Effectiveness of Wetlands for Improving Different Water Quality Parameters in Various Climatic Conditions

Abstract

Engineered wetland has been used as a Best Management Practice (BMP) to remove pollutants and maintain water quality in watersheds. This study is focused on developing models to analyze the impacts of discharges on the efficiency of wetlands to improve water quality downstream. The watershed hydrological Soil & Water Assessment Tool (SWAT) and wetland (Personal Computer Storm Water Management Model—PCSWMM) models were developed to analyze the efficiency of engineered wetlands to remove the pollutants for different basins under three different climatic conditions (i.e., dry, average and wet year). The SWAT was calibrated and validated to simulate discharge and water quality parameters. The wetland model was developed using inflow hydrographs and concentrations of the water quality parameters biochemical oxygen demand (BOD), total suspended solids (TSSs), total nitrogen (TN) and total phosphorous (TP), simulated from a Soil & Water Assessment Tool (SWAT) model. A PCSWMM (wetland) was developed based on the physical and first order decay process within the wetland system for three basins in Prairie du Pont watershed in Illinois, USA. The results showed that pollutant removal efficiencies decreased from low to high discharges (dry to wet climatic conditions) for all watersheds and pollutants (except for BOD) based on trendline analysis. Nevertheless, the efficiencies were highly variable, specifically during low discharges. Furthermore, the sensitivity of the k-parameter (areal rate constant) was pollutant dependent. Overall, this study is helpful to understand the efficacy of wetlands’ pollutant removal as a function of discharge. The approach can be used in watersheds located in other geographic regions for the preliminary design of engineered wetlands to remove non-point source pollution and treat stormwater runoff.

1. Introduction

Best Management Practices (BMPs) are the measures or techniques that are installed in watersheds to reduce the effects of urban non-point source pollution associated with flows and stormwater runoff volumes on the receiving waters [1]. They have been used extensively to manage the quantity and improve the quality of stormwater runoff by replicating natural processes [2,3,4]. It is mainly structural BMPs that are used to treat stormwater at the point of generation or point of discharge to the receiving waters. The selection of types of BMPs affects the efficiency of the structure to enhance the water quality of the system [5]. There are several types of BMP categories according to their fundamental processes and performance. Major BMPs include ponds, wetlands, vegetative bio filters, infiltration trenches, and organic sand filters, etc. The fundamental processes like sedimentation, floatation, filtration, infiltration, adsorption, degradation, biological uptake and conversion are mainly involved in the BMPs [1,6].

The application of chemical fertilizers and animal manure in agricultural land provides nitrogen and phosphorous required for the plant growth; however, all these nutrients are not completely utilized/uptaken by plants [7,8]. These excess nutrients are swept from farm fields and released into waterways during rain and snowmelt events. Furthermore, they can also migrate into groundwater by leaching through soil, affecting the quality [9]. Hydrological (e.g., SWAT, HEC-HMS, SWMM, etc.) and integrated hydrological–ground water (MIKE SHE) models can simulate water and nutrient movement in watersheds [4,10,11,12,13].

SWAT is a semi-distributed watershed-scale model, which is generally used to simulate hydrologic and water quality parameters such as discharge, sediments, nutrients and pollutants for watershed management practices for different time periods at daily and monthly time-steps [10,14,15]. However, there might be some uncertainties in using SWAT for predicting water quality parameters compared to the streamflow and sediment loads [16]. Nonetheless, the sensitivity of the model parameters may vary on multiple factors, such as geography, land use, meteorological and forcing data (i.e., precipitation) and their quality.

The Storm Water Management Model (SWMM) simulates watershed discharges based on the relationship between rainfall and runoff [13]. The SWMM was developed by USEPA to simulate single and continuous rainfall–runoff events. Different BPM models can be designed into SWMM to study impacts on water quantity and quality. Personal Computer Storm Water Management Model (PCSWMM) is a GIS-based version of SWMM further developed by Computational Hydraulics International (CHI).

A high concentration of nutrients in stream water is considered as pollution. The pollutants present in stormwater, if not removed, can have negative impacts on both aquatic and terrestrial ecosystems by promoting the growth of algae and plants, increasing the risk of eutrophication of water bodies and environmental degradation [17,18]. One of the major issues with high nutrient concentration in the water body is that it leads to excessive growth of algae and later the decomposition reduces dissolved oxygen, which may lead to suffocation of fish and aquatic species [19].

The major pollutant in domestic wastewater is biochemical oxygen demand (BOD), chemical oxygen demand (COD) and ammonia, which are regulated under the National Pollutant Discharge Elimination System (NPDES) in watersheds [20]. BOD and COD are the amount of oxygen needed to oxidize organic compounds both biologically and chemically. Wetland helps to remove BOD in a watershed mostly by aerobic microbial degradation and sedimentation and filtration processes.

