1. Introduction
Constructed wetlands (CWs) have been recognized as a reliable wastewater treatment technology [
1,
2]. CWs are engineered systems that are designed and constructed to utilize natural processes that improve water quality in wetlands in a more controlled or modified environment [
1]. Compared to conventional treatment systems, constructed wetlands can be more economical, easily operated and maintained, and therefore have a strong potential for applications in developing countries. Furthermore, CWs as a decentralized wastewater technology, can be applicable in rural and small communities, eco-cities, and individual households that do not have the resources or need for complex and costly centralized wastewater treatment systems [
3,
4].
There are two different types of constructed wetlands: free water surface (FWS) and sub-surface flow (SSF). SSF wetlands may be classified according to the direction of flow, either horizontal or vertical [
5]. In developed temperate-climate countries, the horizontal sub-surface flow constructed wetlands (HSF-CWs) have been successfully used for the treatment of various types of wastewater for more than four decades [
1,
5]. However, to date there has been limited information about CWs in developing countries [
3,
6,
7] particularly those located in arid and semi-arid areas. The adoption of CWs has been surprisingly slow there due to the lack of understanding of CW’s potential benefits, actual performance, and appropriate design features [
7]. In addition, the selection and design of CWs in arid areas requires proper consideration of the particular climatic conditions [
8]. For example, Vera
et al. [
8] considered the hydraulic retention time (HRT) and plant species as the two important design parameters in hyper- and super-arid areas.
Jordan is one of the five most water-deprived countries in the world. The climate is generally arid to semi-arid; 90% of Jordan's land area receives less than 200 mm rainfall per year [
9]. In Jordan, Al-Omari
et al. [
10] investigated the performance of HSF-CWs, and indicated that HSF-CWs were capable of reducing BOD, different forms of nitrogen, TSS, FC, and TC. Whereas, HSF-CWs design parameters such as media size, HRT, bed dimensions, and plantation types have not been studied or reported as yet.
To provide the needed information mentioned above, this study focuses on evaluating the performance and the treatment efficiencies of HSF-CW in arid climates, with an attempt to comprehensively evaluate the various designs of HSF-CW on their removal of organic pollutants, heavy metals, anion, and cations, etc. Two types of media size (fine and coarse), two bed dimensions (long and short) and three plantation types (reed, kenaf, or without vegetation as a control) were tested to compare the proper combinations on their performance, and to subsequently identify the major design criteria and proper systems for HSF-CWs in arid areas. The study was conducted from November 2008 through November 2011.
2. Materials and Methods
2.1. Site Description
This study was implemented in the Al-Samra agricultural research station in central Jordan, 36 km northeast of downtown Amman and at an elevation of 550 m above sea level. The mean precipitation and evaporation of the region is 123 mm·year
−1 and 1500 mm·year
−1, respectively (according to Al-Samra meteorological station data from 2005–2014).
Figure 1 shows the location of our study area.
Figure 1.
Location map of the study area.
Figure 1.
Location map of the study area.
A benchmark site for HSF-CW treatment systems was established in 2008. The system was designed to capture and store the partially-treated municipal wastewater effluent in an influent holding pond, and then re-treat it through HSF-CW beds. The outlet water from HSF-CW beds was stored in another effluent-holding pond and then used for irrigating forest trees. The influent and effluent holding ponds had capacities of 300 m3 and 150 m3, respectively. The HSF-CW system was comprised of 17 sub-surface HSF-CWs. The mean hydraulic residence time was two days. The partially treated municipal wastewater was the only available source of irrigation in this research station and therefore selected as the influent of our HSF-CWs.
