A Numerical Investigation Concerning the Effect of Step-Feeding on Performance of Constructed Wetlands Operating under Mediterranean Conditions

Abstract

The effects of wastewater step-feeding (SF) on the performance of horizontal subsurface flow constructed wetlands (HSF CWs) are numerically investigated. The purpose is to check if this alternative feeding technique increases the ability of HSF CWs to remove pollutants. Two methodologies are used: Initially, the tanks-in-series (TIS) methodology, based on the finite volume method (FVM), is analyzed using the volumetric degradation coefficient λ. In this case, the operation of a CW is similar to a series of continuous stirred-tank reactors (CSTRs) operating under steady conditions. Then, the step-feeding (SF) procedure is presented, in which the CW is operated like a plug flow reactor (PFR). For the numerical investigation, the available experimental data for five existing HSF CWs are used. The results show that SF does not improve the performance of HSF CWs in removing biochemical oxygen demand (BOD) operating under Mediterranean conditions. When the HSF CWs operate without the SF procedure, the performance is between 55 and 81% for the TIS method and 60 and 89% for the PFR method, while the ability of the CW tank to remove the BOD decreases and varies from 48 to 79% (TIS) and from 54 to 86% (PFR), respectively.

1. Introduction

Wastewater treatment is one of the main problems that should be solved in order to ensure the protection of the environment and public health, especially in small cities and settlements where the construction and operation of modern wastewater treatment units are difficult to carry out. An effective, economic and ecological solution for this problem is the construction and operation of constructed wetlands (CWs) [1,2,3]. As these systems have become increasingly popular over the last 25 years, especially in developing countries, it is important to investigate if the use of alternative feeding techniques could have positive or negative effects on their performance. One of these techniques is the step-feeding of the wastewater [4,5,6].

“Step-feeding” (SF) describes the case when wastewater is not entirely introduced at the inlet of the CW, but it is split at several points along the flow path in a gradational way. Most of the studies that have been realized concern the SF procedure in vertical flow (VF) CWs [7,8,9,10,11,12,13]. Fewer studies are available about the implementation of the SF technique in horizontal subsurface flow (HSF) CWs [14,15] with experiments [16] or computer codes [17].

The economic cost for the construction and operation of a CW facility could be useless and have a negative impact if the results do not provide a sustainable solution. This means that the inlet’s contaminated wastewater must have a zero (or almost zero) concentration at the outlet after its “treatment” inside the CW. The operation could also be sustainable when the outlet concentration is significantly smaller in comparison with the inlet concentration. Usually, the CW operation provides positive results, as the outlet wastewater is cleaner than the inlet quantity [1,2].

Alternative feeding techniques, like wastewater step-feeding or effluent recirculation, are used in praxis in order to yield a better result [6]. But, as presented in detail in the present study, sometimes, SF does not improve the performance of a CW in pollutant removal, and in this case, the procedure should not be selected [16,17]. The key factors are usually operation characteristics, like environmental parameters, the type of pollutant and the dimensions of the CW.

In the present study, we numerically investigated if SF could increase the ability of HSF CWs to remove the biochemical oxygen demand (BOD), which is one of the main pollutants monitored in municipal wastewater. This investigation is realized for pilot-scale HSF CWs, which were constructed and operated under Mediterranean conditions. In this case, it is interesting to check the effectiveness of HSF CWs in BOD removal, as the Mediterranean climate is usually characterized by hot, dry summers and cool, wet winters. The term “Mediterranean” is usually used for locations between about 30° and 45° latitude north and south of the Equator and on the western sides of the continents.

Two different methodologies, TIS and PFR, are used. The first method has been used in CWs, but recently, PFR has been preferred. Grismer et al. [18] initially used the TIS method for the analysis of residence time distributions (RTDs) to determine the detention times, water velocities and dispersion coefficients. Data from an HSF CW were used for the treatment of winery effluent when the wetlands were planted initially with cattails while remaining unplanted in the second stage. The porous material was pea gravel. The TIS model can also be applied in VF CWs for the prediction of the effects feeding (continuous and batch) and hydraulic loading rate (HLR) on hydraulic behavior [19] and for assessing pesticide reduction within a Bayesian framework [20].