Total suspended solids (TSSs) are a measure of sediment concentration in surface water that can be used to determine the state of aquatic life in an aquatic ecosystem [21]. TSSs from stormwater discharge are an important variable for assessing pollution in different natural water bodies (e.g., stream, pond, lake, etc.) [22]. High TSS levels in water bodies can have negative biological impacts as the presence of suspended solids increases the turbidity and reduces light availability, which lowers dissolved oxygen concentrations [23]. This phenomenon makes photosynthetic organisms less active and promotes algae blooms and decomposition [24]. Furthermore, substantial harmful pollutants like heavy metals and organic materials are adsorbed onto TSSs and settle on sediment, compromising the aquatic system’s health [25].

Wetlands provide valuable services like flood control, surface water flow maintenance, water quality improvement, erosion control, groundwater recharge and providing a habitat for wide range of plants and animals [26]. Wetlands are the land areas that remain inundated by surface water or groundwater permanently or seasonally, making them transitional between terrestrial and aquatic ecosystems [27,28]. Furthermore, wetlands are wet for sufficiently longer durations to create a favorable habitat for vegetation, which can grow in saturated soils and change the soil conditions due to physical, chemical and biological processes [29,30,31]. Studies have shown that wetlands are one of the richest ecosystems on this planet which support an extensive food chain and biological diversity [32]. Vegetation and soil present in the wetland offer hydraulic resistance that decreases the velocity of runoff water which helps to retain the sediment within the wetland as the floodwater flows along its length [33], and reduces the sediment loads to receiving waters.

Wetland is used as a BMP to remove pollutants from stormwater by the combination of sedimentation, degradation and exporting pollutants to the atmosphere or leaching them to groundwater [34]. Stormwater runoff carries suspended solids, nutrients, bacteria, metals and toxic substances [35], which can be improved (reduced or removed) with wetlands by absorbing or settling those pollutants down naturally. Nutrients such as phosphorous and nitrogen are removed, released and transformed in wetlands by adsorption and denitrification processes [36]. The mechanisms of removal of pollutants like solids, biochemical oxygen demand (BOD), nitrogen and phosphorus in wetland generally include physical, chemical and biological processes.

Pollutant removal in engineered wetlands has been analyzed based on simple-to-process-based models; however, process-based models are complex and data intensive [6]. A first order decay model is the one of the common models used to simulate pollutant removal in engineered wetlands, assuming steady-state plug-flow [29,30,31,37,38,39,40,41,42]. Due to its simplicity, it ignores the full mechanistic complexities that drive pollution removal processes, which are combined into a first order decay rate constant [37,40]. Nonetheless, the decay rate constant varies based on the local field conditions and wetland operations [41,43]. Despite its simplicity, the first order decay model has been used successfully to remove pollutants (e.g., TN), and is comparable with other models and field-based results for constructed wetlands [39,41,42].

Hydraulic performance is a major factor in modeling the treatment processes of engineered wetlands. The performance of wetland is driven by the length-to-width ratio, scale and sizing, shape and configuration and the distribution of the vegetation [44]. Inflow, hydraulic loading rate (HRT) and detention time determine the performance of the wetland for stormwater treatment, which are functions of rainfall depth, intensity and discharge and the area or volume of the wetland [45].

Engineered ponds are designed specifically in urban settings to improve water quality and enhance freshwater ecosystems. Therefore, engineered ponds help to conserve and improve aquatic biodiversity in ecologically poor environments (urban settings) and offer an opportunity for water managers to restore degraded ecosystems [46]. Engineered ponds offer a sustainable approach to counter general issues in water resource management and climate change, focusing on nutrient retention, discharge interception and carbon sequestration [47]. Furthermore, small dams, constructed for different purposes (i.e., water diversion for agricultural use, drinking and municipal water supply, sediment capture, creating recreational areas, etc.), enhance water quality downstream, increasing the retention of agricultural pollutions such as nitrogen and phosphorus [48].

Wetland performance is influenced by wetland structure, hydrology, climate, soils, vegetation and watershed imperviousness. Furthermore, a wetland’s performance also depends on different physical, chemical and biological processes between flow, vegetation and microorganism development in the system [49]. For instance, the presence of plants, the nature of the rooting substrate and the degree of pretreatment have all been demonstrated to influence removal rates [45]. Evapotranspiration (ET), which is dependent on meteorological variables like air temperature, wind speed and solar radiation and plant species, diversity, density and physiology, can affect treatment performance in constructed wetlands [50]. The pollutant removal in wetlands also depends on the watershed hydrology [51,52]. Precipitation affects the detention time and consequently influences the nutrient and solid removal efficiencies in wetlands. Variation in water volume and flow may affect the quantity of pollutants discharged into the treatment system and its capacity to remove the compounds. Nevertheless, information about changes in wetlands’ efficiency at removing pollutants with respect to discharges (i.e., low to high) based on the first order decay process is lacking. Due to economic reasons and practical issues, the continuous monitoring of wetland efficiency for removing pollutants over the long-term may not be feasible.