2.2. Treatment System Design
The test site for the HSF-CW was divided into two main categories: (1) nine long beds with dimensions of 9.5 m × 1.7 m × 0.8 m (L × W × D). Coarse volcanic tuff (10–20 mm in diameter) was used as the wetland media. Three of the beds were planted with the reed plant
Phragmites australis and three of the beds were planted with the kenaf plant
Hibiscus cannabinus. The remaining three beds were left without vegetation and were used as controls. (2) Eight short beds with dimensions of 6.5 m × 2.5 m × 0.8 m (L × W × D); four beds were filled with coarse volcanic tuff media (10–20 mm in diameter) and the other four beds were filled with fine volcanic tuff (4–8 mm in diameter). Two of the beds filled with fine media and two of the beds filled with coarse media were planted with reed plants; the remaining four beds did not have vegetation and were used as controls. The area dimension and field facilities were the major variables that controlled the distribution of the treatments discussed above. A schematic of the experimental layout is shown in
Figure 2. In the middle of each bed, a tube was installed to facilitate daily measurements of the water temperature. All long and short beds were designed to have the same volume (13 m
3 for each bed). The total HSF-CW volume and surface area was 221 m
3 and 275 m
2, respectively. The reeds were planted in November 2008 and removed at the end of the study in November 2011; the planting density was nine rhizomes per square meter. Kenaf is an annual summer crop and was planted in three seasons in 2009, 2010, and 2011 from May until November, with 20 transplants per square meter.
Figure 2.
The horizontal sub-surface flow constructed wetland treatment (HSF-CW) system layout.
Figure 2.
The horizontal sub-surface flow constructed wetland treatment (HSF-CW) system layout.
In this study, we use the following abbreviations: long (L) bed, short (S) bed, coarse (C) media, and fine (F) media. Reeds, kenaf, and no vegetation will be referred to as R, K, and N, respectively. For example, the long bed with coarse media planted with reeds will be referred to as (LCR) and likewise for all of the treatments discussed above.
2.3. Wastewater Quality Monitoring
The HSF-CW beds’ influent and effluent water were sampled and analyzed by the Jordanian National Center for Agricultural Research and Extension on a bimonthly basis which lasted for 18 months from November 2008 until August 2011. In all, 327 water samples were collected and analyzed for 21 water quality parameters.
The chemical and biological characteristics of wastewater used in this study were: Biological Oxygen Demand (BOD
5), Chemical Oxygen Demand (COD), pH, Electrical Conductivity (EC), Total Suspended Solid (TSS), FC, sulfate (SO4
2−), magnesium (Mg
2+), calcium (Ca
2+), chloride (Cl
−), sodium (Na
+), potassium (K
+), zinc (Zn), cadmium (Cd), cobalt (Co), chromium (Cr), lead (Pb), arsenic (As), phosphorus (P), carbonates (CO
3−), and bicarbonates (HCO
3−). The samples were analyzed according to the Standard Methods for the Examination of Water and Wastewater [
11]. Details about the sampling procedure, preparations, and instruments are available in the manual [
12]. For each water quality parameter, the removal efficiency (%) was calculated based on mass flow difference between the effluent and influent relative to the influent.
2.4. Wastewater Flow Rates and Distribution
Two water meters were used to measure the influent flow rates, each of them located in the main influent supply line as shown in
Figure 2. Influent flow rate was controlled to be the same for all the beds through an identical flow rate controller with 17 feed lines, each one installed at the entry point of each bed. Calibration of water flow for even distribution was done before starting up the system. The mean of the two influent water meter readings divided by 17 was used to determine the influent flow rate for each bed. Due to water shortage, our total wastewater share varied from 35 m
3·day
−1 in the winter seasons to 25 m
3·day
−1 in the summer seasons, as regulated by the governmental water conservation policy.
An additional water meter and controller for 17 outlet lines were used for effluent flow rate measurement; each one was installed at the end of the bed. However, since the effluent flow rate from each bed was very low the water meters did not work properly. Therefore, we manually measured the effluent flows, and collected two-hour composite effluent samples for each bed. We repeated this measurement six times for each bed, three times in the winter seasons and the other three in the summer seasons.
2.5. Statistical Analyses
We used the Statistical Package for the Social Sciences (SPSS 16) software to conduct statistical analyses. Coinciding with our experimental design and goals, the Kruskal-Wallis test was used to conduct our comparisons. The first comparison was performed between the influent and effluent of each bed. The effluents of LCN, LCK, and LCR were compared as a group 1. Likewise, SCN was compared to SCR as a group 2, SFN to SFR as a group 3, and SC to SF as a group 4. SC and SF media were calculated as SC equals the mean of (SCN and SCR) and SF equals the mean of (SFN and SFR). Moreover, the effluents of LCR were compared to SCR, and the effluents of LCN were compared to SCN to figure out the differences between long and short beds efficiency. Significance was recognized when α <0.05. All “significant differences” mentioned later in this study indicate statistical differences.
