Method description

StormTac Web is a commercial conceptual low-complexity conceptual model (LCCM) stormwater and recipient (receiving water) model. It is today, after the start of operation in 2000 and with continuous updates, a fully functioning operative Web application. The application was developed by StormTac Corporation. It is used by consulting and construction companies, municipalities and universities for modelling urban stormwater pollutant loads and their effects on receiving waters, such as watercourses, lakes and sea bays.

The model complexity has been adapted to the scope of input data typically available in different stages of urban drainage projects, ranging from the planning to construction of stormwater treatment and flow detention facilities. Only one model is needed for stormwater management analysis, including the calculation of (I) annual transport calculations of stormwater and baseflow, (II) impacts on the chemistry of receiving waters, (III) acceptable (allowable) annual pollutant loads or required pollutant load reductions, and the design of transport, flow detention and pollutant treatment facilities.

A schematic flowchart of StormTac Web is shown in the following figure. The model consists of the 5 interlinked modules (1) Runoff and baseflow, (2) Pollutant transport, (3) Stormwater treatment (4) Receiving water, and (5) Flow detention and transport. The notation “transport” is here understood as the annual transport of runoff and pollutants from the (sub)watershed area to the outlet to the receiving water, or to pollutant treatment or flow detention facilities.

 

 

 

In simple applications, the model requires very little input data: (1) the catchment area (ha) per land use, and (2) annual precipitation data (mm/year). The area and volume of the receiving water body is required to calculate acceptable pollutant loads to the receiving water body and the required load reductions. There are default values for all input data in the application, typical after recommendations from literature findings. These can be changed to reflect site-specific conditions. More site-specific annual runoff flow rates and pollutant loads can be obtained by substituting locally measured input data into the model, e.g., annual precipitation (mm/year), flow weighted stormwater (and baseflow) pollutant concentrations (µg/l) and receiving water pollutant concentrations (µg/l).

The model uses input data forms and presents results in a flowchart and in automatically generated reports, see the following figure.

The model is used as a tool for action planning in urban water management and is suitable for water quantity and quality calculations within watersheds (catchment areas). It integrates processes of runoff, transport, recipient, treatment, and flow detention. The simplicity of a LCCM model such as StormTac Web provides the advantage that it can be used with minimum training. It may offer comparable levels of uncertainty as complex dynamic models, regarding the simulating of annual runoff flows and pollutant concentrations and loads. Urban drainage models can be simplified without losing modelling accuracy according to Dotto et al. (2009), who investigated the model MUSIC which is widely used in Australian stormwater practices. They found that only 2 out of 13 calibration parameters of the rainfall/runoff model matter. The model results were insensitive to the remaining 11 parameters.

The article by Larm et al. (2022) examines simpler stormwater models (LCCM), their methodology, calculated uncertainty, and their usefulness when input data is limited, as in most stormwater investigations and projects with areas planned to be exploited. According to Professor Jiri Marsalek (LTU and Burlington Research Institute, Canada), calibration of stormwater models is a major challenge, especially at the planning level, as the physical system under construction does not yet exist, with which calibration data must be transferred from similar existing areas.

The data-driven approach, implemented within the framework of StormTac Web and associated database, has proven to be a practical tool to facilitate the analysis of stormwater quality and pollution control in urban catchment areas. This is especially true for planned cases that cannot be measured and are therefore necessary to model. This also applies to existing areas as it is very costly to measure flows and pollutants, which would require flow proportional sampling over a longer period (Larm et al. 2022).

The three most important input parameters in the StormTac Web model are precipitation, runoff coefficients and default land-use-specific concentrations. The model may avoid introduction of additional uncertainties caused by over-parametrization and excessive details, compared with high complexity urban drainage models. However, in some projects, more detailed data may be needed, e.g. when addressing acute toxicity of stormwater impacts on the chemistry in receiving waters or when addressing flooding levels in stormwater sewers and upstream infrastructure.

Unique model properties
The unique properties of the model especially refer to the aspects below:

• It is simple to use and consists of an overall system presentation from a flowchart of the entire watershed system (using clickable boxes and input forms in the flowchart). A changed input from the flowchart results in presented changed outputs within the flowchart (watershed system).

• The parameters and methods are continuously being updated with more input data.

• It requires little input data and no large manual or comprehensive education (help notes are included in the file), however a short manual that let you begin quickly (“within one hour”) with the calculations is included in the included guide.

