Evaluation of clogging in full-scale subsurface flow constructed wetlands

Ecological Engineering 95 (2016) 505–513

Contents lists available at ScienceDirect

Ecological Engineering journal homepage: www.elsevier.com/locate/ecoleng

Evaluation of clogging in full-scale subsurface flow constructed wetlands Rosa Aiello a , Vincenzo Bagarello b , Salvatore Barbagallo a , Massimo Iovino b , Alessia Marzo a , Attilio Toscano c,∗ a

Department of Agriculture, Food and Environment, University of Catania, Via S. Sofia 100, 95123 Catania, Italy Department of Agricultural and Forest Sciences, University of Palermo, Viale delle Scienze ed. 4, 90128 Palermo, Italy c Department of Agricultural and Food Sciences, University of Bologna, Viale G. Fanin 50, 40127 Bologna, Italy b

a r t i c l e

i n f o

Article history: Received 16 February 2016 Received in revised form 24 June 2016 Accepted 26 June 2016 Keywords: Constructed wetlands Clogging Hydraulic conductivity Tracer test

a b s t r a c t Treatment processes that occur in constructed wetlands can result in gradual clogging of the porous medium. Clogging may result in hydraulic malfunction and/or reduced treatment performance. The aim of this study was to analyze the hydraulic aspects of horizontal subsurface flow (H-SSF) constructed wetlands (CWs), and, in particular, to investigate the clogging phenomena through in situ measurements of hydraulic conductivity of the gravel bed, quantification of accumulated clog matter and flow paths visualization by means of tracer tests. Removal efficiencies of chemical and physical contaminants were also assessed. Experiments were carried out in three full-scale H-SSF CWs in Sicily (Italy) that are used for tertiary treatment of the effluent of a conventional wastewater treatment plant. One bed had been operating for eight years while the other two are two years old. The oldest CW had lower hydraulic conductivity of the porous media and higher concentrations of total solids, volatile solids and belowground plant biomass than the younger ones. Furthermore, several stagnant zones and preferential flow paths were only detected in the oldest CW. Despite these results should be indicative of some degree of medium clogging, the treatment capacity remained largely unchanged after eight years of operation. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Constructed wetlands (CWs) are used worldwide as a suitable or complementary technology to conventional wastewater treatment plants (WWTPs) for removing contaminants from various types of wastewaters (municipal, agricultural and industrial wastewaters, landfill leachate and stormwater runoff) due to their efficient performance, mechanical simplicity and low energy requirement in comparison to intensive engineered treatment plants (Toscano et al., 2013; Vymazal, 2014; Yalcuk and Ugurlu, 2009). Several types of constructed wetlands have been developed based on the hydraulic characteristics (water level, flow direction) and the vegetation types (emergent, submerged, freefloating macrophytes). Among them, subsurface flow CWs are the most commonly applied, especially in Europe (Fonder and Headley,

∗ Corresponding author. E-mail addresses: [email protected] (R. Aiello), [email protected] (V. Bagarello), [email protected] (S. Barbagallo), [email protected] (M. Iovino), [email protected] (A. Marzo), [email protected] (A. Toscano). http://dx.doi.org/10.1016/j.ecoleng.2016.06.113 0925-8574/© 2016 Elsevier B.V. All rights reserved.

2013). These CW systems consist of waterproof basins, filled with porous media (generally gravel), planted with emergent macrophytes (such as Phragmites australis), in which wastewater flows horizontally (H-SSF) or vertically (V-SSF). As wastewater moves through a constructed wetland, the removal of pollutants occurs due to the interaction of several biological, physical and chemical processes such as decomposition, filtration, accumulation, nitrification/denitrification, adsorption and precipitation, all of them strongly relying on the contact time between wastewater and porous media, biofilm and plant roots (Kadlec and Wallace, 2009; Knowles et al., 2011; Nivala et al., 2012). The interaction among treatment processes, wastewater characteristics and hydraulic loading rates may result, at long times, in a gradual clogging of the substrate that can negatively affect the removal and hydraulic performances of the CW and, definitively, reduce the lifetime of the system (Cooper et al., 2005; CasellesOsorio and Garcıa, 2006; Nivala and Rousseau, 2009). The presence of stagnant zones can be assessed by comparing the actual residence time (aRT) with nominal residence time (nRT), if aRT is much larger than nRT wastewaters stagnate in the reactor and do not participate in reactions (Kadlec and Wallace, 2009).

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

2. Materials and methods 2.1. Description of constructed wetlands The research activity was carried out at San Michele di Ganzaria, 90 km South-West of Catania (Sicily), where the biggest natural treatment plant of South Italy was realized for tertiary treatment of municipal wastewater aimed at agricultural reuse. Urban wastewater is treated in a municipal wastewater treatment plant (WWTP) consisting of a pre-treatment step followed by two parallel water lines (Imhoff tank, trickling filter and a secondary sedimentation tank). The effluent from the WWTP is collected to the natural treatment plant that includes four H-SSF CWs operating in parallel, followed by three batch wastewater storage reservoirs. All CWs were excavated into the ground (for a depth of 1 m) leaned with impermeable membrane and filled with gravel (for a depth of 0.6 m).

