ABSTRACT

This paper, and the associated keynote presentation at ICWMM 2025, review several topics that show how stormwater monitoring has been critical in the development of stormwater quality models. The first example is a timeline of how street cleaning was originally conceived as a stormwater quality control and how subsequent focused monitoring enabled a more accurate representation of its benefits. The next example illustrates scaling issues in monitoring and how data have been used to verify appropriate extrapolations of the information. Finally, a short summary of emerging contaminants of current interest illustrates the need for expanded monitoring to understand these little understood stormwater constituents and how they can be modeled to predict their sources, transport, controls, and fates. Many other topics relating to monitoring and modeling exist obviously, but these examples, mostly from the author’s publications, illustrate the range of some of these issues needing further monitoring and model development. The examples shown are only a small portion of the data collected during the referenced studies, and the full reports should be consulted for further information.

1 Street Cleaning as a Stormwater Control

Street cleaning, originally developed for street debris removal and for aircraft safety at airports, was one of the first stormwater controls suggested for water quality improvement. Detention ponds have also been around a long time, having started as peak flow reduction tools, and early catch basins were originally developed for large debris control to prevent clogging of stormwater drainage systems.

The following discussion summarizes three phases of studies that have investigated street cleaning. The first addresses early street cleaning tests conducted during the cold war as a tool to remove contaminated fallout from surfaces. These included many controlled tests examining removal of particulates by street cleaners for different particle sizes for different operational conditions and loadings. Some of these early researchers were later involved in environmental research and suggested that they could extend the early street cleaning research to examine stormwater control benefits. These data were used in some of the first stormwater models. Additional monitoring research was conducted to verify some of the early models and assumptions. As in most research, it seems that more questions are raised than answered, leading to additional monitoring and model development. The following paragraphs briefly summarize this long history of street cleaning as a stormwater quality control and how model changes are needed to address the better understanding of how street cleaning and stormwater quality interact.

1.1 Part 1. Early street cleaning performance tests

Early street cleaning tests were conducted from 1948 to 1963 as a method to decontaminate areas affected by radioactive fallout. Figure 1 shows an early street cleaning test and a plot of measured before and after street dirt loadings for different equipment and loading rates. Many other tests and several projects were conducted in this period exploring the ability of street cleaners to remove dust and dirt material from paved surfaces (including rooftops!).

Figure 1 Motorized sweeping (Wayne Model 450) on Portland cement concrete (Lee et al. 1959).

Sartor (involved in the early street cleaning research) and Boyd (1972) investigated street dirt loadings and cleaning benefits for a wide range of land uses and cities. This information was (and is) used in many stormwater quality models. Figure 2 is a street dirt loading plot showing measured street dirt loadings on streets since the streets were last reported to be cleaned. Note the zero loadings at zero time since cleaning, which assumed that the street cleaning removed all the street dirt.

Figure 2 Street dirt accumulation rates by land use (Sartor and Boyd 1972).

This plot shows zero loadings right after cleaning, which forces the accumulation rate to be very large to result in the measured loads at specific times after the prior cleaning. As found during later research, the residual loadings on the streets after cleaning can be large, depending on the street texture and equipment operating conditions. This adjustment in the initial slopes of the accumulation curves resulted in much lower (about ten times) actual accumulation rates. Artificially high accumulation rates on streets forces incorrectly lower contributions from other source areas when calibrating models with outfall discharge data.

Sartor and Boyd (1972) also chemically analyzed street dirt material from several land uses in each of eight US cities (solids and particle size, nutrients/BOD/COD, metals, pesticides, and PCBs). These data also illustrated how particulate strengths (the number of pollutants associated with particulates, typically expressed as mg pollutant per gram of particulate solids), changed for different particle sizes, as shown in Figure 3 for total phosphorus. It is common for many of the street dirt pollutants to have greater particulate strengths for the smaller particles.

Figure 3 Variation of total phosphate mass associated with different particle sizes (Sartor and Boyd 1972).

Sartor and Boyd (1972) also conducted tests examining wash-off of street dirt and debris by particle size for different controlled rain intensities, as shown in Figure 4.

Figure 4 Street dirt wash-off tests for smooth asphalt for low and high intensity rains, by particle size (Sartor and Boyd 1972).

Many models assume that the asymptotic constant values on these plots represent the total street loading values, but actually they only represent the loading that can wash-off for that particle size and rain condition (usually only about 10 to 30% of the total load, with the greater yields for small particle sizes, smooth streets, and high rain intensities).

Table 1 compares the removal efficiency (as a percentage of the initial street dirt load) for many street cleaning tests. The in-situ tests used actual street dirt loadings at many cities and land uses, while the equation values are based on the earlier test results with very high street dirt loadings manually applied before the test.

Table 1 Estimated street sweeper efficiency (Sartor and Boyd 1972).

	Removal Efficiency (%)
Particle Size (µm)	In-situ test	Equation	Composite (estimate)
>2,000	78.8	–	79
840–2,000	66.4	–	66
246–840	69.5	49.2	60
104–246	47.7	48.7	48
43–104	<0	22.2	20
<43	<0	15.8	15

The “equation” values represent removal rates for very high street dirt loadings, while the in-situ values represent actual loadings that are much lower. Street cleaning removal efficiency as a percentage of initial load is highly dependent on the initial street dirt loadings, street texture, and operating conditions, with much greater percentage removals for high loads and much reduced percentage removals for low loads.

1.2 Part 2. Street dirt contributions to outfall discharges

The above research led to the reasonable conclusion that there was a very large number of pollutants residing on streets that were available for wash-off during rains. Pitt and Amy (1973) examined street dirt characteristics in detail. The commercial geochemical lab doing mass spectrophotometric analyses on the street dirt samples wanted to know where the mine was! Also, at this time, it was assumed that the vast majority of the runoff in urban areas originated from directly connected impervious areas, of which streets were the most prominent component. It was also assumed that streets were the major source of most stormwater pollutants, and that street cleaning would be an effective method to improve stormwater quality. Therefore Pitt (1979) included full-scale street dirt and outfall stormwater monitoring to measure the actual effectiveness of street cleaning in reducing stormwater pollutant discharges for a range of street cleaning efforts, equipment, and land uses, compared to concurrent outfall monitoring data.

One of the major objectives of this research was to examine the mass balance of street surface particulates, including their deposition and accumulation rates, sources, transport to and through drainage systems, and resuspension of street particulates to the air for redeposition on adjacent areas, and to measure the relationship between street surface contamination and outfall stormwater quality. And, of course, the role that street cleaning has in affecting these discharges.