2.6. Climate Data
Weather data were collected from the Al-Samra meteorological station, which is the closest weather station to our field test site.
3. Results
Influent water was fed continuously to the beds. The HSF-CW overall mean influent flow was 28 m
3·day
−1. After it passed through the beds, the effluent was reduced to 23 m
3·day
−1 with an overall loss of 17%. Mean daily influent and effluent flow rates for each bed are shown in
Table 1. In the summer season, the reed, kenaf, and unplanted beds had water losses of 45%, 30%, and 6%, respectively. During the study period, the mean monthly temperature ranged from 9–29 °C. The long-term mean of the monthly temperature ranged from 11.9–24.2 °C (according to the Al-Samra meteorological station data). The range of the monthly water temperature in the beds was 8–24 °C with a mean of 14 °C. The influent pH ranged from 7.2–8.7 with a mean of 7.9. The mean pH of the effluents ranged from 7.5–7.8.
Table 1.
Mean daily influent and effluent flow rates for each bed.
Table 1.
Mean daily influent and effluent flow rates for each bed.
Type of Bed | Influent (m3/day) | Effluent (m3/day) | Percentage of Losses (%) |
---|
In the Summer Seasons | In the Winter Seasons | In All the Study Period |
---|
Reed bed | 1.62 | 1.15 | 45 | 17 | 29.4 |
Kenaf bed | 1.62 | 1.37 | 30 | 3.2 | 15.5 |
Unplanted bed | 1.62 | 1.56 | 6 | 2.8 | 4.0 |
All the beds | 27.62 | 23.07 | 26.9 | 8.6 | 16.5 |
3.1. HSF-CW Performance
The results of BOD
5, COD, TSS, and P concentration, mass flow, fecal coliform count, and BOD
5/COD are available in
Table A1. Results of removal efficiencies and fecal coliform reduction are shown in
Table 2. The overall removal efficiency of BOD
5 and COD were 55% and 51%, respectively. The different HSF-CW treatments yielded different removal efficiencies. BOD
5 removal efficiency ranged from 37% under SCN conditions to 67% under SFR conditions. In addition, the removal efficiency of COD ranged from 38% for SCN to 64% for SFR. Reed was significantly more efficient than kenaf and unplanted beds in reducing BOD
5, and no significant differences were observed among all three vegetation conditions in reducing COD. There were no significant differences between the long and short beds in removing both parameters. BOD and COD removal rates were not affected by the bed dimensions. A comparison between fine and coarse media indicates that the effluent’s BOD
5 content from the fine media was significantly lower than it was from the coarse media. The fine media was also more efficient than the coarse media in removing COD. However, there were no significant differences between the effluent of the fine and coarse media.
The concentration and mass flow of TSS in the effluent from all of the treatments was significantly lower than that in the influent. The overall removal efficiency of TSS was 67%. TSS removal efficiency ranged from 56% for LCN to 79% in SCR. TSS removal was not affected by the bed dimensions.
The fecal coliform count in the effluent of all HSF-CW treatments was significantly lower than in the influents. The mean log reduction of fecal coliform was 0.8 and ranged from 0.4 for LCN to 1.2 for SFN and SFR. The fecal coliform in the effluent of the fine media was significantly lower than in the coarse media, however there were no significant differences among reed, kenaf, and unplanted beds in removing fecal coliform.
The HSF-CW effluent’s phosphorus concentration and mass flow were significantly reduced compared with that of the influent. The overall phosphorus removal efficiency was 55%. The highest removal efficiency was 75% in SFR; the lowest removal efficiency was 35% in SCN. A comparison between the fine and coarse media indicates that the phosphorus in the effluent from the fine media was significantly lower than from the coarse media. The mass flow of phosphorus in the reed effluent was significantly lower than from the kenaf and unplanted beds. In addition, there were no significant differences between long and short beds in P removal. Phosphorus had a regular behavior during the test period, as revealed by the relatively low standard deviation in
Table A1.
Table 2.
BOD5, COD, TSS, Fecal Coliform, and P removal.
Table 2.
BOD5, COD, TSS, Fecal Coliform, and P removal.