H-SSF4

33

23

3 5

H-SSF2/H-SSF3

45

The growth of the biological film, accumulation of sludge and belowground plant biomass and deposition of chemical precipitates decrease the amount and size of the void spaces over time thus reducing the hydraulic conductivity of the porous medium (Blazejewski and Murat-Blazejewska, 1997; Pedescoll et al., 2009). Therefore, hydraulic conductivity measurements have proven to be one of the most suitable technique to indirectly assess the degree of clogging (Pedescoll et al., 2011; Knowles et al., 2010). For a porous medium, the clogging process, as detected by saturated hydraulic conductivity reductions, was found to be driven by the cumulative loading of total suspended solids (Viviani and Iovino, 2004). Therefore, clogging phenomena observed at long time for low concentrated wastewater could be representative of short term behaviour of CWs treating high concentration wastewater. Bulk hydraulic conductivity can be estimated from the Darcy law by measuring the water table height at different points in the bed. However, when the hydraulic gradients are small as in the H-SSF CW, accuracy of this approach is questionable (Sanford et al., 1995). Laboratory measurements of hydraulic conductivity are hampered by the non-cohesive nature of gravel that makes it difficult to remove a representative sample whereas in situ assessment of hydraulic conductivity requires instruments and methods specifically designed for highly permeable porous materials (Knowles and Davies, 2009). Pedescoll et al. (2009) successfully applied a simple method based on the falling head permeability test (NAVFAC, 1986) to assess the degree of clogging in two full-scale CWs in Spain. Although many studies have been done, no specific protocol exists to assess the degree of clogging. Therefore, comparison of results obtained by different indirect techniques is probably the unique viable approach to understand to what extent the clogging phenomena influence the overall performance of a constructed wetland. This study aimed at evaluating the impact of clogging on the treatment performance and hydraulic properties of three fullscale H-SSF CWs treating wastewater from secondary WWTP. The extent of clogging phenomena was investigated by comparing two CWs with identical design and influent characteristics but different operational life. Since there is no single method that can quantitatively measure clogging, three approaches were used: (i) characterization of the clog material, in terms of total solids, volatile solids and belowground plant biomass, (ii) in situ measurements of hydraulic conductivity of the porous media, and (iii) visualization of water flow dynamic by means of tracer tests. Finally, the effect of clogging on the CW treatment performance was evaluated comparing the removal efficiencies of some chemical and physical contaminants.

5

506

Outlet

Outlet

Hydraulic conductivity TS, VS and BGB Tracer test Piezometer

Fig. 1. Schematic setup of the CW beds with indication of the sampling sites for clog matter quantification, saturated hydraulic conductivity measurements and tracer tests. All dimensions are in meters.

The present study was conducted during spring-summer 2014 in three CWs, namely H-SSF2, H-SSF3 and H-SSF4. H-SSF2 has been operating for eight years (since 2006). It has a surface area of about 2000 m2 , is filled with 10–15 mm volcanic gravel, treats a wastewater discharge of about 2 L s−1 and is planted with P. australis at a density of four rhizomes per m2 . Both H-SSF3 and H-SFF4 have been operating since summer 2012. H-SSF3 has the same design characteristics as H-SSF2 (area, porous medium, flow rate, vegetation type and density). Also the current stem density is similar in both beds, amounting to about 350 stems per m2 . H-SSF4 has a surface area of about 1000 m2 and it is planted with Typha, at a density of four rhizomes per m2 , with a current stem density of about 170 stems per m2 . In all CWs, the influent is distributed at the bed-head through a perforated 200 mm PVC pipe located above the substrate and normal to the flow direction to allow homogenous wastewater distribution into the bed. Wastewater is intercepted downstream by a cross perforated pipe located in the final section at the bottom of the bed and connected to an adjustable outlet to control water level. Electromagnetic flow meters (ISOIL mod. MS2500), installed at the inlet and at the outlet pipes, measure the flow rate (L s−1 ) in continuous. The electromagnetic flow meters have a totalizer that allows to know the volume of wastewater flowed. In each bed, nine piezometers, arranged in a 3 by 3 array, consisting of 20 cm diameter, open-ended perforated plastic tubes inserted into the gravel to the bottom of the bed, are used to measure the water table heights and to collect wastewater samples (Fig. 1). 2.2. Accumulated material (clog matter) Laboratory analyses were carried out on bulk samples of gravel media to quantify the clog matter in terms of concentrations of accumulated total solids (TS), volatile solids (VS) and belowground plant biomass (BGB). For each CW, the granular medium, mixed with plant roots and decomposed organic matter, was sampled in four points along three alignments (Fig. 1). For each alignment, one sampling point was selected close to the inlet and the others in proximity to each piezometer. A total of 12 gravel samples were collected in each CW. At each sampling point, a 20 cm in diameter by 40 cm long sharpened steel tube was used to collect a wet gravel material sample

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

(approximately 6 L in volume). The tube was placed on the bed surface and the granular medium of the unsaturated zone inside the tube was extracted by hand while the tube was progressively inserted into the bed to avoid collapse of lateral wall inside the hole. A depth of 40 cm was explored as most of plant root apparatus concentrated in the upper layer of the bed. All samples were maintained in refrigerated plastic bags until they were processed in the laboratory. The vegetable biomass was separated by the granular medium, oven-dried at 105 ◦ C and weighed. The BGB concentration, expressed as amount of vegetable biomass per surface unit was then calculated. To determine TS and VS concentrations in the granular medium, a volume of 0.5 L of distilled water was mixed with a same volume of granular medium. The samples were shaken by hand (CasellesOsorio et al., 2007; Pedescoll et al., 2009) and the water, containing accumulated solids, was filtered through a 1 mm metal mesh of a sieving machine (AS 200 control, Retsch). Laboratory analyses were carried out according to the conventional methods described in APHA-AWWAWPCF (2001). Statistical significance between two mean values of TS, VS and BGB was tested by a two-tailed t test (P = 0.05) assuming a normal distribution for these variables.

