- Email: [email protected]
Biological sand filter system treating winery effluent for effective reduction in organic load and pH neutralisation
Journal of Water Process Engineering 25 (2018) 118–127
Contents lists available at ScienceDirect
Journal of Water Process Engineering journal homepage: www.elsevier.com/locate/jwpe
Biological sand filter system treating winery effluent for effective reduction in organic load and pH neutralisation G.A. Holtmana,b, R. Haldenwangb, P.J. Welza,
T
⁎
a Biocatalysis and Technical Biology (BTB) Research group, Institute of Biomedical and Microbial Biotechnology, Cape Peninsula University of Technology, PO box 1906, Symphony way, Bellville, 7535, South Africa b Department of Civil Engineering, Cape Peninsula University of Technology, South Africa
A R T I C LE I N FO
A B S T R A C T
Keywords: Biological sand filter Chemical oxygen demand Constructed wetland pH neutralisation Treatment Organics Winery wastewater
Wineries generate 0.2–14 L of wastewater per litre of wine produced, which is often used for irrigation or discharged into aquatic systems. To mitigate adverse environmental impacts, there is a need for low cost wastewater treatment options. The novel biological sand filtration system described in this pilot study is a sustainable off-grid modular system which can be easily retrofitted to current infrastructure. The system was operated with average hydraulic and organic loading rates of 150 L m−3 sand.day-1 and 152 gCOD.m−3 of sand.day−1, respectively. Over 762 days of operation, average removal efficiencies of 79% and 77% in terms of chemical oxygen demand and total phenolic concentrations were achieved. In addition, an average 1.8-fold increase in the calcium concentration was achieved, with a concomitant reduction in the sodium adsorption ratio and similar indices. This pilot study also confirmed in a ‘real world’ setting the results of laboratory-based studies where biological sand filters neutralised acidic synthetic winery wastewater and reduced the organic load.
1. Introduction Wine production can be water-intensive. In addition to vineyard irrigation requirements, an estimated 0.2–14 L of water is needed to produce 1 L of bottled wine [1–3]. Most of this water leaves the cellar in the form of winery wastewater (WWW) that is typically acidic, and contains organic and inorganic fractions associated with seasonal cleaning activities [3,4]. The character and volume of the WWW varies on a temporal basis, with a pattern of high organic to inorganic load during the crush season, and vice-versa during the non-crush period [5]. In many countries, including South Africa, Australia, and the United States, WWW is often re-used for irrigation of pastures or crops in water-stressed areas; unless adequately treated, WWW may pose a threat to the soil environment and/or groundwater [6–8]. Many treatment processes have been researched and/or applied for the treatment of WWW, including: physicochemical processes, biological processes, membrane filtration and separation processes, and advanced oxidation processes [9]. Membrane systems have not been widely adopted by the wine industry because of the high capital outlay
and operating costs, the propensity for membrane fouling, and brine generation [10,11]. Physicochemical processes rely on either sedimentation, precipitation, coagulation/flocculation, or electrocoagulation [9]. Currently, the most effective physicochemical and advanced oxidation processes require skilled to semi-skilled labour, chemical inputs, create solid waste streams, and need to be used in conjunction with biological treatment methods to achieve satisfactory organic removal rates [10,11]. While biological systems such as rotating biological contactors, upflow anaerobic sludge blankets and membrane bioreactors may be suited to larger wineries for biodegradation of WWW organics, there is a global need for simple, low maintenance, cost effective systems at small wineries that do not have the finances or personnel to operate complex systems [9]. Many wineries utilise anaerobic or aerated ponds. These are simple and cheap to operate. However, odour problems are associated with ponds, and long hydraulic retention times (HRTs) are required, which translates into a large spatial footprint [12,13]. Constructed wetlands (CWs) are an effective option if the WWW is diluted or mixed with domestic wastewater (WW) [14,15]. The main drawback of CWs is that (poly)phenolic-rich effluent such
Abbreviations: WWW, winery wastewater; WW, wastewater; HRT, hydraulic retention time; CW(s), constructed wetland; OMWW, olive mill wastewater; COD, chemical oxygen demand; BSF(s), biological sand filter; k, hydraulic conductivity; VOA(s), volatile organic acid; AAC, acetic acid equivalents; SAR, sodium adsorption ratio; CROSS, cation ratio of soil structural stability; Q, flow rate; T, time; V, volume; OLR, organic loading rate ⁎ Corresponding author. E-mail addresses: [email protected] (G.A. Holtman), [email protected] (R. Haldenwang), [email protected] (P.J. Welz). https://doi.org/10.1016/j.jwpe.2018.07.008 Received 9 May 2018; Received in revised form 16 July 2018; Accepted 23 July 2018 2214-7144/ © 2018 Elsevier Ltd. All rights reserved.
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
as olive mill wastewater (OMWW) and WWW is phytotoxic [14,16–19]. For example, Shepherd et al. [19] demonstrated that gravel-filled horizontal subsurface flow CWs were able to achieve 97% COD removal when treating WWW with chemical oxygen demand (COD) concentrations < 5000 mg L−1, but poor removal efficiency was achieved and plants died at COD concentrations of 12 800–16800 mg L−1. It can also be argued that the inclusion of plants in CWs treating WWW is unnecessary because they play a limited role in reducing COD and suspended solids, they require periodic harvesting, and their roots can facilitate the formation of preferential flow paths [16,20]. Biological sand filters (BSFs) can be seen as unplanted CWs. The sand creates a physical substrate for biofilm attachment, can adsorb pollutants, catalyse chemical transformations, and provide metabolic co-factors; the functional microorganisms within the biofilm are responsible for biotransformation, biodegradation and mineralization of chemical pollutants [24]. In a series of fundamental laboratory-based experiments, it was shown that simple BSFs containing ungraded locally-available dune sand [21] were able to cope with the temporal nature of WWW and achieve reliable removal of organics, including (poly)phenolics [22,23]. To determine whether BSFs are viable for the treatment of WWW is a ‘real-world’ setting, a pilot system was designed, installed and operated at a wine farm in the Stellenbosch area close to Cape Town, South Africa. This manuscript details and analyses the performance of the system over approximately 27 months.
2.2.1. Determination of chemical oxygen demand, nutrients, and selected organic fractions The COD, total phosphate and total nitrogen concentrations were determined using a Merck (Merck®, Whitehouse Station, USA) Spectroquant® Pharo instrument and Merck Spectroquant® COD cell tests for low, medium and high range samples (cat. no. 1.14895.0001, 1.14541.0001 and 1.14691.0001) and total nitrogen cell tests 1.14543.0001 and total phosphate as PO4-P 1.14537.0001 according to manufacturer’s instructions., according to the manufacturer’s instructions. The total phenolic concentrations were determined using the Folin-Ciocalteu micro method based on that described by Slinkard and Singleton [25] using Merck®Folin–Ciocalteu reagent (Cat No: 1.09001.0500). The concentrations of volatile organic acids (VOAs) were determined using the Hach (Loveland, USA) esterification method (cat no 8196) in accordance with manufacturer’s instructions with some modifications: 3 standard concentrations of acetic acid were prepared (945.00 mg L−1, 472.50 mg L−1 and 236.25 mg L−1) and were used to prepare a standard graph for the determination of the VOA concentrations in acetic acid equivalents (mgAAC.L−1).
2. Materials and methods
2.2.3. Sodium adsorption ratio, cation ratio of soil structural stability Concentrations of sodium, potassium, calcium and magnesium were determined using a Varian® MPX inductively coupled plasma optical emission spectrophotometer (Agilent Technologies, Santa Clara, USA) at Bemlab (Pty) Ltd. (Strand, South Africa). The sodium adsorption ratio (SAR) and cation ratio of soil structural stability (CROSS) were determined using the equations described by Marchuk and Rengasamy [26] and Oster et al. [8].
2.2.2. Determination of pH The pH of the samples was determined according to the manufacturer’s instructions using a CyberScan pH300 meter and appropriately calibrated pH probe PHWP300/02 K (Eutech instruments, Singapore).
2.1. Set-up and operation of pilot scale biological sand filtration system The BSF system consisted of a number of connected polyethylene (PE) containers: a 5000 L collection tank, a 500 L holding tank, 4 × 1000 L sand filter modules, and 4 × 100 L flow-control tanks. The system was connected to an existing 45 m3 baffled concrete solidssettling delta (Fig. 1A). A fraction of the WWW from the delta was treated and returned to the head of the delta, thereby improving the quality of the WWW which was used to irrigate a sheep pasture. The WWW from the cellar gravitated into the delta inlet and settled effluent was pumped from the delta outlet to the 5000 L collection tank using a Shurflo (Pentair, Minneapolis, USA) 2088-313-145 12 V DC diaphragm premium demand solar pump. The pump was controlled by a liquid level relay and a probe within the 5000 L collection tank, which was set to fill the collection tank to 3800 L when the volume dropped to 2500 L. The flow through the remainder of the system was controlled by gravity, float valves and adjustable flow-control tanks (Fig. 1B). The float valves and flow control tanks enabled the HLR to be automatically altered when the flow rates through the filters increased or decreased. A constant head of 30 cm was maintained across each filter module. The modules were filled with a locally available dune sand, which had previously been well characterised and found to be suitable for use in BSF in terms of treatment capacity and hydraulic conductivity (k) [21]. The system could be operated in parallel or series, and was operated in parallel during the experimental period.
2.3. Calculation of operational parameters 2.3.1. Flow rates and volume of wastewater treated The flow rate was determined by measuring the volume of water collected from the outlet over the period of one minute. Measurements were taken in triplicate, averaged, and converted into the daily flow rate (Q). Q was used to determine the volume of WWW treated and was estimated using 2 methods: (i) the average of all the measured flow rates (Table 1) and (ii) a trapezoidal equation Eq. (1) that assumed a linear decrease or increase in Q between measuring instances (t).
∑ Voln = Qn−1 (tn−1) + 0.5[(Qn−Qn−1 )(tn−1−tn)]
(1)
2.3.2. Hydraulic retention time, hydraulic conductivity, hydraulic loading rate, organic loading rate The saturated cross-sectional area at the discharge point was used in combination with the change in height (h) divided by the change in length (l) to determine k in terms of Darcy’s law Eq. (2) [27,28]. The HRT was determined using the volume (V) of liquid within a packed media divided by the flow rate as shown in Eq. (3) [29], while the volume of liquid was determined by multiplying the porosity of the substrate by the volume of the reactor. The HLR was expressed in two different ways: (i) as the discharge rate of influent divided by the volume of the reactor (L m−3 d-1) Eq. (4), and (ii) as the discharge rate divided by the cross-sectional area of the reactor (L m-2 d-1), which is the formula typically applied for reactors operated in horizontal mode Eq. (5). The organic loading rate (OLR) was determined by multiplying the influent flow rate by the influent COD or BOD concentration divided by the volume of reactor [29] Eq (6).
2.2. Sampling and characterisation of influent and effluent Influent was defined as the grab WW samples from the 5000 L collection tank, and the effluent as the final effluent (Fig. 1). During the final year of the study, the sampling regime was increased for more accurate assessment of BSF performance (Table 1). In addition to regular monthly or bi-weekly analyses, a once-off batch test was performed, which entailed hourly sampling (Table 1). In three scheduled sampling instances, no samples could be taken due to operational problems (Dec 2015, Jan 2016, Mar 2016).