Several thousand street dirt subsamples were obtained during this research in three large study areas (subdivided by street surface texture), focusing on the street dirt loadings in relation to street cleaning or rain events. Figure 5 is a schematic of these loading patterns.

Figure 5 Sawtooth pattern associated with deposition and removal of street dirt particulates (Pitt 1979).

Figure 6 shows the actual street loading changes for different street cleaning operations (different equipment and cleaning frequencies) for a several month test period at one of the study areas, showing the residual loads on the streets immediately after street cleaning and the actual accumulation rates and changes in the median particle sizes of the street dirt samples. These data were obtained by measuring the street dirt loadings many times in the same study areas throughout the multi-year study period, with samples obtained immediately before and after street cleaning, and immediately before and after rains, in addition to other samples during long periods without rains or street cleaning.

Figure 6 Street dirt loads and median particle size as a function of time for the Keyes, good asphalt, test area (Pitt 1979).

The actual loads at any time since cleaned are highly dependent on the texture of the road surface, while the accumulation rates (the slopes of the trends) did not vary as much during this research, as shown on Figure 7 for the different test areas.

Figure 7 Street dirt accumulations as a function of time since last cleaned for each test area (all seasons combined) (Pitt 1979).

Note that the “starting” values (at zero days since street cleaning) are not zero (in contrast to the assumed starting value in the Sartor and Boyd street dirt loading plot which only obtained a single sample for each location).

Many tests were conducted to examine the removal of street dirt particles by size for different street cleaning operations and locations. Figure 8 is an example for a smooth street surface showing much greater removals of large particles compared to small particles.

Figure 8 Street dirt removal by street cleaning on smooth asphalt street by particle size (Pitt 1979).

Several rains were monitored during this research at the outfalls located at the study areas. Figure 9 is a plot of the rainfall and flow patterns for a rain series, showing a classical rainfall-hydrograph plot.

Figure 9 Rainfall and runoff from Keyes study area during rains of March 15 and 16, 1977 (Pitt 1979).

Mass calculations compared the weights of contaminants removed from the streets during rains (from before and after rain street dirt loading measurements) to the total outfall monitored discharges for all monitored rains. Table 2 shows the calculations for the event and location associated with the hydrograph shown in Figure 9.

Table 2 Street surface pollutant removals compared to runoff yields for March 15–16, 1977 Storm (Pitt 1979).

	Oil and Screen		Asphalt		Cumulative
Parameter	kg/curb-km difference	Total kg difference in 3.5 curb-km	kg/curb-km difference	Total kg difference in 4.3 curb-km	Total Keyes area-kg difference	Runoff yield (kg)	Ratio of street surface wash-off to runoff yield
Total solids	34	120	8.2	36	160	430	0.36
COD	6.8	24	0.85	3.7	28	390	0.071
Kjeldahl N	0.09	0.33	1.5	6.4	6.7	23	0.29
Ortho PO4	0.065	0.23	0.0014	0.0060	0.24	9.6	0.025
Pb	0.11	0.40	0.054	0.23	0.63	0.79	0.80
Zn	0.019	0.067	0.0062	0.027	0.094	0.32	0.29
Cr	-0.0024	-0.0084	0.0040	0.017	0.0088	0.029	0.30
Cu	-0.0039	-0.014	0.0068	0.029	0.015	0.059	0.26
Cd	0.000087	0.00031	0.000028	0.00012	0.00043	0.012	0.037

For this event and location, the street wash-off amounts accounted for about 3% (orthophosphate) to 80% (lead) of the total outfall pollutant discharges. This research found that streets are likely a significant source of sediment, heavy metals, oils and grease, and other toxic materials, but there are other important sources of these, and other pollutants in urban areas, as shown on Table 3. The relative street pollutant contributions to outfall discharges also vary for changes in rain depth and other factors.

Table 3 Potential significant urban runoff pollutant sources (Pitt 1979).

	Rain	Roofs	Streets	Parking lots	Litter (incl. animal feces)	Landscaped areas	Vacant land	Construction sites	Road ice control	Other (industrial and solid waste runoff)
Sediment			X				X	X	X
Oxygen demanding matter					X	X
Nutrients	X				X	X	X
Salts									X
Bacteria					X		X
Heavy metals		X	X	X
Pesticides/herbicides						X
Oils and grease			X	X				X
Floating matter					X	X
Other toxic materials		X	X	X					X	X

There were relatively few rainfall events during the Pitt (1979) research in San Jose, CA (a relatively arid climate compounded by being in a drought period). Additional monitoring projects focusing on street cleaning and urban area mass balances were conducted in wetter areas.

1.3 Part 3: Other source contributions to outfall discharges (including urban hydrology issues)

Pitt (1983) conducted research in Ottawa, Ontario to investigate how street cleaning can affect bacteria stormwater discharges. The first question addressed was how much of the area stormwater flows were associated with street runoff. Detailed site investigations and monitoring resulted in the plot shown in Figure 10, indicating that streets were a major source of the outfall runoff for rains up to about 20 mm. Larger rains resulted in relatively constant street runoff contributions of about 10 to 30 percent of the whole watershed areas, depending on the land use and source areas present.

Figure 10 Percentage of total urban area runoff flows originating from street surfaces as a function of rain depth, Ottawa, Ontario (Pitt 1983).

Source area sheet flow grab samples were also obtained from various subareas in the study areas for bacteria analyses, as shown in Table 4.

Table 4 Catchment subarea sheet flow bacteria and lead characteristics (August 15 and September 23, 1981 events) (Pitt 1983).

	Roof runoff	Vacant land, unpaved parking	Paved parking	Gutter flow
Fecal coliforms, geometric mean (#/100mL)	85	5,600	2,900	3,500
Fecal strep., geometric mean (#/100mL)	170	16,500	11,900	22,600
Lead, average (µg/L)	30	30	350	320

The monitored street runoff loadings compared to the total outfall yields were small for fecal coliforms, as shown in Figure 11. Drainage areas having large street areas contributed up to 10% of the total outfall fecal coliform yields.

Figure 11 Fecal coliform street surface particulate loading and runoff yields compared, Ottawa, Ontario (Pitt 1983).