This study focused on assessing the change in efficacy of wetlands with respect to discharges to remove pollutants, i.e., biological oxygen demand (BOD), total nitrogen (TN), total phosphorous (TP) and total suspended solids (TSSs), based on the first order decay process for agricultural-, urban- and forest-dominated watersheds under three different climatic conditions (i.e., dry, average and wet) integrating the process-based SWAT model and PCSWMM. Numerical modeling to predict engineered wetland efficiency to remove pollutants coupled with limited measured data (pollution concentration) would be a more viable approach to help with future engineered wetland design.

2. Methodology

This study focuses on modeling water quality to analyze the impacts of wetland construction on water quality (BOD, TN, TP, TSS) improvement and efficiency (Figure 1). The watershed hydrological (SWAT) and wetland (PCSWMM) models were integrated to simulate water quality in watersheds and downstream of wetlands, respectively. SWAT-simulated discharges and water quality based on meteorological (i.e., precipitation, air temperature, wind speed, relative humidity), land use, soil and Digital Elevation Model (DEM) that represent the watershed topography were input for the wetland (PCSWMM) model. Similarly, PCSWMM output discharges and water quality parameters were simulated considering the initial/background and inlet pollutant concentration, discharge and hydraulic resident time (function of wetland volume and discharge) and areal rate constant for wetland treatment. The differences between input and output wetland pollution concentrations with respect to input concentrations were analyzed to quantify the efficiency of wetlands to remove the pollutants from the water body (Figure 1).

2.1. Study Area

The Prairie Du Pont (PDP) and Judy’s Branch watersheds are sub-watersheds of the Cahokia-Joachim watershed (Figure 2). Both of these watersheds drain into the Mississippi river and their streams are highly channelized/modified due to urbanization and agricultural practices [54]. This study is mainly focused on the Prairie Du Pont watershed located in the St. Clair County, IL, which has an area of 315.5 Km². The Prairie Du Pont creek is highly channelized/modified due to urbanization and agricultural practices and yields an absolute elevation change of about 80 m over approximately 32 km of stream length. It has gradient of 0.25% and drains into the Mississippi river, according to IEPA [55]. Prairie Du Pont watershed is impaired due to industrial waste, as well as due to the non-point source pollutant results of agricultural practices [54]. It has significant (8.6% of total watershed) water bodies (wetland, lakes and ponds), which are intermittent and perennial in nature [54]. These water bodies are important to counter degraded water quality in the watershed and provide opportunities for restoration for future water resource management.

2.2. Functional Scenarios

Three functional scenarios were considered for this study based on climatic conditions of wet, average and dry. Yearly total and average precipitation were estimated based on daily average data from 1982 to 2014 to identify the dry, average and wet climatic conditions. The water year 2006 represents a dry climatic condition (total yearly precipitation of 801.57 mm and yearly average precipitation of 2.20 mm), whereas the water year 2013 (total yearly precipitation of 1127.01 mm and yearly average precipitation of 3.08 mm) is an average climatic condition. The water year 2008 with a total yearly rainfall of 1434.91 mm and yearly average rainfall of 3.93 mm is considered as a wet climatic condition.

2.3. SWAT Development

Watershed hydrological models were developed using the Soil & Water Assessment Tool (ArcSWAT 2012) software for hydrologic analysis involving spatial (land use, soil and elevation) and temporal datasets (precipitation, humidity, air temperature and speed, solar radiation) [15,16,53]. The model simulated watershed hydrology and water quality parameters, which were used as input for PCSWMMs (wetland). We used SWAT to simulate watershed discharges and water quality parameters from 1979 to 2014 with a 2-year initial warmup period. Prairie Du Pont (PDP) watershed does not have any hydrological and water quality measurement locations to calibrate the model. Therefore, we calibrated and validated the hydrological model at the Judy’s Branch watershed (Figure 1 and Figure 2), which is a neighboring watershed and has similar land use and watershed management practices [54]. Judy’s Branch watershed (300 km²) model includes 35 sub-basins, whereas the PDP watershed has 39 sub-basins (315.5 Km²).

For the hydrological model calibration processes, the parameters with the highest sensitivity parameters, namely water capacity (SOL_AWC), soil hydraulic conductivity (SOL_K) and the runoff coefficient for SCS (CN2), were modified in SWAT-CUP. The model was calibrated from 2001 to 2005 and validated from 2006 to 2010 using measured monthly average discharges at the USGS gage station (05588720) in the Judy’s Branch creek (Figure 3). The hydrological model parameters obtained from the calibrated model were transferred to the Prairie Du Pont watershed SWAT model (Figure 1 and Figure 2). The water quality model was run using parameters estimated for Racoon River watershed, Iowa, which is located in the same geographic region and has comparable land use and management practices (dominated with agricultural practices) [53]. The water quality model simulated biochemical oxygen demand (BOD), total nitrogen (TN), total phosphorus (TP) and suspended solids (TSSs).