• It integrates watershed and runoff properties with treatment/detention facilities and impacts on receiving waters; all in one model.

• It includes a database with continuously updated precipitation data, runoff coefficients, concentration data and reduction efficiencies. Within the database, a cost model for stormwater facilities is included.

• It includes a fine specification of both urban and rural land uses (>100 land uses, among them residential areas, e.g. townhouses and apartment blocks, thoroughfares (roads), commercial areas, industrial areas, parks, meadows, forest and agricultural lands) and a large set of nutrients and pollutants (around 80 pollutants, among them substances included in the WFD, such as heavy metals, suspended solids and various organic compounds). The number of land uses and pollutants is increasing with new updates.

Main methodology
The main methodology has been reviewed internationally through scientific papers and a doctoral thesis. The methods are described in more details under publications.

For selecting project specific volumetric runoff coefficients and pollutant concentrations, the users have two options: (1) select typical (default) values from the model application, or (2) use their own site-specific data.

The implemented typical pollutant concentration data for different land-uses in the application are taken from compiled data in the StormTac database, as annual event-mean concentrations (EMCs) obtained from case studies with a single specific land-use, using flow proportional sampling and analysis of stormwater from long periods, at least several months, but preferably one year or more.

The model has been developed to automate the calculations by using land use specific “typical” (default) values. It is best suited for long-term predictions. Site specific yearly precipitation data and rain intensities can be used.

Runoff water flow is calculated from precipitation data and land use specific runoff coefficients and areas.

For runoff and pollutant simulations with StormTac Web, users need to define the runoff contributing (sub)watershed area (A) and the groundwater collection area (the baseflow area; Ab). The runoff and groundwater watershed boundaries are generally assumed to coincide, a common and often justified assumption (SKB, 2003).

Pollutant load rate is quantified from calculated flow and from typical concentrations.

Calculations of pollutant concentrations and loads in StormTac Web are based on land use specific concentrations (Larm, 2000). There are significant differences in concentrations from different types of land-uses (Pitt et al., 2004). There are options for model users to further define and customize the selections of substances and land uses.

The typical concentrations are estimated empirically from a large set of flow proportional field sampling data, which contributes to their general applicability. These are tabled as typical, minimum and maximum values. The data are available in StormTac database.

Base flow, base flow concentrations and loads are calculated using specific coefficients (infiltration rates and coefficients for leakage/connection into ditches, lakes and stormwater sewers) and typical concentrations for base flow (different for different land uses, from measured base flow concentrations). As is the case with typical concentrations and runoff coefficients for runoff, the base flow coefficients and concentration data can also be changed by the user. These data are available in StormTac database.

StormTac Web includes a large amount of sub models and equations for the design of different stormwater facilities. The user can choose between a relatively detailed or a quick and simple design. The resulted dimensions by using different methods and by changing parameter values can easily be reviewed and compared. Examples of included design parameters are runoff coefficients, land use areas, facility permanent water depth, water depth of detention volume, slope, design rain depth, outflow, emptying time and reduction efficiency.

The design methods have been employed for a large number of case studies from pre-studies to final detailed construction drawings.

The employed equations are presented in the flowchart when hovering the mouse over the presented data.

The method used in StormTac Web to calculate the annual pollution load from catchments is based on the product of the annual pollutant concentrations of various land use and annual flow, where the annual flow in turn is calculated from the product of the annual rainfall, area and volume runoff coefficient. The method allows the direct calibration against the flow and concentration, and it is peer reviewed (Larm, 2000).

Observations usually show that the stormwater event mean concentration (EMCs) is poorly or not at all correlated with event stormwater flow or stormwater volume. This indicates that it is adequate to assume that the concentration for a specific pollutant is constant during individual events, at all times, according to ASCE (1994) and Novotny (1995), even if the EMCs can vary from event to event. StormTac Web therefore uses flow-weighted data during long periods, sampled from specific land-use sub-catchments to estimate such data. These are compiled in the StormTac database, including data from e.g. the databases of NURP (USEPA, 1983) and NSQD (Pitt et al., 2004). An annual runoff volume can be multiplied with this constant concentration to produce an annual runoff load. This simplified approach is best suited to estimation of long-term (annual or seasonal) pollutant loads, since “simple prediction methods generally perform better over a long averaging time and poorly at the level of a single storm event” (ASCE, 1994).