Pitt (1985) conducted a study under the US EPA’s Nationwide Urban Runoff Program (NURP) (one of several that examined street cleaning benefits). Two residential areas in Bellevue, WA, were extensively monitored over two years during many rains. The streets in these typical residential areas comprised 10 to 12% of the total surface areas, with most of the areas being landscaped front and back yards (about 60%) and rooftops (about 17 to 19%). Runoff and total solids contributions varied for different rain depths, as shown on Figure 12.

Figure 12 Runoff and total sources for Bellevue, WA study areas, for different rain depths (Pitt 1985).

Even though streets made up large fractions of the runoff volume for the smallest rains, the back and front yards were the major sources of total solids for rains larger than about 2.5 mm.

Many street dirt samples were obtained before and after street cleaning and rain events. Figure 13 shows the wash-off removals of street dirt material by particle size for rains, indicating highly preferential removal of the smallest particle sizes, and wash-on of larger particles from adjacent eroding areas onto the streets.

Figure 13 Percent wash-off by particle size during rains, Bellevue, WA (Pitt 1985).

This is the opposite removal relationship for street cleaning which has highly preferential removal of large particles and little removal of the small particles, as shown on Figure 14.

Figure 14 Percentage Street dirt removal by particle size during street cleaning on very rough (a) and smooth streets (b), Bellevue, WA (Pitt 1985).

Tests were also conducted comparing standard mechanical street cleaners and a modified regenerative air street cleaner. The modified street cleaner was able to better reduce the street dirt loadings after cleaning compared to the standard mechanical street cleaner, especially for low street dirt initial loads, as shown on Figure 15.

Figure 15 Street dirt loads before and after street cleaning for standard and regenerative air street cleaning, Bellevue, WA (Pitt 1985).

Outfall monitoring during rains at the two Bellevue study areas with and without street cleaning indicated insignificant reductions in outfall runoff discharges, as shown in Figure 16 (for normalized runoff loads) when the streets were cleaned and without cleaning.

Figure 16 Monitored outfall total solids discharges for clean vs, dirty streets periods at the Surrey Downs test area (wet season) vs. runoff flow amounts, Bellevue, WA (95% confidence intervals shown) (Pitt 1985).

Pitt and McLean (1986) examined sources of stormwater and snowmelt flow and pollutant sources in Toronto, Ontario. Detailed wash-off and runoff tests were conducted on streets in the study areas. Figure 17 shows that the volumetric runoff coefficient (Rv, the ratio of runoff depth to rain depth) for the urban streets are relatively low, especially for small rains, much lower than the conventionally accepted values of 0.85+ used in most drainage studies for large rains. Rains of at least 50 mm would result in the Rv values usually used for drainage design, as appropriate.

Figure 17 Rainfall vs runoff amounts for smooth and rough streets in Toronto, Ontario (Pitt and McLean 1986).

Figure 18 shows the flow contributions for different source areas in the monitored residential area.

Figure 18 Urban runoff flow sources in the Thistledown residential study area for different rain depths, Toronto, Ontario (Pitt and McLean 1986).

Figure 19 outlines the movement, deposition, and removal opportunities of stormwater constituents. A comprehensive urban stormwater quality model must include a wide variety of surfaces and land uses, along with different processes affecting the yield of the pollutants from the sources into the receiving waters.

Figure 19 Schematic of pollutant deposition and removals in urban areas (Pitt and McLean 1986).

A major outcome of the Toronto research was comparing baseflows, stormwater, and snowmelt discharge concentrations for the residential and industrial study areas, as summarized in Tables 5a and 5b.

Table 5a Median concentrations (mg/L) for selected warm weather stormwater and baseflow discharges (yellow high-lighted cells show the highest median concentrations for all flow phases) (Pitt and McLean 1986).

	Warm Weather Baseflow		Warm Weather Stormwater
Constituent	Residential	Industrial	Residential	Industrial
Total solids	979	554	256	371
Filterable solids	973	454	230	208
Particulate solids	<5	43	22	117
Total phosphorus	0.09	0.73	0.28	0.75
Total Kjeldahl nitrogen	0.9	2.4	2.5	2.0
Phenolics (µg/L)	<1.5	2.0	1.2	5.1
Chemical oxygen demand	22	108	55	106
Fecal coliforms (#/100mL)	33,000	7,000	40,000	49,000
Fecal strep. (#/100mL)	2,300	8,800	20,000	39,000
Chromium	<0.06	0.42	<0.06	0.32
Copper	0.02	0.045	0.03	0.06
Lead	<0.04	<0.04	<0.06	0.08
Zinc	0.04	0.18	0.06	0.19

Table 5b Median concentrations (mg/L) for selected cold weather baseflow and snowmelt discharge (yellow high-lighted cells show the highest median concentrations for all flow phases) (Pitt and McLean 1986).

	Cold Weather Baseflow		Cold Weather Melting Periods
Constituent	Residential	Industrial	Residential	Industrial
Total solids	2,230	1,080	1,580	1,340
Filterable solids	2,210	1,020	1,530	1,240
Particulate solids	21	50	30	95
Total phosphorus	0.18	0.34	0.23	0.50
Total Kjeldahl nitrogen	1.4	2.0	1.7	2.5
Phenolics (µg/L)	2.0	7.3	2.5	15
Chemical oxygen demand	48	68	40	94
Fecal coliforms (#/100mL)	9,800	400	2,320	300
Fecal strep. (#/100mL)	1,400	2,400	1,900	2,500
Chromium	<0.01	0.24	<0.01	0.35
Copper	0.015	0.04	0.04	0.07
Lead	<0.06	<0.04	0.09	0.08
Zinc	0.065	0.15	0.12	0.31

Warm weather flows have much greater bacteria levels than cold weather flows, but the baseflows and snowmelt have greater pollutant concentrations than the stormwater for several pollutants. All flows must be considered when comparing potential control programs managing urban discharges to receiving waters. Warm weather stormwater is not the only source of urban runoff pollutants discharged to receiving waters. Warming climates in the future will shift the importance of these discharges, also with changes in runoff characteristics with warmer waters and likely changes in the performance of stormwater controls.

A number of more recent studies, such as conducted by the WI Department of Natural Resources and the USGS, investigated “modern” mechanical and advanced vacuum street cleaners. These newer street cleaners generally performed better than the older mechanical street cleaners studied in the research summarized earlier, especially the vacuum units. However, none of the outfall monitoring conducted during varying street cleaning efforts resulted in statistically significant improvements in stormwater quality. It is expected that these studies would have been able to identify significant reductions of at least 25%, based on the number of events monitored and the measurement variabilities. Modeling using WinSLAMM with updated street cleaner performance functions indicates expected outfall stormwater suspended solids reductions of up to about 35% with very intensive street cleaning (several times a week) using vacuum street cleansers on smooth streets in residential areas.