2.3. Saturated hydraulic conductivity Measurements of saturated hydraulic conductivity were conducted at the same alignments used for clog matter determinations. In particular, nine measurement points were selected for each alignment at intervals of approximately 5 m in H-SSF2 and H-SSF3 and 3 m in H-SSF4 (Fig. 1). Given the inlet and outlet ends of beds were filled with coarser gravel to provide an even distribution of wastewater, these zones were excluded from measurements. A total of 27 saturated hydraulic conductivity tests were conducted for each constructed wetland. The falling-head test method (NAVFAC, 1986; Caselles-Osorio and Garcıa, 2006; Pedescoll et al., 2009) was used to estimate the saturated hydraulic conductivity, Ks (L T−1 ) of the CW bed. For each measurement point, a small hole was dug in the granular medium using a spade until the water level was reached. The permeameter cell consisted of a steel tube with an internal diameter of 10 cm and a length about of 150 cm, perforated on one side, that was inserted into the medium using a mallet to a depth of 20 cm. A pressure probe (STS – Sensor Technik Sirnach, AG), connected to a laptop by means of a Campbell Scientific data logger (CR200-R), was placed inside the tube on the granular medium surface. The tube was filled with water in a pulse mode using a bucket and the decrease of water height inside the tube was monitored until the water level reached the zero reference plane corresponding to the bed water table. The pressure data collected by the probe were then converted into water heights. For each test, the experimental monitoring time was fixed in 60 s and the pressure probe was configured to collect three water height data per second. Estimation of saturated hydraulic conductivity, Ks [L T−1 ] was obtained from the transient falling head data according to Lefranc’s formula:



Ks =

d2 ln 2L/d 8Lt

 ln

h0 h(t)

(1)

where d [L] is the diameter of the permeameter cell, L [L] is the submerged length of the tube (perforated zone), h0 [L] is the height of water inside the tube at time zero, h(t) [L] is the height of water inside the tube at a given time t [T] during the falling head process.

507

An iterative, nonlinear, fitting techique was implemented via Excel solver (Frontline Systems, Incline Village, NV) to minimize the difference between measured and estimate h(t) data: t=n 

2

(hm (t) − he (t)) → 0

(2)

t=0

where hm (t) is the measured height of the water table level inside the tube at time t and he (t) is the corresponding calculated value by Eq. (1). Probability distributions for Ks were preliminary investigated. In particular, the assumption of data distributed according to the normal (N) and the two parameters lognormal (LN2) distributions was tested by the Probability Plot Correlation Coefficient (PPCC) normality test at a probability level P = 0.10 (Helsel and Hirsch, 1992). Statistics of Ks data were then calculated according to the corresponding theoretical distribution and difference between means was tested by a two-tailed t-test (P = 0.05). 2.4. Tracer test Estimation of the actual residence time (aRT) and other hydraulic parameters was conducted by tracer tests using NaCl as a conservative tracer (Barbagallo et al., 2011; Wang et al., 2014; Suliman et al., 2006; Ronkanen and Kløve, 2007). The saline solution, consisting of 175 kg of NaCl dissolved in 450 L of water, was injected, with a pump, at the inlet of the distribution system of each bed. The injection time, which lasted about 10 min, may be considered practically instantaneous compared to the residence time of the beds. The tracer concentration was determined from wastewater electric conductivity (EC) measured at 10 sampling points (Fig. 1) for each bed (i.e. the nine piezometers and the outlet device). Portable conductivity meters with a data logger (Delta OHM – HD 2106.2) were used to measure and automatically record the EC data at interval of 15 min thus allowing a practically continuous record of tracer concentrations. The EC values measured in mS/cm were then converted to chloride concentrations (mg/L) using a linear calibration curve between NaCl concentration and EC (R2 = 0.99). Before fitting the curve, the initial electrical conductivity was subtracted at each measuring point. The actual residence time, the average time that the water remains in the horizontal SSF, was calculated using the retention time distribution (RTD) curve obtained at CW outlet, after background correction, according to the methodology reported in Levenspiel (1972). No rainfall occurred during execution of the tracer tests at the three CWs. The amount of water loss by evapotranspiration (ET) was evaluated by a water balance, as difference between the measured influent and effluent water volumes. The ET losses were low, ranging from about 3% to 5% of the inflow rate. Similar results were reported by Milani and Toscano (2013) for P. australis in a pilot CW plant located in the same study area. Thus, the enrichment of tracer concentration due to ET losses during the tracer tests was considered negligible. Since aRT is determined directly from the shape of the tracer response curve, it doesn’t take into account any loss of effective detention volume (García et al., 2004; Persson et al., 1999). For this reason, the hydraulic efficiency (␭), that evaluates both the effective volume utilization and the shape of the tracer response curve, was determined. It is calculated as the ratio of the time taken for the tracer to reach the peak of the RTD curve at the CW outlet to the nominal hydraulic retention time, this latter calculated on a plug flow assumption, as reported in IWA (2000), and assuming unchanged the nominal porosity (␧). The higher the ␭, the better the available CW volume is used, and thus the higher is the portion of wastewater efficiently involved in CW retention. Finally, the percentage of tracer mass recovery was calculated in order to provide