We calculated Nash–Sutcliffe Efficiency (NSE), Coefficient of Determination (R²) and mean squared error (MSE) for the model performance analysis for both calibration and validation (Figure 3). NSE, R² and MSE were 0.70, 0.72 and 0.015, respectively, for the calibration period, whereas they were 0.48, 0.57 and 0.036, respectively, for the validation period.

The SWAT model was used to simulate daily discharges (Figure 4) and water quality parameters (BOD, TN, TP and TSSs) to input in the wetland PCSWMM as a boundary condition for three different basins (agriculture, forested and urban) and for dry (2006), average (2013) and wet (2008) climatic conditions. The highest and lowest daily discharges were simulated for forested and agricultural basins, respectively.

2.4. Pollutant Removal in Wetlands

We implemented wetland models to quantify the effectiveness of the removal of pollutants for different climatic conditions in agricultural, urban and forested basins. Wetlands in their natural state provide longer flow paths and the contact with vegetation encourages pollutant removal. Wetlands are not only efficient in the removal of particulate-bound contaminants, including trace metals and nutrients, by sedimentation, but also have the advantage of achieving water quality improvement in combination with biological and chemical treatment mechanisms [56].

The pollutants are removed by natural decay processes and can also be reduced by introducing treatment equations in PCSWMM [38]. Typically, a model with a two-parameter first order decay function is used for the treatment of water quality parameters. A linear function (Equation (1)) represents the reduction in pollutant concentration within the wetland system [30,31].

$$\mathrm{q}{\displaystyle \frac{\mathrm{d}\mathrm{C}}{\mathrm{d}\mathrm{x}}}=-\mathrm{k}\left(\mathrm{C}-{\mathrm{C}}^{\mathrm{*}}\right)$$

(1)

where

• q = hydraulic loading rate (m/yr);
• x = fraction of distance from inlet to outlet;
• C = concentration of water quality parameters;
• C* = background concentration of water quality parameters;
• k = areal rate constant.

2.4.1. The k-C* Model

The first order decay model with background concentration is generally used in predicting the performance of a wetland (Equation (2)) [30,31].

$${\displaystyle \frac{{\mathrm{C}}_{\mathrm{o}}-{\mathrm{C}}^{\mathrm{*}}}{{\mathrm{C}}_{\mathrm{i}}-{\mathrm{C}}^{\mathrm{*}}}}={\mathrm{e}}^{-{\displaystyle \frac{\mathrm{k}}{\mathrm{q}}}}$$

(2)

where C_o is the outlet pollution concentration, C_i is the inlet concentration, C^* is the background concentration and k is the areal rate constant. k and C* are lumped parameters (

Table 1. Aerial rate constants and background concentrations for water quality, which is a dapted from refs. [30,31].

Water Quality Parameter	k (m/yr); C*
Total Suspended Solids (TSSs)	k = 1000 m/yr C* = 5.1 + 0.16Ci
Biochemical Oxygen Demand (BOD)	k = 34 m/yr C* = 3.5 + 0.053 Ci
Total Phosphorous (TP)	k = 12 m/yr C* = 0.02 mg/L
Total Nitrogen (TN)	k = 22 m/yr C* = 1.50 mg/L

), which were used to represent the combined effects of multiple pollutant removal mechanisms and processes in the first order decay model [37,40]. For example, a high value of k yields a higher treatment capacity [37].

We used an alternative approach, based on the water depth and hydraulic residence time (HRT), because it has been used for analyzing the long-term treatment performance of wetlands (Equation (3)) [30,31].

$${\mathrm{C}}_{\mathrm{o}}-{\mathrm{C}}^{\mathrm{*}}=\left({\mathrm{C}}_{\mathrm{i}}-{\mathrm{C}}^{\mathrm{*}}\right){\mathrm{e}}^{-{\displaystyle \frac{\mathrm{k}\theta }{\mathrm{d}}}}$$

(3)

where C* is the background concentration, k is an areal rate constant (length/time), θ is the hydraulic residence time (HRT) and d is water depth. This equation can be re-arranged into a removal function as follows (Equation (4)):

$$\mathrm{r}=1-{\displaystyle \frac{{\mathrm{C}}_{\mathrm{o}}}{{\mathrm{C}}_{\mathrm{i}}}}=\left[1-\exp \left(-{\displaystyle \frac{\mathrm{k}\theta }{\mathrm{d}}}\right)\right]\left[1-{\displaystyle \frac{{\mathrm{C}}^{\mathrm{*}}}{{\mathrm{C}}_{\mathrm{i}}}}\right]$$

(4)

Hydraulic residence time (HRT), the average time spent by water within a completely mixed node, is evaluated by the volume of wetland and flow.

$$\theta \left(\mathrm{t}+∆\mathrm{t}\right)=\left(\theta \left(\mathrm{t}\right)+∆\mathrm{t}\right){\displaystyle \frac{\mathrm{V}\left(\mathrm{t}\right)}{\mathrm{V}\left(\mathrm{t}\right)+{\mathrm{Q}}_{\mathrm{i}\mathrm{n}}∆\mathrm{t}}}$$

(5)

where θ(t) is the hydraulic residence time, V(t) is the volume of stored water at time t, Q_in is the inflow rate to the node and ∆t is the current time step.

