Abstract

The objective of this work is to evaluate the performance of two horizontal subsurface flow constructed wetlands, one planted with cattail (Typha latifolia) and the other unplanted. The distinguishing feature of this study is that it spans a period of more than 10 years, from start-up to a final operation with heavy clogging and full overland flow. For most of the time, starting in June 2007, the system received municipal sewage previously treated in an upflow anaerobic sludge blanket (UASB) reactor, but for one specific period, the pre-treatment was comprised of the UASB reactor and a trickling filter in series. The two constructed wetlands worked in parallel, each serving approximately 50 p.e. and continuously receiving a flow around 7.5 m3 d−1 for most of the time. The beds had a length of 25 m and a width of 3 m and were filled with blast furnace slag. For most of this long operational period, performance was very good in terms of biochemical oxygen demand (BOD), chemical oxygen demand (COD) and total suspended solids (TSS), with median effluent concentrations of 19/18, 46/52 and 12/8, respectively (planted/unplanted units). Clogging was noticeable in the first years of operation, soon leading to overland flow. However, treatment performance was still successful, even when the system's hydraulics were strongly deteriorated. The type of pre-treatment and the applied loads influenced more the performance of the units than the period of operation itself, evidencing the robustness of the system.

INTRODUCTION

Horizontal subsurface flow constructed wetlands (HF CW) are primarily designed to remove organic matter and suspended solids, with thousands of units implemented worldwide, showing successful performances. The major operational concern is clogging, which is an inevitable process that, depending on its extent, may affect the use of such an important treatment system (Vymazal 2005, 2018a). The process of clogging starts next to the inlet end and advances along the bed length over the years (Samsó & García 2014).

Clogging occurs when voids are filled with solids associated with sedimentation and filtration material, chemical precipitation, material wear, and biofilm and roots growth. However, the adoption of suitable filtering materials and applied loads, together with system maintenance, may mitigate this process (Kadlec & Wallace 2009; Vymazal 2018a). A strong manifestation of the clogging process is when the increased head loss leads to surface overland flow, with its possible aesthetic implications. However, another impact may be the possible reduction in treatment efficiency. In the literature there are few studies that cover wetland units over many years of operation, relating the clogging process with the performance of the system. This is endorsed by Vymazal (2018a), who stated that surface runoff is considered a failure in the system and that there have been no reports on assessing the effect of clogging on a long-term treatment for removal of organics and suspended solids.

The objective of this work is to evaluate the performance of two horizontal subsurface flow constructed wetlands, with one filter planted with cattail (Typha latifolia) and the other filter unplanted. Both units received sewage previously treated. The distinguishing feature of this study is that it spans a period of more than 10 years, from start-up to a final operation with heavy clogging and full overland flow. The investigation was undertaken in Brazil, under tropical conditions. Besides monitoring data, a compilation of several studies undertaken in the same system is presented, aiming at relating hydraulic, physical and process aspects with the system performance. In this publication, performance assessment focuses mainly on the traditional constituents representing organic matter (chemical oxygen demand and biochemical oxygen demand) and total suspended solids. Although nitrogen and phosphorus are also presented, their coverage is with less detail, since, as expected, nutrient removal is not high in this and in similar horizontal-flow systems. In a similar way, even though results from the planted and unplanted units are presented here, the objective of this work is not to undertake a full comparison of ‘planted versus unplanted’, since this topic is vastly covered in the wetlands literature. Additional details on this comparison, the role of plants and the removal of nutrients in this system can be found in Costa et al. (2013, 2015).

MATERIAL AND METHODS

The research was carried out at the Centre for Research and Training in Sanitation – CePTS UFMG/COPASA, located at coordinates 19°53′42″S and 43°42′52″W, in the Arrudas Wastewater Treatment Plant in the city of Belo Horizonte, Brazil. The area has a tropical climate, with an average air temperature of 21.8 °C and annual rainfall of 1,602 mm/year. Figure 1 shows data from the Normal Climatological for air temperature and rainfall between the years 1981 to 2010 (INMET 2018).

Figure 1

Air temperature (left) and rainfall (right) of the study area. Source:INMET (2018).

Figure 1

Air temperature (left) and rainfall (right) of the study area. Source:INMET (2018).

Part of the incoming sewage, after preliminary treatment (coarse and fine screens followed by grit removal), was diverted to feed the experimental units. Before going to the wetlands, the sewage underwent a previous treatment, most of the time involving only a UASB (upflow anaerobic sludge blanket) reactor. After that, the effluent was directed to the two HF CW units, each one designed for a population equivalent of 50 inhabitants, occupying a per capita area of around 1.5 m2/inhabitant. One of the units was planted (PU) with Typha latifolia, popularly known as cattail, while the other unit remained unplanted (UPU). The wetlands have been in operation since 2007, working in parallel, each continuously receiving a flow of around 7.5 m3/d.

Typha latifolia was planted with a density of four plants/m2, a density suggested by Reed et al. (1995). Cutting was done manually, after flowering of the plants (at an average of approximately 3 months), leaving a stem height of 20 cm, relative to the soil.

The units were filled with blast furnace slag as substrate, with diameter d10 equal to 19.1 mm and uniformity coefficient (d60/d10) equal to 1.2 (Dornelas et al. 2009), except for the inlet and outlet zones, which had larger stones (between 10 to 15 cm) to facilitate the distribution and collection of sewage. Table 1 shows the most important design, construction and operational aspects of each unit.