Measured values from the study area can otherwise be used (ASCE, 1994). These measured values can replace the default land use specific values implemented in the StormTac Web model.

A large number of data from river basins in the United States in the 1980s were used and the pollution load was calculated in different models. The two most significant variables for the calculation proved to be the total annual rainfall and basin area. For some of the models were also among other variables impermeable area and land use significant (Novotny, 1995; Sing, 1995).

Encouragingly, rainfall and runoff functions show good potential for predicting pollutant loads (Francey, 2010).

Required input data
The model requires very little input data. Watershed area (ha) per land use (e.g. residential area, roads and woodland) is the only obligatory input data. Information on the traffic intensity (vehicles/day) is needed if studying the loads from larger roads within the catchment area. The area and volume of the receiving water are needed for estimating allowable loads. The included databases help to make more accurate analyses by letting you change other input data such as precipitation, runoff coefficients and water depths or slopes of facilities.

The model parameters can be calibrated to measured data to ensure that site specific conditions are being considered. In such cases further input data consist of measured flow, precipitation, rain intensity and sampled concentration (mg/l or μg/l) in stormwater, base flow and/or the receiving water.

Water quantity
An increased urbanization and climate effects may cause an increasing number of floods. StormTac Web can calculate the capacity of the transport system and required detention volumes for the design rain return time and rain duration, including the implementation of climate factors.

StormTac Web quantity calculations include:

• The quantification of yearly average water flows (yearly runoff volumes of stormwater, base flow and groundwater) and runoff flows during average rain events.

• Calculation of design flows for different return times, including climate factors.

• Design of stormwater transport systems, e.g. sewers, ditches and channels.

• Calculation of the flow capacity for new designed transport systems, as well as for existing systems.

• Design of stormwater flow detention facilities, e.g. dry/wet ponds and detention basins.

Water quality
Metals and nutrients are examples of pollutants in stormwater that may cause toxic and eutrophic effects in the receiving waters. StormTac Web is the tool that can be of great use in the development of a more sustainable stormwater management.

StormTac Web can be used as a simple-to-use forecast tool (and as such requiring little input data) for water quality and Action plans for stormwater and surface water, e.g. to be used within the EU Water Framework Directive (WFD). It includes a large number (>80) of substances, of which several are included in the WFD.

StormTac Web quality calculations include:

• The quantification of yearly average pollutant concentrations and loads in the discharge points and from different land uses.

• Comparison of measured concentration data to calculated values.

• Identification of the largest pollutant sources and discharge locations to a recipient, presenting loads from different land uses and loads from different materials (such as copper roofs) if these specific areas have been set up as input.

• Presentation of data from up to 99 sub watershed areas in each project, to be used in e.g. Action plans for stormwater management in a whole municipality or for different lakes or water courses.

• Design of stormwater treatment facilities (e.g. areas and volumes of wet ponds, constructed wetlands, biofilters, swales, ditches, underground filter basins, underground detention basins and filter strips) regarding used criteria for stormwater concentrations in the discharges and/or allowable loads and surface water quality criteria.

• Calculation of treatment reduction efficiencies (% or in- and outlet concentrations and loads) for designed or existing facilities for site specific and more reliable calculations, such as the effects of inlet and outlet concentrations, the share of water vegetation, flow detention, hydraulic efficiency (length:width ratio etc.) and temperature.

• Setting up water and mass balances for receiving waters (lakes, sea bays and water courses), including calculation of net internal loading from the sediments (kg/year) or net sedimentation load to the sediments (kg/year).

• Calculation of required treatment load to reach allowable (acceptable) loads to the receiving waters, considering water quality criteria in the receiving waters (μg/l), as those stated in the WFD.

• Calculation of the new concentrations in the receiving water after reduced load after a designed treatment facility or after changed land use in the watershed area, e.g. after a planned residential area on an existing woodland area. Comparison to water quality criteria, presenting need for more treatment (larger facility, different facilities or more facilities in other sub watershed areas).

Land use specific concentrations
Pollutant calculations in StormTac Web is based on estimated land use specific concentrations, recommended as default values. These values have been developed by performing a critical assessment of the raw statistical data compiled in the StormTac database, including analyzing temporal trends, performing calibrations in case studies, and comparisons of data from similar land uses. Therefore, the default concentrations generally do not represent median values of the data of specific land use sub catchments. Some of the references to the data included in the database may comprise several case studies. Therefore, that data may have been assigned a greater weight, compared to the data from other references. In the case of land uses, for which no flow-proportional data are available for certain pollutants, the default concentrations are partly based on calibrations against the data available from case studies, and/or based on weighing of the reliable available data for other comparable land uses.