In contrast to TIS, the PFR method in CWs has been used more recently. Brito-Espiso et al. [21] developed a mathematical model for primary treatment using HSF CWs for ammonia (NH₃) removal. This model was based on the finite element method (FEM) and provided useful design information. Similarly, the PFR model has been used in order to show the importance of vetiver grass for secondary wastewater treatment in HSF CWs for the removal of antibiotics and nutrients [22]. Finally, the impact of HLR on VF CW performance when a combination of the PFR and CSTR models is chosen for the removal of various pollutants (BOD, NH₄, NO₃) has been investigated [23].

The above analysis shows that there are no available studies that evaluate SF techniques in CWs using the TIS or PFR methodology. This is applied in the present study, where a numerical investigation is followed, and the results are compared with other studies, which conduct experiments or use computer codes. The advantage of numerical analysis is that it is an economical form of research that does not need to construct a whole facility (experimental part) or to buy/develop a computer code (computational procedure). The results of the present study could be an important and economical tool in order to predict the sustainability of these systems when they operate under Mediterranean conditions. They also provide the possibility to avoid the (construction and operation) costs if the conclusion is that the SF procedure does not really improve the performance of CWs.

2. Methods and Materials

2.1. The Numerical Formulation of Step-Feeding

The problem of SF is formulated for the case of a pilot-scale HSF CW, which is presented in Figure 1.

Five similar HSF CWs were constructed in total and are still in operation, in the facilities of the Laboratory of Ecological Engineering and Technology, Department of Environmental Engineering, Democritus University of Thrace, Xanthi, Greece. The dimensions of each CW tank are 3 m in length, 0.75 m in width and 1 m in depth, while the thickness of the porous media is 0.45 m.

A different combination of vegetation and porous medium appears in each tank. The types of porous materials were medium gravel (MG), fine gravel (FG) and cobbles (CO). The tanks were planted with common reeds (Phragmites australis—R) or cattails (Typha latifolia—C), while one tank was kept unplanted (Z). Therefore, the symbols for the five tanks were MG-R, MG-C, MG-Z, FG-R and CO-R. The units were equipped with vertical perforated plastic pipes (50 mm diameter each), which could be used in praxis during the SF procedure for the introduction of the wastewater at several points along the length (each pipe per 1 m), as shown in Figure 1.

The main operational characteristics are presented in

Table 1. Operation characteristics of HSF CWs.

CW Tank	HRT (Days)	ε (%)	Κ (m/sec)	Vp (m3)	Tav (°C)	Cin (mg/L)	Cout (mg/L)	λ (Day−1)
MG-R	6	0.35	0.7196	0.3544	5.7	321.5	141.8	0.1212
	8	0.355	0.7640	0.3594	11.8	333.0	87.7	0.1992
	14	0.38	0.9743	0.3848	18.7	332.0	60.9	0.2546
	20	0.35	0.7196	0.3544	22.4	345.3	48.3	0.3852
MG-C	6	0.33	0.5620	0.3341	25.7	327.0	26.0	0.3852
	8	0.335	0.5980	0.3392	4.5	350.0	124.0	0.1155
	14	0.34	0.6366	0.3443	8.8	358.6	72.4	0.1799
	20	0.33	0.5620	0.3341	13.1	348.5	50.8	0.2178
MG-Z	6	0.37	0.9120	0.3746	16.3	380.0	34.3	0.2740
	8	0.37	0.9120	0.3746	19.2	373.7	22.6	0.3217
	14	0.37	0.9120	0.3746	11.2	372.0	65.3	0.1126
	20	0.37	0.9120	0.3746	17.9	358.7	40.3	0.1426
FG-R	6	0.29	0.0199	0.2936	21.2	461.5	30.7	0.1784
	8	0.29	0.0199	0.2936	8.7	355.8	45.5	0.0923
	14	0.31	0.0262	0.3139	13.8	323.3	22.9	0.1210
	20	0.29	0.0199	0.2936	16.3	364.3	34.8	0.1066
CO-R	6	0.28	3.8710	0.2835	21.9	381.4	35.0	0.1085
	8	0.28	3.8710	0.2835	5.7	321.5	141.8	0.1212
	14	0.28	3.8710	0.2835	11.8	333.0	87.7	0.1992
	20	0.28	3.8710	0.2835	18.7	332.0	60.9	0.2546