2.4.2. Wetland Model (PCSWMM) Development

A numerical model to simulate changes in water quality based on the physical, chemical and biological processes within the wetland system was developed using PCSWMM (https://www.pcswmm.com/). PCSWMM is re-developed from the USEPA Storm Water Management Model (EPA SWMM, version 5) with an enhanced Graphical User Interface (GUI) and Geographic Information System (GIS) [57]. We used the K-C* model [30,31] to simulate BOD, TN, TP and TSS concentrations based on HRT, water depth, a background concentration (C*) and an areal rate constant (k), using PCSWMM for three different watersheds, agricultural, urban and forested, under three different climatic conditions (i.e., dry, average and wet). The output concentration depended on the areal rate constant k, which largely depended on the water temperature, type of vegetation and biological activities. Nevertheless, we chose the k- and C*-values based on the past literature [30,31]. The changes in concentrations between the inlet and outlet of the wetland were modeled as a first order decay process using Equations (3)–(5).

The catchments and streamlines developed in the SWAT model and GIS were imported as sub-catchments and conduit layers, respectively, in the PCSWMM. Wetlands were represented by a 1D single storage node, which received the external user-defined time series of discharge and pollutants in the PCSWMM. Existing wetland data were utilized for representing wetlands and to analyze the efficiency to remove pollutants BOD, TN, TP and TSSs. We assumed a constant area for variable depths up to 1.5 m and developed a stage–area relationship to input in the model. The initial depth was assumed 0.5 m, which was controlled by a weir at an outlet structure of the wetland. The average elevation of the wetland and downstream outfall node were estimated based on the topography (DEM) of the modeled area. Since the main objective of the study was to analyze impacts of discharges on the efficiency of the wetland to remove pollutants, the same wetland area (747,560 m²) was considered for all three basins.

Discharge hydrographs and pollutant concentrations (daily average) of BOD, TN, TP and TSSs (simulated from the SWAT) for dry (2006), average (2008) and wet (2013) climatic conditions by water year (WY) (1 October to 31 September) were input in wetland models (Figure 1). The PCSWMMs were run for a year to simulate changes in pollutant concentrations.

2.4.3. Parameter Sensitivity Analysis

We conducted a sensitivity analysis for the k-value (areal rate constant), changing the value by increasing (high) and decreasing (low) ±25% to the original (medium) value. We did not consider C* for sensitivity analysis because it is also a function of Ci (inlet concentration) for BOD and TSSs. We simulated pollutants based on low, medium and high k-values and compared them to analyze the sensitivity of the parameter using a two-tail t-test at the significance level of 5%. We selected data from the agricultural basin of the average year 2013 for this analysis.

3. Results and Discussion

3.1. Sensitivity Analysis

The areal rate constant with ±25% change did not have significant difference for BOD, while total nitrogen and total phosphorous output concentrations varied significantly with changes. The sensitivity analysis between medium and low k-values resulted in the annual average BOD concentrations of 4.18 mg/L and 4.44 mg/L, respectively, resulting in the p-value of 0.08. Similar cases were seen with the medium and high ranges of k-values that resulted in the 4.18 mg/L and 4.11 mg/L annual average BOD concentrations, resulting in a p-value of 0.127. In both comparisons, p-values were higher than 0.05, which suggests no statistically significant difference in the output concentrations. The annual concentration of nitrogen with high, medium and low k-values yielded annual average concentrations of 1.72, 1.68 and 1.81 mg/L. The two-tailed t-test among these values resulted in a p-value less than 0.05. The low, medium and high k-values resulted in annual average output concentrations of total phosphorus of 0.103, 0.086 and 0.078 mg/L. The t-tests between high–medium and low–medium values yielded p-values less than 0.05, which signified that the output pollutant concentrations for total nitrogen and total phosphorous were sensitive to k-values.

3.2. Pollutant Removal

We implemented the use of wetlands to quantify the effectiveness for removal of BOD, sediment and nutrient pollutants in agricultural, urban and forested basins for different climatic conditions (Figure 5). In general, wetlands provide longer flow paths and the contact with vegetation encourages pollutant removal. Wetlands are not only efficient in the removal of particulate-bound contaminants, including trace metals and nutrients, by sedimentation, but also have the advantage of achieving water quality improvement in combination with biological and chemical treatment mechanisms [56].