Table 1

Design, construction and operational characteristics of each constructed wetlands unit

ParameterUnitValue
Total medium height (ht0.40 
Useful (liquid) height (design value) (hu0.30 
Length (top) (L25.0 
Width (top) (B3.0 
Length/width ratio (L/W) 8.3 
Longitudinal bottom slope (i0.5 
Total bed volume (Vtm3 30.0 
Bed surface area (top) (Asm² 75.0 
Porosity of the filter medium during construction (ɛm3/m3 0.40 
Influent design flow (Qinm3/d1 7.50 
Theoretical hydraulic retention time (HRT1.20 
ParameterUnitValue
Total medium height (ht0.40 
Useful (liquid) height (design value) (hu0.30 
Length (top) (L25.0 
Width (top) (B3.0 
Length/width ratio (L/W) 8.3 
Longitudinal bottom slope (i0.5 
Total bed volume (Vtm3 30.0 
Bed surface area (top) (Asm² 75.0 
Porosity of the filter medium during construction (ɛm3/m3 0.40 
Influent design flow (Qinm3/d1 7.50 
Theoretical hydraulic retention time (HRT1.20 

The units were constructed after adaptation of a formerly existing maturation pond. Because of this, there was a resulting high length/width ratio (L:W = 8.3:1), much above the recommendations of 2:1 to 4:1 for HF CW working in secondary treatment (IWA Task Group 2017). This aspect ratio, although positive from a reactor's engineering point of view, led to a small cross-sectional area, which induces higher headlosses.

Over more than 10 years (September 2007 to August 2018), the system operated with four different phases, depending on the pre-treatment configuration and the loading conditions, in order to allow research into different operational settings: (a) Phase 1: effluent from UASB reactor (total of 47 months); (b) Phase 2: effluent from trickling filter following UASB reactor (total of 26 months); (c) Phase 3: back to effluent from UASB reactor, but with the units showing clogging and overland flow (total of 45 months); (d) Phase 4: same configuration as in Phase 3, but doubling the influent flow (about 15 m3/d) in order to increase the applied loads and speed up the clogging process (13 months).

Monitoring was undertaken with a typical frequency of once a week, leading to hundreds of results for each parameter. Grab samples were collected at four points: raw sewage, effluent from UASB reactor, effluent from trickling filter (only during Phase 2) and effluent from planted and unplanted units. The parameters monitored comprised a vast array of water constituents. Those of special interest here are biochemical oxygen demand (BOD), chemical oxygen demand (COD), total suspended solids (TSS), total Kjeldahl nitrogen (TKN) and total phosphorus (TP). The physicochemical analyses were undertaken at the Department of Sanitary and Environmental Engineering of the Federal University of Minas Gerais, Brazil, following the procedures of Standard Methods for the Examination of Water and Wastewater (APHA/AWWA/WEF 2005).

The impact of clogging on the system was categorized on a subjective qualitative scale in terms of the two following aspects:

  • impact on filter hydraulics: (i) no impact, when overland flow was not noticed; (ii) deterioration, when overland flow was observed; (iii) collapse (failure), when overland flow was coming close to the outlet structures;

  • impact on system performance: (i) no impact, when removal efficiencies remained without noticeable decreases; (ii) deterioration, when removal efficiencies had been substantially decreased; (iii) collapse (failure), when removal efficiencies were very low and could be considered unacceptable.

With the objective to evaluate the difference between the phases, the non-parametric Kruskal-Wallis statistical test was performed for the comparison of effluent concentrations and removal efficiencies based on loads. The tests were interpreted in terms of a significance level of 5% using the software® Statistica. As mentioned in the Introduction, the focus of this work is not the comparison between the planted and unplanted units, but rather the operational phases of the units. Therefore, no statistical comparison between ‘planted’ versus ‘unplanted’ is presented.

RESULTS AND DISCUSSION

Over the full monitoring period, the median BOD, COD, TSS, TKN and TP concentration values in the influent to the treatment plant (raw sewage) were 247, 406, 190, 29 and 4.2 mg/L, respectively, and the corresponding median concentrations in the effluent from the treatment ahead of the wetlands (either UASB reactor alone or UASB reactor followed by trickling filter); that is, the influent to the wetlands units, were 58, 138, 45, 31 and 4.1 mg/L.

Table 2 shows the descriptive statistics related to the five quality parameters and Figure 2 the box-plot graphs of the effluent concentrations and removal efficiencies of BOD, COD and TSS in the planted and unplanted units, during the four operational phases. The results of the statistical tests, comparing the four phases, are presented in Table 3. In general, a good performance and a certain stability among the first three operational phases (118 months, almost 10 years) can be observed, highlighting the robustness of the system. Only from Phase 4 (the last 13 months), when the system was highly clogged, with overland flow occupying virtually the full filter areas and influent flow had been doubled, a clear tendency to have increased effluent concentrations and reduced removal efficiencies can be noticed.

Table 2

Descriptive statistics of effluent concentrations and removal efficiencies of planted and unplanted units in the four operational phases