508

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

Table 1 Operational characteristics of the three CWs and experimental settings for the tracer experiments. Constructed wetlands

H-SSF2 H-SSF3 H-SSF4

Bed characteristics

Tracer test design parameters

Operational time (year)

Area (m2)

␧

nRT (h)

Vegetation

Qmean (L/s)

Cin (g/L)

Min (kg)

duration (h)

8 2 2

2000 2000 1000

0.47 0.47 0.47

77 77 54

Phragmites sp. Phragmites sp. Typha latifolia

2 1.9 1.2

333 333 333

175 175 150

128 109 115

␧: nominal porosity; Qmean (L/s): mean flow rate = (Qin + Qout )/2; nRT: nominal retention time (h); Cin: tracer concentration at CW inlet (g/L); Min : amount of the salt injected (kg).

3. Results and discussion 3.1. Solid and belowground plant biomass concentrations in the granular medium

1,0 0,9 0,8 Cumulative frequency

a quality check of the reliability of the tracer tests. In particular, the amount of tracer recovery was calculated multiplying the outflowing wastewater volume (measured by the flow meter) by the tracer concentration (measured by portable conductivity meters). According to Headley and Kadlec (2007), a tracer test is considered acceptable if at least 80% of the mass of tracer added at the inlet is recovered at the outlet. In Table 1, the operational characteristics of the three CWs and the experimental setting for the tracer studies are reported.

0,7 0,6 0,5 0,4 0,3

H-SSF2

0,2

H-SSF3

0,1 Table 2 reports statistics of the total solids (TS) concentration, volatile solids (VS) concentration and belowground plant biomass (BGB) concentration accumulated in the gravel media of the three CWs. Analysis of data is mainly focused on comparison of gravel properties measured in H-SSF2 and H-SSF3 given that these beds have the same design characteristics but different operating lives. Comparison between H-SFF4 and H-SFF3, that were put in operation at the same data, is also reported in order to detect possible influence of bed dimensions and plant species on the clogging phenomena. In the newly operating H-SSF3 constructed wetland, the TS concentration varied between 1070 and 6650 mg L−1 with a mean value of 2805 mg L−1 (CV = 59.6%). The VS concentration varied from 173 to 1157 mg L−1 with a mean value of 355 mg L−1 (CV = 81.1%) and, on average, the volatile fraction accounted for 13% of the total solids concentration. The ratio between the maximum and the minimum values was 6.2 and 6.7, respectively for TS and VS, denoting high spatial variability of solid accumulation in the gravel medium even after two years of operation. Pedescoll et al. (2009) also reported ratios up to 4.5 and 9.5, respectively, for TS and VS concentrations in two CWs in Spain. As expected, the average concentration of TS in the old CW (H-SSF2) was higher than in the new H-SSF3 (mean value of TS = 4371 mg/L) even if the difference, corresponding to a factor of 1.6, was not statistically significant (Table 2). Spatial variability of TS after eight years of operation was comparable (CV = 63%, ratio maximum/minimum = 5.9) to that of the two years old H-SSF3 (CV = 60%, ratio maximum/minimum = 6.2). Mean concentration of VS was higher in H-SSF2 than in H-SSF3 by a significant factor of 4.3 but the increase in VS concentrations was not associated to a change in the coefficients of variation (Table 2). Considering that the sample arrangement was identical in the two beds (Fig. 1), it was concluded that the overall spatial variability of VS was similar in the two CWs. A close examination of the empirical frequency distribution of VS values measured in the two CWs, showed that the minimum value of VS increased by a factor of 2.5 with operating life whereas the maximum VS value increased by a factor of 4.0 and, also, five out of twelve VS values in the old CW were higher than the maximum VS determined in the new CW (Fig. 2). Compared to

0,0 0

100 0

200 0 30 00 VS (mg/L)

40 00

5000

Fig. 2. Empirical frequency distributions of volatile solids.

H-SSF3, the relative fraction of volatile solids in H-SSF2 increased up to 33% of the total solids. The observed increase in VS during CW operation is an expected consequence of organic matter accumulation in the granular medium due to wastewater supply (Pedescoll et al., 2009). However, plant roots played a major role in determining the organic matted accumulation in H-SSF2 as can be deduced by the observed increase in belowground biomass concentration. Even if not statistically significant, mean BGB concentration after eight years of operation was more than twice the one detected in the relatively new bed (Table 2). Independently of the operating life, spatial variability of BGB concentration was high, denoting a great heterogeneity in the root distribution across the granular medium. However, in close similarity to VS, the ratio maximum to minimum BGB increased from H-SSF3 (ratio = 152) to H-SSF2 (ratio = 185). Moreover, the linear regression between BGB and VS in the old H-SSF2 resulted in a statistically significant regression coefficient (R2 = 0.504) whereas no regression was found for H-SSF3. It was concluded that the observed increase in VS concentration observed during CW operating was due to the increase of BGB concentration. No statistical significant difference was found between the solid concentrations in the two new operating CWs (H-SSF3 and HSSF4) (Table 2), but BGB concentrations differed by a factor of 18 as consequence of the different planted species (i.e., P. australis in H-SSF3 and Typha in H-SSF4). In particular, VS concentrations in the two beds were practically identical (355.3 and 345.6 mg/L, respectively) thus confirming that in newly operating CWs the average value of VS concentration is mostly influenced by wastewater characteristics. Conversely, decomposed root material that progressively accumulates in the granular medium is the main mechanism of long-term increase of organic matter in CWs planted with P. australis. Data collected in the H-SSF2 and H-SSF3 at S.