119
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 1. Set-up of biological sand filtration system showing the connection to the existing settling delta (A) and the operational components and sampling points of the system (B).
The functional (saturated) volume of sand was calculated using a formulated trapezoidal equation to determine the cross-sectional area of the filter Eq. (7). The cross-section was calculated by drawing a straight line between the height of the saturated zone at the inlet and the outlet. The height of the saturated zones were determined by inserting pipes into the sand.
Q = −k A
dh dl
HRT =
(2)
120
Vliquid Q
(3)
HLRV =
Q Vreactor
(4)
HLRA =
Q A Cross section
(5)
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Table 1 Sampling schedule and analytical parameters. Sampling points
Period
Occurrence
Analyses
Collection tank (influent), holding tank, filter outlets, final effluent Influent and effluent
Aug 2015
Once-off
COD, total phenolics, VOA, pH, sodium, potassium, calcium, magnesium, flow rate (Q)
Oct 2015-Feb 2016 Mar 2016-Mar 2017
Monthly
COD, total phenolics, VOA, total phosphorus, total nitrogen, pH, sodium, potassium, calcium, magnesium, flow rate (Q) COD, total phenolics, VOA, total phosphorus, total nitrogen, pH, sodium, potassium, calcium, magnesium, flow rate (Q)
Bi-weekly
COD = chemical oxygen demandVOA = volatile organic acids. Table 2 Operation parameters and organic removal rates in comparison to similar systems treating winery or olive mill wastewater. System
Wastewater (type)
Influent COD (mg L−1)
COD RE (%)
HLR (L/m3.day−1)
OLR (gCOD/m3.day−1)
tHRT (days)
1. Experimental SF 2. Full scale HSSF CW
OMWW & DW WWW
3. Experimental BSF 4. Full-scale CW VF 5. Full-scale 2-stage CW
WWW WWW & DW WWW & DW WWW
70–90 60 80 98 98.9 VF: 29–70 HF: 13–79 28-98
5.0–68.2a 23a 45a 7.1a 22a VF: 55–154a HF: 1.3–3.6a 150
2102.6–54.1a 107–218 35–277a 16.4a 58a VF: 30.7–332.9a HF: 0.4–5.5a 152
7–101a 7 14 22 NG NG
This study: Pilot BSF
30830 ± 1690 1: 2000–12000 2: 2000–12000 2304 ± 629 2580 VF: 1558 ± 1023 HF: 711 ± 769 1138
1.8
RE = removal efficiency HLR = hydraulic loading rate OLR = organic loading rate HRT = theoretical hydraulic retention time SF = sand filter CW = constructed wetland SF = biological sand filter HSSF = horizontal subsurface flow VF = vertical Flow OMWW = olive mill wastewater WWW = winery wastewater DW = domestic wastewater NG = Not given. References: 1. Achak et al. [16]; 2. Mulidzi [42]; 3. Ramond et al. [43]; 4. Rozema et al. [15]; 5. Serrano et al. [13]. a Calculated from information provided in journal article.
Fig. 2. Comparisons of the hydraulic and organic loading rates from October 2015 to October 2017.
OLR =
CODInfluent × Q Vreactor
A = 1070x−10925 [mm2]
the system is still fully functional to date (July 2018). (6) 3.1. Operational parameters
(7)
3.1.1. Hydraulic capacity In wineries, WWW production spikes during the crush season, with small spikes at other times of the year when cellar activities such as bottling are taking place. Two methods (2.3.1) were used to estimate the average daily volume of WWW that was treated [427 L.day−1 and 402 L.day-1 (137 L m-3 of sand−1)]. The latter values (calculated using Eq. (1), were used in all subsequent calculations. The pilot system was not intended to treat the entire volume of WWW produced by the cellar. Nevertheless, 22% was treated. In addition, the treated effluent was circulated back into the system, thereby improving the quality of all the WWW used for irrigation by diluting the fresh WWW entering the delta with final effluent from the BSF system.
3. Results and discussion The BSF system is intended to improve the suitability of WWW for irrigation or discharge to the environment. The system was commissioned in January 2015, and WWW was slowly added and drained intermittently over a 3 month start-up period in order to allow the microbial communities to acclimate. This incremental priming with WWW during start-up has been shown to increase the treatment performance in experimental BSFs [30]. A batch test was performed in August 2015, and regular sampling commenced in October 2015. The operational results over a further 24 months are presented in this manuscript, but 121
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 3. The chemical oxygen demand (A), volatile organic acid (B) and total phenolic (C) concentrations measured in influent and effluent samples, and respective removal efficiencies.
porous WW treatment systems [33]. Some researchers have developed models to predict the extent of changes in k but none have been properly validated [e.g. 33,34]. This is probably because the extent of biomass accumulation differs according to the type of physical substrate, the type of effluent being treated, and in different spatial locations within the systems [35]. In a previous study using sand-filled columns treating synthetic WWW, a 51% reduction in the k of dune sand from an initial value of 0.284 mm s−1 was determined experimentally in column experiments using the constant head method [21]. As no solids were present in the synthetic WWW, the reduction was attributed to biomass alone. In this pilot study, there was an 80% reduction in the calculated k of the dune sand from the initial value of 0.199 mm s−1, to the operational average of 0.040 mm s−1. The start-up HRT for a single BSF module of 0.53 days was calculated using the initial flow rate (Q = 1 900 L day−1) and porosity values (0.292 ± 0.02) determined experimentally for the ‘fresh’ dune sand. Thereafter, the initial porosity and the average system flow rate
In terms of size, SAWIS [31] broadly divides wineries into 6 categories depending on the amount of grapes crushed each year (< 100 tons ≥ 500 tons ≥ 1000 tons ≥ 10,000 tons <). The infrastructure of the BSF system shown in Fig. 1 can support 1 to 12 filter modules, after which the capacity (and footprint) would need to be increased. Using Eq. (2) combined with a safety factor of 1.1 [32], it was calculated that on an annual basis the infrastructure of the pilot BSF system is capable of treating the WWW from a cellar crushing 47 tons of grapes, which could be extended to a cellar crushing 141 tons by maximising the number of filters to 12.
Water = 4037.5 × T 0.9243
(8)
3.1.2. Hydraulic conductivity and hydraulic retention time Decreases in permeability and porosity, and changes in mass transport parameters due to the build-up of functional biomass and solids retention cause increased k leading to decreased HRT in free-flow 122
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 4. The pH measured in influent and effluent samples.
(Q = 402 L.day−1) was used to calculate the average operational HRT of a single BSF module [2.1 days, adjusted to 1.8 days to account for partial saturation of the sand, and 11.5 days when including the collection tank and other infrastructure in the calculation]. This HRT was significantly shorter than the HRT of other biological systems used to treat WWW (Table 2), but did not account for any decrease in porosity. It was not possible to accurately determine the operational k values because the changes in porosity were in a state of flux due to the constant formation and degradation of functional biomass.
state of flux (Fig. 2). Overall, the pilot BSF system demonstrated resilience to synthetic WWW fluctuations, which was also demonstrated in experimental systems treating synthetic WWW [22]. 3.2. Performance evaluation 3.2.1. Introduction The variable character of WWW may lead to temporal sampling bias when collecting influent and effluent [22]. In this study, the incoming WWW was balanced in the collection and holding tanks. The HRT within the BSF filters was only 1.8 days (Section 3.1.1). It was therefore expected that there would be no significant sampling bias and shortterm temporal system fluctuations. To confirm that the measured parameters were a true reflection of transformations effected within the BSFs (for example, removal efficiencies), a once-off batch test was performed. This consisted of taking samples every hour for 4 h (5 sampling instances; Supplementary material 1). For the batch study, no significant temporal differences (ANOVA: p > 0.5) were seen in the conductivity, COD, VOA, and total phenolic measurements in: (i) the t = 0 to t = 4hr influent samples, (ii) the t = 0 to t = 4hr effluent samples (iii) the t = 0 to t = 4hr samples taken separately from each of the individual BSF modules. This indicated that the performance assessment results were valid. However, when comparing the results from the different BSF modules, a significant difference in the COD and VOA measurements (ANOVA: p < 0.05) was noted (data not shown). Although the modules were set up in an identical manner and received the same WWW, inter-BSF module performance variability in terms of organic degradation was demonstrated.
3.1.3. Hydraulic and organic loading rates The HLR and OLR in CWs are typically calculated using the surface or cross-sectional area, but the total volume gives a better indication of the treatment efficiency of the medium [21]. For example, a slender vessel may have a high surface loading rate but relatively low overall loading rate, while the converse is true for a wide, shallow vessel. The initial HLR values for overall and cross-sectional loading of the sand filters were 646 L m−3sand.day−1 and 1084 L m−2sand.day−1, respectively. In comparison, the average operational values were 150 L m−3sand.day−1 (range: 12–313 L.m−3sand.day−1) and 252 L.m−2 sand.day−1 (range: 21-526 L m−2sand.day−1), respectively. The 77% decrease from initial to operational HLR was due to accumulation of bio-solids, including functional biomass, and was accompanied by increased COD removal rates (3.2). Over a 10 week period, Achak et al. [16] also described a reduction (50%) in the HLR, and improvement of organic removal rates in a vertical flow sand filter treating diluted OMWW with a starting HLR of 68.2 L.m3sand.day-1. In this case, the authors terminated the study because they felt that the HLR of 5.0 L m3sand.day−1 was inadequate. Some degree of biomass clogging is essential, because the microbial population is the most important functional component for organic biodegradation [36,37]. In reality, there is a balance between the HLR and system performance. If the HLR is too high, treatment may be inefficient, but if it is too low, the spatial footprint and capital costs may render the system uneconomical. High HLRs can be overcome by setting up systems that allow hydraulic flexibility, and strategies such as intermittent operation and adjustment of loading rates have been proposed to minimise biomass clogging and increase the HLR [16,36–39]. The variable nature of the OLR seen in this study resulted from the WWW being generated from different seasonal cellar activities, and is typical of all wineries. The average OLR over the study period was 152 gCOD m−3sand.day−1 (range: 23–469 gCOD m−3sand.day-1), or 256 gCOD m−2sand.day−1 (range: 39–788 gCOD m−2sand.day−1) for the total volume and surface area of the sand, respectively. After the start-up period, the HLR also fluctuated, which suggested that the accumulation and degradation of organic solids and/or biomass was in a
3.2.2. Removal of chemical oxygen demand and selected organics The system achieved an average COD removal efficiency of 79% (range: 28–98%). The average COD concentration in the final effluent (n = 35) was 287 mg L−1 (range: 24–1382 mg L−1, Fig. 3A). VOAs and ethanol constitute the highest fraction of the COD in WWW [5]. VOAs can also be formed in BSFs from other organic molecules, including phenolics, sugars and ethanol [22]. In this study, the average VOA concentrations in the influent and effluent were 472.8 mgAAE.L−1 (range: 9.4–1018.4 mgAAE.L−1) and 186.3 mgAAE.L−1 (range: 0.2–662.0 mgAAE.L−1), respectively, with ‘removal’ rates of -93% to 99.9% (Fig. 3B). The intermittent negative removal rates indicated formation of VOAs within the system. Phenolics in OMWW and WWW may be toxic to plants and microbes and should therefore be removed/reduced before discharge of the WW to the environment [16,40]. Removal of these aromatic molecules in BSFs takes place via biotic and abiotic mechanisms, and it has been shown that over 88% removal of (poly)phenolics with COD 123
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 5. The sodium and calcium concentrations (A), pH and change in calcium concentration (B), and sodium adsoption and cation ratio of soil structural stability ratios (C) measured in influent and effluent samples.