ItemEffluent concentration (mg/L)
Removal efficiency in the filter units (%)
Phase 1Phase 2Phase 3Phase 4Overall periodPhase 1Phase 2Phase 3Phase 4Overall period
BOD 
Number of data 93/96 69/76 73/71 23/22 258/265 84/87 47/48 43/47 36/35 210/217 
 Mean 25/23 13/12 22/22 83/71 26/24 65/64 53/54 58/56 30/37 55/56 
Median 20/19 10/11 19/19 88/58 19/18 72/72 58/65 65/63 38/50 62/66 
 Minimum 5/5 1/1 7/6 6/1 1/1 −23/−21 −83/−27 −11/5 −47/−68 −83/−68 
 Maximum 90/103 51/32 115/133 155/164 155/164 94/92 97/96 87/92 88/93 97/96 
 Standard deviation 19/17 10/8 15/16 51/41 28/24 21/24 36/33 22/23 35/39 30/30 
COD 
Number of data 111/109 70/74 58/56 27/28 266/267 99/98 53/54 31/38 39/39 222/229 
 Mean 50/55 26/31 65/76 92/109 51/58 71/65 68/63 59/48 42/35 64/56 
Median 50/52 25/27 57/66 82/102 46/52 73/68 70/67 62/55 50/44 70/62 
 Minimum 11/4 2/4 7/2 8/47 2/2 10/19 −8/−14 −7/−12 −28/−59 −28/−59 
 Maximum 106/139 86/80 194/218 174/192 194/218 93/99 97/93 89/98 92/76 97/99 
 Standard deviation 19/25 15/17 43/43 47/41 35/38 13/17 22/25 25/26 33/32 24/26 
TSS 
Number of data 156/153 79/79 30/43 17/17 282/275 118/116 54/55 17/21 17/18 206/210 
 Mean 11/9 17/13 26/24 17/25 15/12 69/74 62/74 61/65 63/49 66/71 
Median 9/6 13/10 17/12 13/17 12/8 77/84 76/85 71/76 70/71 76/82 
 Minimum 1/0.3 4/2 1/1 2/4 1/8 −60/−60 −69/−57 −48/−33 −16/−111 −69/−111 
 Maximum 40/44 69/72 206/170 34/74 206/170 97/99 97/97 98/98 95/96 98/99 
 Standard deviation 9/8 12/12 36/30 9/22 16/15 29/26 38/30 39/36 29/57 33/32 
TKN 
Number of data 123/126 69/79 42/43 NM 230/244 95/94 57/64 22/24 NM 167/175 
 Mean 30/31 21/22 25/25 NM 27/27 21/16 40/34 44/30 NM 32/28 
Median 29/31 22/24 27/26 NM 26/26 24/24 36/35 48/30 NM 31/29 
 Minimum 6/9 2/2 6/5 NM 2/2 −16/−37 2/−53 9/−51 NM −16/−53 
 Maximum 55/52 37/37 48/46 NM 55/52 67/54 80/83 82/85 NM 82/85 
 Standard deviation 10/9 8/8 10/11 NM 10/9 20/19 19/26 24/38 NM 21/23 
TP 
Number of data 65/65 37/49 NM NM 102/114 43/44 25/32 NM NM 64/72 
 Mean 2.3/2.3 1.6/1.4 NM NM 2.1/1.9 43/37 69/71 NM NM 54/53 
Median 1.4/1.8 1.1/0.9 NM NM 1.4/1.4 52/37 83/85 NM NM 59/59 
 Minimum 0.0/0.1 0.1/0.1 NM NM 0.0/0.1 −48/−24 5/−71 NM NM −48/−71 
 Maximum 7.1/8.0 8.4/5.2 NM NM 8.4/8.0 94/93 97/99 NM NM 97/99 
 Standard deviation 1.9/1.7 1.6/1.2 NM NM 1.8/1.6 33/30 31/41 NM NM 31/35 
ItemEffluent concentration (mg/L)
Removal efficiency in the filter units (%)
Phase 1Phase 2Phase 3Phase 4Overall periodPhase 1Phase 2Phase 3Phase 4Overall period
BOD 
Number of data 93/96 69/76 73/71 23/22 258/265 84/87 47/48 43/47 36/35 210/217 
 Mean 25/23 13/12 22/22 83/71 26/24 65/64 53/54 58/56 30/37 55/56 
Median 20/19 10/11 19/19 88/58 19/18 72/72 58/65 65/63 38/50 62/66 
 Minimum 5/5 1/1 7/6 6/1 1/1 −23/−21 −83/−27 −11/5 −47/−68 −83/−68 
 Maximum 90/103 51/32 115/133 155/164 155/164 94/92 97/96 87/92 88/93 97/96 
 Standard deviation 19/17 10/8 15/16 51/41 28/24 21/24 36/33 22/23 35/39 30/30 
COD 
Number of data 111/109 70/74 58/56 27/28 266/267 99/98 53/54 31/38 39/39 222/229 
 Mean 50/55 26/31 65/76 92/109 51/58 71/65 68/63 59/48 42/35 64/56 
Median 50/52 25/27 57/66 82/102 46/52 73/68 70/67 62/55 50/44 70/62 
 Minimum 11/4 2/4 7/2 8/47 2/2 10/19 −8/−14 −7/−12 −28/−59 −28/−59 
 Maximum 106/139 86/80 194/218 174/192 194/218 93/99 97/93 89/98 92/76 97/99 
 Standard deviation 19/25 15/17 43/43 47/41 35/38 13/17 22/25 25/26 33/32 24/26 
TSS 
Number of data 156/153 79/79 30/43 17/17 282/275 118/116 54/55 17/21 17/18 206/210 
 Mean 11/9 17/13 26/24 17/25 15/12 69/74 62/74 61/65 63/49 66/71 
Median 9/6 13/10 17/12 13/17 12/8 77/84 76/85 71/76 70/71 76/82 
 Minimum 1/0.3 4/2 1/1 2/4 1/8 −60/−60 −69/−57 −48/−33 −16/−111 −69/−111 
 Maximum 40/44 69/72 206/170 34/74 206/170 97/99 97/97 98/98 95/96 98/99 
 Standard deviation 9/8 12/12 36/30 9/22 16/15 29/26 38/30 39/36 29/57 33/32 
TKN 
Number of data 123/126 69/79 42/43 NM 230/244 95/94 57/64 22/24 NM 167/175 
 Mean 30/31 21/22 25/25 NM 27/27 21/16 40/34 44/30 NM 32/28 
Median 29/31 22/24 27/26 NM 26/26 24/24 36/35 48/30 NM 31/29 
 Minimum 6/9 2/2 6/5 NM 2/2 −16/−37 2/−53 9/−51 NM −16/−53 
 Maximum 55/52 37/37 48/46 NM 55/52 67/54 80/83 82/85 NM 82/85 
 Standard deviation 10/9 8/8 10/11 NM 10/9 20/19 19/26 24/38 NM 21/23 
TP 
Number of data 65/65 37/49 NM NM 102/114 43/44 25/32 NM NM 64/72 
 Mean 2.3/2.3 1.6/1.4 NM NM 2.1/1.9 43/37 69/71 NM NM 54/53 
Median 1.4/1.8 1.1/0.9 NM NM 1.4/1.4 52/37 83/85 NM NM 59/59 
 Minimum 0.0/0.1 0.1/0.1 NM NM 0.0/0.1 −48/−24 5/−71 NM NM −48/−71 
 Maximum 7.1/8.0 8.4/5.2 NM NM 8.4/8.0 94/93 97/99 NM NM 97/99 
 Standard deviation 1.9/1.7 1.6/1.2 NM NM 1.8/1.6 33/30 31/41 NM NM 31/35 

