Non-point source pollution (NPS) besides point source pollution (PS) has contributed to pollutant loading into natural receiving water bodies. Due to the nature of NPS, the quantification of pollutant loading from NPS is very challenging but crucial to riverine water quality management, especially for the river reach flowing through urban areas. The water quality in the river reach of the Bow River flowing through the City of Calgary in Alberta, Canada, is affected by both PS and NPS. Thus, understanding and characterizing water quality of discharges (affected by NPS) into the river reach is necessary for better managing riverine water quality and preventing water quality degradation. In the paper, monitored event mean concentrations (EMCs) of stormwater runoff and mean concentrations of snowmelt and baseflow of seven common pollutants from sub-catchments, which are categorized into four land use types including commercial, industrial, residential and on-going development land uses, were used to investigate the linkage between land use and water quality. Statistical analysis techniques were adopted to identify differences or similarities in water quality among different flow types, different land use types, and among/between catchments of same land use. The results indicated that EMCs of many water quality parameters vary among different land use types and among/between catchments of same land use. The results also showed median EMCs of pollutants of snowmelt and baseflow are, in general, lower than those of stormwater runoff. In addition, Stormwater Management Model was employed to investigate the physical process that would affect water quality response to storm events for two typical land uses, industrial and residential land uses. The modeling results supported that wash-off of particulate matters might primarily affect water quality response of catchments between different land uses. All the results shed the light on the necessity of quantifying pollutant loading considering the characteristics of land uses.
Urbanization occurring across the world alters urban hydrologic cycle and it largely increases water quantity due to the increase of impervious area, such as roads, parking lots, and rooftops. On the other hand, urbanization also affects generation and mobilization of both point source and non-point source pollutants from urban settings. Therefore, urbanization can lead to the increase of water quantity and the degradation of water quality of natural receiving water bodies [
Owing to the recent rapid urbanization, water discharged from urban settings has attracted more attention in water management. Water released from stormwater drains of a typical urban area/catchment includes baseflow, snowmelt and stormwater runoff. Stormwater runoff and snowmelt are discharges from storm drains resulting from storm and snowmelt events, respectively; while baseflow is the discharge accounting for groundwater seeping into stormwater drains and surface water connected to storm drains. The different types of flow all contribute pollutant loading from nonpoint pollution sources (NPS). Among these flow types, stormwater runoff has been claimed to be one of the most common sources of water pollution, for example in U.S. [
A large body of studies (e.g., [
The City of Calgary, Alberta, Canada, which has a separate storm sewer system, has put efforts to quantify pollutant loading discharged into the Bow River at a city-wide scale aiming to understand and consequently formulate strategies to prevent water quality degradation in the river. To achieve the goal, Calgary has developed a monitoring program, which targets to understand water quality of different types of flow including baseflow, snowmelt and stormwater runoff from storm drains in catchments of various land uses. Considering the cost, resources, and time required by a monitoring program to cover the entire city, modeling tool has been considered to compensate the limitations of monitoring. Therefore, the two-fold objectives of this paper were to: 1) investigate the differences/or similarities in water quality concentrations of three flow types (baseflow, snowmelt, and stormwater runoff) from catchments of different land uses; and 2) use modeling approach to study which process (either pollutant buildup and/or wash-off) might largely affect the quality of stormwater runoff for catchments of different land uses. All the results would shed light on how to accurately quantify and thus efficiently mitigate pollutant loading from urban settings.