Periodically, researchers have considered quantifying street cleaning benefits by measuring the weight of debris captured by the street cleaners, assuming that this material was prevented from being washed off during rains. However, as indicated above, most of this material is larger than the sizes that would be washed off the streets during rains. Table 6 shows a hypothetical calculation for street cleaning benefits associated with hopper content measurements by particle size:

Table 6 Example calculation of stormwater benefits associated with street cleaning using hopper content measurements.

	Street dirt in street cleaner hopper, by particle size (kg)	Particles washed off by rains (%)	Potential particle mass washed off by rains (kg)
>6370	20	0	0
2000 - 6370	60	0	0
850 - 2000	85	0	0
600 - 850	75	7	5
250 - 600	65	10	7
106 - 250	30	17	5
45 - 106	10	33	3
<45	5	45	2
Total	350		22

In this hypothetical example, about 350 kg were captured by the street cleaner for a cleaning route. Subsamples of this material were dried and sieved into particle size ranges. Most of this material was in the 250 to 2,000 µm size range. However, only about 22 kg was likely available for wash-off during rains and contributed to the stormwater discharge quantity. This can be compared to the outfall discharges to estimate how effective the street cleaning operations were in reducing stormwater discharges. This also assumes that there were no further particulate reductions between the streets and the outfall. Sedimentation in piped drainage systems and in catch basin inlets can be an issue for large particles (usually larger than about 100 or 350 µm), but catch basins have minor effects for the smaller particles also. However, with grass swales, all particles can be retained. During monitoring, we have found that if catch basin inlets are effectively being used in the area, street cleaning operations have no additional measurable effect on outfall discharges (they are removing the same class of particles).

When modeling stormwater particulates, it is critical to monitor several particle sizes separately and quantify how they are transported, captured, or scoured, and not to apply a simple overall percentage value.

2 How can Small-Scale Monitoring be Applied to Total Watershed Outfall Discharge Calculations?

The previous discussion on street cleaning included a number of issues associated with scale, such as how to extrapolate the results from small test plots to watershed scale benefits. In the early street cleaning projects, only street dirt removal on test strips was examined with no concurrent outfall or other larger scale measurements. When later projects examined both street cleaning and outfall discharges, it was found that earlier assumptions on extrapolating the small-scale results to large scale watershed outfalls were not entirely valid.

There has also been much effort over the years investigating other individual stormwater controls, such as individual biofilters, hydrodynamic separators, media cartridge filters, etc., that have discrete influent and effluent sampling locations. These results are usually extrapolated to large-scale implementations of the controls in a watershed. These extrapolations are accurate if the effluent is not further affected before the outfalls (such as by drainage controls like grass swales or outfall controls like detention ponds), and if the fraction of the stormwater being affected by the controls to the total outfall discharges is known. Several projects have monitored complete watersheds having numerous distributed stormwater controls. These efforts have used suitable control watersheds that were monitored for comparison, either simultaneously with adjacent areas without controls, or if retrofitting the controls, monitoring of pre-retrofitting conditions. The following discussion summarizes some of these studies.

2.1 Large scale monitoring of watersheds having dense applications of stormwater controls

Selbig and Bannerman (2008) monitored parallel control and test watersheds over a 7-year period in Cross Plains, Wisconsin. One watershed was developed conventionally (55 ha in total with 13 ha developed) and the other was developed with extensive stormwater controls (78 ha in total with 15 ha developed). The undeveloped portions in each watershed were steep forested areas. The conventionally developed watershed had curbs and gutters, 12 m streets, and a fully connected stormwater-conveyance system. The test watershed had grass swales, reduced impervious areas (10 m street widths), a detention pond, and an infiltration basin. Figure 20 shows these two adjacent areas.

Figure 20 Comparison of test and control watersheds at Cross Plains, Wisconsin (Selbig and Bannerman 2008).

Rainfall and stormwater runoff volume and quality (TSS and PO₄) monitoring was conducted at these locations. Only six events with precipitation depths less than or equal to 10 mm produced measurable discharges from the test watershed having stormwater controls. In the conventionally developed watershed, the number of discharge events associated with these small rains was 180. In contrast, there was very little difference in the number of runoff events associated with the larger rains (greater than about 38 mm, or 1.5 inches) as shown on Figure 21. Total annual unit area discharge volumes measured from the conventionally developed control watershed were 1.3 to 9.2 times greater than from the test watershed with stormwater controls. Unfortunately, the differences in the annual mass discharges of sediment and phosphate from these two watersheds were not as expected; the differences were small and were not statistically significant due to other sources of these materials in the watersheds.

Figure 21 Histograms showing the frequency of runoff events resulting in discharges from the test and control watershed as a function of precipitation (Selbig and Bannerman 2008).

In another example of watershed scaling, Talebi (2014) evaluated several watershed monitoring projects in New Jersey, Ohio, and Missouri, that contained “green infrastructure” (GI) stormwater controls, as part of her Ph.D. dissertation research. She stated that “Numerous studies have evaluated the performance of different types of individual green infrastructure components at relatively small scales, indicating that these features are expected to be effective in reducing runoff volumes (important for combined and separately sewered areas) as well as enhancing the stormwater runoff quality (important for separately sewered areas). Yet, there are few demonstrations of integrated green infrastructure facilities at large scales with explicit real time evaluations of its impacts on reducing flows in combined sewers.” She also states that “The main objective of this research is to examine the effectiveness of retrofitted green infrastructure stormwater controls in several small- and large-scale developed urban watersheds in areas served either by separated or combined sewer systems. At small scales, infiltration measurements from individual stormwater controls, including drywells and biofilters, were used to quantify the benefits (such as runoff volume reductions). In addition, the effects of soil types, storm event characteristics, and land development characteristics on the infiltration behavior of individual green infrastructure stormwater controls were examined. At large scales, direct measurements of flow from in‐system flow monitors placed in combined or separate sewers affected by individual green infrastructure devices were evaluated. Real-time rainfall and runoff data from combined and separate sewer systems affected by GI stormwater controls in upstream areas were analyzed both before and after construction of the stormwater controls. The runoff characteristics of the pre- and post-construction conditions were then statistically compared to measure the benefits of integrated GI stormwater controls at the large scales.”