Values shown in the table ‘x/y’ refer to the values of the planted unit (x) and unplanted unit (y).

Removal efficiency is calculated on the basis of the removed load (flow x concentration) and comprises only the filter units (pre-treatment on the UASB reactor and on the trickling filter is not included in these calculations).

NM: not measured.

Phase 1: effluent from UASB reactor; Phase 2: effluent from trickling filter following UASB reactor; Phase 3: back to effluent from UASB reactor, but with the units showing clogging and overland flow; Phase 4: same configuration as in Phase 3, but doubling the influent flow in order to increase the applied loads and speed up the clogging process.

Table 3

p-value of the Kruskal-Wallis test of concentrations and removal efficiencies of BOD, COD and TSS of the planted and unplanted wetlands in the four phases of operation

Effluent concentration
Comparison between phasesParameterPhase 1Phase 2Phase 3
Phase 2 BOD
COD
TSS 
0.000015*/0.000003*
0.000000*/0.000000*
0.000041*/0.004949* 
– – 
Phase 3 BOD
COD
TSS 
1.000000/1.000000
1.000000/0.030246*
0.000154*/0.038196* 
0.000063*/0.000025*
0.000000*/0.000000*
1.000000/1.000000 
– 
Phase 4 BOD
COD
TSS 
0.000016*/0.000002*
0.002450*/0.000000*
0.024993*/0.000425* 
0.000000*/0.000000*
0.000000*/0.000000*
1.000000/0.232824 
0.000025*/0.000003*
0.073336/0.019956*
1.000000/0.967006 
Removal efficiency
Comparison between phasesParameterPhase 1Phase 2Phase 3
Phase 2 BOD
COD
TSS 
0.387914/0.807673
1.000000/1.000000
1.000000/1.000000 
– – 
Phase 3 BOD
COD
TSS 
0.503199/0.268804
0.109591/0.000823*
1.000000/0.821230 
1.000000/1.000000
0.518550/0.005942*
1.000000/0.948584 
– 
Phase 4 BOD
COD
TSS 
0.000000*/0.000347*
0.000001*/0.000000*
1.000000/0.243169 
0.000770*/0.095367
0.000183*/0.000004*
1.000000/0.310519 
0.000745*/0.287205
0.245207/0.747803
1.000000/1.000000 
Effluent concentration
Comparison between phasesParameterPhase 1Phase 2Phase 3
Phase 2 BOD
COD
TSS 
0.000015*/0.000003*
0.000000*/0.000000*
0.000041*/0.004949* 
– – 
Phase 3 BOD
COD
TSS 
1.000000/1.000000
1.000000/0.030246*
0.000154*/0.038196* 
0.000063*/0.000025*
0.000000*/0.000000*
1.000000/1.000000 
– 
Phase 4 BOD
COD
TSS 
0.000016*/0.000002*
0.002450*/0.000000*
0.024993*/0.000425* 
0.000000*/0.000000*
0.000000*/0.000000*
1.000000/0.232824 
0.000025*/0.000003*
0.073336/0.019956*
1.000000/0.967006 
Removal efficiency
Comparison between phasesParameterPhase 1Phase 2Phase 3
Phase 2 BOD
COD
TSS 
0.387914/0.807673
1.000000/1.000000
1.000000/1.000000 
– – 
Phase 3 BOD
COD
TSS 
0.503199/0.268804
0.109591/0.000823*
1.000000/0.821230 
1.000000/1.000000
0.518550/0.005942*
1.000000/0.948584 
– 
Phase 4 BOD
COD
TSS 
0.000000*/0.000347*
0.000001*/0.000000*
1.000000/0.243169 
0.000770*/0.095367
0.000183*/0.000004*
1.000000/0.310519 
0.000745*/0.287205
0.245207/0.747803
1.000000/1.000000 

Values shown in the table ‘x/y’ refer to the p-values of the planted unit (x) and unplanted unit (y). The symbol * indicates where there is a significant difference.

p ≤ 0.05: medians of the phases are significantly different. p > 0.05: medians of the phases are not significantly different.

Phase 1: effluent from UASB reactor; Phase 2: effluent from trickling filter following UASB reactor; Phase 3: back to effluent from UASB reactor, but with the units showing clogging and overland flow; Phase 4: same configuration as in Phase 3, but doubling the influent flow in order to increase the applied loads and speed up the clogging process.

Figure 2

Box-plot of the concentrations and removal efficiencies (calculated on the basis of the removed load) of BOD, COD and TSS of the planted and unplanted units in the four phases of operation. Phase 1: effluent from UASB reactor; Phase 2: effluent from trickling filter following UASB reactor; Phase 3: back to effluent from UASB reactor, but with the units showing clogging and overland flow; Phase 4: same configuration as in Phase 3, but doubling the influent flow in order to increase the applied loads and speed up the clogging process.