concentrations of 5842 mg.L−1 can be achieved [24]. Achak et al. [16] also demonstrated similar removal of phenolics from OMWW (87–97%) with an influent COD concentration of 4590 mg L−1. In this study, the average total phenolic concentration in the influent and effluent were 18.4 mgGAE.L−1 (range: 5.1–43.7 mgGAE.L−1) and 3.9 mgGAE.L−1 (range: ≤0.09–16.6), respectively, with average removal rates of 77% (range: 16–100%) (Fig. 3C). In comparison to other simple WWW treatment systems, the system performed well in terms of COD removal (Table 2). Montalvo et al. [41] used two adjacent pilot-scale batch-fed aeration lagoons to treat 790 L.day−1 and 170 L.day−1 of WWW with a similar OLR to the pilot BSF. Both ponds removed 91% of the COD, but the system required aeration, which is costly. Mulidzi [42] achieved 60% and 80% COD
removal for a CW with retention times of 7 and 14 days, respectively. However, sub-surface flow CWs are more complex, require additional maintenance, and WWW is potentially toxic to CW plants [16,19]. Rozema et al. [15] co-treated domestic and WWW (ratio 1:3) in a vertical flow CW and achieved 99% COD removal. The system operated at a relatively low OLR compared to the BSF in this study. Serrano et al. [13] also co-treated WWW with domestic WW, with a COD reduction of 73% across a two-stage CW fed by a hydrolytic upflow sludge bed followed by a horizontal flow CW. It is difficult to compare these systems with the BSF used in this study because of the presence of domestic WW. Domestic WW has a high carbon to nitrogen ratio in comparison to WWW. Therefore, an advantage of this type of cotreatment is that the domestic WW may enhance microbial activity 124
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 6. The total nitrogen and total phosphorous concentrations measured in influent and effluent samples.
concentrations of potassium were 121.7 mg L−1 (range: −1 −1 17.3–284.9 mg L ) and 126.8 mg L (range: 25.9–302.9 mg L−1), respectively (data not shown). The average sodium concentrations of the influent and effluent samples (Fig. 5A) were 29.6 mg L−1 (range: 10.6–70.8 mg L−1) and 29.8 mg L−1 (range: 10.0–62.9 mg L−1), respectively. In CWs containing clay, salts may be removed by physicochemical processes [53], but in this instance the sand did not contain clay particles, and the data showed that neither magnesium, potassium, nor sodium were removed by the system. In contrast, the calcium concentration (Fig. 5A) showed an average 1.8 fold increase in concentration from influent to effluent of 36.2 mg L−1 (range: 96.6–104.0 mg L−1) to 65.7 mg L−1 (range: 14.7–200.3 mg L−1). This was attributed to the dissolution of calcite as discussed in 3.2.3. There was a highly significant (p < 0.005) negative correlation (-0.37) between the pH of the influent samples and the change in calcium concentration from influent to effluent (Fig. 5B). This strongly suggested that the calcium dissolution rates increased with an increase in acidity. The BSF system achieved an average decrease in the SAR (13%) and CROSS (12%) from influent to effluent (n = 31) (Fig. 5C). This decrease in SAR and CROSS improved the quality of the final effluent for soil application. The CROSS values in the effluent were all below the limit of 20 suggested by Laurenson et al. [48] for irrigation with WWW. Crops such as banana (Musasapientum) and lucerne (Medicago sativa) have a remarkable ability to accumulate potassium [54]. Recent studies have also shown the ability of sugar beet to take up Na [55]. It is envisaged that WWW should be used for beneficial irrigation of these, or other halophytic crops after treatment in BSFs.
because of the addition of nutrients to the WWW. A disadvantage is the potential introduction of pathogens. 3.2.3. Neutralization of acidic winery wastewater The pH in the influent ranged from 4.55 to 7.95, while the pH of the effluent ranged from 6.63 to 8.69 (Fig. 4). Even the formation of VOAs in the system did not lead to an increase in the acidity of the effluent. This effective neutralisation of the acidic WWW was attributed to both abiotic and biotic mechanisms: The mineral calcite (formed from calcium carbonate) was present in the dune sand, with calcium constituting 17.6% (wt.wt) of the sand [21]. In this study, there was an average increase of 82% from the influent to the effluent which provided evidence of calcite dissolution as an abiotic neutralisation mechanism [44]. WWW was also neutralised in experimental BSFs containing river sand that did not contain calcite or aluminosilicates [43], strongly suggesting that biotic neutralisation mechanisms also exist. 3.2.4. Inorganics Sodium hydroxide is widely used as a cleaning agent and disinfectant in wineries and WWW can contain high concentrations of sodium [21]. The cation binds to one negatively charged soil particle, causing a tight arrangement which reduces infiltration of water into soils [45]. Potassium has a similar effect, but not to the same magnitude [46]. Calcium and magnesium are divalent cations, which bind to more than one negatively charged soil particle, creating a matrix within the soil which allows for aeration and infiltration [45,47]. The presence of divalent cations can therefore offset the detrimental effects of monovalent cations such as sodium. The SAR is a parameter that is widely used to assess the potential of water to cause sodicity, and takes into account the concentration of sodium, calcium and magnesium [48]. According to the U.S. Salinity Laboratory [49], soil with an SAR > 13 is considered sodic [50], and Horneck et al. [51] state that soils with a SAR < 5, 5–13, and > 13, pose low, medium and high risks, respectively. However, the SAR does not take into account the effect of potassium on soil dispersion, which is relevant for WWW that can contain elevated levels of potassium, sodium, calcium and magnesium [45]. The CROSS is a sodicity indicator that was derived by Rengasamy and Sumner [52] to overcome this limitation. This ratio is more suitable for monitoring the effects of WWW on the soil structure [45]. Laurenson et al. [48] suggested a maximum CROSS of 20 for disposal of WWW to the environment. The authors found an average CROSS of 9.2 (range: 2.5–13.3) in WWW from 8 different wineries in Australia. The average influent and effluent concentrations of magnesium were 4.5 mg L−1 (range: 1.6–10.2 mg L−1) and 4.5 mg L−1 (range: 1.3–10.1 mg L−1), respectively, while the average influent and effluent
3.3. Major nutrients The average total nitrogen (N) and phosphate (P) (Fig. 6) in the influent was 1.6 mg L−1 (range < 0.5-10.4 mg L−1) and 2.4 mg L−1 (range: 0.86–3.98 mg L−1), respectively. WWW is known to be deficient in N and P [5], and these may be added to assist bioremediation [56–58]. For example, researchers have dosed raw WWW with nutrients to attain COD:N:P ratios of 502:5:1 [59], 400:7:1 [12] and 3060:2:1 [60]. In this study, no nutrients were added. The influent COD:N:P ratio was 1138:1.6:2.4 (n = 17), exhibiting a particular N deficiency [high carbon (C):N ratio]. It has been shown that bacterial N fixation can take place in systems treating high C:N WW, such as that from the paper and pulp industry [61] and synthetic WW [62]. This could explain why the COD removal efficiency in the BSF system was high despite the nutrient limitation in the influent.
125
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
flow constructed wetlands, Ecol. Eng. 57 (2013) 205–209. [21] P.J. Welz, W. Mbasha, I. Smith, G. Holtman, G. Terblanche, M. Le Roes-Hill, R. Haldenwang, The influence of grain physicochemistry and biomass on hydraulic conductivity in sand-filled treatment wetlands, Ecol. Eng. 116 (2018) 21–30. [22] P.J. Welz, M. Le Roes-Hill, Biodegradation of organics and accumulation of metabolites in experimental biological sand filters used for the treatment of synthetic winery wastewater: a mesocosm study, J. Water Process Eng. 3C (2014) 155–163. [23] P.J. Welz, Z. Palmer, S. Isaacs, B. Kirby, M. Le Roes-Hill, Analysis of substrate degradation, metabolite formation and microbial community responses in sand bioreactors treating winery wastewater: a comparative study, J. Environ. Manage. 145 (2014) 147–156. [24] P.J. Welz, J.-B. Ramond, D.A. Cowan, S.G. Burton, Phenolic removal processes in biological sand filters, sand columns and microcosms, Bioresour. Technol. 119 (2012) (2012) 262–269. [25] K. Slinkard, V.L. Singleton, Total phenol analysis: automation and comparison with manual methods, Am. J. Enol. Vitic. 28 (1977) 49–55. [26] A. Marchuk, P. Rengasamy, Threshold electrolyte concentration and dispersive potential in relation to CROSS in dispersive soil, Soil Res. 50 (2011) 473–481. [27] H. Darcy, Les fontaines publiques de la ville de Dijon, Dalmont, Paris, 1856. [28] E. Wilson, Engineering Hydrology, 4th ed., Macmillan Press, London, U.K, 1990. [29] Metcalf and Eddy, Inc, Wastewater Engineering: Treatment, Disposal and Reuse, 4th edition, The McGraw-Hill Companies, New York, New York, 2004. [30] P.J. Welz, J.B. Ramond, D.A. Cowan, A. Prins, S.G. Burton, Ethanol degradation and the value of incremental priming in pilot scale constructed wetlands, Ecol. Eng. 37 (2011) 1453–1459. [31] SAWIS, South African Wine Industry Statistics (No.42), [Online] Available: http:// www.sawis.co.za/info/download/Book_2017_statistics_year_english_final.pdf [2018, July 13] (2017). [32] C. Sheridan, A Critical Process Analysis of Wine Production to Improve Cost Efficiency, Wine Quality and Environmental Performance. Unpublished Masters of Science Thesis, University of Stellenbosch, Stellenbosch, 2003. [33] M. Peszynska, A. Trykozko, G. Iltis, S. Schlueter, D. Wildenschild, Biofilm growth in porous media: experiments, computational modeling at the porescale, and upscaling, Adv. Water Resour. 95 (2016) (2016) 288–301. [34] R. Samsó, J. Garcia, P. Molle, N. Forquet, Modelling bioclogging in variably saturated porous media and the interactions between surface/subsurface flows: application to constructed wetlands, J. Environ. Manage. 