The following summary focuses on the EPA green infrastructure demonstration project in Kansas City, Missouri. Small scale monitoring examined infiltration of stormwater at several biofilters and rain gardens, while large scale monitoring included flow monitoring in a 40-ha area served with combined sewers having many retrofitted green infrastructure controls. Parallel large-scale monitoring was also conducted in an adjacent 37-ha control watershed with no stormwater controls. Figure 22 shows the test watershed and the placement of the stormwater controls which were restricted to the public rights-of-way to allow continued maintenance by the city. Table 7 shows the breakdown for the 179 controls used in this area.

Figure 22 Stormwater controls in the 40-ha test (pilot) study area in Kansas City, MO (Talebi 2014).

Table 7 Summary of stormwater controls in test area, by type (Talebi 2014).

Design plan component	Structural description	Number of stormwater control type
Bioretention	Bioretention without curb extension	24
	Curb extensions with bioretention	28
	Shallow bioretention	5
Bioswale	Vegetated swale infiltrates to background soil	1
Cascade	Terraced bioretention cells in series	5
Porous sidewalk or pavement	With underdrain	18
	With underground storage cubes	5
Rain garden	Rain garden without curb extension	64
	Curb extensions with rain gardens	8
Below grade storage	Retains stormwater control overflow and underdrain outflow from selected bioretention cells or porous pavement	21

Ten small-scale stormwater controls (curb extension biofilters, rain gardens, and cascade swales) were also monitored individually on small scales. An initial pre-retrofitting monitoring period included 69 events, and 37 events were monitored after completion of the retrofitting (including re-lining of the combined sewer). Analyses separated the direct runoff associated with rains from the base dry weather sanitary sewage flows. Figure 23 shows the schematic for the rainfall direct flow separation (based on the monitored sanitary flows for different days of the week and for different months). This example was for the Cincinnati portion of the project, but the process was the same for Kansas City.

Figure 23 Example of monitored flows after a storm event in August 2012 for Cincinnati State College combined sewer system at manhole number 29613032 (Talebi 2014).

Figure 24 plots the volumetric runoff coefficients (Rv, the ratio of the runoff depth to the rainfall depth) for the two sets of data, before and after the retrofitting of the controls. The Rv variations are large and mostly explained by the rain depths, as shown on Figure 25.

Figure 24 Volumetric runoff coefficients at UNKC01 before and after green infrastructure retrofitting (Talebi 2014).

Figure 25 Rain vs. runoff plots for UMKC01 during different monitoring periods as a function of rain depth (Talebi 2014).

The differences in the Rv values during the different monitoring periods were highly significant for all rain categories <38 mm, but larger rains did not indicate statistically significant differences. The stormwater quality model, WinSLAMM, was calibrated using the results of the small-scale stormwater controls and resulted in highly significant agreements using the verification test data at the downstream end of the test watershed.

The benefits of the retrofitted controls on the whole watershed discharges were directly related to the amount of runoff directed to the individual controls. During small rains, most of the runoff entered the controls and was infiltered. For larger rains, the individual controls were overwhelmed, with reduced benefits. Also, with retrofitted controls, not all the watershed flows can be directed to stormwater controls due to interferences with existing infrastructure and structures, and regulations restricting the use of the controls in specific areas (such as placement only on public lands). New construction can more effectively direct most of the runoff to stormwater controls as the interferences can be considered during the design process. Modeling must accurately consider the specific drainage areas to each stormwater control within the watershed to better calculate bypass overflow conditions.

2.2 Long term and large-scale monitoring of controls

The previous section described two research projects that included monitoring of large watersheds that had extensive use of stormwater controls. These indicated the variable success of applying information from monitoring of individual stormwater controls to large-scale applications of many controls throughout a watershed. Another scale process of concern in stormwater management modeling is the time scale, specifically how well stormwater controls keep functioning with time, with appropriate maintenance. Few data sets are available that have included long-term monitoring, especially when applied to large scales. It’s important that stormwater models consider changing performance with time due to clogging/siltation, scour, and breakthrough of stormwater controls, along with performance recovery with maintenance.

Pitt et al. (2022) examined possible changes in performance for 10 full-sized media treatment stormwater controls over a 6-to-9-year monitoring period. The ten stormwater controls were grouped into four types (“culvert modifications” which are media filters installed at drainage road crossings, detention bioswales having large subsurface storage, a large sedimentation pond/biofilter treatment train, and a sedimentation tank/media filter treatment train). The stormwater controls are located at the Santa Susana Field Laboratory (SSFL) in Ventura County, California, an industrial site being restored which had historic rocket engine testing (until 2006) and nuclear energy research (until 1988) activities. Current activities on the site include demolition, remediation, and restoration. Stormwater is regulated by the Los Angeles Regional Water Quality Control Board through a NPDES (National Pollutant Discharge Elimination System) permit that includes numeric effluent limits for many constituents, including metals, organic solvents, dioxins, and radionuclides, at multiple outfalls. The primary purpose of the stormwater controls is to reduce the occurrence of exceedances of NPDES permit limits at the regulated outfalls and to product the health and welfare of the nearby residences. Long-term monitoring at the outfalls have shown success in meeting this objective.

As part of the stormwater control strategy, many locations within the site are sampled to identify appropriate locations for distributed stormwater controls. This process has been on-going for more than ten years, along with performance monitoring at influent and effluent locations at selected stormwater controls to verify their performance and to identify maintenance issues. In almost all cases, the effluent concentrations tracked the influent concentrations, with no significant performance or effluent concentration differences with time. Silt clogging at one facility occurred as predicted based on initial laboratory tests, and the stormwater control was restored. Figure 26 shows an example of the accumulation of sediment in the pre-treatment sediment pond and in the final biofilter treatment unit over a period of four years. This shows the ability of the pre-treatment pond to protect the biofilter media over an extended period.

Figure 26 Lower lot sediment pond and biofilter sediment accumulation with years of operation (Pitt et al. 2022).

Many statistical analyses were used to identify and quantify changes in performance of these stormwater controls with time, stressing effluent quality for the critical constituents of concern. Effluent trends were mostly associated with influent trends due to changing site conditions. Particulate strength trends of influent and effluent samples also did not indicate any significant changes with time, supporting the lack of breakthrough of fine particulates during the monitoring period.