Figure 2

Box-plot of the concentrations and removal efficiencies (calculated on the basis of the removed load) of BOD, COD and TSS of the planted and unplanted units in the four phases of operation. Phase 1: effluent from UASB reactor; Phase 2: effluent from trickling filter following UASB reactor; Phase 3: back to effluent from UASB reactor, but with the units showing clogging and overland flow; Phase 4: same configuration as in Phase 3, but doubling the influent flow in order to increase the applied loads and speed up the clogging process.

Analysing in more detail, lower effluent concentrations of BOD and COD in Phase 2, when the units received effluent from a better pre-treatment (UASB + trickling filter), and higher effluent concentrations in Phase 4, when the inflow was doubled and the system was highly clogged, were observed in both units. The results of the statistical tests showed that the differences in concentrations between the phases were significant, except for COD in the planted unit, between Phases 3 and 4. TSS had lower effluent concentrations during the first phase and, contrary to what could be expected, was not influenced by the increase of the load and the heavy clogging in Phase 4. As expected, the median removal efficiencies of BOD, COD and TSS were lower in Phase 4, when compared to the other phases, in both units. However, the statistical tests showed that the differences were significant for BOD and COD, except for BOD in UPU, when comparing Phase 4 with Phases 2 and 3.

Phase 2 demonstrated better removal efficiencies of TP, which was expected due to the addition of the trickling filter. Nevertheless, the median values of TKN removal for the whole monitored period were low, 31% and 29% in PU and UPU, respectively, reinforcing the inability of horizontal units in supporting nitrification. The removal of TP in the overall period was 59% in PU and 59% in UPU. Costa et al. (2015) performed an N and P balance in the same system and found a 42% removal of N and 66% of P in the planted unit, and attributed this removal to several possible intervening mechanisms. Plant absorption accounted for only 7% removal of N and 6% of P. The rest of the P removal could have been due to adsorption by the steel slag used as support medium. A water balance was performed by Costa (2013) over one year in this system and the author found a water loss of only 5% due to evapotranspiration in PU and a water loss due to evaporation of only 2% in UPU. Therefore, water losses are expected to have had little impact on the effluent concentrations.

In the first phase of the system monitoring, when the units were still relatively new (two years of operation), von Sperling & De Paoli (2013) determined the dispersion number (d) from tracer (Br) tests. The results indicated a low to moderate dispersion, with values of d = 0.084 for the planted unit and d = 0.079 for the unplanted unit. These values of the dispersion number d correspond to the equivalent of 6.5 and 6.9 tanks in series (N) for the planted and unplanted units, respectively. These equivalent numbers of tanks in series are large but are coherent with the elongated aspect ratio of both units, with a high length:width ratio of 8.3:1. Von Sperling & De Paoli (2013) also tested different hydraulic models of COD decay, based on measurements along the filters' length. The best simulations were obtained with the first-order dispersed-flow and tanks-in-series models, both with residual (C*). Another feature studied was the incorporation of water losses in the models, due to wetlands evapotranspiration. The inclusion of this adjustment has improved even further the performance of all models. The resulting values of the COD decay coefficient (K) in the dispersed-flow model with residual C* was 1.31 d−1 for the planted unit and 1.29 d−1 for the unplanted unit. As expected, similar K values were obtained with the tanks-in-series model with residual C* (1.32 d−1 for the planted unit and 1.29 d−1 for the unplanted unit).

This tracer test indicated a real mean HRT (hydraulic retention time) of 1.30 days for the planted unit and 1.43 days for the unplanted unit (de Paoli & von Sperling 2013). However, Matos et al. (2015) undertook other tracer tests (this time with NaCl) in the same units after seven years of operation, when clogging was noticeable and overland flow was advanced. The values of the real mean HRT were 1.45 days at the planted unit at the growth stage of the plants (before their cutting) and 1.38 days at another growing stage of the plants (after their cutting). At the unplanted unit the same authors found a lower real mean HRT (1.06 days). These results demonstrated that throughout the five-year interval between the two tracer tests there was a higher reduction of the volume of void space at the unplanted unit, indicating that it was in a more advanced stage of clogging than the planted unit.

Overland flow (surface runoff) began to be observed in 2009 (still in Phase 1), two years after the start of operation. De Paoli & von Sperling (2013) observed a larger area of surface runoff in the planted unit, but from Phase 3, in the year 2013 (six years of operation), the unplanted unit showed larger areas. After almost 11 years of operation, the planted unit presented 88% runoff on its surface, while the unplanted unit had 100% of its surface fully covered by liquid. Despite the progress of the runoff, the performance of both units remained satisfactory until Phase 3 (10 years of operation). Cooper (2009) conducted a survey in the UK on 255 tertiary HF systems and identified that 76 systems suffered from overland flow on most of the bed surface. However, surface runoff did not impair treatment efficiency.

Hydraulic conductivity (kf) tests were carried out in order to allow a better understanding of the flow inside the beds. De Paoli & von Sperling (2013) obtained higher values of hydraulic conductivity in the unplanted unit when compared to the planted unit. The hydraulic conductivity at the inlet zone (3 and 6 m from inlet) ranged from 7 to 41 m/d in the planted unit and from 67 to 166 m/d in the unplanted unit. At the outlet zone (18 m from inlet), values were higher, as expected, ranging from 30 to 107 m/d in the planted unit and from 198 to 324 m/d in the unplanted unit. Matos et al. (2017) performed new measurements of kf at the end of 2013 (after seven years of operation) and obtained values in the inlet zone (4 and 7 m) from 5 to 22 m/d in the planted unit and from 4 to 40 m/d in the unplanted unit. The variation in the outlet zone (22 m) was from 6 to 61 m/d for the planted unit and from 31 to 125 m/d for the unplanted unit. It can be seen that, with the passage of time, the values of the hydraulic conductivity decreased, indicating pore obstruction associated with clogging. As expected, several authors also identified that clogging was the most severe within the first few meters along the length of the bed (Fisher 1990; Tanner & Sukias 1995; Tanner et al. 1998; Vymazal 2003).