165 (2016) 271–279. [35] A. Caselles-Osorio, J. García, Performance of experimental horizontal subsurface flow constructed wetlands fed with dissolved or particulate organic matter, Water Res. 40 (19) (2006) 3603–3611. [36] A. Juwarkar, S. Singh, A. Mudhoo, Comprehensive overview of elements in bioremediation, Rev. Environ. Sci. Biotechnol. 9 (3) (2009) 215–288. [37] K. Semple, L. Doick, K. Wick, H. Harms, Microbial interactions with organic contaminants in soil: definitions, processes and measurement, Environ. Pollut. 150 (1) (2007) 166–176. [38] P. Knowles, G. Dotro, J. Nivalla, J. Garcia, Clogging in subsurface-flow treatment wetlands: occurrence and contributing factors, Ecol. Eng. 37 (2011) 99–112. [39] J.W. McKinley, R.L. Siegrist, soil clogging genesis in soil treatment units used for onsite wastewater reclamation: a review, Environ. Sci. Technol. 41 (24) (2011) 2186–2209. [40] C. Sheridan, D. Hildebrand, D. Glasser, Turning wine (waste) into Water: toward technological advances in the use of constructed wetlands for winery effluent treatment, AIChE J. 60 (2) (2014) 420–431. [41] S. Montalvo, L. Guerrero, E. Rivera, R. Borja, A. Chica, A. Martin, Kinetic evaluation and performance of pilot-scale feed-batch aerated lagoon treating winery wastewater, Bioresour. Technol. 101 (2010) (2010) 3452–3456. [42] A.R. Mulidzi, Winery and distillery wastewater treatment by constructed wetland with shorter retention time, Water Sci. Techno. 61 (10) (2010) 2611–2615. [43] J.-B. Ramond, P.J. Welz, M.I. Tuffin, S.G. Burton, D.A. Cown, Assessment of temporal and spatial evolution of bacterial communities in a biological sand filter mesocosm treating winery wastewater, J. Appl. Microbiol. 115 (2013) (2013) 91–101. [44] O. Sulpis, C. Lix, A. Mucci, B.P. Boudreau, Calcite dissolution kinetics at the sediment-water interface in natural seawater, Mar. Chem. 195 (2017) (2017) 70–83. [45] P. Rengasamy, A. Marchuk, Cation ratio of soil structural stability (CROSS), Soil Res. 49 (3) (2011) 280–285. [46] M. Arienzo, E.W. Christen, W.C. Quayle, A. Kumar, A review of the fate of potassium in the soil-plant system after land application of wastewaters, J. Hazard. Mater. 164 (2009) 415–422. [47] F. Ranjbar, M. Jalali, The combination of geostatistics and geochemical simulation for the site-specific management of soil salinity and sodicity, Comput. Electron. Agric. 121 (2016) 301–312. [48] S. Laurenson, D. Houlbrooke, T. Styles, Determination of Soil Dispersion in Response to Changes in Soil Salinity Under Winery Wastewater Irrigation, [Online]. Available: http://www.marlborough.govt.nz/sitecore/shell/Controls/Rich%20Text %20Editor/∼/media/Files/MDC/Home/Environment/Land/Soils/AgResearch_ Client_Report_for_Marlborough_District_Council.ashx [2015, February 8] (2012). [49] U.S. Salinity Lab. Staff, Diagnosis and Improvement of Saline and Alkali Soils. Agricultural Handbook No. 60, United State Department of Agriculture, 1954. [50] B.D. Seelig, Salinity and Sodicity in North Dakota Soils. EB-57, North Dakota State University, Fargo, ND, 2000. [51] D.A. Horneck, J.W. Ellsworth, B.G. Hopkins, D.M. Sullivan, R.G. Stevens, Managing Salt-Affected Soils for Crop Production. Oregon State University, November 2007, Oregon State University, 2007, p. 24 Pacific Northwest Extension publication. [52] P. Rengasamy, M.E. Sumner, M.E. Sumner, R. Naidu (Eds.), Processes Involved in Sodic Behaviour, Sodic Soils. Distribution, Properties, Management, and
4. Conclusion This study showed that the BSF system is a viable option for the treatment of WWW at small cellars. The capex and running costs were low, the system was sustainable, required minimal maintenance, and did not need skilled labour to operate. Future designs will focus on creating greater treatment capacity and removing salts. For example, different modes of operation will be tested (continuous v/s batch mode, vertical flow), and the treated effluent will be used for beneficial irrigation of salt accumulating crops in order to reduce the sodium content. Acknowledgements The authors would like to thank the National Research Foundation of South Africa for funding this project (P. Welz grant number CSUR13091742638), Jacques Rossouw and Reckson Mulidzi from the Agricultural Research Council and Distell, respectively, for assistance with site selection, and the (unnamed) winery involved. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jwpe.2018.07.008. References [1] D. Bolzonella, D. Rosso, Narbonne, France, Proceedings of the 6th IWA Conference on Viticulture and Winery Wastes (2013). [2] M. Oliveira, E. Duarte, Treatment and use of non-conventional organic residues in agriculture, RAMIRAN International Conference, Lisboa, Portugal, 2010, pp. 12–15. [3] A.G. Vlyssides, E.M. Barampouti, S. Mai, Wastewater characteristics from Greek wineries and distilleries, Water Sci. Techno. 51 (2005) 53–60. [4] A. Bories, Y. Sire, Impacts of winemaking methods on wastewaters and their treatment, S. Afr. J. Enol. Vitic 31 (2010) 38–44. [5] P.J. Welz, G. Holtman, R. Haldenwang, Le Roes-Hill, Characterisation of winery wastewater from continuous flow settling basins and waste stabilisation ponds over the course of 1 year: implications for biological wastewater treatment and land application, Water Sci. Technol. 74 (9) (2016) 2036–2050. [6] D.R. Hirzel, K. Steenwoerth, S.J. Parikh, A. Oberholster, Impact of winery wastewater irrigation on soil, grape and wine composition, Agric. Water Manage. 180 (2017) 178–189. [7] K.P.M. Mosse, A.F. Patti, R.J. Smernik, E.W. Christen, T.R. Cavagnaro, Physicochemcial and microbiological effect of long- and short-term winery wastewater application to soils, J. Hazard. Mater. 201–202 (2011) 219–228. [8] J.D. Oster, G. Sposito, C.J. Smith, Accounting for potassium and magnesium in irrigation water quality assessment, Calif. Agric. 70 (2016) 71–76. [9] L.A. Ioannou, G. Li Puma, D. Fatta-Kassino, Winery wastewater treatment by physicochemical, biological and advanced processes: a review, J. Hazard. Mater. 286 (2015) 343–368. [10] G. Lofrano, S. Meric, A comprehensive approach to winery wastewater treatment: a review of the state-of-the-art, Desalin. Water Treat. 57 (2016) 3011–3028. [11] A.A. Zorpas, A. Saranti, Multi-criteria analysis of sustainable clean technologies for the treatment of winery’s wastewater, Int. J International Journal of Global Environ. Issues 15 (2016) 151–168. [12] R. Ganesh, R. Rajagopal, J.V. Thanikal, R.A. Ramanujin, M. Torrijos, Anaerobic treatment of winery wastewater in a fixed bed reactor, Bioprocess. Biosyst. Eng. 33 (5) (2010) 619–628. [13] L. Serrano, D. de la Varga, I. Ruiz, M. Soto, Winery wastewater treatment in a hybrid constructed wetland, Ecol. Eng. 37 (2011) 744–753. [14] M. Arienzo, E.W. Christen, W.C. Quayle, Phytotoxicity testing of winery wastewater for constructed wetlands treatment, J. Hazard. Mater. 169 (2009) 94–99. [15] E.R. Rozema, L.R. Rozema, Y. Zheng, A vertical flow constructed wetland for the treatment of winery process water and domestic sewage in Ontario, Canada: six years of performance, Ecol. Eng. 86 (2016) 262–268. [16] M. Achak, l. Mandi, N. Quazzani, Removal of organic pollutants and nutrients from olive mill wastewater by a sand filter, J. Environ. Manage. 90 (2009) 2771–2779. [17] A. Mekki, A. Dhouib, S. Sayadi, Polyphenols dynamics and phytotoxicity in a soil amended by olive mill wastewaters, J. Environ. Manage. 84 (2007) 134–140. [18] S. Wu, S. Wallace, H. Brix, P. Kuschk, W. Kipkemoi Kirui, F. Masi, R. Dong, Treatment of industrial effluents in constructed wetlands: challenges, operational strategies and overall performance, Environ. Pollut. 201 (2015) 107–120. [19] H.L. Shepherd, M.E. Grismer, G. Tchobanoglous, Treatment of high-strength winery wastewater using a subsurface-flow constructed wetland, Water Environ. Res. 73 (4) (2001) 394–403. [20] M. von Sperling, A.C. de Paoli, First-order COD decay coefficients associated with different hydraulic models applied to planted and unplanted horizontal subsurface-
126
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Environmental Consequences, New York Press, New York, 1998, pp. 35–50. [53] J.M. Jesus, A.C. Cassoni, A.S. Danko, A. Fiuza, T. Borges, Role of three different plants on simultaneous salt and nutrient reduction from saline synthetic wastewater in lab-scale constructed wetlands, Sci. Total Environ. 576 (2017) 447–455. [54] J. Noy, A. Feinmesser, The use of wastewater for agricultural irrigation, in: H.I. Shuval (Ed.), Water Renovation and Reuse, Academic Press, New York, 1997, pp. 73–92. [55] P.A. Myburgh, C.L. Howell, Use of boundary lines to determine effects of some salinity-associated soil variables on grapevines in the Breede River Valley, S. Afr. J. Enol. Vitic. 35 (2014) 234–241. [56] P. Artiga, M. Carballa, J.M. Garrido, R. Mendez, Treatment of winery wastewaters in a membrane submerged bioreactor, Water Sci. Techno. 56 (2) (2007) 63–69. [57] G. Andreottola, P. Foladori, P. Nardelli, A. Denicolo, Treatment of winery wastewater in a full-scale fixed bed biofilm reactor, Water Sci. Techno. 54 (1) (2005) 71–73. [58] S. Kalyuzhnyi, M. Gladchenko, V. Sklyar, Y. Kizimenko, S. Shcherbakov, One and
[59]
[60]
[61] [62]
127
two stage upflow anaerobic sludge bed reactor pre-treatment of winery wastewater, Appl. Biochem. Biotechnol. 90 (2) (2001) 107–123. C. Da Ros, C. Cavinato, P. {avan, D. Bolzonella, Mesophilic and thermophilic anaerobic co-digestion of winery wastewater sludge and wine lees: an integrated approach for sustainable wine production, J. Environ. Manage. 203 (2) (2017) 745–752. Grape and Winer Research and Development Cooperation, Winery Wastwater Management and Recycling Operational Guidelines, [Online]. Available https:// www.wineaustralia.com/getmedia/72627da6-d28a-42f2-b600-28fdd5a6c85c/ Operational-Guidelines.pdf [2015, April 15] (2011). M.A. Dennis, M.L. Cotter, A.H. Slade, D.J. Gapes, The performance of a nitrogenfixing SBR, Water Sci. Techno. 50 (10) (2004) 269–278. P.J. Welz, J.B. Ramond, L. Braun, S. Vikram, M. Le Roes-Hill, Bacterial nitrogen fixation in sand bioreactors treating winery wastewater with a high carbon to nitrogen ratio, J. Environ. Manage. 207 (2018) (2018) 192–202.