, for a total operation period of two years. For each CW tank and for the used values of hydraulic residence time HRT (6, 8, 14 and 20 days), the experimental values of porosity ε, hydraulic conductivity K, volume of porous material V_p, average temperature T_av, inlet and outlet concentrations of the pollutant (BOD) C_in and C_out, respectively, and first-order decay coefficient λ are presented. A more detailed description of the calculation and explanation for each one of these parameters is given by [24]. The impact of the size of porous materials is obvious in the values of hydraulic conductivity, especially between the FG-R and CO-R tanks.

The total inflowing wastewater flow rate Q = Q_in, in L/day, is distributed in three inflowing points along the CW tank: x₀ = 0, x₁ = 1 m and x₂ = 2 m. The following equations give the inflowing wastewater flow rates for each one of these points:

$$Q_1=xQ$$

(1)

$$Q_2=yQ$$

(2)

$$Q_3=zQ$$

(3)

where x, y and z are the corresponding percentages of Q in each position; therefore, they fulfill the following condition:

$$x+y+z=1$$

(4)

When the SF procedure is present during the operation of the wetlands, i.e., x#0, y#0 and z#0, the aim is to determine the optimal values of these percentages in order to minimize the concentration of the pollutant at the outlet C_out, in mg/L. Then, this value should be compared with the corresponding outlet concentration for the case that CW is operated without SF, i.e., x = 1 and y = z = 0. The results of this comparison give the answer to whether the performance of the tank was increased or decreased by using the SF procedure. The performance R, in %, is given by the following equation:

$$\mathit{R}=\frac{{\mathit{C}}_{\mathit{i}\mathit{n}}-{\mathit{C}}_{\mathit{o}\mathit{u}\mathit{t}}}{{\mathit{C}}_{\mathit{i}\mathit{n}}}$$

(5)

2.2. Analysis by Using the TIS Methodology

For the numerical investigation of the problem, first the tanks-in-series (TIS) methodology is used [25], which is based on finite volume method [26,27]. It is considered that the wetlands are operated as continuous stirred-tank reactors (CSTRs), under steady conditions, and the volumetric degradation coefficient λ is used. According to the experimental procedure [16], the tank is divided into three equal cells (Figure 2), with corresponding volumetric degradation coefficients λ₁, λ₂ and λ₃.

The above cells have equal volume, which is

$$V_1=V_2=V_3=\frac{V_p}{3}$$

(6)

where V_p = εLWd; L is the length of the CW, in m; W is the width of the CW, in m; and d is the depth of the porous material, in m.

As shown in Figure 2, the operation of the above system begins with the entering flow rate Q₁ at the inlet of the first cell. The outlet flow rate of the first cell Q₁ is mixed with the entering flow rate of the second cell Q₂. Similarly, the outlet flow rate of the second cell Q₁ + Q₂ is mixed with the entering flow rate of the third cell Q₃. Finally, at the end of the operation, the concentration C_out,3 is the outlet concentration of the system C_out which should be minimized.

The numerical formulation of the whole procedure is the following: For the first cell K1, the outlet concentration C_out,1 and the hydraulic residence time HRT₁ are

$$C_{out,1}=\frac{C_{in}}{1+{\lambda }_{1}HR{T}_1}$$

(7)

$$HR{T}_1=\frac{V_1}{Q_1}=\frac{V_1}{xQ}$$

(8)

The pollutant mass balance between the first and the second cell is used in order to determine the inlet concentration C_in,2:

$$Q_1{C}_{out,1}+Q_2{C}_{in}=(Q_1+Q_2)C_{in,2}$$

(9)