The City of Calgary is located in the transition between the Canadian Rockies Foothills and the great Prairies. The City is bounded between 113˚54'54"W and 114˚16'34''W and between 50˚50'38''N and 51˚12'44''N. It is recognized as Canadian hub for oil and gas industry. The City is one of the most fast growing cities in Canada with a population of 1.23 million currently and the most populated community centre in Alberta. The City is situated on the confluence of the Bow and Elbow Rivers. These two rivers supply water to more than one million populations residing in the City. The Bow River also supports the blue ribbon fishery. To protect the water quality of the river, the City of Calgary has conducted water monitoring in the last decades. To fulfill the objectives of the paper, the data sets collected by the water monitoring program during 2001-2005 were selected and revisited because: 1) water quality was monitored in three types of flow (baseflow, snowmelt, and stormwater runoff) separately; and 2) monitoring was conducted in catchments of four typical types of urban land use including residential, industrial, commercial, and on-going development land uses. Considering the data availability, several sub-catchments were selected from the monitoring program for each land use type: two commercial sub- catchments (Eau Claire and Rundle), two industrial sub-catchments (Bonnybrook and Wigmore East), three on-going development sub-catchments (69th St. West, Cranston, and Crestmont West), and five residential sub-catchments (68th St. East, 68th St. West, 69th St. East, Rocky Ridge Inlet, and McKenzie Towne). Note that the land use categorization is based on the dominant land use of the sub-catchments. The locations of the monitoring sites shown in
5-day BOD was determined in accordance with Standard Methods. At the 12 selected sub-catchments, EMCs of the water quality parameters were reported in a total of 570 storm events, 166 snowmelt events, and 321 baseflow events during the monitoring program. Considering the less variation in the quantities of baseflow and snowmelt, their mean concentrations were considered equivalent to EMCs. Thus, the terminology of EMC was used for all three types of flow throughout the paper for convenience.
When monitoring stormwater runoff, discrete samples were collected up to five hours in each storm. Water samples were composited and EMCs were reported in three phases in an event: first flush (FF) in the first 30 minutes, first remainder of the event (ROE1) in the following 2.25 hours, and second remainder of the event (ROE2) in the another 2.25-hour duration. Note that if a storm is shorter than five hours, ROE1 and/or ROE2 were absent. EMCs for an event were calculated using the reported EMCs in the three phases based on the flow-weighted average approach. Baseflow was manually sampled and water quality was reported as average concentrations during a dry period. Similar to baseflow, water samples of snowmelt were collected manually and average water quality was reported during a snowmelt event. In the presence of baseflow and snowmelt at the outlets of the sub-catchments, one to five samples per month were collected.
Statistical analysis was conducted to investigate the differences/or similarities in water quality of different types of flow from catchments of different types of land use. Non-parametric statistical analysis was selected since water quality datasets are not normally distributed according to the Kolmogorov-Smirnov test as expected. To compare two or more than two water quality datasets, Wilcoxon rank sum test and Kruskal-Wallis (K-W) test followed by multiple comparisons were applied. All these analyses were performed at the significance level of 5%.
One of popular modeling tools for urban stormwater runoff is SWMM, which is applicable for both event-based and continuous simulation. Details on the model can be found in reference [
The calibration for stormwater runoff quantity was preceded with the calibration of stormwater runoff quality. In the model calibration for TSS, the parameters/or coefficients for sediments buildup and wash-off were determined. In the SWMM, the buildup and wash-off processes of sediments were modeled using exponential functions, which have been commonly adopted to model TSS in practices (e.g., [
where C1 is the maximum possible buildup (mass/area); C2 is the buildup rate constant (1/day); and t denotes the time to accumulate pollutants; Cw1 and Cw2 are the wash-off coefficient and exponent, respectively; q is runoff rate per unit area.