3 Need to Monitor Characteristics and Treatability of Emerging Contaminants of Concern

Stormwater monitoring has historically included flow rates and volumes, particulate solids (TSS and now also SSC), nutrients, bacteria, and various heavy metals. However, regulatory agencies are responding to concerns about emerging contaminants, including the inclusion of numeric effluent limits in stormwater discharge permits for a variety of constituents that have limited information. Information lacking includes discharge concentrations for different regions and land uses, their sources, and treatability, all topics of concern for stormwater model users.

The availability of information for stormwater emerging contaminants varies, including:

• Some information (periodic detailed research or specialized studies for basic stormwater discharges, but not comprehensive for all locations, seasons, or land uses): pesticides and PAHs (polycyclic aromatic hydrocarbons).
• Rare sources of information: PCBs (polychlorinated biphenyls) and dioxins.
• Very little information: microplastics, PFAS (per- and polyfluoroalkyl substances) compounds, and pharmaceuticals/personal care products.

To effectively model sources, treatability, transport, fates, and effects of these little understood contaminants, much continued monitoring is needed. The most critical data need is knowing associations of the contaminants with filtered (“dissolved”) portions of stormwater and associations with a range of particle sizes, for different land uses and source areas.

It is well known that stormwater characteristics vary considerably. Geographical area and land use have been identified as important factors affecting base flow and stormwater runoff quality, for example. Bochis (2010), during her Ph.D. research, evaluated stormwater quality as a function of numerous factors, including land use, season, and geographical area, using data from more than 9,000 events contained in the National Stormwater Quality Database (NSQD), available at: https://bmpdatabase.org/national-stormwater-quality-database. Bochis (2010) compared the main effects and interactions of these factors affecting stormwater quality, as shown Table 8 for different common pollutants. Yellow and green cells note statistically significant relationships (p<0.05), and yellow interactions should be used for predictive modeling purposes. Land use is a consistent factor affecting the observed variation in stormwater quality, while location (EPA rain zones) is also very important, while season alone had little effect.

Table 8 Main factors and interactions affecting stormwater quality (Bochis 2010).

Constituent	Land Use (LU)	Season (SN)	EPA Rain Zone (EPA)	LU*SN	LU*EPA	SN*EPA	LU*EPA*SN
TSS mg/L	<0.0001	0.74	<0.0001	0.017	<0.0001	0.18	<0.0001
BOD5 mg/L	<0.0001	0.16	<0.0001	0.0008	<0.0001	0.0011	0.22
COD mg/L	<0.0001	0.13	<0.0001	0.034	<0.0001	0.014	0.0085
TP mg/L	<0.0001	0.69	<0.0001	0.055	<0.0001	0.0004	<0.0001
NO2 + NO3 mg/L	<0.0001	0.11	<0.0001	0.052	<0.0001	0.034	0.057
TKN mg/L	0.0026	0.024	<0.0001	0.99	<0.0001	<0.0001	0.17
Cu mg/L	<0.0001	0.11	<0.0001	0.62	<0.0001	0.038	0.14
Pb mg/L	<0.0001	0.76	<0.0001	0.42	<0.0001	0.29	0.011
Zn mg/L	<0.0001	0.91	<0.0001	0.94	<0.0001	0.014	<0.0001

When evaluating emerging contaminants for general stormwater characteristics, it is important that these factors all be considered. A single monitoring area, with a small number of different land uses, can provide critical preliminary information, but a much more comprehensive monitoring program is needed to better understand the range and variability of constituent characteristics. Suitable numbers of sampled events are also necessary for each condition considering appropriate data quality objectives (power and confidence). Within each land use, source areas (roofs, parking areas, streets, landscaped areas, etc.) can also have different attributes and need to be evaluated when developing modeling tools.

Pitt et al. (2018) evaluated stormwater particles sizes in relation to their fates after discharge to open waters. Settling rates for different particulate size ranges were calculated using Newton’s (turbulent) and Reynold’s (laminar) settling equations. Figure 27 plots the approximate settling times needed for four particle size ranges examined, for 3 m to 30 m water depths.

Figure 27 Settling times (hours) for different particle size ranges and water depths (Pitt et al. 2018).

Near field effects are associated with the largest stormwater particles (>63 μm) and would need about 1 hour, or less, to settle in 30 m of water, and only about 5 minutes, or less, to settle in 3 m of water. These particles have the greatest potential of affecting areas close to a discharge location and would not be widely dispersed. Far field effects are associated with intermediate particles (20 to 63 μm) and would require about 50 hours to settle in 30 m of water and about 5 hours to settle in 3 m of water. These particles would affect sediments located further from the discharge location or be dispersed throughout an enclosed bay or lake. The smallest particles (<20 μm) would require even longer times to settle, about 500+ hours in 30 m of water and 50+ hours to settle in 3 m of water. Unless impounded, these particles would likely be transported a large distance beyond the discharge location. Filtered (or “dissolved”) forms of the constituents would also be widely dispersed and generally travel with the discharged water and may undergo biochemical reactions with other water constituents.

Knowledge of particulate associations (and filtered forms) of emerging contaminants are needed to understand their transport and fate in receiving waters, and how they can be controlled using different stormwater treatment processes.

Pitt et al. (2018) presented monitoring data for a range of pollutants in a mixed residential/commercial area and at a military base in San Diego, CA. The focus was on heavy metals, PAHs, and PCBs, specifically investigating particle size relationships for determining fates after discharge. Figure 28 shows the total PCB particulate strength values for all samples combined, by particle size range. It is apparent that the smallest particle sizes (0.7 to 2.7 μm) have a wider range with larger observed values than the larger particle sizes. These small particles have minimal near shore effects on the marine sediments but are also harder to reduce using conventional stormwater controls.

Figure 28 Sum of PCB particulate strengths by particle size (Pitt et al. 2018).

Pitt et al. 2026 also evaluated monitoring data for PFAS compounds at several San Diego, CA, Puget Sound, WA, and Lubbock, TX military bases having a variety of land uses. The focus of these monitoring activities was to characterize PFAS compounds and to measure their control using several types of stormwater controls. Table 9 summarizes the observed pollutant reductions for media filters and biofilters for the monitored PFAS compounds. PFOS (perfluorooctane sulfonic acid) was generally the most abundant but also had poor removals with the stormwater controls tested.

Table 9 Removal performance for PFAS compounds (Pitt et al. 2026).