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

509

Table 2 Statistics of total solid concentration (TS), volatile solid concentration (VS), belowground biomass concentration (BGB) and saturated hydraulic conductivity (HC) measured in the three constructed wetlands. H-SSF2

H-SSF3

H-SSF4

TS (mg/L)

VS (mg/L)

BGB (g/m2 )

Ks (m/d)

TS (mg/L)

VS (mg/L)

BGB (g/m2 )

Ks (m/d)

TS (mg/L)

VS (mg/L)

BGB (g/m2 )

Ks (m/d)

Min Max Mean

1683 9927 4371 a

433 4593 1514 (a)

9 1665 551 a

38 366 212 (a)

1070 6650 2805 ab

173 1157 355 (a)b

7 1065 229 a(b)

251 549 390 (a)b

633 3880 1956 b

83 1357 346 b

2 51 12 (b)

76 613 407 b

SD CV

2759 63

1238 82

605 109

103 48

1672 60

288 81

351 153

103 27

947 48

329 95

13 107

135 33

Mean value followed by the same letter enclosed in parenthesis are different according to a two-tailed t-test (P = 0.05). SD = standard deviation, CV = coefficient of variation (%).

Fig. 3. Application of the PPCC normality test to untransformed saturated hydraulic conductivity data.

Michele Ganzaria suggest that significant increase in VS requires a relatively long (i.e., eight years) accumulation period.

3.2. Saturated hydraulic conductivity According to the PPCC normality test, the normal distribution hypothesis was never rejected for Ks data (Fig. 3), whereas log-normal distribution was rejected in two out of three cases. Therefore, a normal distribution for Ks was assumed and all statistical comparisons were conducted accordingly. Log-normal distribution is usual for field soil saturated hydraulic conductivity data. However, normal distribution could describe more adequately Ks data for relatively uniform porous media, like repacked soil samples or granular medium. Saturated hydraulic conductivity in CW HSSF3 varied between 215 and 549 m/d with a mean value of 390 m/d (CV = 26.5%) (Table 2). In CW H-SSF2, Ks was lower and more variable than in H-SSF3 (mean Ks = 212 m/d; CV = 48.3%). Therefore, a significant reduction of the saturated hydraulic conductivity by a factor of approximately two was observed as a consequence of clogging of the granular medium that occurred in the old CW (Table 2).

Due to the very high spatial variability of solids concentration a clear trend was not detected between TS (or VS) and Ks . However, distribution of Ks in the direction of flow highlighted a different behavior between H-SSF2 and H-SSF3. Indeed, the average Ks values calculated for a transect (N = 3) at a given distance from the inlet of the bed were always higher for the H-SSF2 as compared to H-SSF3 and tended to increase from the inlet to the outlet (Fig. 4). This is an expected result given that, in the early operating stages, occlusion phenomena are more probably localized close to the inlet zone thus explaining the relatively similar results for the two beds whereas relatively larger difference in Ks were observed in the outlet zone as consequence of clogging phenomena that exclusively affected the old operating bed (Fig. 4). Neither of the two trends was statistically significant due to the high variability that was observed at each measurement transect. However, when the Ks data were aggregated according to two bulk zones that included an inlet zone (i.e. from the inlet to the middle of the CW) and an outlet zone (i.e., from the middle to the outlet), a clear influence of the operating life on the localization of the clogging phenomena was observed. In particular, in the new operating H-SSF3 the mean inlet hydraulic conductivity (Ks = 360.3 m/d, N = 12) was not different from the outlet one (Ks = 401.6 m/d, N = 15).

510

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513 Table 3 Mean influent and effluent pollutant concentrations and pollutant removal efficiencies (R) during the period Jan 2013–Oct 2014. Standard deviation values are also reported. TSS Influent H-SSF2 H-SSF3 H-SSF4

Fig. 4. Distribution of mean Ks values measured along transects at increasing distance from the inlet. Error bars are also plotted.

In the old H-SSF, the mean hydraulic conductivity significantly decreased from the inlet zone (Ks = 305.1 m/d, N = 12) to the outlet zone (Ks = 215.0 m/d, N = 15). Therefore, the second half of the older CW was characterized by a saturated hydraulic conductivity 1.4 times lower than that of the inlet zone. Assuming the average Ks value of the outlet zone of H-SSF3 as a reference unclogged value, the reduction of saturated hydraulic conductivity after eight years of bed operating ranged up to a factor close to two. It worth saying that the falling head technique explores a relatively superficial layer of the granular bed and that further analyses are probably necessary to confirm if the clogging phenomena also occurred downward in the filtering bed. Notwithstanding this, a rough confirmation of the clogging phenomena that occurred in HSSF2 can be obtained by the hydraulic performance of the two CWs as retrieved from tracer tests (see below). 3.3. Tracer test Fig. 5 shows the three breakthrough curves at the CW outlet. To make comparison easier, the concentration curves vs. normalized time (aRT/nRT) obtained in the three tests are also presented. From the tracer response curves the percentage of mass tracer recovery was obtained, as well as some parameters that define the hydraulic behavior of the beds: RTD function, actual RT and hydraulic efficiency. The percentages of tracer recovery, 82% for H-SSF2, 71% for HSSF3 and 88% for H-SSF4, indicate that the tracer experiments are acceptable. The lower mass recovery in H-SSF3 was probably due to the relatively shorter duration of the tracer tests (Table 1). The concentration curves versus normalized time (aRT/nRT) obtained in the tests performed in the new operating beds (HSSF3 and H-SSF4) showed nearly identical shapes (Fig. 5d). Both curves displayed a sharp rise to a peak, followed by an exponential decrease with a tail. Different was the trend of the RTD curve for the old bed (H-SSF2). As for the above discussed quantities, it is interesting the comparison between H-SSF2 and H-SSF3 having the same design characteristics (Table 1) and similar flow rate (Table 2) but different operating life. In the H-SSF2, the tracer peak concentrations was recorded much sooner (after 36 h of the tracer injection) than in the H-SSF3 (after 76 h) (Fig. 5a,b), and the actual retention time was shorter than the nominal one (Fig. 5a). This could be due to the establishment of preferential flow paths resulting in short-circuiting (Alcocer et al., 2012; Giraldi et al., 2009) and in a poor utilization of