The IWA Task Group (2017) recommends using a maximum surface organic loading rate of 8 gBOD5/m2 · d for horizontal wetlands performing secondary treatment of domestic sewage. In this study, the applied rates were higher only in Phase 4 (double the recommended maximum value – see Table 4), as a consequence of the doubling of the influent flow in order to study the acceleration of the clogging process. Regarding the organic loading rate in the cross-section, the maximum recommended value by the IWA Task Group (2017) is 250 gBOD5/m2 · d, and in this study the values were much higher, given the elongated shape of the units and the resulting small cross-sectional area. This may explain the fast starting of the clogging process and the resulting overland flow after only a few years of operation. Only in Phase 2, when the system received effluent from the UASB reactor followed by trickling filter, were the loading rates below the recommended level. In Phase 4 the values were much higher, exceeding 1,000 gBOD5/m2 · d. Wallace (2014) recommends conservative rates of 100 gBOD5/m2 · d; that is, 10 times lower than those used in Phase 4 of this study. The organic loading rate in the cross-section is directly related to the bed clogging and, the lower the adopted rates, the longer should be the system life. In tropical countries, as is the case for this research, it is expected that it could be possible to adopt higher rates than those used in temperate climates.

Table 4

Basic information regarding operational conditions and treatment performance of the system in the four phases

ItemPhase 1Phase 2Phase 3Phase 4Overall period
Treatment configuration UASB + wetland UASB + TF + wetland UASB + wetland UASB + wetland UASB (+TF) + wetland 
Duration of each phase (months) 47 26 45 13 131 
Cumulative time (from start-up to end of phase) (years) ∼4 ∼6 ∼10 ∼11 ∼11 
Mean real hydraulic retention time, based on tracer tests (d) 1.30/1.43 NM 1.45a;1.38b/1.06 NM – 
Surface hydraulic loading rate (m3/m2 · d) 0.107/0.104 0.102/0.104 0.079/0.111 0.213/0.219 0.107/0.107 
Surface organic loading rate (gBOD5/m2 · d) 6.9(6.4)/7.3(6.0) 3.2(2.2)/3.8(2.5) 5.8(3.3)/6.9(3.1) 16.0(12.5)/16.6(12.3) 5.6(7.4)/6.5(7.2) 
Cross-sectional organic loading rate (gBOD5/m2 · d) 414(383)/429(361) 192(133)/226(150) 345(194)/412(184) 1,019(751)/1,090(736) 345(445)/388(433) 
Surface organic loading rate (gCOD/m2 · d) 17.2(10.7)/17.8(10.3) 10.8(7.8)/10.9(9.0) 12.1(8.8)/14.9(9.5) 40.1(16.6)/35.4(17.3) 15.1(12.9)/15.6(12.6) 
Surface solids loading rate (gTSS/m2 · d) 3.0(10.8)/3.3(10.9) 4.6(5.6)/4.9(5.2) 3.3(4.0)/4.6(5.3) 20.5(8.1)/22.4(8.8) 3.4(9.3)/3.8(9.4) 
Percentage of surface area covered with overland flow (%) 0–60/0–40 23/21 80/87 88/100 – 
Impact of clogging in terms of hydraulics No impact Deterioration Deterioration Collapse – 
Impact of clogging in terms of performance No impact No impact No impact Deterioration – 
Effluent concentration (mg/L) 
 BOD 20(19)/19(17) 10(10)/11(8) 19(15)/19(16) 88(51)/58(41) 19(28)/18(24) 
 COD 50(19)/52(25) 25(15)/27(17) 57(43)/67(43) 91(47)/102(41) 46(35)/52(38) 
 TSS 9(9)/6(8) 13(12)/10(12) 17(36)/10(30) 10(9)/26(22) 11(16)/8(15) 
Removal efficiency at wetlands stagec (%) 
 BOD 72(21)/72(24) 58(30)/65(33) 65(22)/63(23) 38(35)/50(39) 62(29)/66(30) 
 COD 73(13)/68(17) 70(22)/67(25) 62(25)/55(26) 50(33)/44(32) 70(24)/62(26) 
 TSS 77(29)/84(26) 66(38)/85(30) 71(54)/76(36) 70(29)/71(57) 76(34)/82(32) 
Overall efficiency: pre-treatment + wetlandsc (%) 
 BOD 89(13)/90(13) 94(4)/94(4) 91(4)/90(5) 84(11)/90(11) 91(10)/91(10) 
 COD 88(8)/89(9) 94(7)/92(14) 86(15)/79(22) 78(26)/73(21) 88(17)/86(18) 
 TSS 94(6)/94(5) 91(22)/95(11) 90(12)/91(25) 93(4)/95(8) 92(15)/94(12) 
ItemPhase 1Phase 2Phase 3Phase 4Overall period
Treatment configuration UASB + wetland UASB + TF + wetland UASB + wetland UASB + wetland UASB (+TF) + wetland 
Duration of each phase (months) 47 26 45 13 131 
Cumulative time (from start-up to end of phase) (years) ∼4 ∼6 ∼10 ∼11 ∼11 
Mean real hydraulic retention time, based on tracer tests (d) 1.30/1.43 NM 1.45a;1.38b/1.06 NM – 
Surface hydraulic loading rate (m3/m2 · d) 0.107/0.104 0.102/0.104 0.079/0.111 0.213/0.219 0.107/0.107 
Surface organic loading rate (gBOD5/m2 · d) 6.9(6.4)/7.3(6.0) 3.2(2.2)/3.8(2.5) 5.8(3.3)/6.9(3.1) 16.0(12.5)/16.6(12.3) 5.6(7.4)/6.5(7.2) 
Cross-sectional organic loading rate (gBOD5/m2 · d) 414(383)/429(361) 192(133)/226(150) 345(194)/412(184) 1,019(751)/1,090(736) 345(445)/388(433) 
Surface organic loading rate (gCOD/m2 · d) 17.2(10.7)/17.8(10.3) 10.8(7.8)/10.9(9.0) 12.1(8.8)/14.9(9.5) 40.1(16.6)/35.4(17.3) 15.1(12.9)/15.6(12.6) 
Surface solids loading rate (gTSS/m2 · d) 3.0(10.8)/3.3(10.9) 4.6(5.6)/4.9(5.2) 3.3(4.0)/4.6(5.3) 20.5(8.1)/22.4(8.8) 3.4(9.3)/3.8(9.4) 
Percentage of surface area covered with overland flow (%) 0–60/0–40 23/21 80/87 88/100 – 
Impact of clogging in terms of hydraulics No impact Deterioration Deterioration Collapse – 
Impact of clogging in terms of performance No impact No impact No impact Deterioration – 
Effluent concentration (mg/L) 
 BOD 20(19)/19(17) 10(10)/11(8) 19(15)/19(16) 88(51)/58(41) 19(28)/18(24) 
 COD 50(19)/52(25) 25(15)/27(17) 57(43)/67(43) 91(47)/102(41) 46(35)/52(38) 
 TSS 9(9)/6(8) 13(12)/10(12) 17(36)/10(30) 10(9)/26(22) 11(16)/8(15) 
Removal efficiency at wetlands stagec (%) 
 BOD 72(21)/72(24) 58(30)/65(33) 65(22)/63(23) 38(35)/50(39) 62(29)/66(30) 
 COD 73(13)/68(17) 70(22)/67(25) 62(25)/55(26) 50(33)/44(32) 70(24)/62(26) 
 TSS 77(29)/84(26) 66(38)/85(30) 71(54)/76(36) 70(29)/71(57) 76(34)/82(32) 
Overall efficiency: pre-treatment + wetlandsc (%) 
 BOD 89(13)/90(13) 94(4)/94(4) 91(4)/90(5) 84(11)/90(11) 91(10)/91(10) 
 COD 88(8)/89(9) 94(7)/92(14) 86(15)/79(22) 78(26)/73(21) 88(17)/86(18) 
 TSS 94(6)/94(5) 91(22)/95(11) 90(12)/91(25) 93(4)/95(8) 92(15)/94(12) 