	Low "constant" effluent concentrations	High reductions	Moderate reductions	Low reductions	No change in concentrations with treatment
Media Filters			PFPeA filt	PFBA filt	PFBA part
				PFOS filt	PFDA filt
				PFOS part	PFDA part
					PFHpA part
					PFHxA part
					PFNA part
					PFOA filt
					PFOA part
Biofilter	PFHpA part	PFDoA part	PFDA filt		PFOS part
		PFHpA filt	PFHxA part
			PFNA filt
			PFNA part
			PFOA part

The media filter removals of PFAS compounds were mixed, with many showing no reductions in effluent concentrations compared to influent concentrations, with only a few having moderate and low reductions. The biofilters generally performed better and had moderate to high removals of several PFAS compounds, but one (particulate PFOS) had no reduction in effluent concentrations, while one (particulate PFHpA, perfluoroheptanoic acid) had constant low effluent concentrations.

Pitt et al. (1999) showed that full-scale projects in the field using actual stormwater are needed to best evaluate removals of emerging contaminants. The use of “simulated” stormwater can be misleading in the absence of interfering compounds and unrepresentative particle size characteristics of the pollutants of concern. However, long-term laboratory pilot-scale studies (again using actual stormwater) can be very useful when comparing and selecting treatment unit processes to be verified in the field using full-scale installations (Pitt et al. 2021).

4 Conclusions

This paper presented a case history of street cleaning as an example of how early data and assumptions have been tested and updated with corresponding modifications in stormwater modeling addressing pollutant sources, transport, and control. Monitoring showed how street cleaning varies for different operating conditions (street cleaner type and speed) and environmental conditions (street dirt loading, particle sizes, and pavement texture). In addition, monitoring revealed actual accumulation and wash-off rates for street surface particulates. This knowledge led to a better understanding of other sources of stormwater pollutants in urban areas and pollutant transport processes, resulting in more accurate pollutant mass balance calculations.

The role of scale is also briefly described in this paper. Two case studies are summarized indicating if models can accurately represent combinations of distributed stormwater controls throughout an area. Both monitoring programs were able to successfully represent stormwater runoff reductions associated with the infiltrating controls, but the results for stormwater quality were mixed, probably due to the new development having unstable soils and other uncontrollable sources of the contaminants. The retrofitted study was hindered as stormwater controls could only be used in the public rights-of-way, and few effluent measurements were available at the small stormwater controls due to their success in infiltrating most of the influent flows. Micro-scale flow pattern analyses were found to be important to accurately model the combined outfall flows for these examples.

Other scales of interest are associated with seasonal variations in stormwater characteristics, especially in northern areas experiencing large amounts of snowmelt. Warm weather stormwater may only contribute a portion of the total flows from some areas. Monitoring, and associated models, need to consider these other flows when protecting receiving waters (both groundwater and surface water). Other scales associated with time relate to changes in performance of stormwater controls with time. Too few monitoring programs have examined these changes over many years to recognize failures and to determine needed maintenance. Models typically do not consider these changes with time and therefore cannot accurately calculate future conditions well.

A critical monitoring need, and concurrent model development, relates to poorly understood emerging contaminants of concern, such as microplastics, PFAS compounds, and pharmaceuticals, while slightly more information is available for stormwater PCBs and dioxins, and more information is available for pesticides and PAHs. Most of these are only present in very low concentrations in most stormwater, but even these low concentrations can be cause of concern. Some (dioxins, PCBs, and PAHs) are strongly associated with stormwater particulates, but others can have significant filterable portions. Sources, their associations with different particle sizes, and their treatability, need to be better understood for these compounds, and appropriate model enhancements are needed to address the range of characteristics of these compounds of concern.

There are many other examples where monitoring is still needed to verify prior assumptions that have been incorporated into models, in addition to the need to address future issues that are not recognized yet that may have significant detrimental effects on human and environmental health. Methods to extrapolate current monitoring data to future climate conditions are needed, along with modifications to stormwater models to consider these changes.

ACKNOWLEDGEMENTS

Many research sponsors have supported our research over the years, some described in this paper, including: US EPA, US Navy, NSF, Metropolitan Sewer District of Greater Cincinnati, Kansas City, Toronto, Ontario Ministry of the Environment, City of Tuscaloosa, TetraTech, Geosyntec, HydroInternational, Institute of Scrap Recycling Industries, The Boeing Co., and Beijing Municipal Government.

Most importantly, many students have contributed to research during their graduate studies, including, but not limited to: Dr. Laith Alfaqih, Dr. Humberto Avila, Dr. Jejal Bathi, Dr. Ryan Bean, Yezhao Cai, Dr. Vijay Eppakayala, Dr. Kenya Goodson, Dr. Alex Maestre, Dr. Renee Morquecho, Yukio Nara, Dr. Olga Ogburn, Dr. Radahegn Sileshi, Dr. Noburo Togawa, and Dr. Brad Wilson.

Also, colleagues and their students at other institutions that participated in research activities included: Penn State Harrisburg (Dr. Shirley Clark), University of Missouri, Kansas City (Dr. Deborah O’Bannon), Universidad de los Andes (Dr. Mario Diazganados Ortiz), and Texas Tech University (Dr. Danny Reible).

Colleagues at URS Research Co, Woodward Clyde Consultants, and the Wisconsin Department of Natural Resources also participated on past stormwater research projects.

Many of our research reports, student’s theses, and dissertations are available at my teaching and research web site at: https://www.winslamm.net/dr_pitt_presentations_and_publications.html