(mg/L) out (mg/L) R (%) out (mg/L) R (%) out (mg/L) R (%)

BOD5 52 15 74 10 79 12 74

± ± ± ± ± ± ±

COD 30 13 12 8 10 11 13

28 10 64 11 58 11 54

± ± ± ± ± ± ±

TN 13 7 15 5 19 6 23

51 17 67 20 58 19 57

± ± ± ± ± ± ±

24 12 19 8 19 8 20

20 10 51 11 42 11 44

± ± ± ± ± ± ±

5 6 26 3 17 4 23

available volume due to clogging phenomena. This is confirmed by the results of hydraulic efficiency parameter. According to Persson et al. (1999), the hydraulic behaviour of CWs can be categorized into the following three groups: (1) good hydraulic efficiency when ␭ > 0.75, 2) satisfactory efficiency when 0.5 < ␭ ≤ 0.75 and 3) poor efficiency when ␭ ≤ 0.5. The hydraulic efficiency of the H-SSF2 was 0.4, which is classified as poor. The RTDs measured in the piezometers located inside the HSSF2 CW supported the presence of preferential flow paths that excluded large zones of the gravel bed from the reaction (Fig. 6a). As can be observed, the tracer regularly flowed through the bed almost exclusively along the direction of piezometers 3, 6 and 9 (where very high EC values have been detected) while none or minimum tracer was detected along the other direction (i.e., piezometers 1, 4, 7 and piezometers 2, 5, 8) (Fig. 6a), where also stagnant zones are evident (e.g., piezometer 2). Sufficiently different was the hydraulic behavior of H-SSF3 in which the tracer clearly moved more uniformly along the three directions (Fig. 6b) thus confirming that almost the total available volume of the bed was involved in the reaction. Furthermore, the actual residence time similar to the nRT (Fig. 5b) as well as the good hydraulic efficiency (␭ = 0.8), suggested that limited clogging problems affected this bed. However, also in H-SSF3 a preferential flow along two directions seems to occur, with 1/3 of the plant that works with a smaller flow. 3.4. Removal efficiency Table 3 reports the average influent and effluent pollutant concentrations and the mean removal efficiencies for the three CWs. Water quality was assessed about two times per month during the period from January 2013 to October 2014 that includes the field experimental campaign. At each time, samples were collected at the inlet of CW system (i.e., following WWTP treatments) and at the outlet of CW units and analysed according to standard methods (APHA et al., 2005). Despite the differences in the hydraulic behaviour, all the constructed wetlands have proved to be efficient in removing the main chemical and physical pollutants from the secondary effluent of urban wastewater treatment plants and no great difference in the removal performances was observed. Total Suspended Solids (TSS) decreased by 74% from a mean of 52 mg/L to a mean concentration of 15 mg/L after treatment through the H-SSF2, and to a mean concentration 10 mg/L after treatment through both H-SSF3 and H-SSF4. The average removal efficiencies achieved for BOD5 in the three CWs (H-SSF2, H-SSF3 and H-SSF4) were 64, 58 and 54%, respectively. COD was reduced of 67% in H-SSF2, 58% in H-SSF3 and 57% in H-SSF4, with mean discharge concentrations of 17, 20 and 19 mg/L, respectively. Probably H-SSF2 removes more organic matter due to a greater development of the roots, that implies a longer surface for biofilm growth.

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

511

Fig. 5. Tracer concentration as function of time at the outlet of a) H-SSF2 bed; b) H-SSF3 bed; c) H-SSF4 bed and d) comparison among the three beds (normalized time = aRT/nRT). Values of actual retention time (aRT) and nominal retention time (nRT) are also showed. TRM = tracer mass recovery.

The mean total ammonia removal (51% for H-SSF2, 42% for H-SSF3 and 44% for H-SSF4) was lower than those for organic matter and TSS (Table 3), as usually in H-SSF systems (Stefanakis Alexandros and Tsihrintzis Vassilios, 2009; Toscano et al., 2015), but quite similar for all the beds. The results of this study confirm the high reliability of CWs for tertiary wastewater treatment given that the H-SSF2 treatment capacity remained largely unchanged after eight years of operation.