parameters (except TDP and
Given a flow type, the water quality data from the sub-catchments were pooled together according to land use type. A comparison of EMCs for sub-catchments with different types of land use was then conducted for each flow type. In the analysis, the EMCs of FF of storm events were used considering that the EMCs of FF are significantly higher than those of ROE1 and ROE2 and the initial stormwater runoff during a storm event is often of great interest for managing urban stormwater runoff. The analysis results are presented in
Parameter | Baseflow | Snowmelt | Stormwater runoff (FF) |
---|---|---|---|
TP | Ind = Res < Com | Com = Ind = Res < Dev | Com = Res < Ind < Dev |
TDP | Ind = Dev < Com = Res | Dev < Ind = Res | Dev < Res |
Com < Res | Com < Ind | Ind < Res | |
TKN | Com = Ind = Dev = Res | Com = Res < Dev | Res < Dev |
NO2‾/NO3‾-N | Ind < Dev < Res Ind < Com | Ind < Com = Res Dev < Com | Dev < Ind |
TSS | Com = Ind = Res < Dev | Com = Ind = Res < Dev | Com = Res < Ind < Dev |
BOD | Res < Com = Ind = Dev | Com = Res < Ind | Dev = Res < Com = Ind |
the significant difference between residential and commercial land uses). As for other water quality parameters, the qualitative relationships between water quality levels and land uses (e.g., in terms of the order of the land uses) appear to vary among the three types of flow. For instance, the median EMCs of TP is highest for commercial land use in baseflow; whereas it was reported to be highest for on-going development land use in both stormwater FF and snowmelt. A previous study by reference [
Given a type of land use, the EMCs were pooled according to flow type and the comparison results from the K-W test for each water quality parameter among the three flow types are summarized in
Parameter | Results | Parameter | Results |
---|---|---|---|
Commercial land use | |||
TP | Base < Melt < FF | NO2‾/NO3‾-N | FF < Melt < Base |
TDP | FF < Base = Melt | TSS | Base < Melt < FF |
Base < Melt = FF | BOD | Base < Melt < FF | |
TKN | Base < Melt < FF | ||
Industrial land use | |||
TP | Base < Melt = FF | NO2‾/NO3‾-N | Base = Melt = FF |
TDP | Base < Melt = FF | TSS | Base < Melt = FF |
Base < FF < Melt | BOD | Base < Melt = FF | |
TKN | Base < Melt = FF | ||
On-going development land use | |||
TP | Base < Melt = FF | NO2‾/NO3‾-N | FF < Melt = Base |
TDP | Base = Melt = FF | TSS | Base < Melt = FF |
Base < FF < Melt | BOD | Base < Melt = FF | |
TKN | Base < Melt = FF | ||
Residential land use | |||
TP | Base < Melt < FF | NO2‾/NO3‾-N | FF < Melt < Base |
TDP | Melt < Base | TSS | Base < Melt < FF |
Base < Melt = FF | BOD | Base < Melt < FF | |
TKN | Base < Melt < FF |
stormwater runoff FF and snowmelt are significantly higher than those of baseflow for all water quality parameters except the dissolved constituents including
Given a flow type and a land use type, the comparison of the median EMCs among/or between sub-catchments was conducted using the K-W or Wilcoxon rank sum tests to investigate whether significant differences among/or between sub-catchments exist. As demonstrated in
As shown by the results obtained from above statistical analysis, water quality levels are functions of types of both flow and land use. The detected significant differences in water quality among/or between sub-catchments suggest the necessity and importance of selecting representative sub-catchments for water quality monitoring. All these detected differences complicate the formulation of an effective monitoring scheme for accurately quantifying pollutant loading into receiving water bodies from an urban setting. An efficient water quality monitoring program should be capable of capturing the variations of water quality. Apart from classifying land use based on the major categories (commercial, industrial, on-going development, and residential land uses), more elaborate classification, for example taking account more geophysical characteristics (slope, soil type, and percentage of impervious area) and hydrology and water quality
Flow | Land use | TP | TDP | TKN | NO2‾/NO3‾-N | TSS | BOD | |
---|---|---|---|---|---|---|---|---|
Baseflow | Com | NEQ | EQ | EQ | NEQ | NEQ | NEQ | NEQ |
Ind | NEQ | NEQ | EQ | NEQ | NEQ | EQ | NEQ | |
Dev | NEQ | NEQ | NEQ | NEQ | NEQ | EQ | NEQ | |
Res | NEQ | NEQ | NEQ | NEQ | EQ | EQ | EQ | |
Snowmelt | Com | NEQ | EQ | NEQ | NEQ | NEQ | NEQ | NEQ |
Ind | EQ | EQ | EQ | EQ | NEQ | EQ | EQ | |
Dev | EQ | NEQ | NEQ | EQ | EQ | EQ | EQ | |
Res | NEQ | NEQ | NEQ | NEQ | NEQ | NEQ | NEQ | |
Stormwater runoff (FF) | Com | EQ | EQ | EQ | EQ | EQ | EQ | EQ |
Ind | EQ | EQ | EQ | EQ | NEQ | EQ | EQ | |
Dev | EQ | NEQ | EQ | EQ | NEQ | EQ | EQ | |
Res | NEQ | NEQ | EQ | NEQ | EQ | NEQ | NEQ | |
Stormwater runoff (ROE1) | Com | EQ | EQ | EQ | EQ | NEQ | EQ | EQ |
Ind | NEQ | NEQ | EQ | EQ | EQ | NEQ | EQ | |
Dev | EQ | NEQ | EQ | NEQ | EQ | EQ | EQ | |
Res | NEQ | EQ | NEQ | EQ | EQ | NEQ | EQ | |
Stormwater runoff (ROE2) | Com | EQ | NEQ | EQ | EQ | EQ | EQ | NEQ |
Ind | NEQ | NEQ | EQ | EQ | EQ | NEQ | NEQ | |
Dev | -- | -- | -- | -- | -- | -- | -- | |
Res | EQ | EQ | EQ | EQ | EQ | NEQ | EQ |
response of catchments, might be necessary from the point of view of quantifying pollutant loading at a city-wide scale.