3.2.1. Biochemical Oxygen Demand (BOD)

The yearly average BOD input for wetlands varied from 30 to 53 mg/L for all three basins (i.e., agricultural, forested and urbanized) and climatic conditions (i.e., dry, average and wet), whereas output reduced noticeably and varied from 4 to 5 mg/L (Figure 5,

Table 2. Characteristics of pollutants’ biological oxygen demand (BOD), total nitrogen (TN), total phosphorous (TP) and total suspended solids (TSSs) for different watersheds and dry, average and wet climatic conditions. * yearly average concentration; ** standard deviation of concentration; *** trendline slope (-ve sign indicates downward trendline); ^ɸ bold and underlined p-value indicates statistically significant trendline.

			BOD (mg/L)			TN (mg/L)			TP (mg/L)			TSS (mg/L)
			Dry	Average	Wet	Dry	Average	Wet	Dry	Average	Wet	Dry	Average	Wet
Agricultural	Input	* Average	33.3	49.4	35.7	7.8	7.6	6.2	0.9	1.9	1.4	877.2	1259.5	854.6
		** SD	16.9	32.4	32.4	1.3	1.6	1.9	0.5	1.5	1.5	444.0	814.3	702.4
	Output	Average	3.6	4.4	4.3	1.5	1.7	1.6	0.0	0.1	0.1	13.5	30.4	25.2
		SD	1.3	5.2	5.7	0.0	0.6	0.6	0.0	0.4	0.5	39.8	111.8	92.0
		Efficiency (%)	81	80	73	80	78	73	92	93	87	97	97	96
		*** Slope	−20.3	−3.6	5.1	−1.7	−47.7	−36.3	−12.2	−77.2	−67.1	−38.6	−32.8	−29.3
		ɸp-value	0.07	0.87	0.82	0.40	0.00	0.00	0.20	0.00	0.00	0.00	0.00	0.00
Forested	Input	Average	47.8	52.3	30.1	5.8	4.3	3.8	0.5	0.8	0.4	1352.7	1409.0	765.1
		SD	20.1	43.8	43.8	2.1	1.6	2.1	0.2	0.6	0.5	609.7	1223.1	654.4
	Output	Average	3.9	4.6	4.7	1.5	1.6	1.4	0.0	0.1	0.1	35.5	60.7	36.8
		SD	2.0	5.1	7.4	0.1	0.5	0.5	0.0	0.2	0.2	104.8	172.8	97.1
		Efficiency (%)	90	84	76	70	60	56	93	91	86	97	96	96
		Slope	−10.7	−5.6	−11.8	−11.9	−28.0	−38.9	−20.3	−66.7	−71.8	−28.3	−24.0	−36.1
		p-value	0.01	0.59	0.29	0.02	0.01	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Urban	Input	Average	49.0	53.3	31.0	6.9	5.4	5.6	0.5	0.7	0.5	383.3	375.9	223.2
		SD	24.0	43.5	43.5	5.0	3.3	4.4	0.3	0.5	0.4	180.2	318.2	168.1
	Output	Average	3.9	4.7	4.7	1.5	1.7	1.4	0.0	0.1	0.1	10.5	15.8	11.0
		SD	1.4	4.2	6.9	0.2	1.0	0.6	0.0	0.1	0.2	23.3	40.3	25.1
		Efficiency (%)	87	84	78	66	62	64	92	90	88	97	95	94
		Slope	−11.6	−12.1	−23.6	−15.4	−46.4	−58.2	−23.8	−80.2	−103.2	−34.3	−23.6	−26.6
		p-value	0.16	0.21	0.00	0.12	0.01	0.00	0.00	0.00	0.00	0.00	0.00	0.00

). The yearly average BOD removal efficiencies of wetlands ranged from 73 to 81%, 76 to 90% and 78 to 87% in different climatic conditions for agricultural, forested and urbanized basins, respectively, (Figure 5, Table 2). The efficiencies of wetlands in our study were higher than 50%, similar to those reported in past studies [58,59], where efficiencies of BOD were 62–69%. Nonetheless the efficiency of the wetland to remove BOD in this study was comparable (~80%) to artificial constructed wetland in a greenhouse [20]. The biological processes for BOD removal, nitrification and denitrification are temperature dependent and it is concluded that the temperature variation affects the wetland’s performance and operation. Nevertheless, Kadlec and Reddy [60] found atmospheric temperature to have minimal effect on biochemical oxygen demand and phosphorus removal but a higher significant effect on nitrogen removal.

The wetland was highly efficacious for BOD removal, mainly during the dry climatic condition, and gradually decreased from dry to wet climatic conditions for all watersheds. However, this trend was not statistically significant (p > 0.05) (Figure 6 and Table 2). The efficiencies were the lowest for agricultural basin compared to forested and urbanized basins in all climatic conditions.