UASB, upflow anaerobic sludge blanket reactor; TF, trickling filter; NM, not measured.

aBefore plant cutting.

bAfter plant cutting.

cCalculations based on loads. Values before parentheses are medians and inside parentheses () are standard deviations: median planted (standard deviation planted)/median unplanted (standard deviation unplanted).

Table 4 presents an overall summary of the main information related to the four operational phases in terms of loading conditions and treatment performance.

Vymazal (2018b) evaluated the performance of 17 HF CW operating between 20 and 27 years in the Czech Republic and concluded that the wetlands provided a very stable removal in terms of BOD5, COD and TSS, if the applied surface loading rates were below 5 gBOD5/m2 · d, 15 gCOD/m2 · d and 10 gTSS/m2 · d. Another conclusion was that the treatment performance was very stable, and no remodeling was necessary if the systems were not overloaded. In the current study (see Table 4), the applied loading rates of BOD and COD remained below those recommended by Vymazal (2018b) only during Phase 2. The applied TSS loads were below those recommended in Phases 1, 2 and 3.

The characterization of the material accumulated in the pores of the bed may assist in understanding the mechanisms associated with the clogging process. De Paoli & von Sperling (2013) performed a sampling in Phase 1 (after two years and four months of operation) and identified that the planted unit had higher concentrations of solids in the inlet zone, with values ranging from 122 to 113 g/L of total solids (interstitial and attached solids). At the unplanted unit the concentrations were much lower, from 43 to 48 g/L. In another sampling by Matos (2015), at the same zone in Phase 3, there was a reduction in solids concentrations only in the planted unit, with values of 21 and 52 g/L in the planted unit and 21 to 97 g/L in the unplanted unit. The unplanted unit showed a clear relationship between the reduction of hydraulic conductivity and the accumulated solids in Phase 1. De Paoli & von Sperling (2013) stated that the presence of solids in the unplanted unit was a predominant factor in the hydrodynamic changes, but in the planted unit there was a low variation. It was verified that the mass of clogging in the planted unit was predominantly made up of dead plant tissue, as was also observed by Pedescoll et al. (2011).

It may be seen that the hydraulic deterioration took place before leading to a deterioration in the system's performance. Even with a portion of the units having overland flow, this surface component could still be infiltrated inside the medium and undergo treatment under subsurface-flow conditions. Later on, hydraulic collapse (failure) was evident, based on the fact that the overland flow was covering most of the units' surface areas, leading to a deterioration in terms of performance, but which could not yet be considered a performance collapse. This shows the robustness of the wetland's treatment in terms of organic matter and suspended solids removal: even with adverse hydraulic conditions, conversion and removal processes still took place in an acceptable way, and only with extremely poor hydraulic conditions could an important deterioration in terms of performance be ascertained. Vymazal (2018a) obtained similar results, based on an evaluation of four HFs operating for more than 20 years and observed surface runoff over a long period of operation. However, surface runoff did not cause deterioration in effluent quality and system performance remained stable over the years.

