the effect of uncertainty in the collected data, resulting in tighter confidence intervals for the regressions. Selbig and others (2020) also asserted that depending on the distribution and abundance of tree species, concentrations of phosphorus in runoff can be elevated even in areas with less street canopy. This finding was supported in laboratory tests by Wang and others (2020) and Dorney (1986), indicating the amount of leachable phosphorus can vary considerably depending on tree species. This phenomenon was observed by Selbig and others (2020) in one study catchment having relatively low street tree canopy cover (21 percent) but a mean TP concentration that was similar to catchments with twice the amount of tree canopy. The difference was attributed to the larger percentage of honey locust, a species that has a higher concentration of leachable phosphorus than other species of trees in the study catchment (Dorney, 1986).

All mean and median concentrations of phosphorus in the test catchment were lower during the treatment phase (table 2). Unexpectedly, concentrations in the control catchment also decreased, although to a lesser degree than in the test catchment (about a 30-percent reduction in the control compared to 40 to 50 percent in the test). It is unclear why concentrations of TP and DP decreased in the control, particularly because no changes were made to street cleaning or leaf collection in both phases of the study. One explanation may again be attributed to the small number of samples representing the calibration phase. The event on October 25, 2022, resulted in phosphorus concentrations that were more than double what was measured for any event during the calibration phase in both catchments (table 2). Leaf litter accumulated for nearly 2 weeks before the October 25 event, longer than any other interevent dry period during the calibration phase. Wang and others (2020) determined phosphorus load in stormwater runoff is highly correlated to the leaf mass on streets before a storm event. Concentrations of phosphorus in runoff were likely elevated because of the large mass of leaf litter that accumulated on streets before the October 25 event. Censoring this event from the dataset would result in mean concentrations of TP and DP in the control catchment to be within 1 and 5 percent, respectively, of those measured during the treatment phase, whereas the test catchment still indicated appreciable reductions of 22 and 30 percent for TP and DP, respectively. However, no reasonable justification exists to censor the October 25 event from table 2.

Mean concentrations of nitrogen during the calibration phase ranged from 2.88 to 3.28 mg/L in the test and control catchments for TN, respectively, and 1.98 to 2.11 mg/L in the test and control catchments for DN, respectively (table 2). In general, a positive correlation between mean concentrations of phosphorus and nitrogen was observed in each catchment; higher amounts of phosphorus also would produce higher amounts of nitrogen (fig. 2). Like phosphorus, mean concentrations of TN and DN were generally within the range of values reported in Selbig and others (2020) for residential catchments but higher overall than those with a similar percentage of street tree canopy. Although leaf litter is recognized as a primary contributor of nitrogen in stormwater, other factors, such as residential fertilizer application, may have elevated concentrations. Hobbie and others (2017) determined household fertilizer resulted in 37 to 59 percent of nitrogen inputs in urban drainage basins in St. Paul, Minnesota (not shown).

Changes in Nutrient Loads as a Result of Leaf Collection and Street Cleaning

Paired loads from the control and test catchments (table 3) were evaluated to determine if changes in the relation established during the calibration phase of the study can be attributed directly to activities related to leaf collection and street cleaning during the treatment phase. Because the frequency and method of collecting leaf piles did not change between phases, differences in nutrient load reflect the amount of organic material removed from streets by street cleaning. As shown in figure 3, the relations in fall loads of TP, DP, TN, and DN developed between the control and test catchments during the calibration and treatment phase. Differences in these relations were quantified if results of the ANCOVA test for differences in slopes or intercepts exceeded an 80-percent statistical significance (probability less than or equal to 0.20).