Contents lists available at ScienceDirect
Journal of Water Process Engineering journal homepage: www.elsevier.com/locate/jwpe
Biological sand filter system treating winery effluent for effective reduction in organic load and pH neutralisation G.A. Holtmana,b, R. Haldenwangb, P.J. Welza,
T
⁎
a Biocatalysis and Technical Biology (BTB) Research group, Institute of Biomedical and Microbial Biotechnology, Cape Peninsula University of Technology, PO box 1906, Symphony way, Bellville, 7535, South Africa b Department of Civil Engineering, Cape Peninsula University of Technology, South Africa
A R T I C LE I N FO
A B S T R A C T
Keywords: Biological sand filter Chemical oxygen demand Constructed wetland pH neutralisation Treatment Organics Winery wastewater
Wineries generate 0.2–14 L of wastewater per litre of wine produced, which is often used for irrigation or discharged into aquatic systems. To mitigate adverse environmental impacts, there is a need for low cost wastewater treatment options. The novel biological sand filtration system described in this pilot study is a sustainable off-grid modular system which can be easily retrofitted to current infrastructure. The system was operated with average hydraulic and organic loading rates of 150 L m−3 sand.day-1 and 152 gCOD.m−3 of sand.day−1, respectively. Over 762 days of operation, average removal efficiencies of 79% and 77% in terms of chemical oxygen demand and total phenolic concentrations were achieved. In addition, an average 1.8-fold increase in the calcium concentration was achieved, with a concomitant reduction in the sodium adsorption ratio and similar indices. This pilot study also confirmed in a ‘real world’ setting the results of laboratory-based studies where biological sand filters neutralised acidic synthetic winery wastewater and reduced the organic load.
1. Introduction Wine production can be water-intensive. In addition to vineyard irrigation requirements, an estimated 0.2–14 L of water is needed to produce 1 L of bottled wine [1–3]. Most of this water leaves the cellar in the form of winery wastewater (WWW) that is typically acidic, and contains organic and inorganic fractions associated with seasonal cleaning activities [3,4]. The character and volume of the WWW varies on a temporal basis, with a pattern of high organic to inorganic load during the crush season, and vice-versa during the non-crush period [5]. In many countries, including South Africa, Australia, and the United States, WWW is often re-used for irrigation of pastures or crops in water-stressed areas; unless adequately treated, WWW may pose a threat to the soil environment and/or groundwater [6–8]. Many treatment processes have been researched and/or applied for the treatment of WWW, including: physicochemical processes, biological processes, membrane filtration and separation processes, and advanced oxidation processes [9]. Membrane systems have not been widely adopted by the wine industry because of the high capital outlay
and operating costs, the propensity for membrane fouling, and brine generation [10,11]. Physicochemical processes rely on either sedimentation, precipitation, coagulation/flocculation, or electrocoagulation [9]. Currently, the most effective physicochemical and advanced oxidation processes require skilled to semi-skilled labour, chemical inputs, create solid waste streams, and need to be used in conjunction with biological treatment methods to achieve satisfactory organic removal rates [10,11]. While biological systems such as rotating biological contactors, upflow anaerobic sludge blankets and membrane bioreactors may be suited to larger wineries for biodegradation of WWW organics, there is a global need for simple, low maintenance, cost effective systems at small wineries that do not have the finances or personnel to operate complex systems [9]. Many wineries utilise anaerobic or aerated ponds. These are simple and cheap to operate. However, odour problems are associated with ponds, and long hydraulic retention times (HRTs) are required, which translates into a large spatial footprint [12,13]. Constructed wetlands (CWs) are an effective option if the WWW is diluted or mixed with domestic wastewater (WW) [14,15]. The main drawback of CWs is that (poly)phenolic-rich effluent such
Abbreviations: WWW, winery wastewater; WW, wastewater; HRT, hydraulic retention time; CW(s), constructed wetland; OMWW, olive mill wastewater; COD, chemical oxygen demand; BSF(s), biological sand filter; k, hydraulic conductivity; VOA(s), volatile organic acid; AAC, acetic acid equivalents; SAR, sodium adsorption ratio; CROSS, cation ratio of soil structural stability; Q, flow rate; T, time; V, volume; OLR, organic loading rate ⁎ Corresponding author. E-mail addresses: [email protected] (G.A. Holtman), [email protected] (R. Haldenwang), [email protected] (P.J. Welz). https://doi.org/10.1016/j.jwpe.2018.07.008 Received 9 May 2018; Received in revised form 16 July 2018; Accepted 23 July 2018 2214-7144/ © 2018 Elsevier Ltd. All rights reserved.
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
as olive mill wastewater (OMWW) and WWW is phytotoxic [14,16–19]. For example, Shepherd et al. [19] demonstrated that gravel-filled horizontal subsurface flow CWs were able to achieve 97% COD removal when treating WWW with chemical oxygen demand (COD) concentrations < 5000 mg L−1, but poor removal efficiency was achieved and plants died at COD concentrations of 12 800–16800 mg L−1. It can also be argued that the inclusion of plants in CWs treating WWW is unnecessary because they play a limited role in reducing COD and suspended solids, they require periodic harvesting, and their roots can facilitate the formation of preferential flow paths [16,20]. Biological sand filters (BSFs) can be seen as unplanted CWs. The sand creates a physical substrate for biofilm attachment, can adsorb pollutants, catalyse chemical transformations, and provide metabolic co-factors; the functional microorganisms within the biofilm are responsible for biotransformation, biodegradation and mineralization of chemical pollutants [24]. In a series of fundamental laboratory-based experiments, it was shown that simple BSFs containing ungraded locally-available dune sand [21] were able to cope with the temporal nature of WWW and achieve reliable removal of organics, including (poly)phenolics [22,23]. To determine whether BSFs are viable for the treatment of WWW is a ‘real-world’ setting, a pilot system was designed, installed and operated at a wine farm in the Stellenbosch area close to Cape Town, South Africa. This manuscript details and analyses the performance of the system over approximately 27 months.
2.2.1. Determination of chemical oxygen demand, nutrients, and selected organic fractions The COD, total phosphate and total nitrogen concentrations were determined using a Merck (Merck®, Whitehouse Station, USA) Spectroquant® Pharo instrument and Merck Spectroquant® COD cell tests for low, medium and high range samples (cat. no. 1.14895.0001, 1.14541.0001 and 1.14691.0001) and total nitrogen cell tests 1.14543.0001 and total phosphate as PO4-P 1.14537.0001 according to manufacturer’s instructions., according to the manufacturer’s instructions. The total phenolic concentrations were determined using the Folin-Ciocalteu micro method based on that described by Slinkard and Singleton [25] using Merck®Folin–Ciocalteu reagent (Cat No: 1.09001.0500). The concentrations of volatile organic acids (VOAs) were determined using the Hach (Loveland, USA) esterification method (cat no 8196) in accordance with manufacturer’s instructions with some modifications: 3 standard concentrations of acetic acid were prepared (945.00 mg L−1, 472.50 mg L−1 and 236.25 mg L−1) and were used to prepare a standard graph for the determination of the VOA concentrations in acetic acid equivalents (mgAAC.L−1).
2. Materials and methods
2.2.3. Sodium adsorption ratio, cation ratio of soil structural stability Concentrations of sodium, potassium, calcium and magnesium were determined using a Varian® MPX inductively coupled plasma optical emission spectrophotometer (Agilent Technologies, Santa Clara, USA) at Bemlab (Pty) Ltd. (Strand, South Africa). The sodium adsorption ratio (SAR) and cation ratio of soil structural stability (CROSS) were determined using the equations described by Marchuk and Rengasamy [26] and Oster et al. [8].
2.2.2. Determination of pH The pH of the samples was determined according to the manufacturer’s instructions using a CyberScan pH300 meter and appropriately calibrated pH probe PHWP300/02 K (Eutech instruments, Singapore).
2.1. Set-up and operation of pilot scale biological sand filtration system The BSF system consisted of a number of connected polyethylene (PE) containers: a 5000 L collection tank, a 500 L holding tank, 4 × 1000 L sand filter modules, and 4 × 100 L flow-control tanks. The system was connected to an existing 45 m3 baffled concrete solidssettling delta (Fig. 1A). A fraction of the WWW from the delta was treated and returned to the head of the delta, thereby improving the quality of the WWW which was used to irrigate a sheep pasture. The WWW from the cellar gravitated into the delta inlet and settled effluent was pumped from the delta outlet to the 5000 L collection tank using a Shurflo (Pentair, Minneapolis, USA) 2088-313-145 12 V DC diaphragm premium demand solar pump. The pump was controlled by a liquid level relay and a probe within the 5000 L collection tank, which was set to fill the collection tank to 3800 L when the volume dropped to 2500 L. The flow through the remainder of the system was controlled by gravity, float valves and adjustable flow-control tanks (Fig. 1B). The float valves and flow control tanks enabled the HLR to be automatically altered when the flow rates through the filters increased or decreased. A constant head of 30 cm was maintained across each filter module. The modules were filled with a locally available dune sand, which had previously been well characterised and found to be suitable for use in BSF in terms of treatment capacity and hydraulic conductivity (k) [21]. The system could be operated in parallel or series, and was operated in parallel during the experimental period.
2.3. Calculation of operational parameters 2.3.1. Flow rates and volume of wastewater treated The flow rate was determined by measuring the volume of water collected from the outlet over the period of one minute. Measurements were taken in triplicate, averaged, and converted into the daily flow rate (Q). Q was used to determine the volume of WWW treated and was estimated using 2 methods: (i) the average of all the measured flow rates (Table 1) and (ii) a trapezoidal equation Eq. (1) that assumed a linear decrease or increase in Q between measuring instances (t).
∑ Voln = Qn−1 (tn−1) + 0.5[(Qn−Qn−1 )(tn−1−tn)]
(1)
2.3.2. Hydraulic retention time, hydraulic conductivity, hydraulic loading rate, organic loading rate The saturated cross-sectional area at the discharge point was used in combination with the change in height (h) divided by the change in length (l) to determine k in terms of Darcy’s law Eq. (2) [27,28]. The HRT was determined using the volume (V) of liquid within a packed media divided by the flow rate as shown in Eq. (3) [29], while the volume of liquid was determined by multiplying the porosity of the substrate by the volume of the reactor. The HLR was expressed in two different ways: (i) as the discharge rate of influent divided by the volume of the reactor (L m−3 d-1) Eq. (4), and (ii) as the discharge rate divided by the cross-sectional area of the reactor (L m-2 d-1), which is the formula typically applied for reactors operated in horizontal mode Eq. (5). The organic loading rate (OLR) was determined by multiplying the influent flow rate by the influent COD or BOD concentration divided by the volume of reactor [29] Eq (6).
2.2. Sampling and characterisation of influent and effluent Influent was defined as the grab WW samples from the 5000 L collection tank, and the effluent as the final effluent (Fig. 1). During the final year of the study, the sampling regime was increased for more accurate assessment of BSF performance (Table 1). In addition to regular monthly or bi-weekly analyses, a once-off batch test was performed, which entailed hourly sampling (Table 1). In three scheduled sampling instances, no samples could be taken due to operational problems (Dec 2015, Jan 2016, Mar 2016).