Thus, it is

$$C_{in,2}=\frac{Q_1{C}_{out,1}+Q_2{C}_{in}}{Q_1+Q_2}$$

(10)

Equations (7) and (10) show

$$C_{out,1}\le C_{\mathit{in},2}$$

(11)

Similarly, for the second cell K2, the outlet concentration C_out,2 and the hydraulic residence time HRT₂ are

$$C_{out,2}=\frac{C_{in,2}}{1+{\lambda }_{2}HR{T}_2}$$

(12)

$$HR{T}_2=\frac{V_2}{Q_1+Q_2}=\frac{V_2}{(x+y)Q}$$

(13)

The pollutant mass balance between the second and the third cell is used:

$$(Q_1+Q_2)C_{out,2}+Q_3{C}_{in}=Q{C}_{in,3}$$

(14)

Thus, the inlet concentration in the third cell C_in,3 is

$$C_{in,3}=\frac{(Q_1+Q_2)C_{out,2}+Q_3{C}_{in}}{Q}$$

(15)

Equations (12) and (15) show

$$C_{out,2}\le C_{\mathit{in},3}$$

(16)

Finally, for the third cell K3, the outlet concentration C_out,3 (which is the outlet concentration for the whole system C_out) and the hydraulic residence time HRT₃ are

$$C_{out,3}=C_{out}=\frac{C_{in,3}}{1+{\lambda }_{3}HR{T}_3}$$

(17)

$$HR{T}_3=\frac{V_3}{Q_1+Q_2+Q_3}=\frac{V_3}{(x+y+z)Q}=\frac{V_3}{Q}$$

(18)

The main problem could be expressed numerically as follows: Which are the optimal values of x, y and z in order to minimize the function C_out = f (x, y, z)?

2.3. Analysis by Using the PFR Methodology

The pilot-scale HSF CWs could be considered to operate like plug flow reactors (PFRs). This means that the contribution of the dispersion during the operation of the tanks is negligible. By following the same steps as in the previous paragraph, the inlet and outlet concentrations are determined.

Thus, for the first cell K1,

$$HR{T}_1=\frac{V_1}{xQ}$$

(19)

$$C_{out,1}=C_{in}{e}^{-{\lambda }_{1}HR{T}_1}$$

(20)

Similarly, for the cell K2,

$$HR{T}_2=\frac{V_2}{(x+y)Q}$$

(21)

$$C_{in,2}=C_{in}\frac{y+x{e}^{-{\lambda }_{1}HR{T}_1}}{x+y}$$

(22)

$$C_{out,2}=C_{in}\frac{y+x{e}^{-{\lambda }_{1}HR{T}_1}}{x+y}{e}^{-{\lambda }_{2}HR{T}_2}$$

(23)

Finally, for the last cell K3,

$$HR{T}_3=\frac{V_3}{Q}$$

(24)

$$C_{in,3}=C_{in}\left[z+y{e}^{-{\lambda }_{2}HR{T}_2}+x{e}^{-{\lambda }_{3}HR{T}_3\left(\frac{x+y}{x(x+y)}\right)}\right]$$

(25)

$$C_{out,3}=C_{out}=C_{in}\left[z+y{e}^{-{\lambda }_{2}HR{T}_2}+x{e}^{-{\lambda }_{3}HR{T}_3\left(\frac{x+y}{x(x+y)}\right)}\right]e^{-{\lambda }_{3}HR{T}_3}$$

(26)

3. Results and Discussion

3.1. Numerical Investigation

The numerical investigation is presented below in order to examine whether the SF procedure increases or decreases the performance in BOD removal of the five HSF CWs (MG-R, MG-C, MG-Z, FG-R and CO-R) which were described previously. For the main parameters (C_in, V_p, λ), average values from the available experimental data presented in Table 1 are used and presented in

Table 2. Impact of SF on HSF CW performance in BOD removal.