3.2.2. Total Nitrogen (TN)

The yearly average TN input ranged from 4 to 8 mg/L under various climatic conditions, whereas TN output ranged from 1 to 2 mg/L across all three watersheds (Figure 5, Table 2). Total nitrogen (TN) removal efficiency ranged from 60 to 80% for agricultural, forested and urbanized basins for three climatic conditions (Figure 5, Table 2). Total nitrogen input was highest for the agricultural basin compared to the forested and urbanized basins. Consequently, the agricultural basin had the highest removal efficiency (for the dry climatic condition). In the forested basin, the removal efficiency was lowest during the wet year. The TN removal efficiency was the lowest in all three watersheds under different climatic conditions compared to other pollutants (Figure 5, Table 2). In general, TN removal efficiency decreased significantly (p < 0.05) with higher discharges for all basins and climatic conditions except for a few cases (Figure 6, Table 2). The analysis found higher total nitrogen concentration in an agricultural basin compared to the urban and forested basins, which was attributed to the use of fertilizer in the farm areas [61,62].

3.2.3. Total Phosphorous (TP)

The annual average total phosphorous inflow for three basins (agricultural, forested and urbanized) ranged from 0.4 to 1.9 mg/L under three climatic conditions, while total phosphorous outflows were less than 0.1 mg/L for all three basins (Table 2). The agricultural basin had a higher pollutant concentration for total phosphorus, which is supported by past studies e.g., [61,62]. They found higher watershed nitrogen exports in agricultural basins followed by suburban area, and lowest exports were found in the forest watersheds. The wetland’s efficacy was highest during the dry condition, followed by the average and wet climatic conditions, implying that effectiveness decreased significantly (p < 0.05) as discharge increased (Figure 6, Table 2). Wetland removal efficiency ranged from 87 to 92% for agricultural basins, 86 to 93% for forested basins and 88 to 92% for urbanized watersheds. The TP removal in another study varied from 3% to 88% [45]. The analysis conducted by Sovann, Irvine, Suthipong, Kok and Chea [13] observed a lower treatment efficiency for total phosphorus of around 31%, while the treatment efficiency for total nitrogen was around 71%. Contrary to the past study [13], our study resulted in a higher TP (86% to 93%) removal efficiency than TN. This difference in result could be attributable to differences in input concentrations in our study from that of Sovann et al. [13], where higher TP concentrations were observed in both model and field measurements.

3.2.4. Total Suspended Solids (TSSs)

The annual average TSS inflow to wetlands varied from 223.18 to 1408.95 mg/L while output ranged from 10.45 to 60.69 mg/L (Figure 5, Table 2). The annual average TSS input was lowest in the urbanized basin in all considered years. Wetland pollutant removal efficiency was different for various pollutants and climatic conditions. The removal efficiency of the TSSs was the highest overall for all three basins compared to the other pollutants (Figure 5, Table 2). For agricultural and forested basins, the efficiency ranged from 96 to 97% and for urban it was 94 to 97%. The wetland’s efficacy was highest during dry climatic conditions and effectiveness decreased significantly (p < 0.05) as discharge increased (Figure 6, Table 2). Other studies indicated that the removal efficiencies of TSSs were between 22 and 99 percent, with typical removal efficiencies of 65 to 95% [63], which is consistent with our findings. The wetland area (747,560 m²) was considerably large for all three sub-basins in our study, which contributed to the higher hydraulic residence time, allowing most of the sediments to settle. This may have resulted in the wetlands having the highest efficiency for TSS removal. Although this study did not specifically analyze this, generally higher rainfall intensity and the resulting large flow volumes through the wetland may contribute to the resuspension of fine-grained sediment and the transport of suspended particles out of the wetland [45].

3.3. General Observation

Our study showed that the efficiency of wetlands for pollutant removal varies with pollution types (Figure 5, Table 2), which is consistent with past studies [13,45,58,59,63,64]. For example, Simpson and Weammert [64] reported that removal efficiencies varied 61–91% for TSSs, 19–56% for TN and 16–69% for TP. Generally, pollutant removals were higher for the dry climatic condition in our study, while they were lowest for the wet condition irrespective of the basin with different land uses. The removal efficiency showed an inverse relationship (downward trendline) with the higher rainfall and discharge which was consistent with past studies [52,65], but was highly variable specifically for the low discharges (Figure 5 and Figure 6 and Table 2). Nonetheless, trendline analysis could oversimplify relationships and mislead the interpretation of results. These results (low efficiency during high discharges) may not be universally applicable. For example, Griffiths and Mitsch [17] reported higher TP and TN removal during a wet season than a dry.

Our study approach and results should be interpreted cautiously because the models (i.e., SWAT and PCSWMM) may raise some uncertainties. Accuracy of model prediction of stream water quality is challenging due to the complex hydrological, physical, chemical and biological processes among various watershed and meteorological drivers [66]. This study yielded NSEs ranging from 0.70 to 0.48 during SWAT model calibration and validation processes, which is consistent with past studies [16,53]. We calibrated and validated the SWAT model using monthly averaged precipitation and discharges analyzing NSEs, R² and MAE in a nearby watershed. Our results on simulated water qualities may have some uncertainties, as reported in past studies [15,16,53]. Therefore, it is important to improve the SWAT model performance to predict sediment, nutrient and water quality modeling. However, if discharges and water quality measurements are available, they can be input in the wetland pollutant removal model (PCSWMM) to assess the wetland’s performance for pollutant removal. In case measured water quality data is unavailable, a calibrated and validated SWAT model can be used in the study area or a nearby watershed with similar characteristics [18,67].