As discussed previously, stormwater runoff is more contaminated as the median EMCs of many water quality parameters are higher than and equivalent to those of baseflow and snowmelt, respectively (
A total of 12 and six storm events, in which data of flow, EMC of TSS and rainfall are available, were used for developing the models for BB and MK sub-cat- chments, respectively. At BB sub-catchment, eight events observed during 2003- 2004 were used in model calibration and other four events observed in 2005 were used in model validation. At MK sub-catchment, four events in 2002 were adopted to calibrate the model, while the other two events from the same year were used to validate the model. Among the 12 storm events for BB sub-catch- ment, the return period of three storm events were longer than two years return period; while at MK sub-catchment, the return periods of all storm events (except event on July 27, 2002) were equal to or less than two years.
Variable | BB sub-catchment | MK sub-catchment |
---|---|---|
Area (km2) | 3.754 | 1.478 |
Width (m) | 1033.00 | 506.25 |
Slope (%) | 3.61 | 3.55 |
Impervious area (%) | 35.67 | 32.00 |
N of impervious area | 0.01 | 0.01 |
N of pervious area | 0.31 | 0.24 |
Impervious depression storage (mm) | 0.52 | 0.23 |
Pervious depression storage (mm) | 1.73 | 2.07 |
Bonnybrook | Simulated (Storm by Storm) | Simulated (Average all) | ||
---|---|---|---|---|
Event Date | EV (%) | EP (%) | EV (%) | EP (%) |
Model calibration | ||||
06/01/2003 | +18.71 | −0.32 | +15.79 | −1.29 |
06/18/2003 | +22.61 | −30.76 | +25.68 | −23.17 |
07/05/2003 | +32.97 | +9.37 | +47.36 | +31.93 |
08/19/2003 | +36.99 | −9.36 | +16.44 | −24.77 |
06/06/2004 | +5.22 | +9.55 | −18.74 | +10.51 |
07/07/2004 | +6.79 | +9.59 | −0.41 | +10.70 |
08/04/2004 | +13.60 | +4.93 | +2.56 | +5.28 |
08/15/2004 | −11.14 | +4.97 | +1.17 | +3.86 |
Model validation | ||||
07/19/2005 | -- | -- | −23.10 | −13.00 |
07/24/2005 | -- | -- | −27.46 | −43.40 |
08/02/2005 | -- | -- | −16.81 | −25.11 |
08/23/2005 | -- | -- | +36.96 | −0.66 |
to 32% when the average calibrated parameters were used. The coefficients of determination (R2) (between observed and simulated values) are 99% and 88% for total runoff volume and peak flow, respectively, in the model calibration. As shown in
At BB sub-catchment, the modeled and observed hydrographs of the storm event on July 5, 2003, which was reported to have highest errors among the calibration events, are shown in
Compared to the modeling results at BB sub-catchment, the calculated relative errors of total runoff volume and peak flow spanned wider ranges at MK sub- catchment. The errors in total runoff volume ranged from −15% to 116% and the errors in peak flow varied from −60% to 72% in the model calibration at MK sub-catchment. However, better model performance was achieved in the model validation, in which the errors in total runoff volume and peak flow are in the range of −19% to −9% and −40% to 5%, respectively. The inferior performance at MK sub-catchment, especially in the model calibration, might be ascribed to less number of events available for model development. As illustrated in
August 10, 2002 in the model calibration, the calibrated model was capable of capturing the dynamic behavior of flow.