CW Tank	Cin,av (mg/L)	λav (Day−1)	Vp,av (m/sec)	SF Scenario	Cout/Cin		HSF CW Performance (%)
					TIS	PFR	TIS	PFR
MG-R	333	0.240	363	(A) (B) (C) (D) (E)	Cout = 0.1945Cin Cout = 0.2079Cin Cout = 0.2254Cin Cout = 0.2736Cin Cout = 0.3130Cin	Cout = 0.1132Cin Cout = 0.1374Cin Cout = 0.1391Cin Cout = 0.2033Cin Cout = 0.2216Cin	80.55 79.20 77.46 72.63 68.66	88.69 86.27 86.11 79.69 77.84
MG-C	346	0.225	339	(A) (B) (C) (D) (E)	Cout = 0.2284Cin Cout = 0.2412Cin Cout = 0.2583Cin Cout = 0.2809Cin Cout = 0.3259Cin	Cout = 0.1481Cin Cout = 0.1746Cin Cout = 0.1792Cin Cout = 0.1940Cin Cout = 0.2544Cin	77.17 75.88 74.17 71.91 67.41	85.18 82.54 82.09 80.60 74.56
MG-Z	371	0.210	375	(A) (B) (C) (D) (E)	Cout = 0.2202Cin Cout = 0.2332Cin Cout = 0.2503Cin Cout = 0.2775Cin Cout = 0.3232Cin	Cout = 0.1396Cin Cout = 0.1655Cin Cout = 0.1677Cin Cout = 0.1858Cin Cout = 0.2465Cin	77.98 76.68 74.96 72.24 67.68	86.04 83.45 83.23 81.42 75.35
FG-R	376	0.125	297	(A) (B) (C) (D) (E)	Cout = 0.4456Cin Cout = 0.4535Cin Cout = 0.4646Cin Cout = 0.4929Cin Cout = 0.5189Cin	Cout = 0.3955Cin Cout = 0.4086Cin Cout = 0.4207Cin Cout = 0.4242Cin Cout = 0.4593Cin	55.41 54.65 53.54 50.71 48.11	60.45 59.14 57.93 57.58 54.07
CO-R	342	0.170	285	(A) (B) (C) (D) (E)	Cout = 0.3613Cin Cout = 0.3711Cin Cout = 0.3848Cin Cout = 0.4128Cin Cout = 0.4520Cin	Cout = 0.2974Cin Cout = 0.3270Cin Cout = 0.3295Cin Cout = 0.3353Cin Cout = 0.3799Cin	63.87 62.89 61.52 58.72 54.80	69.97 67.30 67.05 66.47 62.01

(C_in,av, V_p,av, λ_av). The range of the values’ first-order decay coefficient was determined in [24], in which the dependence of λ on HRT and temperature T for the same CW tanks of the present study was analyzed and relative empirical relations were developed. The selected average values of λ_av for the numerical investigation of the present study are based, for each HSF CW, on these results [24].

Four scenarios are investigated, and the outlet concentration (C_out) for each one of these scenarios is compared with the corresponding outlet concentration when the wetlands are operating without SF. Apart from scenario (A), in which the CWs are operated without SF (x = 1 and y = z = 0), four more scenarios are investigated: (B) The wastewater is introduced at two inflowing points along the HSF CW length: 80% at x₀ = 0 and 20% at x₁ = 1 m, i.e., x = 0.80, y = 0.20 and z = 0. (C) The wastewater is introduced at two inflowing points along the HSF CW length: 60% at x₀ = 0 and 40% at x₁ = 1 m, i.e., x = 0.60, y = 0.40 and z = 0. (D) The wastewater is introduced at three inflowing points along the HSF CW length: 60% at x₀ = 0, 25% at x₁ = 1 m and 15% at x₂ = 2 m, i.e., x = 0.60, y = 0.25 and z = 0.15. (E) The wastewater is introduced at three inflowing points along the HSF CW length: 33.33% at x₀ = 0, 33.33% at x₁ = 1 m and 33.33% at x₂ = 2 m, i.e., x = y = z = 0.333.