We adopted a first order decay-based K-C* model (Kadlec and Knight (1996) [30] and only considered the physical processes of the wetland system (i.e., HRT and water depth of wetland) in this study, although other physical, chemical and biological processes impact the wetland’s performance at removing pollutants [45,49,68]. The model ignored different mechanics of pollution removal processes and lumped them together into overall first order decay rate constants. Furthermore, two constant parameters k and C*, which are a reflection of intersystem variability attributed to differences in wetlands’ physical, ecological and pollutant characteristics, were adopted in the model and also possess an uncertainty in their results [31,41]. The sensitivity of k depends on pollutants based on the analysis performed in this study. However, this approach has been used in several past studies successfully to analyze pollutant removal processes in engineered wetland, where pollutant removal processes were comparable to other models and field observations [29,30,31,37,38,39,40,41,42].

Additionally, ET can decrease volumetric flow, thereby increasing hydraulic retention time and increasing concentrations of dissolved constituents [69]. In our study, water temperature, evapotranspiration and type of vegetation have not been considered, which is one of the caveats. Shallow wetlands are strongly influenced by the weather or atmospheric temperature in their pollution removal efficiency [70].

Although some uncertainty exists in models (i.e., SWAT and PCSWMM), as discussed above, we argue that the approach we adopted is reasonable to analyze the impacts of discharge on wetlands’ efficiency (relative) at removing pollutants in a small watershed. The uncertainties in the model will be consistent across all different discharges [71]. Therefore, the model results should not be significantly different when analyzing the efficiency trend as a function of different discharges (i.e., low to high).

3.4. Study Application

Wetlands have been widely used to address hydrology and water quality issues in both agricultural and urban areas [13,18,49,52,58], specifically used to reduce non-point source pollution. But wetland systems are very dynamic; therefore designing these structures based on first order decay processes is highly challenging [39]. The results of this study highlight the advantages of SWAT and wetland pollutant removal modeling to analyze the impacts of wetland on water quality and the role of discharges on the efficiency of pollutant removal. Although the model in this study is not calibrated and validated in the study area, this approach can be effective (and cost-effective) to improve water quality in a watershed. The modeling should be supported by field sampling (even limited), because a full sampling approach may not be practical considering effort and cost [18]. However, it is important to calibrate the model in future studies considering local conditions, specifically to define k- and C*-parameters for designing engineered wetlands. The typical values of areal rate constants and background concentrations vary with local conditions and processes [30,31,41,43]. Furthermore, C* changes from event to event; therefore, it is important to consider lower–upper values for engineered wetland design using a model and to analyze its performance [39].

This approach (integration of SWAT and PCSWMM) tested in a small watershed located in Southern Illinois, USA, is universally transferable in other geographic regions for the preliminary design of wetlands and to analyze their performance at reducing runoff and improving water quality. Therefore, the integration of SWAT and PCSWMM can be useful tools for watershed and water managers in the preliminary design of engineered wetlands to restore watershed water quality and reduce runoff in watersheds, considering future climate change [4,13,18,31,45,46]. According to Tanner [68], the climate of the study area is extremely important in the effluent treatment systems since it affects the biological processes that regulate nutrient removal in the wetlands.

4. Conclusions

An integrated SWAT and PCSWMM (wetland model based on first order decay process) was applied to analyze the impacts of discharge on wetlands’ performance at removing pollutants. Different pollutants (i.e., BOD, TN, TP, TSS) in watersheds were simulated for dry, average and wet climatic conditions to assess the impact of discharge among agricultural, urbanized and forested basins. Hydraulic retention time (HRT) is the main controlling factor for pollution treatment in engineered wetlands in our study, which impacts pollution removal efficiency.

The pollution removal efficiencies of wetlands varied from 56% to 97% considering different pollutants and basins. A higher removal efficiency was observed for total suspended solids and lower for total nitrogen, among all of the pollutants. Overall, the efficiency of pollutant removal was higher during dry climatic conditions compared to the average and wet conditions, which is attributed to higher flows in wet climatic conditions. One of the important findings of this study is that wetland efficiency (to remove pollutants) decreases as discharges increase. However, the efficiencies were highly variable, specifically for low discharges. Furthermore, the sensitivity of the k-parameter (areal rate constant) was pollutant dependent. The study results and approaches can be applicable to other watersheds in the region for the preliminary design of engineered wetlands to remove watershed pollutants, but wetland models need to be calibrated to define k- and C*-parameters based on local conditions. Therefore, this study can be beneficial to the watershed managers in planning and constructing (designing) wetlands and restoring existing wetlands to maintain water quality.