4. Conclusions Measurements of accumulated clog matter, saturated hydraulic conductivity and breakthrough curve were performed in the gravel bed of three CWs with the same design characteristics but different operational lives (H-SSF2 and HSSF3) and with the same operational life but different vegetation types (H-SSF3 and H-SSF4). After eight years of operation, the mean TS concentration increased by a non-significant factor of 1.6 whereas the VS concentration significantly increased 4.6 times, thus resulting in a marked accumulation of organic matter fraction in the granular medium of the old HSSF2. Clogging phenomena occurring in H-SSF2 were confirmed by reduction of Ks up to a significantly factor of two. The two CWs with the same operational live but different plant species showed almost identical mean VS concentrations and Ks values. Therefore, in newly operating CWs, wastewater characteristics control organic matter accumulation. The belowground plant biomass (BGB) concentration increased more than twice during the six-year spell of operational life in CWs planted with P. australis. Despite other factors, like growth of microbial biomass or accumulation of non-degradable waste material in the biofilm, may affect the long term clogging phenomenon, our results highlight the role that accumulation of decomposed root material may have on longterm clogging of gravel bed material. Given the BGB concentration resulted 18 times higher for the P. australis than for Typha, it can be

supposed that CW planted with the latter species is less susceptible to clogging due to decomposed root accumulation. Saturated hydraulic conductivity showed high spatial variability thus preventing precise localization of the clogging phenomena. However, when data were aggregated at a larger spatial scale, a clear influence of the operating life was observed given that the second half of the old CW was characterized by a 1.4 times lower Ks value than the inlet zone. No statistical difference was observed for the new CW, thus concluding that occlusion phenomena tend to affect the outlet part of the gravel bed. Non-uniform localization of clogging phenomena was confirmed by tracer tests. Breakthrough curve for H-SSF2 showed that the tracer peak concentration was recorded much sooner than in the H-SSF3 and the actual residence time was shorter than the nominal one thus denoting a lower hydraulic efficiency. Furthermore, large zones of the gravel bed were excluded from reactions as consequence of establishment of preferential paths and short-circuits. Despite the occurrence of clogging phenomena and the decline in hydraulic performance, efficiency of the H-SSF2 in removing chemical and physical pollutants was comparable and even better than that of the newly operating beds. The results of this study confirmed that clogging in constructed wetlands involves different processes and, probably, none of the single applied techniques can be recommended for a complete assessment of the phenomena, while the combined methodology here proposed could be considered a viable approach for full-scale investigations of clogging phenomena in constructed wetlands. It was recognized that HSSF CWs planted with P. australis are susceptible to clogging phenomena but, due to high hydraulic flexibility, they maintain largely unchanged performances for a relatively long period. Therefore, HSSF CWs confirm to be reliable and economic solutions for sustainable tertiary treatment of urban wastewaters. It is worth noting that results were obtained for CWs operating as tertiary treatment, with low organic and solid loading rates.

512

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513

Fig. 6. Breakthrough curves measured in piezometers of H-SSF2 (a) and H-SSF3 (b).

Given the wastewater contaminant load has a significant effect on the clogging development, further studies are needed to assess the clogging phenomena also in full-scale CWs operating with higher loading rates.

Acknowledgements This research activity was carried within the WATER4CROPS European Union FP7 Project – Integrating biotreated wastewater reuse and valorization with enhanced water use efficiency to support the Green Economy in European Union and India, financed by EU (FP7 grant No. 311933).

References APHA, AWWA, AEF, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. Baltimore: American Public Health Association (APHA), AmericanWaterWorks Association (AWWA), and American Environment Federation (AEF), Washington, DC, USA. Alcocer, D., Vallejos, G., Champagne, P., 2012. Assessment of the plug flow and dead volume ratios in a sub-surface horizontal-flow packed-bed reactor as a representative model of a sub-surface horizontal constructed wetland. Ecol. Eng. 40, 18–26. Barbagallo, S., Cirelli, G.L., Marzo, A., Milani, M., Toscano, A., 2011. Hydraulic behavior and removal efficiencies of two H-SSF constructed wetlands for wastewater reuse with different operational life. Water Sci. Technol. 64 (5), 1032–1039.