First, the TIS methodology is followed. According to Equation (6), V₁ = V₂ = V₃ = V = V_p/3. By using Equations (1)–(4) and (7)–(18), the values of inlet concentrations, outlet concentrations and hydraulic residence time are as follows:

Cell K1:

$$HR{T}_1=\frac{V}{xQ}$$

(27)

$$C_{out,1}=\frac{xQ{C}_{in}}{xQ+V{\lambda }_{av}}$$

(28)

Cell K2:

$$HR{T}_2=\frac{V}{(x+y)Q}$$

(29)

$$C_{in,2}=\left(\frac{\frac{x^{2}Q}{xQ+V{\lambda }_{av}}+y}{x+y}\right)C_{in}$$

(30)

$$C_{out,2}=\left(\frac{\frac{x^{2}Q}{xQ+V{\lambda }_{av}}+y}{x+y+V{\lambda }_{av}}\right)C_{in}$$

(31)

Cell K3:

$$HR{T}_3=\frac{V}{Q}$$

(32)

$$C_{in,3}=\left[(x+y)\frac{x^{2}Q}{(xQ+V{\lambda }_{av})(x+y+V{\lambda }_{av})}+z\right]C_{in}$$

(33)

$${ C}_{out,3}=C_{out}=\frac{Q}{Q+V{\lambda }_{av}}\left[(x+y)\frac{x^{2}Q}{(xQ+V{\lambda }_{av})(x+y+V{\lambda }_{av})}+z\right]C_{in}$$

(34)

Thus, by following the limitations of Equation (4), the purpose is to determine the optimal values of x, y and z which minimize the function C_out = f(x, y, z) of the above Equation (34).

Similarly, the PFR methodology is used. For the same values of V_P, HRT and λ_av, Equations (1)–(4) and (19)–(26) give the following results:

Cell K1:

$$HR{T}_1=\frac{V}{xQ}$$

(35)

$$C_{out,1}=C_{in}{e}^{-\frac{V{\lambda }_{av}}{xQ}}$$

(36)

Cell K2:

$$HR{T}_2=\frac{V}{\left(x+y\right)Q}$$

(37)

$${ C}_{in,2}=C_{in}\left(\frac{y+x{e}^{-\frac{V{\lambda }_{av}}{xQ}}}{x+y}\right)$$

(38)

$$C_{out,2}=C_{in}\left(\frac{y+x{e}^{-\frac{V{\lambda }_{av}}{xQ}}}{x+y}\right)e^{-\frac{V{\lambda }_{av}}{(x+y)Q}}$$

(39)

Cell K3:

$$HR{T}_3=\frac{V}{Q}$$

(40)

$$C_{in,3}=C_{in}\left[z+y{e}^{-\frac{V{\lambda }_{av}}{(x+y)Q}}+x{e}^{-\frac{V{\lambda }_{av}}{xQ}\left(\frac{x+y}{x(x+y)}\right)}\right]$$

(41)

$$C_{out,3}=C_{out}=C_{in}\left[z+y{e}^{-\frac{V{\lambda }_{av}}{(x+y)Q}}+x{e}^{-\frac{V{\lambda }_{av}}{xQ}\left(\frac{x+y}{x(x+y)}\right)}\right]e^{-\frac{V{\lambda }_{av}}{xQ}}$$

(42)

Similarly, the optimal values x, y and z which minimize the function C_out = f(x, y, z) of Equation (42) should be determined, when the limitations of Equation (4) are followed.

The results of the numerical investigation for each one of the two methodologies (TIS and PFR) and for each SF scenario are presented in Table 2. Apart from the performances of CW tanks in BOD removal, the relationships between the inlet C_in and the outlet C_out concentrations are shown.

3.2. Discussion and Comparison with Other Studies

For both methodologies (TIS and PFR) and for all the HSF CWs, the results show that the minimum values of outlet concentration C_out (therefore, the maximum performance in BOD removal) were found for the scenario according to which the CWs are operating without SF. This confirms that this procedure does not increase the performance of the HSF CWs, at least for wetlands operating in similar Mediterranean conditions to those in the present study. On the contrary, when the number of inlet points of the wastewater among the CW tanks is increased to more than one inlet point, see scenarios (D) and (E) of the previous paragraph, the ability of the wetlands to remove BOD becomes lower. The main explanation for this is that the parts of the effluent entering at the points x₁ = 1 m and x₂ = 2 m have lower HRT; thus, the total time of the operation of the HSF CW is shorter.