Quantification of the overall effect of leaf collection and street cleaning frequency and method of removing leaf litter and other organic detritus is presented as a percentage reduction in nutrient load (table 4). The weekly use of a mechanical-broom street cleaner and biweekly leaf collection using a transfer method were able to reduce fall loads for all measured nutrients in stormwater runoff. Nutrients in the dissolved phase had the greatest reductions at 51 and 52 percent for DP and DN, respectively. This result was expected given that most of the phosphorus and nitrogen are in the dissolved phase (Selbig, 2016; Selbig and others, 2020). Reductions in TP and TN were slightly lower, but still significant, at 46 and 42 percent, respectively. Comparison of slopes indicates a clear difference between phases for TN and DN but not TP and DP (fig. 3). This difference is due to a single paired load during the calibration phase that seems to hold leverage on the resulting trendlines, affecting both slopes and intercepts. This point again corresponds to the October 25, 2022, event. Additional samples representing the calibration phase of the study across the same range of loads measured during the treatment phase would potentially reduce the effect of outliers and improve statistical outcomes.

Unlike nitrogen, paired loads of TP and DP during the calibration phase were more evenly distributed across the range of values measured during the treatment phase. Although results of the ANCOVA test did not indicate any significant changes in slope between phases, an overall downward vertical shift during the treatment phase resulted in a significant difference in intercept (table 4). Although additional samples during the calibration phase would have reduced variability and improved statistical confidence for TP and DP, the overall conclusion that a mechanical-broom street cleaner is able to reduce nutrient loads in runoff would likely not change.

Comparison to Previous Municipal Leaf Collection and Street Cleaning Programs

Results presented in table 4 represent the last in a series of evaluations on the effectiveness of municipal leaf collection and street cleaning programs (fig. 4). Although the frequency and technology presented in figure 4 are not inclusive of all municipal options, they represent practices that are more commonly used by cities in the Midwest and Northeast. From figure 4, significant reductions in TP and DP loads are not observed until tree canopy approximates 30-percent coverage over streets. The exception is the V2 study treatment, which represents biweekly leaf pile collection only using a vacuum-mulch collection system and no street cleaning. Despite having the highest overhead tree canopy in the series of evaluations presented in figure 4, the lack of street cleaning allowed the accumulation of leaf litter in the street to overcome any reduction in phosphorus that may have resulted from leaf pile collection on the side of the street.

The frequency and method of street cleaning rather than leaf pile collection have more effect on phosphorus load reduction in runoff. Selbig and others (2020) previously determined regenerative-air street cleaners operating at a frequency of once per week (indicated by “R1” in fig. 4) have the greatest percentage of TP and DP reduction regardless of form and frequency of leaf pile collection. The exception was in a study catchment having less than 10-percent street tree canopy, an indication that the mass of leaf litter on streets in this catchment likely was not enough to detect a significant difference in phosphorus load compared to no street cleaning. Selbig and others (2020) also reported mechanical-broom street cleaners, having less capacity to removed solids from street surfaces than regenerative-air cleaners (Selbig and Bannerman, 2007), resulted in lower reductions of TP and DP loads. In that study, catchments using the transfer method for leaf pile removal every other week and a mechanical-broom street cleaner operating once every other week (indicated by “T2–M2” in fig. 4) reduced TP and DP loads by 40 and 45 percent, respectively. Increasing the frequency of street cleaning to once per week (indicated by “T2–M1” in fig. 4), as was tested as part of this study, elevated the percentage reduction of TP and DP loads to 46 and 51 percent, respectively. This result indicates the reduction of phosphorus in stormwater runoff from residential areas with about 30-percent or more overhead street tree canopy is more a function of the cleanliness of the street before a storm event than the removal of leaf piles on the side of the street.

Selbig and others (2020) determined that unlike phosphorus, reductions in nitrogen loads were not as prevalent and, in some cases, increased after treatment (fig. 4). The T2–M1 treatment described as part of this study was only one of two leaf collection and street cleaning combinations that resulted in a reduction of TN and DN, the other being T2–R1. With one exception, catchments undergoing weekly street cleaning, regardless of technology, indicated reductions in TN, DN, or both. When the frequency of street cleaning decreased to every other week, nitrogen load in runoff increased or did not significantly change from a condition with no street cleaning.