The estimations of default concentrations in StormTac Web are based primarily on land use, which raises the question, whether there is another method that would better describe the stormwater quality. One example of such an alternative method, incorporated in StormTac Web, is the empirical equations describing pollutant concentrations as functions of Average Daily Traffic intensities (ADT). The literature on this topic indicates that a number of other factors (besides land use and ADT) could be considered as determinants of stormwater quality. Examples of such factors are the catchment area, geographical location, seasonal impacts, percentage of imperviousness and the type of drainage conveyance (Pitt et al., 2004).

Two drainage areas with the same size, percentage of imperviousness, ground slope, sampling methods, and stormwater controls will produce different stormwater concentrations if the main activity in one watershed is for example an automobile manufacturing facility (industrial land use) while the other is for example a shopping Centre (commercial land use). Previous studies indicated that there are significant differences in stormwater constituents for different land use categories (Pitt et al., 2004). One question to be addressed here is however if there is a different classification method that better describes stormwater quality, possibly by also considering such factors as geographical area, season, percentage of imperviousness, watershed area, type of conveyance, controls in the watershed, sampling method, and type of sample compositing, and possible interactions between these factors.

The Nationwide Urban Runoff Program (NURP) (EPA, 1983) concluded that concentrations for different land uses were not significantly different, so all their data were pooled into a single category. The National Stormwater Quality Database (NSQD, version 1.1, 2004) is unique in that detailed descriptions of the test areas and sampling conditions are being tabulated. This project also involved extensive quality assurance/quality control evaluations of these data; and performing statistical analyses and summaries of these data. The very large number of samples represented in the NSQD resulted in statistically significant differences of land use concentrations being identified. The NSQD is much more representative of a broader range of land uses, while almost all of the NURP data was obtained from residential areas (Pitt and Maestre, 2005). Statistical analyses found significant differences for land use categories for all pollutants. This is notable because National Urban Runoff Program (NURP) findings showed no significant differences in urban runoff concentrations as a function of common urban land uses (EPA, 1983), likely because they had few data from non-residential areas.

The National Stormwater Quality Database (NSQD) is an urban stormwater runoff characterization (quality) database developed under the direction of Dr. Robert Pitt, P.E., of the University of Alabama and the Center for Watershed Protection, under support from the U.S. Environmental Protection Agency. Originally released in 2001, followed by several updates by Dr. Pitt and Dr. Alexander Maestre (also at University of Alabama), it has moved to a companion project to the International Stormwater BMP Database. The updated database (Pitt, 2015) includes data from the databases NURP, BMP Database and NSQD. It consists of compiled measured data from 10 000 events throughout the US. Mean data from the automatic flow proportional samples from the updated database have been calculated for the different land uses and have been compiled in StormTac database, used among other studies for estimating the land use specific typical concentrations in StormTac Web.

The conclusion is that many of the constituents do have significant concentration differences by land uses, shown by analyzing the large set of data from the NSQD database. Statistical ANOVA analyses for all land use categories (using SYSTAT) found significant differences for land use categories for all pollutants.

The open space COD concentrations are the lowest, and the freeway COD concentrations are the largest for most of the data range.

Total Kjeldahl Nitrogen (TKN), copper, lead, and zinc observations are lowest for open space areas, as for most constituents (Pitt and Maestre, 2005).

Freeway locations generally had the highest median values, except for phosphorus, nitrates, fecal coliforms, and zinc.

Industrial and institutional sites had the highest reported zinc concentrations (Pitt, Maestre and Morquecho, 2004).

One of the conclusions of the final NURP report was that the Event Mean Concentrations (EMCs) of stormwater constituents were described by lognormal distributions. This finding has been re-evaluated, with the conclusion that not all stormwater constituents were adequately described by lognormal distributions (Van Buren, 1997; Beherra, 2000). Most of the stormwater constituents, however, can be assumed to follow a lognormal distribution with little error (Maestre and Pitt, 2004).