When comparing the planted and unplanted units, their behaviour was very similar in terms of performance, as can be seen from Table 2 and Figure 2. Regarding their hydraulic behaviour, there were also similarities, although the unplanted unit had a stronger clogging in the first years of operation, but showed less overland flow in the final years.

CONCLUSIONS

The system performed very well for most of the period of more than 10 years of operation. The wetlands stage made a very important contribution to the overall system performance, with its high removal efficiencies and the low concentrations in the final effluent.

The planted and unplanted wetlands showed good performance, even under advanced conditions of clogging. The type of pre-treatment and the applied loads influenced more the performance of the units than the time of operation itself, evidencing the robustness of the system. A better performance was observed in the second operational phase, when the system received effluent from UASB reactor followed by trickling filter, while the worst performance was observed in the last phase, when the applied influent loads doubled and the units were heavily clogged.

Hydraulic deterioration followed by collapse, with the increase in the surface runoff due to clogging, came before the deterioration of the effluent quality, and was observed only in the last operational phase. To avoid collapse or even deterioration in constructed wetlands it is necessary to follow the recommended design, construction, operation and maintenance criteria.

ACKNOWLEDGEMENTS

The authors would like to thank the Brazilian agencies CNPq, CAPES, FAPEMIG and FUNASA, and IHE-Delft for their support to the research, COPASA for providing the site for the experimental units, CNPq for the scholarship of the first author and the Federal University of Minas Gerais.

REFERENCES

REFERENCES
APHA/AWWA/WEF
2005
Standard Methods for the Examination of Water and Wastewater
, 21st edn.
American Public Health Association
,
Washington, DC
,
USA
.
Costa
J. F.
2013
Remoção de poluentes em um sistema de alagados construídos atuando como pós-tratamento de efluente de reator UASB e de filtro biológico percolador (Pollutant Removal in a Constructed Wetland System Treating the Effluent From UASB Reactor and Trickling Filter)
.
PhD thesis
.
Federal University of Minas Gerais
,
Belo Horizonte
,
Brazil
(
in Portuguese
).
Fisher
P. J.
1990
Hydraulic characteristics of constructed wetlands at Richmond, NSW, Australia
. In:
Constructed Wetlands in Water Pollution Control
(
Cooper
P. F.
&
Findlater
B. C.
, eds).
Pergamon Press
,
Oxford
,
UK
, pp.
21
31
.
INMET
2018
National Institute of Meteorology
.
Normal climatological of Brazil (1981–2010). Available at: http://www.inmet.gov.br/portal/index.php?r=clima/graficosClimaticos
.
IWA Task Group
2017
Treatment Wetlands. Volume 7. Biological Wastewater Treatment Series. IWA Task Group on Mainstreaming the Use of Treatment Wetlands (Dotro, G., Langergraber, G., Molle, P., Nivala, J., Puigagut, J., Stein, O. & Von Sperling, M.). IWA Publishing, London. 154 pp. Available for download at: http://www.iwapublishing.com/open-access-ebooks/3567
.
Kadlec
R. H.
&
Wallace
R. D.
2009
Treatment Wetlands
, 2nd edn.
CRC Press
,
Boca Raton, Florida, USA
,
p. 435
.
Matos
M. P.
2015
Clogging in Subsurface Horizontal Flow Constructed Wetlands: Main Factors and Methods of Identification in Planted and Unplanted Units
.
PhD Thesis in Sanitation
,
Environment and Water Resources. Federal University of Minas Gerais
,
Belo Horizonte
,
Brazil
, p.
301
(in Portuguese)
.
Matos
M. P.
,
Von Sperling
M.
,
Matos
A. T.
&
Passos
R. G.
2015
Use of saline tracer for evaluating the clogging and hydrodynamic conditions in horizontal subsurface flow constructed wetland
.
Engenharia Agrícola
85
(
6
),
1137
1148
(in Portuguese)
.
Matos
M. P.
,
Barreto
A. B.
,
Vasconcellos
G. R.
,
Matos
A. T.
,
Simões
G. F.
&
Von Sperling
M.
2017
Difficulties and modifications in the use of available methods for hydraulic conductivity measurements in highly clogged horizontal subsurface flow constructed wetlands
.
Water Science and Technology
76
(
7
),
1666
1675
.
Reed
S. R.
,
Crites
R. W.
&
Middlebrooks
J. E.
1995
Natural Systems for Waste Management and Treatment
, 2nd edn.
McGraw-Hill
,
New York
,
USA
, p.
448
.
Tanner
C. C.
&
Sukias
J. P. S.
1995
Accumulation of organic solids in gravel-bed constructed wetland
.
Water Science and Technology
32
(
3
),
229
239
.
Vymazal
J.
2003
Distribution of iron, cadmium, nickel and lead in a constructed wetland receiving municipal sewage
. In:
Wetlands – Nutrients, Metals and Mass Cycling
(
Vymazal
J.
, ed.).
Backhuys Publishers
,
Leiden
,
The Netherlands
, pp.
341
363
.
Vymazal
J.
2018b
Is wastewater treatment in horizontal subsurface flow constructed wetlands a sustainable technology?
In:
Proceedings of the 16th IWA Specialist Conference on Wetland Systems for Water Pollution Control
,
30 September–4 October 2018
.
IWA
,
Valencia
,
Spain
, pp.
469
472
.
Wallace
S. D.
2014
Reducing wetland area requirements by using intensification strategies
. In:
Proceedings of the 14th IWA Specialist Group Conference on Wetland Systems for Water Pollution Control
,
4–8 October 2010
.
IWA
,
Shanghai
,
China
, pp.
54
68
.