References

1. Bochis, E-C. 2010. Characteristics of Urban Development and Associated Stormwater Quality. Ph.D. dissertation. Department of Civil, Construction, and Environmental Engineering, the University of Alabama. https://www.winslamm.net/assets/files/presentations_and_publications/17%20Theses%20and%20Dissertations/2010%20Bochis%20dissertation%20Land%20development%20and%20stormwater%20characteristics.pdf
2. Lee, H., J.D. Sartor, and W.H. Van Horn. 1959. Stoneman II. Test of Reclamation Performance, Vol, III. Performance Characteristics of Dry Decontamination Procedures. Research and Development Technical Report USNRDL-TR-336. US Army. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1959%20Lee%20et%20al%20Stoneman%20II%20Test%20Of%20Reclamation%20Performance.pdf
3. Pitt, R. 1979. Demonstration of Nonpoint Pollution Abatement Through Improved Street Cleaning Practices. EPA-600/2-79-161, U.S. Environmental Protection Agency, Cincinnati, Ohio. pp 270. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1979%20Pitt%20San%20Jose%20street%20cleaning%20EPA%20report.pdf
4. Pitt, R. 1983. Urban Bacteria Sources and Control in the Lower Rideau River Watershed, Ottawa, Ontario. Ontario Ministry of the Environment, ISBN 0-7743-8487-5. pp 165. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1983%20Pitt%20Rideau%20River%20Storwater%20Management%20Study%20Technical%20Report.pdf
5. Pitt, R. 1985. Characterizing and Controlling Urban Runoff through Street and Sewerage Cleaning. U.S. Environmental Protection Agency, Storm and Combined Sewer Program, Risk Reduction Engineering Laboratory.  EPA/600/S2-85/038. PB 85-186500. Cincinnati, Ohio. pp 467. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1985%20Pitt%20Bellevue%20Characterization,%20Source,%20and%20Control%20of%20Urban%20Runoff%20by%20Street%20and%20Sewerage%20Cleaning%20EPA%20report.pdf
6. Pitt, R. and G. Amy. 1973. Toxic Materials Analyses of Street Surface Contaminants. EPA-R2-73-283, U.S. Environmental Protection Agency, Washington, D.C. pp 134. https://nepis.epa.gov/Exe/ZyNET.exe/9100RVDW.TXT?ZyActionD=ZyDocument&Client=EPA&Index=Prior+to+1976&Docs=&Query=&Time=&EndTime=&SearchMethod=1&TocRestrict=n&Toc=&TocEntry=&QField=&QFieldYear=&QFieldMonth=&QFieldDay=&IntQFieldOp=0&ExtQFieldOp=0&XmlQuery=&File=D%3A%5Czyfiles%5CIndex%20Data%5C70thru75%5CTxt%5C00000014%5C9100RVDW.txt&User=ANONYMOUS&Password=anonymous&SortMethod=h%7C-&MaximumDocuments=1&FuzzyDegree=0&ImageQuality=r75g8/r75g8/x150y150g16/i425&Display=hpfr&DefSeekPage=x&SearchBack=ZyActionL&Back=ZyActionS&BackDesc=Results%20page&MaximumPages=1&ZyEntry=1&SeekPage=x&ZyPURL
7. Pitt, R. and J. McLean. 1986. Humber River Pilot Watershed Project. Ontario Ministry of the Environment, Toronto, Canada. pp 483. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1986%20Pitt%20and%20McLean%20Toronto%20Area%20Watershed%20Management%20Strategy%20Study%20(Volume%20I%20and%20II).pdf
8. Pitt, R., B. Robertson, P. Barron, A. Ayyoubi, and S. Clark. 1999. Stormwater Treatment at Critical Areas: The Multi-Chambered Treatment Train (MCTT). U.S. Environmental Protection Agency, Wet Weather Flow Management Program, National Risk Management Research Laboratory. EPA/600/R-99/017. Cincinnati, Ohio. 505 pp. March 1999. https://www.winslamm.net/assets/files/presentations_and_publications/11f%20treatment%20trains%20including%20upflow%20and%20MCTT/1999%20Pitt%20et%20al%20MCTT%20EPA%20report.pdf. 
9. Pitt, R., M. Otto, A. Questad, S. Isaac, M. Coyer, B. Steets, R. Gearheart, J. Jones, M. Josselyn, M. Stenstrom, S. Clark, P. Costa, and J. Wokurka. 2021. “Laboratory media test comparisons to long-term performance of biofilter, media filter, and treatment-train stormwater controls.” Journal of Sustainable Water in the Built Environment 7(4). https://doi.org/10.1061/JSWBAY.0000956
10. Pitt, R., M. Otto, A. Questad, S. Isaac, M. Coyer, B. Steets, R. Gearheart, J. Jones, M. Josselyn, M. Stenstrom, P. Costa, and J. Wokurka. 2022. “Changes in performance during long-term monitoring of full-scale media filter stormwater controls at an industrial site.” Journal of Sustainable Water in the Built Environment 8 (1). https://doi.org/10.1061/JSWBAY.0000965
11. Pitt, R., B. Rao, B. Steets, C. Gomez-Avila, H. Zhou, T. Hussain, et al. 2026 “Heavy Metals, PAHs, and PFAS Stormwater Concentrations and Treatment at Seven Current and Recommissioned Military Facilities.” In review. Journal of Watershed Management Modeling.
12. Pitt, R., S. Schal, M. Otto, and B. Steets. 2018. Appendix V: Paleta Creek, San Diego, Stormwater Monitoring and Data Analysis Report. Assessment and Management of Stormwater Impacts on Sediment Recontamination. ORSP Number 13-PAF05133. Strategic Environmental Research and Development Program (SERDP). pp 187. https://www.winslamm.net/assets/files/presentations_and_publications/20%20SERDP%20Stormwater%20monitoring%20and%20modeling%20reports%202018%20to%202023/2018%20Pitt%20et%20al%20Serdp%20first%20phase%20NBSD%20SW%20report.pdf
13. Sartor, J.D. and G.B. Boyd. 1972. Water Pollution Aspects of Street Surface Contaminants. US Environmental Protection Agency. EPA-R2-72-081. https://www.winslamm.net/assets/files/presentations_and_publications/11a%20street%20cleaning/1972%20Sartor%20and%20Boyd%20Water%20Pollution%20Aspects%20Of%20Street%20Surface%20Contaminants%20EPA%20report.pdf
14. Selbig, W.R. and R.T. Bannerman. 2008. A Comparison of Runoff Quantity and Quality from Two Small Basins Undergoing Implementation of Conventional- and Low-Impact-Development (LID) Strategies: Cross Plains, Wisconsin, Water Years 1999–2005. U.S. Geological Survey Scientific Investigations Report 2008–5008. https://www.winslamm.net/assets/files/presentations_and_publications/11%20Stormwater%20Treatment/2008%20Selbig%20et%20al%20Cross%20Plains%20LID%20sir_2008-5008.pdf
15. Talebi, L. 2014. Assessment of Integrated Green Infrastructure-Based Stormwater Controls in Small to Large Scale Developed Urban Watersheds. Ph.D. dissertation.  Department of Civil, Construction, and Environmental Engineering, the University of Alabama. https://www.winslamm.net/assets/files/presentations_and_publications/17%20Theses%20and%20Dissertations/2014%20Talebi%20dissertation%20Stormwater%20green%20infrastructure%20at%20different%20watershed%20scales.pdf