Description and use of typical concentrations
The default “typical values” in the StormTac Web model are not “standard values” but are updated reliably sampled measurement data from areas with specified land use. They should not be compared against grab samples that only give instantaneous values. The typical values are based on median values from flow proportionally updated long-term sampling compiled in StormTac Database.
They have been adjusted after comparison with data from similar land uses, calibration against measurement data and after a study of time trends. They refer to normal areas per land use. The typical values are continuously adjusted to new data and feedback from users as well as calibration to new measurement data.
The values for the project can be changed by studying current data in StormTac Database, where references and info from the case studies are available, and where data can be filtered. The values will not change in the specific opened project file but will change when starting new projects according to continuously added data in the database. Different values are differently safe, which is stated in the database and in the results report
More information about the methodology around the values can be found in the guide and scientifically reviewed articles that can be downloaded from the website. Feedback on suggestions for adjustments is gratefully received via support@stormtac.com.

Typical concentration data can be downloaded, and the corresponding file presents typical, minimum and maximum concentrations for different urban and rural land uses. The typical concentrations should only be used when the stormwater pollutant load from the studied land use is considered to be of average quantity, else values closer to the presented min- or max-values should be used. The background colors in the tables indicate the level of uncertainty, based on the number of data values and their uncertainties.

Observe that when using runoff coefficients and typical concentrations for runoff, only the runoff (stormwater) part is calculated. In StormTac Web the base flow part is also calculated, see method description above and a separate table for base flow concentrations in the same data file.

The typical concentrations for each land use in StormTac Web are mainly based on long-term flow proportional sampling and refer to annual average. References for these samplings are compiled in the StormTac data base.

For some few specific land uses, the only data available is from grab samples. Grab samples are not comprehensive and often tend to underestimate the concentration. This is considered when using such data. See below under the headline “Comparison between calculated and concentration data” for the differences between grab samples and flow proportional sampling.

The typical concentrations are updated continuously. When new reliable data for a land use is implemented in the database, data for other comparable land uses is being revaluated. The typical concentrations are also calibrated to consider different time trends, see below. Therefore, in the StormTac data base, the typical values are not always a median value based on flow proportional sampling but may be a combination of the above described methodology.

Use of typical concentrations
The calculations can be performed for both small and large areas, e.g. different sub areas in a multi-family housing area (e.g. roof, courtyard, local street, parking and park area) or for a whole municipality or a whole river basin divided into sub-basin areas where the areas consist of entire residential areas, forest areas, thoroughfares, etc. Which kind of land use characterization to use depends on the substances to be investigated and the amount of data available for these from different land use in the area, and the purpose for the calculations.

If you only calculate the pollutant load from a forest area to be exploited into a multi-family residential area, it is recommended to calculate for the entire forest area before exploitation and then calculate for the entire multi-family area. If the stormwater from the entire multi-family housing area is planned to be led to a stormwater treatment facility downstream, it is also recommended not to divide the area into sub areas.

If you need to calculate the load for different local measures in the multi-family housing area, you need to divide the area into different areas that are led to different local facilities (BMPs, SUDs etc.), and in areas where the stormwater is not planned to be treated.

For detailed calculations, site-specific typical concentrations should be used where values are adjusted between the default typical, minimum and maximum values. StormTac Web contains factors (0-10) that will be used to calculate land use specific concentrations closer to minimum or maximum values in the database, depending on the site-specific conditions. A factor 5 indicates normal conditions for the land use, while a factor <5 indicates that the levels are reduced towards the minimum value and a factor> 5 towards the maximum value in the database. Copper roofs in a residential area can increase the copper concentration closer to the maximum value. For more densely populated residential areas and more polluting industrial areas, values closer to the maximum are also used. In areas where it is planned to limit the use of copper and zinc as building materials in e.g. roofs, hangers, dumpers, rails and lamp posts, a lower factor of e.g. factor 3 can be used.

StormTac Web also has features that describe pollutant content as a function of traffic intensity, indicating the traffic intensity for the transit routes to be calculated.

The seasonal variations for the example residential data are not as obvious as were geographical variations, except that the bacteria values appear to be lowest during the winter season and highest during the summer and fall (a similar conclusion was obtained during the NURP, EPA 1983, data evaluations) (Pitt, Maestre and Morquecho, 2004; Pitt and Maestre, 2005).

The influence of location on measured stormwater concentrations was studied by Hernandez et al. (2013) for SS, oil, Cu, COD and Zn. The locations in America, Asia and Europe had generally no significant influence on the concentrations of the studied substances, with exception of the event mean Zn concentrations being higher in Asia than in America and Europe. This implies that the land use specific typical concentrations can be used in these three continents but to increase the Zn data when the application is used in Asia. Further studies will be investigated regarding these and other substances, as well as for other locations.