The fact that this conclusion remains the same for different combinations of plants and porous materials is also interesting. The real time of the operation of the HSF CWs, which gave us the experimental data used, was 2 years and included all the seasons of the year. This means that all the climate and meteorological conditions were taken into account; thus, the SF procedure was examined for high and low temperatures (and corresponding plant growth, which increases the BOD removal). Therefore, in the present study, the SF procedure fails to increase the ability of the CW to remove pollutants, irrespective of the growth of the plants and the size of the porous medium. The reliability of the present study is confirmed by the fact that the conclusions are the same when (for the same CW tanks) the SF procedure is followed by using experiments and computer codes.

The conclusion of the present work confirms the results of a relevant computational study [17]: For the same pilot-scale HSF CWs, and by using the computer code Visual MODFLOW, it was proven that SF acts negatively on the performance of these tanks in removing BOD. Schematically, a part of the results from this computational study is shown in Figure 3. This diagram concerns the HSF CW tank which contains medium gravel as porous material and is planted with common reed (Phragmites australis). The values of the main parameters are HRT = 14 days, C_in = 360 mg BOD/L, λ = 0.06 day⁻¹ and ε = 0.38. For the SF procedure, scenario (D) is investigated, i.e., x = 0.60, y = 0.25 and z = 0.15.

In Figure 3, the outlet BOD concentration without SF is C_out,A = 8.42 mg/L, while that with SF is noticeably higher, C_out,D = 17.47 mg/L. The inequalities (11) and (16) could explain the occurrence of the “leaps” in the longitudinal concentration distribution profile at the feeding positions x₁ = 1 m and x₂ = 2 m.

Stefanakis et al. [16] also presented an experimental procedure for the same tanks, and they investigated the above SF scenarios (D) and (E), for the pollutants BOD, chemical oxygen demand (COD) and total Kjeldahl nitrogen (TKN). While an agreement has been observed between these experimental results and the numerical results of the present work concerning scenario (E), i.e., x = y = z = 0.333, a small difference appears for scenario (D), i.e., x = 0.60, y = 0.25 and z = 0.15. Stefanakis et al. [16] concluded that the SF slightly increases the performance of the HSF CWs in BOD removal, for this scenario. This can be explained by the fact that the experiments were realized during the summer months, when the ability of CWs to remove pollutants is always greater, due to high temperatures and the growth of the plants.

Li et al. [15] experimentally investigated the removal of COD, total nitrogen (TN) and ammonium nitrogen (NH4⁺-N). As the results show, the HRT could influence the effectiveness of the SF procedure. Especially for COD and NH4⁺-N removal, the SF procedure acted negatively on CW performances for a low HRT, HRT = 5 days; however, it increased the performance of the wetlands for a higher HRT, HRT = 8.4 days. Concerning the TN, SF seems to increase the performance of HSF CWs significantly, but the removal of nitrogen requires special investigation as it is governed by the phenomenon of adsorption.

In contrast with HSF CWs, more studies have been realized concerning SF in VF CWs. Fan et al. [12] investigated an SF scenario where half of the wastewater is introduced at the inlet and the other half is introduced in the middle of a VF CW. The results showed that the COD removal was increased slightly, while the increase in the performance in TN removal was remarkable. Hu et al. [6] succeeded in showing higher performance in nitrogen removal in VF CWs for the SF scenario (B), i.e., x = 0.80, y = 0.20 and z = 0.

Moreover, Khajah et al. [7] investigated the effects of SF concerning COD and TN removal in a VF CW, for scenarios (B) and (C). The performance was increased concerning COD but remained stable for TN removal. Similarly, positive effects on the ability of VF CWs to remove pollutants were recorded by Saeed et al. [8], concerning BOD and total phosphorus (TP) removal for the SF scenario (F).