119
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 1. Set-up of biological sand filtration system showing the connection to the existing settling delta (A) and the operational components and sampling points of the system (B).
The functional (saturated) volume of sand was calculated using a formulated trapezoidal equation to determine the cross-sectional area of the filter Eq. (7). The cross-section was calculated by drawing a straight line between the height of the saturated zone at the inlet and the outlet. The height of the saturated zones were determined by inserting pipes into the sand.
Q = −k A
dh dl
HRT =
(2)
120
Vliquid Q
(3)
HLRV =
Q Vreactor
(4)
HLRA =
Q A Cross section
(5)
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Table 1 Sampling schedule and analytical parameters. Sampling points
Period
Occurrence
Analyses
Collection tank (influent), holding tank, filter outlets, final effluent Influent and effluent
Aug 2015
Once-off
COD, total phenolics, VOA, pH, sodium, potassium, calcium, magnesium, flow rate (Q)
Oct 2015-Feb 2016 Mar 2016-Mar 2017
Monthly
COD, total phenolics, VOA, total phosphorus, total nitrogen, pH, sodium, potassium, calcium, magnesium, flow rate (Q) COD, total phenolics, VOA, total phosphorus, total nitrogen, pH, sodium, potassium, calcium, magnesium, flow rate (Q)
Bi-weekly
COD = chemical oxygen demandVOA = volatile organic acids. Table 2 Operation parameters and organic removal rates in comparison to similar systems treating winery or olive mill wastewater. System
Wastewater (type)
Influent COD (mg L−1)
COD RE (%)
HLR (L/m3.day−1)
OLR (gCOD/m3.day−1)
tHRT (days)
1. Experimental SF 2. Full scale HSSF CW
OMWW & DW WWW
3. Experimental BSF 4. Full-scale CW VF 5. Full-scale 2-stage CW
WWW WWW & DW WWW & DW WWW
70–90 60 80 98 98.9 VF: 29–70 HF: 13–79 28-98
5.0–68.2a 23a 45a 7.1a 22a VF: 55–154a HF: 1.3–3.6a 150
2102.6–54.1a 107–218 35–277a 16.4a 58a VF: 30.7–332.9a HF: 0.4–5.5a 152
7–101a 7 14 22 NG NG
This study: Pilot BSF
30830 ± 1690 1: 2000–12000 2: 2000–12000 2304 ± 629 2580 VF: 1558 ± 1023 HF: 711 ± 769 1138
1.8
RE = removal efficiency HLR = hydraulic loading rate OLR = organic loading rate HRT = theoretical hydraulic retention time SF = sand filter CW = constructed wetland SF = biological sand filter HSSF = horizontal subsurface flow VF = vertical Flow OMWW = olive mill wastewater WWW = winery wastewater DW = domestic wastewater NG = Not given. References: 1. Achak et al. [16]; 2. Mulidzi [42]; 3. Ramond et al. [43]; 4. Rozema et al. [15]; 5. Serrano et al. [13]. a Calculated from information provided in journal article.
Fig. 2. Comparisons of the hydraulic and organic loading rates from October 2015 to October 2017.
OLR =
CODInfluent × Q Vreactor
A = 1070x−10925 [mm2]
the system is still fully functional to date (July 2018). (6) 3.1. Operational parameters
(7)
3.1.1. Hydraulic capacity In wineries, WWW production spikes during the crush season, with small spikes at other times of the year when cellar activities such as bottling are taking place. Two methods (2.3.1) were used to estimate the average daily volume of WWW that was treated [427 L.day−1 and 402 L.day-1 (137 L m-3 of sand−1)]. The latter values (calculated using Eq. (1), were used in all subsequent calculations. The pilot system was not intended to treat the entire volume of WWW produced by the cellar. Nevertheless, 22% was treated. In addition, the treated effluent was circulated back into the system, thereby improving the quality of all the WWW used for irrigation by diluting the fresh WWW entering the delta with final effluent from the BSF system.
3. Results and discussion The BSF system is intended to improve the suitability of WWW for irrigation or discharge to the environment. The system was commissioned in January 2015, and WWW was slowly added and drained intermittently over a 3 month start-up period in order to allow the microbial communities to acclimate. This incremental priming with WWW during start-up has been shown to increase the treatment performance in experimental BSFs [30]. A batch test was performed in August 2015, and regular sampling commenced in October 2015. The operational results over a further 24 months are presented in this manuscript, but 121
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 3. The chemical oxygen demand (A), volatile organic acid (B) and total phenolic (C) concentrations measured in influent and effluent samples, and respective removal efficiencies.
porous WW treatment systems [33]. Some researchers have developed models to predict the extent of changes in k but none have been properly validated [e.g. 33,34]. This is probably because the extent of biomass accumulation differs according to the type of physical substrate, the type of effluent being treated, and in different spatial locations within the systems [35]. In a previous study using sand-filled columns treating synthetic WWW, a 51% reduction in the k of dune sand from an initial value of 0.284 mm s−1 was determined experimentally in column experiments using the constant head method [21]. As no solids were present in the synthetic WWW, the reduction was attributed to biomass alone. In this pilot study, there was an 80% reduction in the calculated k of the dune sand from the initial value of 0.199 mm s−1, to the operational average of 0.040 mm s−1. The start-up HRT for a single BSF module of 0.53 days was calculated using the initial flow rate (Q = 1 900 L day−1) and porosity values (0.292 ± 0.02) determined experimentally for the ‘fresh’ dune sand. Thereafter, the initial porosity and the average system flow rate
In terms of size, SAWIS [31] broadly divides wineries into 6 categories depending on the amount of grapes crushed each year (< 100 tons ≥ 500 tons ≥ 1000 tons ≥ 10,000 tons <). The infrastructure of the BSF system shown in Fig. 1 can support 1 to 12 filter modules, after which the capacity (and footprint) would need to be increased. Using Eq. (2) combined with a safety factor of 1.1 [32], it was calculated that on an annual basis the infrastructure of the pilot BSF system is capable of treating the WWW from a cellar crushing 47 tons of grapes, which could be extended to a cellar crushing 141 tons by maximising the number of filters to 12.
Water = 4037.5 × T 0.9243
(8)
3.1.2. Hydraulic conductivity and hydraulic retention time Decreases in permeability and porosity, and changes in mass transport parameters due to the build-up of functional biomass and solids retention cause increased k leading to decreased HRT in free-flow 122
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 4. The pH measured in influent and effluent samples.
(Q = 402 L.day−1) was used to calculate the average operational HRT of a single BSF module [2.1 days, adjusted to 1.8 days to account for partial saturation of the sand, and 11.5 days when including the collection tank and other infrastructure in the calculation]. This HRT was significantly shorter than the HRT of other biological systems used to treat WWW (Table 2), but did not account for any decrease in porosity. It was not possible to accurately determine the operational k values because the changes in porosity were in a state of flux due to the constant formation and degradation of functional biomass.
state of flux (Fig. 2). Overall, the pilot BSF system demonstrated resilience to synthetic WWW fluctuations, which was also demonstrated in experimental systems treating synthetic WWW [22]. 3.2. Performance evaluation 3.2.1. Introduction The variable character of WWW may lead to temporal sampling bias when collecting influent and effluent [22]. In this study, the incoming WWW was balanced in the collection and holding tanks. The HRT within the BSF filters was only 1.8 days (Section 3.1.1). It was therefore expected that there would be no significant sampling bias and shortterm temporal system fluctuations. To confirm that the measured parameters were a true reflection of transformations effected within the BSFs (for example, removal efficiencies), a once-off batch test was performed. This consisted of taking samples every hour for 4 h (5 sampling instances; Supplementary material 1). For the batch study, no significant temporal differences (ANOVA: p > 0.5) were seen in the conductivity, COD, VOA, and total phenolic measurements in: (i) the t = 0 to t = 4hr influent samples, (ii) the t = 0 to t = 4hr effluent samples (iii) the t = 0 to t = 4hr samples taken separately from each of the individual BSF modules. This indicated that the performance assessment results were valid. However, when comparing the results from the different BSF modules, a significant difference in the COD and VOA measurements (ANOVA: p < 0.05) was noted (data not shown). Although the modules were set up in an identical manner and received the same WWW, inter-BSF module performance variability in terms of organic degradation was demonstrated.
3.1.3. Hydraulic and organic loading rates The HLR and OLR in CWs are typically calculated using the surface or cross-sectional area, but the total volume gives a better indication of the treatment efficiency of the medium [21]. For example, a slender vessel may have a high surface loading rate but relatively low overall loading rate, while the converse is true for a wide, shallow vessel. The initial HLR values for overall and cross-sectional loading of the sand filters were 646 L m−3sand.day−1 and 1084 L m−2sand.day−1, respectively. In comparison, the average operational values were 150 L m−3sand.day−1 (range: 12–313 L.m−3sand.day−1) and 252 L.m−2 sand.day−1 (range: 21-526 L m−2sand.day−1), respectively. The 77% decrease from initial to operational HLR was due to accumulation of bio-solids, including functional biomass, and was accompanied by increased COD removal rates (3.2). Over a 10 week period, Achak et al. [16] also described a reduction (50%) in the HLR, and improvement of organic removal rates in a vertical flow sand filter treating diluted OMWW with a starting HLR of 68.2 L.m3sand.day-1. In this case, the authors terminated the study because they felt that the HLR of 5.0 L m3sand.day−1 was inadequate. Some degree of biomass clogging is essential, because the microbial population is the most important functional component for organic biodegradation [36,37]. In reality, there is a balance between the HLR and system performance. If the HLR is too high, treatment may be inefficient, but if it is too low, the spatial footprint and capital costs may render the system uneconomical. High HLRs can be overcome by setting up systems that allow hydraulic flexibility, and strategies such as intermittent operation and adjustment of loading rates have been proposed to minimise biomass clogging and increase the HLR [16,36–39]. The variable nature of the OLR seen in this study resulted from the WWW being generated from different seasonal cellar activities, and is typical of all wineries. The average OLR over the study period was 152 gCOD m−3sand.day−1 (range: 23–469 gCOD m−3sand.day-1), or 256 gCOD m−2sand.day−1 (range: 39–788 gCOD m−2sand.day−1) for the total volume and surface area of the sand, respectively. After the start-up period, the HLR also fluctuated, which suggested that the accumulation and degradation of organic solids and/or biomass was in a
3.2.2. Removal of chemical oxygen demand and selected organics The system achieved an average COD removal efficiency of 79% (range: 28–98%). The average COD concentration in the final effluent (n = 35) was 287 mg L−1 (range: 24–1382 mg L−1, Fig. 3A). VOAs and ethanol constitute the highest fraction of the COD in WWW [5]. VOAs can also be formed in BSFs from other organic molecules, including phenolics, sugars and ethanol [22]. In this study, the average VOA concentrations in the influent and effluent were 472.8 mgAAE.L−1 (range: 9.4–1018.4 mgAAE.L−1) and 186.3 mgAAE.L−1 (range: 0.2–662.0 mgAAE.L−1), respectively, with ‘removal’ rates of -93% to 99.9% (Fig. 3B). The intermittent negative removal rates indicated formation of VOAs within the system. Phenolics in OMWW and WWW may be toxic to plants and microbes and should therefore be removed/reduced before discharge of the WW to the environment [16,40]. Removal of these aromatic molecules in BSFs takes place via biotic and abiotic mechanisms, and it has been shown that over 88% removal of (poly)phenolics with COD 123
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 5. The sodium and calcium concentrations (A), pH and change in calcium concentration (B), and sodium adsoption and cation ratio of soil structural stability ratios (C) measured in influent and effluent samples.