Receiving water model
The EU Water Framework Directive, adopted in 2000, were placed to protect water on natural geographical formations in Europe. To meet the goals of good status in surface water there is a need that urban stormwater management is designed considering the required treatment need for the recipient according to the Water Frame Directive.
The receiving water model is a tool to estimate how the pollutant concentration in the recipient is affected by land uses, possible point loads, treatment facilities etc. within its catchment area. Measured and/or calculated concentration in the recipient is compared to limit concentration (critical concentration in the recipient). The limit concentrations are regulated by receiving water quality criteria concentrations such as Environmental Quality Standards (EQS). The recipient model calculates the total annual pollution load (kg/year) that can be transported to the recipient without exceeding these concentrations, so-called, acceptable load (kg/year). Based on the recipient’s acceptable load (kg/year) and the total pollution load (kg/year), a treatment requirement (kg/year) is calculated. The treatment requirement (kg/year) shows how much the pollution load (kg/year) to the recipient must be reduced to achieve an acceptable load (kg/year) to the recipient. The net retention is important to consider, the retention can affect the required treatment need for the recipient. In StormTac Web the net retention is calculated as residual term in the water mass balance and is defined as the difference between the total load in and out from the recipient.
The applied method for calculating the load reduction and acceptable load in StormTac Web has been scientifically reviewed via a PhD thesis (Larm, 2000) and following papers (Larm, 2003; Larm, 2005; Larm and Pirard, 2013). Since the method was developed, it has been continuously updated and calibrated against new measurements in watercourses and lakes.

Quantification of uncertainties
In StormTac Web each concentration for each substance have been categorized in three levels of uncertainty, based on the number and variance of input data. The corresponding uncertainty classification has been performed for stormwater treatment facilities.

There are data of standard deviations of stormwater and baseflow concentrations, as well as for reduction efficiencies of designed stormwater treatment facilities, for each type of land use per substance in StormTac database that also are presented in the result report of StormTac Web.

Calculation/estimation of uncertainties for each input parameter have been taken from literature studies, using relative uncertainties (%) and calculated absolute uncertainties (+/-).

For uncertainty analysis, the method of the Law of Propagation of Uncertainties (LPU) (Taylor and Kuyatt, 1994) was adopted for analyzing the effects of uncertainties in individual input data on model results.

Sensitivity analysis
The Morris Screening Method, a so-called one-step-at-a-time method (OAT) (Welch et al., 1992) has been used for sensitive analysis for calculated flow and pollutant calculations in StormTac Web. The results showed that (sub)watershed area per land use, baseflow (groundwater) area per land use and precipitation intensity were the three most sensitive inputs in calculating total annual water flows (runoff flow + baseflow). These parameters were closely followed land use specific volumetric runoff coefficients. The most sensitive inputs for calculation of pollutant loads were the annual volumetric runoff coefficients and default pollutant concentrations for various land uses.

Recognizing that annual precipitation data are available from national hydrometeorological agencies and that annual volumetric runoff coefficients can be verified by applications of other simple methods, the determination of the land use specific default concentrations of different pollutants is the most challenging in StormTac Web. Consequently, these default concentrations are also one of the main sources of uncertainties.

First flush
First flush refers to an assumed elevated load of pollutants discharged in the first part of a runoff event. First flush effect was not present in all the land uses, and certainty not for all constituents. The first flush effect has been observed more often in small catchments than in large catchments (Thompson et al, 1995, cited by WEF and ASCE 1998) (Maestre and Pitt; Pitt, Maestre and Morquecho, 2004).

It is expected that peak concentrations generally occur during periods of peak flow (and highest rain energy). On relatively small paved areas, however, it is likely that there will always be a short period of relatively high concentrations associated with washing off of the most available material near the beginning of the runoff event (Pitt 1987; Pitt, Maestre and Morquecho, 2004).

The example investigation of first flush conditions indicated that a first flush effect (increased concentrations at the beginning of an event) was not present in all the land uses, and certainly not for all constituents. Commercial and residential areas were more likely to show this phenomenon, especially if the peak rainfall occurred near the beginning of the event. It is expected that this effect will be more likely to occur in a watershed with a high level of imperviousness, but the data indicated first flushes less than 50% of the time for the most impervious areas.