Parameters of pollutant (here TSS) buildup and wash-off equations were determined in the model calibration. Initial assumptions of these parameters were made referring to previous studies. In the absence of pollutograph, EMCs of TSS were used to develop the water quality models. In the calibration, three storm events and one event were not included for BB and MK sub-catchments, respectively, as these events are more intensive and measured TSS EMCs in these events largely deviate from those of the rest calibration events. Thus for BB sub- catchment, five and four events were applied to calibrate and validate model, respectively; while three and two events were used for model calibration and validation, respectively, for MK sub-catchment.
At BB sub-catchment, TSS EMCs were overestimated in three out of five calibration events. The relative errors in EMC ranged from −10% to 59% and R2
Variable | BB | MK | Literature | |
---|---|---|---|---|
Buildup function | Maximum buildup (kg/ha) | 56 | 56 | 5.0 - 351,2 (kg/ha-day) |
Buildup rate (/day) | 1 | 1 | 13 | |
Wash-off function | Wash-off coefficient | 0.098 | 0.087 | 0.11 - 0.192 |
Wash-off exponent | 1.79 | 1.53 | 0.0 - 32 |
1 [
between the observed and simulated EMCs was 0.81 (
In the model validation, the performance of developed model for BB sub- catchment is inferior to that in the model calibration. The results might reflect the stochastic nature of pollutant buildup and/or wash-off on and from land surface. The other possible explanation is that the pollutant deposited on the land surface is dependent on not only the antecedent dry period prior to a storm event but also the remaining pollutants on the land surface after its previous event. However, the developed model for MK sub-catchment performed well in the model validation as the errors in TSS EMCs ranged within ±20%.
Compared to the modeling of stormwater runoff quantity, the modeling of stormwater runoff quality is more challenging as illustrated by the results from the model calibration and validation. Although the results suggest that improvement on stormwater runoff quality modeling is needed, they provide insight into the process which largely governs TSS level in stormwater runoff. For these two sub-catchments of different types of land use, same buildup parameters were determined; whereas the wash-off coefficient and exponent were higher for BB sub-catchment (industrial land use) than those for MK sub-catchment (residential land use) (
function of land use and regional environment, thus water quality response to storms can be site-specific and even storm-specific. Under this circumstance, caution should be paid to the water quality response of catchments when characterizing stormwater runoff quality and quantifying pollutant loading.
This paper investigated the linkage between land use and water quality in an urban setting, the City of Calgary, Alberta, Canada using two different approaches: statistical analysis and modeling. The statistical analysis was conducted to study the differences or similarities in water quality in three types of flow (baseflow, snowmelt and stormwater runoff) generated from catchments grouped into four types of lands uses (commercial, residential, industrial, and on-going development). The analysis identified significant differences in many water quality parameters among the investigated types of flow and the different types of land use, and among/or between catchments categorized into same land use type. In addition, the modeling results of stormwater runoff from two sub-catchments of different land use types demonstrated that sediments wash-off might be the dominant process governing TSS level of stormwater runoff, although the governing process could be site-specific and also a function of event magnitude. All these results suggest that water quality level vary with flow types, land use types, and even catchments of a same land use type. Therefore caution should be paid to the selection of monitoring sites/catchments, which should be representative for capturing the variability of water quality level in different flows, land uses, and catchments of similar land use, and consequently accurately quantifying pollutant loading at a city-wide scale.
This research was financially support by Urban Alliance (a research partnership between the City of Calgary and the University of Calgary) and NSERC. The authors would like to acknowledge the contribution of the personnel, including Mr. Khizar Mahmood, Ms. Stacey Zhao, Mr. Lei Chen, and Ms. Lily Ma (former staff), of the City of Calgary.
Shrestha, D. and He, J.X. (2017) Characterization and Modeling of Urban Water Quality in the City of Calgary, Canada. Natural Resources, 8, 513-530. https://doi.org/10.4236/nr.2017.88032