Wang et al. [10] did not find an important effect of SF on organics removal for scenarios (E) and (F), as the performance with and without the SF procedure remained almost stable. Finally, Torrijos et al. [9] investigated the effects of SF on the performance of a hybrid system, which included both an HSF and a VF CW. The removal of COD was increased from 96% to 99% for the SF scenario (F).

The following table (

Table 3. Performance (%) of HSF and VF CWs for various pollutants and SF scenarios.

Ref.	CW Type	HRT (Days)	Pollutant	Step-Feeding Scenario
				(A)	(B)	(C)	(D)	(E)	(F)
[16]			BOD	84.2	—	—	88.7	—	—
	HSF	6	COD	82.7	—	—	87.4	—	—
			TKN	76.5	—	—	75.5	—	—
			BOD	89.4	—	—	—	83.3	—
		14	COD	89.4	—	—	—	84.0	—
			NH4+-N	71.8	—	—	—	38.5	—
[17]	HSF	6-20	BOD	92.5	91.7	90.3	88.2	82.1	—
[15]			COD	87.2	67.8	—	—	—	—
	HSF	5	NH4+-N	99.0	93.5	—	—	—	—
			TN	55.5	60.6	—	—	—	—
		8.4	COD NH4+-N TN	83.2	90.9	—	—	—	—
				98.9	99.1	—	—	—	—
				56.4	88.1	—	—	—	—
[5]	VF	3	COD	88.0	—	—	—	—	97.0
			TN	44.0	—	—	—	—	82.0
[6]	VF	9	TN	85.0	90.2	—	—	—	—
[7]	VF	0.33	COD	75.0	94.0	90.0	—	—	—
			TN	73.5	70.0	75.0	—	—	—
[8]	VF	1.5	BOD	87.4	—	—	—	—	93.9
			TP	92.4	—	—	—	—	94.2
[10]	VF	0.5	BOD	97.2	—	—	—	95.9	96.8
			COD	96.5	—	—	—	95.9	99.9
[9]	HSF+VF	15	COD	96.0	—	—	—	—	99.0

) presents the results of all these studies. The performance of the HSF and VF CWs for various pollutants and for all the different SF scenarios described in the previous paragraphs are shown. These results show that the SF procedure generally increases the ability of VF CWs to remove pollutants, whereas it has a negative effect on the performance of HSF CWs. The results indicate that the flow of the inlet wastewater is a key factor for the implementation of this alternative feeding technique in full-scale CW facilities. In any case, the environmental operation parameters (like temperature, rainfall and growth of plants) could give non-stable results, especially in a comparison between experimental studies. Especially for the present study, the negative effect of the SF technique on the HSF CW operation could be explained by the fact that the part of the wastewater that is introduced among the tanks decreases the total hydraulic residence time (HRT) of the pollutant, in comparison with the case when the wastewater is introduced entirely at the inlet. Therefore, a part of the wastewater does not remain inside the CW facility long enough in order to be “cleaned”, and the outlet pollutant concentration is higher than that in the case where SF is not applied.

4. Conclusions

The numerical investigation of the present study confirmed that the step-feeding (SF) technique acts negatively on the performance of HSF CWs, especially when more than two inlet positions along the CW for the wastewater are chosen. Thus, this operation technique does not improve the ability of HSF CWs to remove pollutants, at least when they operate under Mediterranean conditions. Therefore, the operation of (pilot-scale or full-scale) CWs by using the SF technique is not recommended, at least for facilities with similar characteristics to those in the present study. The comparison of these results with other similar studies showed the importance not only of the wastewater direction (HSF or VF CWs) but also of the operational characteristics of the CWs (kind of vegetation and porous material, climatic parameters, hydraulic residence time, type of pollutant). It is important to investigate the effect of these characteristics on SF techniques in future relevant studies. The proposed methodology also provides the possibility for an economic prediction of the effectiveness of CWs, by avoiding the operation and maintenance cost of a real facility. As a general conclusion, the results of the present study could be useful for a better understanding of the use of alternative feeding techniques (like wastewater step-feeding or recirculation of the effluent) in constructed wetlands.