Groups of constituents showed different behavior for different land uses. All the heavy metals evaluated showed higher concentrations at the beginning of the event in the commercial land use category. Similarly, all the nutrients show a higher concentration in the residential land use except for total nitrogen and ortho-P. This phenomenon was not found in the bacteria analyses. None of the land uses showed a higher number of colonies during the beginning of the event. Conventional constituents showed elevated concentrations in commercial, residential and institutional land uses (Maestre and Pitt, 2004).

Relationships of common pollutants such as suspended solids, phosphorus, fecal coliforms, and total zinc concentrations for different rain depths show little variation, implying there is no strong “first flush” effect at stormwater outfall locations. About 70% of the constituents in the commercial land use category, about 60% of the constituents in the residential, institutional and the mixed (mostly commercial and residential) land use categories, and about 45% of the constituents in the industrial land use category, had first flushes. In contrast, no constituents were found to have first flushes in the open space category. COD, BOD5, TDS, TKN, and Zn had first flushes in all areas (except for the open space category). In contrast, turbidity, pH, fecal coliforms, fecal strep., total N, dissolved and ortho-P showed no statistically significant first flushes in any category (Pitt and Maestre, 2005; Pitt, Maestre and Morquecho, 2004).

The conflict with TKN and total N implies that there may be other factors involved in the identification of first flushes besides land use. If additional paired data becomes available during later project periods, it may be possible to extend this analysis to consider rain effects, drainage area, and geographical location (Maestre and Pitt, 2004). If first flush effects are present, manual sampling may likely miss these more concentrated flows due to delays in arriving at the site to initiate sampling (Maestre and Pitt, 2005).

Comparison between calculated and measured concentration data
The typical concentrations for different land uses are calibrated continuously against measured concentrations from areas with the same type of land use. This is done partly for individual land use, such as sampling directly downstream of a residential area or from a stormwater well (gully pot, drain inlet) from a road with a certain measured traffic intensity. Secondly, they are carried out from different large (sub)watershed areas through calibration. This has been carried out and are carried out regularly for areas in the size of a roof area that may not be greater than 100 m2 (0.01 ha) to areas greater than 10 km2 (> 1 000 ha) from which there is good flow proportional sampling (Larm, 2000). This means that certain reductions of concentrations in technical and natural transport systems are included, see below.

For measured concentrations to be comparable to the typical concentrations, they must be taken with automatic flow proportional sampling during longer periods (several months in different seasons to one or more years). The typical concentration data are calibrated continuously against flow proportional samples from different types of case studies.

If calculated concentrations are higher than measured concentrations, it can often be explained that the sampling has not been taken flow-proportional over a long period of time. Grab samples usually yield significantly lower levels than the average stormwater levels contain. A larger proportion of samples of in leaking groundwater and connecting drainage water is taken than the flow-weighted mixture contains on average. It can be very large difference. The grab samples may show up to 10 times lower pollutant concentration (also depending on analyzed substance) than the corresponding flow-weighted sample. If you take grab samples or time-weighted samples and start these manually, the risk is high to miss the highest concentrations in the first part of the runoff (first flush), due to delays in arriving at the site to initiate sampling (Pitt and Maestre, 2005). The samples will then not be sufficiently representative of a yearly flow which should be based on long-term volume-weighted samples.

The mean values better represent long‐term mass discharges than median values in the case of measuring stormwater from flow proportional sampling (Pitt, 2011) and in the case of measuring grab samples from water courses (HVMFS, 2013). Median values artificially reduce the effects of the periodic unusually high concentrations that do occur in stormwater (Pitt, 2011). However, median values are recommended for representing longer periods of grab sampling in lake and sea water (HVMFS, 2013).

Another explanation that calculated concentrations may be higher than measured concentrations is that there may be greater retention in the transport systems upstream than included in the specified land uses and their typical concentration data. There may also be existing treatment facilities upstream that may not be included in the calculations. In such cases, these need to be simulated to get a more site-specific calculation.

Differences between calculated and measured concentrations may also be due to differences in building materials, traffic intensities, etc. If, for example, there are copper roofs in the area, so the employed typical concentrations need to be increased to take this into account. In this way, more site-specific calculations are obtained.