Parameter | Long Beds | Short Beds | Short Beds |
---|
Coarse Media | Coarse Media | Fine Media | Coarse 1 Media | Fine 2 Media |
---|
No Plant | Kenaf | Reed | No Plant | Reed | No Plant | Reed |
---|
Group 1 | Group 2 | Group 3 | Group 4 |
---|
BOD5 | | | | | | | | | |
Mass flow removal * efficiency (%) | 51 | 56 | 66G1 | 37 | 62G2 | 50 | 67G3 | 50 | 59G4 |
Number of samples | 41 | 42 | 41 | 22 | 22 | 22 | 22 | 40 | 40 |
COD | | | | | | | | | |
Mass flow removal * efficiency (%) | 42 | 49 | 58 | 38 | 56 | 47 | 64 | 47 | 55 |
Number of samples | 38 | 38 | 38 | 24 | 24 | 24 | 23 | 48 | 47 |
TSS | | | | | | | | | |
Mass flow removal* efficiency (%) | 56 | 64 | 73 | 67 | 79 | 65 | 64 | 73 | 63 |
Number of samples | 53 | 54 | 53 | 34 | 34 | 34 | 34 | 68 | 68 |
P | | | | | | | | | |
Mass flow removal * efficiency (%) | 38 | 46 | 64G1 | 35 | 61G2 | 58 | 75G3 | 49 | 67G4 |
Number of samples | 53 | 54 | 53 | 34 | 34 | 34 | 34 | 68 | 68 |
FC | | | | | | | | | |
Log reduction * | 0.4 | 0.5 | 0.5 | 0.6 | 0.9 | 1.2 | 1.2 | 0.8 | 1.1G4 |
Number of samples | 47 | 48 | 47 | 26 | 26 | 26 | 26 | 52 | 52 |
3.2. Changes in the EC, Major Anions, and Cations
The EC in the effluent from all of the treatments was significantly higher than in the influent. The overall mean EC increase was 16% which matched with the water loss. EC analysis results are given in
Figure 3. The EC in the effluents of the long beds planted with reed was significantly higher than those planted with kenaf or unplanted beds. Moreover, there were no significant differences between the EC values for the fine or coarse media.
The results of the calcium, magnesium, sodium, potassium, chloride, sulfate, carbonates, and bicarbonates mass flow changes are shown in
Table 3, whereas the results of concentration and mass flow are available in
Table A2. The most pronounced changes were the levels of calcium and sulfate.
Calcium mass flow and concentration in the effluent from all of the treatments was significantly higher than in the influent. The overall mean increase in calcium mass flow was 30%. The calcium concentration in the effluent of the beds planted with reed was significantly higher than that in beds planted with kenaf or unplanted. Moreover, calcium mass flow and concentration in the effluent of SF was significantly higher than in SC.
The sulfate concentration and mass flow in the effluent from all of the treatments were significantly higher than those in the influent, except for LCN, LCK, and SCR showing no significant increases in their effluent’s sulfate mass flow. The concentration of sulfate in the effluent of the long and short beds planted with reed was significantly higher than in beds planted with kenaf or unplanted beds. Moreover, there were no significant differences between the effluent’s sulfate levels in the fine and coarse media or even among reed, kenaf, and unplanted effluents.
Relatively, all other elements analyzed showed some changes after treatment, but the differences were not pronounced or significant as compared to calcium and sulfate. The water loss, plant uptakes and releases, have all contributed to many of the changes, along with other unknown factors.
Figure 3.
The mean electrical conductivity (EC) in influent and effluent of various treatments. (1) EC in all the effluents were significantly (p < 0.05) higher than the influent. (2) G1: significant differences (p < 0.05) among (LCN, LCK and LCR) effluents; G2: significant differences (p < 0.05) between (SCR and SCN) effluents; G3: significant differences (p < 0.05) between (SFN and SFR) effluents. (3) Abbreviation in the X-axis are available in “Treatment System Design” section.
Figure 3.
The mean electrical conductivity (EC) in influent and effluent of various treatments. (1) EC in all the effluents were significantly (p < 0.05) higher than the influent. (2) G1: significant differences (p < 0.05) among (LCN, LCK and LCR) effluents; G2: significant differences (p < 0.05) between (SCR and SCN) effluents; G3: significant differences (p < 0.05) between (SFN and SFR) effluents. (3) Abbreviation in the X-axis are available in “Treatment System Design” section.