concentrations of 5842 mg.L−1 can be achieved [24]. Achak et al. [16] also demonstrated similar removal of phenolics from OMWW (87–97%) with an influent COD concentration of 4590 mg L−1. In this study, the average total phenolic concentration in the influent and effluent were 18.4 mgGAE.L−1 (range: 5.1–43.7 mgGAE.L−1) and 3.9 mgGAE.L−1 (range: ≤0.09–16.6), respectively, with average removal rates of 77% (range: 16–100%) (Fig. 3C). In comparison to other simple WWW treatment systems, the system performed well in terms of COD removal (Table 2). Montalvo et al. [41] used two adjacent pilot-scale batch-fed aeration lagoons to treat 790 L.day−1 and 170 L.day−1 of WWW with a similar OLR to the pilot BSF. Both ponds removed 91% of the COD, but the system required aeration, which is costly. Mulidzi [42] achieved 60% and 80% COD
removal for a CW with retention times of 7 and 14 days, respectively. However, sub-surface flow CWs are more complex, require additional maintenance, and WWW is potentially toxic to CW plants [16,19]. Rozema et al. [15] co-treated domestic and WWW (ratio 1:3) in a vertical flow CW and achieved 99% COD removal. The system operated at a relatively low OLR compared to the BSF in this study. Serrano et al. [13] also co-treated WWW with domestic WW, with a COD reduction of 73% across a two-stage CW fed by a hydrolytic upflow sludge bed followed by a horizontal flow CW. It is difficult to compare these systems with the BSF used in this study because of the presence of domestic WW. Domestic WW has a high carbon to nitrogen ratio in comparison to WWW. Therefore, an advantage of this type of cotreatment is that the domestic WW may enhance microbial activity 124
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Fig. 6. The total nitrogen and total phosphorous concentrations measured in influent and effluent samples.
concentrations of potassium were 121.7 mg L−1 (range: −1 −1 17.3–284.9 mg L ) and 126.8 mg L (range: 25.9–302.9 mg L−1), respectively (data not shown). The average sodium concentrations of the influent and effluent samples (Fig. 5A) were 29.6 mg L−1 (range: 10.6–70.8 mg L−1) and 29.8 mg L−1 (range: 10.0–62.9 mg L−1), respectively. In CWs containing clay, salts may be removed by physicochemical processes [53], but in this instance the sand did not contain clay particles, and the data showed that neither magnesium, potassium, nor sodium were removed by the system. In contrast, the calcium concentration (Fig. 5A) showed an average 1.8 fold increase in concentration from influent to effluent of 36.2 mg L−1 (range: 96.6–104.0 mg L−1) to 65.7 mg L−1 (range: 14.7–200.3 mg L−1). This was attributed to the dissolution of calcite as discussed in 3.2.3. There was a highly significant (p < 0.005) negative correlation (-0.37) between the pH of the influent samples and the change in calcium concentration from influent to effluent (Fig. 5B). This strongly suggested that the calcium dissolution rates increased with an increase in acidity. The BSF system achieved an average decrease in the SAR (13%) and CROSS (12%) from influent to effluent (n = 31) (Fig. 5C). This decrease in SAR and CROSS improved the quality of the final effluent for soil application. The CROSS values in the effluent were all below the limit of 20 suggested by Laurenson et al. [48] for irrigation with WWW. Crops such as banana (Musasapientum) and lucerne (Medicago sativa) have a remarkable ability to accumulate potassium [54]. Recent studies have also shown the ability of sugar beet to take up Na [55]. It is envisaged that WWW should be used for beneficial irrigation of these, or other halophytic crops after treatment in BSFs.
because of the addition of nutrients to the WWW. A disadvantage is the potential introduction of pathogens. 3.2.3. Neutralization of acidic winery wastewater The pH in the influent ranged from 4.55 to 7.95, while the pH of the effluent ranged from 6.63 to 8.69 (Fig. 4). Even the formation of VOAs in the system did not lead to an increase in the acidity of the effluent. This effective neutralisation of the acidic WWW was attributed to both abiotic and biotic mechanisms: The mineral calcite (formed from calcium carbonate) was present in the dune sand, with calcium constituting 17.6% (wt.wt) of the sand [21]. In this study, there was an average increase of 82% from the influent to the effluent which provided evidence of calcite dissolution as an abiotic neutralisation mechanism [44]. WWW was also neutralised in experimental BSFs containing river sand that did not contain calcite or aluminosilicates [43], strongly suggesting that biotic neutralisation mechanisms also exist. 3.2.4. Inorganics Sodium hydroxide is widely used as a cleaning agent and disinfectant in wineries and WWW can contain high concentrations of sodium [21]. The cation binds to one negatively charged soil particle, causing a tight arrangement which reduces infiltration of water into soils [45]. Potassium has a similar effect, but not to the same magnitude [46]. Calcium and magnesium are divalent cations, which bind to more than one negatively charged soil particle, creating a matrix within the soil which allows for aeration and infiltration [45,47]. The presence of divalent cations can therefore offset the detrimental effects of monovalent cations such as sodium. The SAR is a parameter that is widely used to assess the potential of water to cause sodicity, and takes into account the concentration of sodium, calcium and magnesium [48]. According to the U.S. Salinity Laboratory [49], soil with an SAR > 13 is considered sodic [50], and Horneck et al. [51] state that soils with a SAR < 5, 5–13, and > 13, pose low, medium and high risks, respectively. However, the SAR does not take into account the effect of potassium on soil dispersion, which is relevant for WWW that can contain elevated levels of potassium, sodium, calcium and magnesium [45]. The CROSS is a sodicity indicator that was derived by Rengasamy and Sumner [52] to overcome this limitation. This ratio is more suitable for monitoring the effects of WWW on the soil structure [45]. Laurenson et al. [48] suggested a maximum CROSS of 20 for disposal of WWW to the environment. The authors found an average CROSS of 9.2 (range: 2.5–13.3) in WWW from 8 different wineries in Australia. The average influent and effluent concentrations of magnesium were 4.5 mg L−1 (range: 1.6–10.2 mg L−1) and 4.5 mg L−1 (range: 1.3–10.1 mg L−1), respectively, while the average influent and effluent
3.3. Major nutrients The average total nitrogen (N) and phosphate (P) (Fig. 6) in the influent was 1.6 mg L−1 (range < 0.5-10.4 mg L−1) and 2.4 mg L−1 (range: 0.86–3.98 mg L−1), respectively. WWW is known to be deficient in N and P [5], and these may be added to assist bioremediation [56–58]. For example, researchers have dosed raw WWW with nutrients to attain COD:N:P ratios of 502:5:1 [59], 400:7:1 [12] and 3060:2:1 [60]. In this study, no nutrients were added. The influent COD:N:P ratio was 1138:1.6:2.4 (n = 17), exhibiting a particular N deficiency [high carbon (C):N ratio]. It has been shown that bacterial N fixation can take place in systems treating high C:N WW, such as that from the paper and pulp industry [61] and synthetic WW [62]. This could explain why the COD removal efficiency in the BSF system was high despite the nutrient limitation in the influent.
125
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
flow constructed wetlands, Ecol. Eng. 57 (2013) 205–209. [21] P.J. Welz, W. Mbasha, I. Smith, G. Holtman, G. Terblanche, M. Le Roes-Hill, R. Haldenwang, The influence of grain physicochemistry and biomass on hydraulic conductivity in sand-filled treatment wetlands, Ecol. Eng. 116 (2018) 21–30. [22] P.J. Welz, M. Le Roes-Hill, Biodegradation of organics and accumulation of metabolites in experimental biological sand filters used for the treatment of synthetic winery wastewater: a mesocosm study, J. Water Process Eng. 3C (2014) 155–163. [23] P.J. Welz, Z. Palmer, S. Isaacs, B. Kirby, M. Le Roes-Hill, Analysis of substrate degradation, metabolite formation and microbial community responses in sand bioreactors treating winery wastewater: a comparative study, J. Environ. Manage. 145 (2014) 147–156. [24] P.J. Welz, J.-B. Ramond, D.A. Cowan, S.G. Burton, Phenolic removal processes in biological sand filters, sand columns and microcosms, Bioresour. Technol. 119 (2012) (2012) 262–269. [25] K. Slinkard, V.L. Singleton, Total phenol analysis: automation and comparison with manual methods, Am. J. Enol. Vitic. 28 (1977) 49–55. [26] A. Marchuk, P. Rengasamy, Threshold electrolyte concentration and dispersive potential in relation to CROSS in dispersive soil, Soil Res. 50 (2011) 473–481. [27] H. Darcy, Les fontaines publiques de la ville de Dijon, Dalmont, Paris, 1856. [28] E. Wilson, Engineering Hydrology, 4th ed., Macmillan Press, London, U.K, 1990. [29] Metcalf and Eddy, Inc, Wastewater Engineering: Treatment, Disposal and Reuse, 4th edition, The McGraw-Hill Companies, New York, New York, 2004. [30] P.J. Welz, J.B. Ramond, D.A. Cowan, A. Prins, S.G. Burton, Ethanol degradation and the value of incremental priming in pilot scale constructed wetlands, Ecol. Eng. 37 (2011) 1453–1459. [31] SAWIS, South African Wine Industry Statistics (No.42), [Online] Available: http:// www.sawis.co.za/info/download/Book_2017_statistics_year_english_final.pdf [2018, July 13] (2017). [32] C. Sheridan, A Critical Process Analysis of Wine Production to Improve Cost Efficiency, Wine Quality and Environmental Performance. Unpublished Masters of Science Thesis, University of Stellenbosch, Stellenbosch, 2003. [33] M. Peszynska, A. Trykozko, G. Iltis, S. Schlueter, D. Wildenschild, Biofilm growth in porous media: experiments, computational modeling at the porescale, and upscaling, Adv. Water Resour. 95 (2016) (2016) 288–301. [34] R. Samsó, J. Garcia, P. Molle, N. Forquet, Modelling bioclogging in variably saturated porous media and the interactions between surface/subsurface flows: application to constructed wetlands, J. Environ. Manage. 