Pollutant load (kg/year) and pollutant concentrations to and in surface waters (streams, rivers, lakes and sea bays) cannot be directly compared to calculated concentrations in the emission points of stormwater (and baseflow) to the receiving surface waters. In surface waters, a dilution occurs, atmospheric deposition is added to its surface and processes in the surface water affect the amounts and concentrations. There may be either a net retention on an annual basis or a net leakage (release) of contaminants from the sediments to the water mass in the surface water. These processes affect the calculated load and concentration in the outlet from the receiving water. The model takes these processes into account and calculates an output pollutant load and concentration from the surface water, which are presented in the recipient section of the result report. Thus, the calculated load and concentration in the outlets from the catchment areas to a recipient cannot be directly compared to the calculated concentration in the recipient and the load out of the recipient.

Explanations why flow proportional sampling should be used instead of grab sampling on stormwater:
– Measured concentrations and flow vary greatly during each runoff event and between events.
– Measured concentrations from grab samples do not represent a yearly average concentration value and are generally lower than the measured concentrations based on flow proportional sampling.
– Baseflow normally contains much lower pollutant concentrations than the stormwater.
– The reduction efficiency is underestimated if calculated from grab samples.

Case studies
StormTac Web has been used for example in the following case studies, where * indicates that calibration or comparison to measured data has been performed: Nybohov*, Stockholm (residential); Essingeleden*, Stockholm (road); Sätra*, Stockholm (residential); Lake Flaten*, Salem (residential); Flemingsbergsviken*, Huddinge (mixed); Tyresö municipality (mixed); Upplands Väsby municipality* (mixed); Lake Edsviken and Lake Norrviken, Sollentuna (mixed), Lidingö municipality* (mixed); Karlstad municipality (mixed); Fittja, Botkyrka (residential); Reykjavik, Iceland (residential); Kaliningrad, Russia (road) and Lake Titicaca, Peru and Bolivia (mixed).

Calibration to measured data has also been performed for a large number of roads and treatment facilities where StormTac Web has been used.

References

Francey M. (2010). Characterising urban pollutant loads. PhD thesis, Monash university.

Hernandez J.R., Valeri V.C.A., Barrera A.H.F. and Fresno D.C. (2013). Relationship between urban runoff pollutant and catchment characteristics. Journal of irrigation and drainage engineering. October 2013.

HVMFS (2013). Swedish Agency for Marine and Water Management (2013:19).

Larm, T. Wahlsten, A. Marsalek, J and Viklander, M. (2022). A Data-Driven Approach to Stormwater Quality Analysis in Two Urban Catchments. Sustainability 2022, 14, 2888. https://doi.org/10.3390/su14052888

Larm T. and Pirard J. (2013). Using a planning tool and the software StormTac for the development of stormwater management plans. StormTac and Sweco Environment publication, March 2013, Stockholm, Sweden.

Larm T. (2005). Designing BMPs considering water quality criteria. Paper for presentation. 10th International Conference on Urban Drainage, Copenhagen/Denmark, 21-26 August 2005.

Larm T. (2003). An operative watershed management model for estimating existing and acceptable pollutant loads on receiving waters and for the design of the corresponding required treatment facilities. International Conference on Urban Drainage and Highway Runoff in Cold Climate. Riksgränsen, Sweden, 2003, p. 235-245.

Larm T. (2000). Watershed-based design of stormwater treatment facilities: model development and applications. PhD Thesis, Dep Civil & Environmental Engineering, KTH, Stockholm, Sweden.

Maestre A. and Pitt R.E. (2005). Identification of Significant Factors Affecting Stormwater Quality Using the NSQD. Draft.

Maestre A. and Pitt R.E. (2004). Stormwater quality descriptions using the three parameter lognormal distribution. Draft.

Maestre A. and Pitt R.E. Nonparametric Statistical Tests Comparing First Flush and Composite Samples from the National Stormwater Quality Database.

Novotny V. (1995). Non point pollution and urban stormwater management. Volume 9.
Sing V.P. (1995). Environmental hydrology.

Pitt R.E. and Maestre A. (2005). Stormwater quality as described in the National Stormwater Quality Database (NSQD). 10th International Conference on Urban Drainage, Copenhagen/Denmark, 21-26 August 2005.

Pitt R.E., Maestre A. and Morquecho R. (2004). The National Stormwater Quality Database (NSQD, version 1.1) February 16, 2004.

Pitt R.E. (2011). The National Stormwater Quality Database, Version 3.1. March 8, 2011.