Table 3.
Mass flow changes (%) for major anions, cations, micronutrients, and heavy metals.
Table 3.
Mass flow changes (%) for major anions, cations, micronutrients, and heavy metals.
Parameters | Long Beds | Short Beds | Short Beds |
---|
Coarse Media | Coarse Media | Fine Media | Coarse1 Media | Fine2 Media |
---|
No Plant | Kenaf | Reed | No Plant | Reed | No Plant | Reed |
---|
Group 1 | Group 2 | Group 3 | Group 4 |
---|
Ca2+ | +25 * | +23 * | +35 * | +27 * | +9 * | +38 * | +45 * | +20* | +47 *,G4 |
Mg2+ | 0 | −7 | −10 | +6 | −24 *,G2 | +11 | −5 | − 9 | +4 G4 |
Na+ | +6 | −1 | −4 | +9*G2 | −14* | +10 | −3 | −2 | +6 |
K+ | +17 *,G1 | +5 | −18 *,G1 | +15 *,G2 | −17* | +26 *,G3 | −11 | 0 | +8 |
Cl− | +6 | 0 | −1 | +10 *,G2 | −12* | +14* | +4 | 0 | +12 |
SO42− | +46 | +40 | +74 * | +50 * | +35 | +77* | +85 * | +46 * | +87 * |
HCO3− | −5 G1 | −10 * | −23 *,G1 | −5 | −27 *,G2 | −10 | −23 *,G3 | −15* | −15 * |
CO3− | +11 | +3 | −23 G1 | +16 | −20 G2 | +22 | −19G3 | −2 | 0 |
Cd | −19 | −33 * | −48 *,G1 | −24 | −45 * | −9 | −38 * | −35 * | −24 a |
Co | −19 | −42 | −54 *,G1 | −21 | −37 | −13 | −28 | −28 | −19 |
Zn | −40 * | −59 * | −50 * | −30 | −57 * | −59* | −69 * | −44 * | −64 * |
3.3. Changes in Micronutrients and Heavy Metals
The removal efficiency of cadmium, cobalt, and zinc in the influent and effluent are shown in
Table 3, whereas results of the concentration and mass flow are available in
Table A2. There were no significant differences between the cadmium concentration in the effluent and influent of each bed. Cadmium mass flow in the effluent from the reed and kenaf was significantly lower than in the influent. The overall mass flow removal efficiency of cadmium was 31% and ranged from 9% in the SFN to 48% in the LCR. Cobalt mass flow in the effluent from LCR was significantly lower than in the influent, cobalt removal efficiency in that bed reached 54%. The overall removal efficiency of cobalt was 29% and ranged from 13% in SFN to 54% in LCR. Zinc mass flow in the effluent from all of the HSF-CWs was significantly lower than in the influent, except the case of SCN that had insignificant reduction. The overall removal efficiency of zinc was 52% and ranged from 30% in SCN to 69% in SFR. There was no significant difference among all other parameters (
i.e., bed type, media type, and vegetation type [including unplanted beds]) in terms of removing zinc.
There was also no significant difference between the influent and effluent chromium, lead, or arsenic in either concentration or mass flow levels.
4. Discussion
This study has provided long-term field test data with statistical analysis on various HSF-CW in an arid environment. The data show that all HSF-CW beds were moderately efficient in removing BOD
5, COD, TSS, and phosphorus. The mass flow reduction, instead of concentration changes, clearly shows the actual performance of all beds. The highest BOD
5 and COD removal percentages were achieved in the case of reed with fine media. This result corresponds well with the results of Reed [
13] and Rani [
14], who found that fine media provides more surface area for treatment and better development of plant roots. Previous studies revealed that the BOD
5 and COD removal are through aerobic and anoxic degradation of organic compounds by microorganisms on plant roots and the media surface [
5,
14]. Sedimentation and filtration also may contribute to BOD
5 removal [
15]. Our BOD
5 and COD removal efficiencies were generally lower than those found in earlier studies; for example, Göçmez [
4] found BOD
5 and COD removal efficiencies of 71% and 63%, respectively. A worldwide survey conducted by Puigagut [
16] found that the BOD
5 removal efficiencies varied between 75% and 93%. This may be due to the lower levels of degradable organic matter in our influent (low BOD
5/COD ratios are shown in
Appendix Supplementary Table S1), as well as the anoxic condition inside the beds. Most of the easily degradable organic matters were removed before entering HSF-CW since our influent was partially biodegraded municipal wastewater.