165 (2016) 271–279. [35] A. Caselles-Osorio, J. García, Performance of experimental horizontal subsurface flow constructed wetlands fed with dissolved or particulate organic matter, Water Res. 40 (19) (2006) 3603–3611. [36] A. Juwarkar, S. Singh, A. Mudhoo, Comprehensive overview of elements in bioremediation, Rev. Environ. Sci. Biotechnol. 9 (3) (2009) 215–288. [37] K. Semple, L. Doick, K. Wick, H. Harms, Microbial interactions with organic contaminants in soil: definitions, processes and measurement, Environ. Pollut. 150 (1) (2007) 166–176. [38] P. Knowles, G. Dotro, J. Nivalla, J. Garcia, Clogging in subsurface-flow treatment wetlands: occurrence and contributing factors, Ecol. Eng. 37 (2011) 99–112. [39] J.W. McKinley, R.L. Siegrist, soil clogging genesis in soil treatment units used for onsite wastewater reclamation: a review, Environ. Sci. Technol. 41 (24) (2011) 2186–2209. [40] C. Sheridan, D. Hildebrand, D. Glasser, Turning wine (waste) into Water: toward technological advances in the use of constructed wetlands for winery effluent treatment, AIChE J. 60 (2) (2014) 420–431. [41] S. Montalvo, L. Guerrero, E. Rivera, R. Borja, A. Chica, A. Martin, Kinetic evaluation and performance of pilot-scale feed-batch aerated lagoon treating winery wastewater, Bioresour. Technol. 101 (2010) (2010) 3452–3456. [42] A.R. Mulidzi, Winery and distillery wastewater treatment by constructed wetland with shorter retention time, Water Sci. Techno. 61 (10) (2010) 2611–2615. [43] J.-B. Ramond, P.J. Welz, M.I. Tuffin, S.G. Burton, D.A. Cown, Assessment of temporal and spatial evolution of bacterial communities in a biological sand filter mesocosm treating winery wastewater, J. Appl. Microbiol. 115 (2013) (2013) 91–101. [44] O. Sulpis, C. Lix, A. Mucci, B.P. Boudreau, Calcite dissolution kinetics at the sediment-water interface in natural seawater, Mar. Chem. 195 (2017) (2017) 70–83. [45] P. Rengasamy, A. Marchuk, Cation ratio of soil structural stability (CROSS), Soil Res. 49 (3) (2011) 280–285. [46] M. Arienzo, E.W. Christen, W.C. Quayle, A. Kumar, A review of the fate of potassium in the soil-plant system after land application of wastewaters, J. Hazard. Mater. 164 (2009) 415–422. [47] F. Ranjbar, M. Jalali, The combination of geostatistics and geochemical simulation for the site-specific management of soil salinity and sodicity, Comput. Electron. Agric. 121 (2016) 301–312. [48] S. Laurenson, D. Houlbrooke, T. Styles, Determination of Soil Dispersion in Response to Changes in Soil Salinity Under Winery Wastewater Irrigation, [Online]. Available: http://www.marlborough.govt.nz/sitecore/shell/Controls/Rich%20Text %20Editor/∼/media/Files/MDC/Home/Environment/Land/Soils/AgResearch_ Client_Report_for_Marlborough_District_Council.ashx [2015, February 8] (2012). [49] U.S. Salinity Lab. Staff, Diagnosis and Improvement of Saline and Alkali Soils. Agricultural Handbook No. 60, United State Department of Agriculture, 1954. [50] B.D. Seelig, Salinity and Sodicity in North Dakota Soils. EB-57, North Dakota State University, Fargo, ND, 2000. [51] D.A. Horneck, J.W. Ellsworth, B.G. Hopkins, D.M. Sullivan, R.G. Stevens, Managing Salt-Affected Soils for Crop Production. Oregon State University, November 2007, Oregon State University, 2007, p. 24 Pacific Northwest Extension publication. [52] P. Rengasamy, M.E. Sumner, M.E. Sumner, R. Naidu (Eds.), Processes Involved in Sodic Behaviour, Sodic Soils. Distribution, Properties, Management, and
4. Conclusion This study showed that the BSF system is a viable option for the treatment of WWW at small cellars. The capex and running costs were low, the system was sustainable, required minimal maintenance, and did not need skilled labour to operate. Future designs will focus on creating greater treatment capacity and removing salts. For example, different modes of operation will be tested (continuous v/s batch mode, vertical flow), and the treated effluent will be used for beneficial irrigation of salt accumulating crops in order to reduce the sodium content. Acknowledgements The authors would like to thank the National Research Foundation of South Africa for funding this project (P. Welz grant number CSUR13091742638), Jacques Rossouw and Reckson Mulidzi from the Agricultural Research Council and Distell, respectively, for assistance with site selection, and the (unnamed) winery involved. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jwpe.2018.07.008. References [1] D. Bolzonella, D. Rosso, Narbonne, France, Proceedings of the 6th IWA Conference on Viticulture and Winery Wastes (2013). [2] M. Oliveira, E. Duarte, Treatment and use of non-conventional organic residues in agriculture, RAMIRAN International Conference, Lisboa, Portugal, 2010, pp. 12–15. [3] A.G. Vlyssides, E.M. Barampouti, S. Mai, Wastewater characteristics from Greek wineries and distilleries, Water Sci. Techno. 51 (2005) 53–60. [4] A. Bories, Y. Sire, Impacts of winemaking methods on wastewaters and their treatment, S. Afr. J. Enol. Vitic 31 (2010) 38–44. [5] P.J. Welz, G. Holtman, R. Haldenwang, Le Roes-Hill, Characterisation of winery wastewater from continuous flow settling basins and waste stabilisation ponds over the course of 1 year: implications for biological wastewater treatment and land application, Water Sci. Technol. 74 (9) (2016) 2036–2050. [6] D.R. Hirzel, K. Steenwoerth, S.J. Parikh, A. Oberholster, Impact of winery wastewater irrigation on soil, grape and wine composition, Agric. Water Manage. 180 (2017) 178–189. [7] K.P.M. Mosse, A.F. Patti, R.J. Smernik, E.W. Christen, T.R. Cavagnaro, Physicochemcial and microbiological effect of long- and short-term winery wastewater application to soils, J. Hazard. Mater. 201–202 (2011) 219–228. [8] J.D. Oster, G. Sposito, C.J. Smith, Accounting for potassium and magnesium in irrigation water quality assessment, Calif. Agric. 70 (2016) 71–76. [9] L.A. Ioannou, G. Li Puma, D. Fatta-Kassino, Winery wastewater treatment by physicochemical, biological and advanced processes: a review, J. Hazard. Mater. 286 (2015) 343–368. [10] G. Lofrano, S. Meric, A comprehensive approach to winery wastewater treatment: a review of the state-of-the-art, Desalin. Water Treat. 57 (2016) 3011–3028. [11] A.A. Zorpas, A. Saranti, Multi-criteria analysis of sustainable clean technologies for the treatment of winery’s wastewater, Int. J International Journal of Global Environ. Issues 15 (2016) 151–168. [12] R. Ganesh, R. Rajagopal, J.V. Thanikal, R.A. Ramanujin, M. Torrijos, Anaerobic treatment of winery wastewater in a fixed bed reactor, Bioprocess. Biosyst. Eng. 33 (5) (2010) 619–628. [13] L. Serrano, D. de la Varga, I. Ruiz, M. Soto, Winery wastewater treatment in a hybrid constructed wetland, Ecol. Eng. 37 (2011) 744–753. [14] M. Arienzo, E.W. Christen, W.C. Quayle, Phytotoxicity testing of winery wastewater for constructed wetlands treatment, J. Hazard. Mater. 169 (2009) 94–99. [15] E.R. Rozema, L.R. Rozema, Y. Zheng, A vertical flow constructed wetland for the treatment of winery process water and domestic sewage in Ontario, Canada: six years of performance, Ecol. Eng. 86 (2016) 262–268. [16] M. Achak, l. Mandi, N. Quazzani, Removal of organic pollutants and nutrients from olive mill wastewater by a sand filter, J. Environ. Manage. 90 (2009) 2771–2779. [17] A. Mekki, A. Dhouib, S. Sayadi, Polyphenols dynamics and phytotoxicity in a soil amended by olive mill wastewaters, J. Environ. Manage. 84 (2007) 134–140. [18] S. Wu, S. Wallace, H. Brix, P. Kuschk, W. Kipkemoi Kirui, F. Masi, R. Dong, Treatment of industrial effluents in constructed wetlands: challenges, operational strategies and overall performance, Environ. Pollut. 201 (2015) 107–120. [19] H.L. Shepherd, M.E. Grismer, G. Tchobanoglous, Treatment of high-strength winery wastewater using a subsurface-flow constructed wetland, Water Environ. Res. 73 (4) (2001) 394–403. [20] M. von Sperling, A.C. de Paoli, First-order COD decay coefficients associated with different hydraulic models applied to planted and unplanted horizontal subsurface-
126
Journal of Water Process Engineering 25 (2018) 118–127
G.A. Holtman et al.
Environmental Consequences, New York Press, New York, 1998, pp. 35–50. [53] J.M. Jesus, A.C. Cassoni, A.S. Danko, A. Fiuza, T. Borges, Role of three different plants on simultaneous salt and nutrient reduction from saline synthetic wastewater in lab-scale constructed wetlands, Sci. Total Environ. 576 (2017) 447–455. [54] J. Noy, A. Feinmesser, The use of wastewater for agricultural irrigation, in: H.I. Shuval (Ed.), Water Renovation and Reuse, Academic Press, New York, 1997, pp. 73–92. [55] P.A. Myburgh, C.L. Howell, Use of boundary lines to determine effects of some salinity-associated soil variables on grapevines in the Breede River Valley, S. Afr. J. Enol. Vitic. 35 (2014) 234–241. [56] P. Artiga, M. Carballa, J.M. Garrido, R. Mendez, Treatment of winery wastewaters in a membrane submerged bioreactor, Water Sci. Techno. 56 (2) (2007) 63–69. [57] G. Andreottola, P. Foladori, P. Nardelli, A. Denicolo, Treatment of winery wastewater in a full-scale fixed bed biofilm reactor, Water Sci. Techno. 54 (1) (2005) 71–73. [58] S. Kalyuzhnyi, M. Gladchenko, V. Sklyar, Y. Kizimenko, S. Shcherbakov, One and
[59]
[60]
[61] [62]
127
two stage upflow anaerobic sludge bed reactor pre-treatment of winery wastewater, Appl. Biochem. Biotechnol. 90 (2) (2001) 107–123. C. Da Ros, C. Cavinato, P. {avan, D. Bolzonella, Mesophilic and thermophilic anaerobic co-digestion of winery wastewater sludge and wine lees: an integrated approach for sustainable wine production, J. Environ. Manage. 203 (2) (2017) 745–752. Grape and Winer Research and Development Cooperation, Winery Wastwater Management and Recycling Operational Guidelines, [Online]. Available https:// www.wineaustralia.com/getmedia/72627da6-d28a-42f2-b600-28fdd5a6c85c/ Operational-Guidelines.pdf [2015, April 15] (2011). M.A. Dennis, M.L. Cotter, A.H. Slade, D.J. Gapes, The performance of a nitrogenfixing SBR, Water Sci. Techno. 50 (10) (2004) 269–278. P.J. Welz, J.B. Ramond, L. Braun, S. Vikram, M. Le Roes-Hill, Bacterial nitrogen fixation in sand bioreactors treating winery wastewater with a high carbon to nitrogen ratio, J. Environ. Manage. 207 (2018) (2018) 192–202.