R. Aiello et al. / Ecological Engineering 95 (2016) 505–513 Blazejewski, R., Murat-Blazejewska, S., 1997. Soil clogging phenomena in constructed wetlands with subsurface flow. Water Sci. Technol. 35 (5), 183–188. Caselles-Osorio, A., Garcıa, J., 2006. Performance of experimental horizontal subsurface flow constructed wetlands fed with dissolved or particulate organic matter. Water Res. 40, 3603–3611. Caselles-Osorio, A., Puigagut, J., Segu´ı, E., Vaello, N., Granes, F., Garcıa, D., Garcıa, J., 2007. Solids accumulation in six full-scale subsurface flow constructed wetlands. Water Res. 41, 1388–1398. Cooper, D.J., Griffin, P., Cooper, P.F., 2005. Factors affecting the longevity of subsurface horizontal flow systems operating as tertiary treatment for sewage effluent. Water Sci. Technol. 51 (9), 127–135. Fonder, N., Headley, T., 2013. The taxonomy of treatment wetlands: a proposed classification and nomenclature system. Ecol. Eng. 5, 203–211. García, J., Chiva, J., Aguirre, P., Álvarez, E., Sierra, J., Mujeriego, R., 2004. Hydraulic behaviour of horizontal subsurface flow constructed wetlands with different aspect ratio and granular medium size. Ecol. Eng. 23, 177–187. Giraldi, D., de’Michieli Vitturi, M., Zaramella, M., Marion, A., Iannelli, R., 2009. Hydrodynamics of vertical subsurface flow constructed wetlands: tracer tests with rhodamine WT and numerical modelling. Ecol. Eng. 35 (2), 265–273. Headley, T.R., Kadlec, R.H., 2007. Conducting hydraulic tracer studies of constructed wetlands: a practical guide. Ecohydrol. Hydrobiol. 7 (3–4), 269–282. Helsel, D.R., Hirsch, R.M., 1992. Statistical 1 Methods in Water Resources. Elsevier Science 2 Publishers, Amsterdam, The Netherlands, ISBN 0-444-81463-9, 522 pp. IWA, 2000. Constructed wetlands for pollution control. In: Scientific and Technical Report No. 8. IWA Publishing, London, England. Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed. CRC Press, Boca Raton, FL, New York, USA. Knowles, P.R., Davies, P.A., 2009. A method for the in-situ determination of the hydraulic conductivity of gravels as used in constructed wetlands for wastewater treatment. Desalin. Water Treat. 5, 257–266. Knowles, P., Griffin, P., Davies, P.A., 2010. Complementary methods to investigate the development of clogging within a horizontal sub-surface flow tertiary treatment wetland. Water Res. 44 (1), 320–330. Knowles, P., Dotro, G., Nivala, J., García, J., 2011. Clogging in subsurface-flow treatment wetlands: occurrence and contributing factors. Ecol. Eng. 37, 99–112. Levenspiel, O., 1972. Chemical Reaction Engineering, 2nd edition. Wiley, New York, USA. Milani, M., Toscano, A., 2013. Evapotranspiration from pilot-scale constructed wetlands planted with Phragmites australis in a Mediterranean environment. J. Environ. Sci. Health A 48, 568–580. NAVFAC, 1986. Soil Mechanics Design Manual 7.01. Naval Facilities Engineering Command, Alexandria, Virginia, USA.

513

Nivala, J., Rousseau, P.L., 2009. Reversing clogging in subsurface-flow constructed wetlands by hydrogen peroxide treatment: two case studies. Water Res. 59, 2037–2046. Nivala, J.A., Knowles, P., Dotro, G., García, J., Wallace, S., 2012. Clogging in subsurface-flow treatment wetlands: measurement, modeling and management. Water Res. 46, 1625–1640. Pedescoll, A., Uggetti, E., Llorens, E., Granés, F., Garcia, D., García, J., 2009. Practical method based on saturated hydraulic conductivity used to asses clogging in subsurface flow constructed wetlands. Ecol. Eng. 35, 1216–1224. Pedescoll, A., Corzo, A., Alvarez, E., Garcıa, J., Puigagut, J., 2011. The effect of primary treatment and flow regime on clogging development in horizontal subsurface flow constructed wetlands: an experimental evaluation. Water Res. 45, 3579–3589. Persson, J., Somes, N., Wong, T., 1999. Hydraulic efficiency of constructed wetlands and ponds. Water Sci. Technol. 40 (3), 291–300. Ronkanen, A.-K., Kløve, B., 2007. Use of stabile isotopes and tracers to detect preferential flow patterns in a peatland treating municipal wastewater. J. Hydrol. 347 (3–4), 418–429. Sanford, W.E., Steenhuis, T.S., Parlange, J.-Y., Surface, J.M., Peverly, J.H., 1995. Hydraulic conductivity of gravel and sand as substrates in rock-reed filters. Ecol. Eng. 4 (4), 321–336. Stefanakis Alexandros, I., Tsihrintzis Vassilios, A., 2009. Effect of outlet water level raising and effluent recirculation on removal efficiency of pilot-scale, horizontal subsurface flow constructed wetlands. Desalination 248, 961–976. Suliman, F., Futsaether, C., Oxaal, U., Haugen, L.E., Jenssen, P., 2006. Effect of the inlet outlet positions on the hydraulic performance of horizontal subsurface-flow wetlands constructed with heterogeneous porous media. J. Contam. Hydrol. 87 (1–2), 22–36. Toscano, A., Hellio, C., Marzo, A., Milani, M., Lebret, K., Cirelli, G.L., Langergraber, G., 2013. Removal efficiency of a constructed wetland combined with ultrasound and UV devices for wastewater reuse in agriculture. Environ. Technol. 34 (13–16), 2327–2336. Toscano, A., Marzo, A., Milani, M., Cirelli, G.L., Barbagallo, S., 2015. Comparison of removal efficiencies in Mediterranean pilot constructed wetlands vegetated with different plant species. Ecol. Eng. 75, 155–160. Viviani, G., Iovino, M., 2004. Wastewater reuse effects on soil hydraulic conductivity. J. Irrig. Drain. Eng. 130 (6), 476–484. Vymazal, J., 2014. Constructed wetlands for treatment of industrial wastewaters: a review. Ecol. Eng. 73, 724–751. Wang, Y., Song, X., Liao, W., Niu, R., Wang, W., Ding, Y., 2014. Impacts of inlet–outlet configuration, flow rate and filter size on hydraulic behavior of quasi-2-dimensional horizontal constructed wetland: NaCl and dye tracer test. Ecol. Eng. 69, 177–185. Yalcuk, A., Ugurlu, A., 2009. Comparison of horizontal and vertical constructed wetlands systems for landfill leachate treatment. Bioresour. Technol. 100, 2521–2526.