The removal of suspended solids is accomplished via sedimentation and filtration in addition to biodegradation, therefore the void space might have some influence [
2,
17]. TSS removal was not affected by the HSF-CW bed dimensions, this result agrees well with the results of Reed [
13]. In previous studies, Rani [
14] compared horizontal- and vertical-subsurface wetlands and found that TSS removal was slightly higher in the horizontal case but the removal efficiency exceeded 96% in both systems. In contrast, the removal performances off TSS ranged from 56.62% and 91.04% in Göçmez’s four year study [
4]. However, the different media types and sizes must have impacted TSS results in various studies.
In our study, the fecal coliform in the effluent of the fine media was significantly lower than in the coarse media, and there were no significant differences among the reed, kenaf, and unplanted beds in removing fecal coliform. Although other studies have shown higher FC reduction, the differences could be due to test variations [
18,
19,
20].
The removal of phosphorus in wetlands is achieved by plant uptake, adsorption, bacterial removal, and precipitation [
14,
21]. The highest phosphorus removal efficiency in our HSF-CWs was in the case of reed with fine media. We expected decreases in phosphorus, both in concentration and in mass flow because reed uptakes it as a nutrient. In addition, the large surface area and high contact between the solid phase and the water column in the fine media enhanced phosphorus removal. Phosphorus might precipitate in the form of calcium phosphate, and this could be caused by calcium from dissolved media in our study. Some other studies have shown the importance of contact and contact time between the water phase and solid media [
2,
13,
14,
21,
22].
Previous studies had higher removal rates of heavy metals than our study [
23]. The change of concentration of heavy metals may be due to, in our case, the media adsorption or release, precipitation, pH decreases, and evapoconcentration. Furthermore, plant uptake plays a major role in planted beds since there were significant differences between the planted and unplanted beds in removing cadmium and cobalt. The volcanic tuff media was efficient at removing the zinc from the effluent. This result was also previously shown [
24].
Evaporative concentration was a major concern in our study, as it has led to an increase of soluble mineral and EC in the effluents. The EC in the reed bed’s effluent was approximately 20% higher than the other bed’s effluent, which is reasonable since the mean water losses in the reed beds was 29%. In a previous study, EC has changed from −0.60% to −13% after passing through CW with various hydraulic retention times [
25], which was attributed to evapotranspiration and/or plant uptake. This attribution could explain our results as well.
Both calcium and sulfate ions increased more significantly than all other ions. In this study, the main source of water-soluble metals is the media material. Our volcanic tuff media, also known as zeolitic tuff and phillipsite is the primary mineral in the tuff [
26]. The phillipsite crystal has variable amounts of silicon, aluminum, sodium, potassium, and high calcium oxide level up to 11.51% [
26]. Plant uptake led to significant decreases in the potassium mass flow in the effluent of the long and short beds planted with reed, as compared with the beds planted with kenaf or the unplanted beds. Potassium increases in the unplanted beds might have come from the media itself [
26]. The higher sulfate level in the HSF-CW’s effluent was probably caused by the dissolution of gypsum [
27] and we expected that the gypsum was mixed with our media from its source. To overcome this problem, we suggest to choose a more suitable media, and to avoid using reed plants in HSF-CW beds if used in arid areas, for reed beds show a mean water loss of almost 29%, a figure that reaches 45% in the summer season. We also suggest not to use kenaf plants in HSF-CWs as it is an annual crop with about 16% water loss and its water effluent quality is not any better than that of unplanted beds. The unplanted beds with either fine or coarse media have performed better than the other combinations. Since that combination was vegetation free, it can be categorized as a granular filtration system.
Regarding bed dimensions, our study showed that the bed length did not significantly affect the general performance given the same bed volume and the dimensional ratios we used.