Clustering the Adsorbents of Horizontal Series Filtration in Greywater Treatment

: One of the important alternative water resources for non-potable purposes is greywater (GW), which must be cleaned of contaminants. In this regard, the clustering analysis of materials consisting of sand (S), zeolite (Z), peat (P) and granular activated carbon (GAC) within a horizontal series ﬁlter (HSF) was used for removal of chemical oxygen demand (COD), biochemical oxygen demand (BOD), total dissolved solids (TDS), and turbidity in GW taken from the Fasa University Student Hostel, Iran. The hierarchical clustering technique was applied to classify the adsorbents. The ﬁndings indicated that there were signiﬁcant di ﬀ erences (more than 95%) between these materials. According to the similarity of level 95%, for COD, BOD, TDS, and turbidity removal, these adsorbents could be separately clustered in three, three, two, and three clusters, respectively. In addition, by considering the simultaneous changes of COD, BOD, TDS, and turbidity together, these adsorbents could be clustered in three di ﬀ erent clusters. This paper proposed an e ﬃ cient method to select the best combination of adsorbents for eliminating of COD, BOD, TDS, and turbidity from GW. Generally, based on the quality of treated greywater and literature, reusing greywater can be implemented for agriculture, artiﬁcial recharge of aquifers, desertiﬁcation, and preventing the dust creation in arid areas such as southern Iran.


Introduction
Today, reducing the overall urban water demand has become a vital subject for water utilities and regulatory bodies. Usually indoor domestic water demand (excluding landscape, irrigation, toilet flush tank and other purposes that do not require using freshwater) in developed countries varies from 100 to 180 L/d per capita or from 36 to 66 m 3 /y per capita [1,2]. In this way, a "new" alternative resource is reusing the water. When regarding urban water reuse, on-site greywater (GW) reuse owns the potential to perform a significant role [3].
The various technologies have been examined for GW treatment like coagulation and magnetic ion exchange resin [4], flocculation [5], septic tank followed by intermittent sand filter [6], a moving bed biofilm reactor [7], trickling filters with suspended plastic media [8], slow sand filter and slate waste followed by granular activated carbon [9], drawer compacted sand filter [10], pelletised mine watersludge [11], aerobic attached-growth biomass [12], green roof-top water recycling system constructed wetland [13], biofilter system [14], compact hybrid filter systems [15], a physical treatment system containing coagulation, sedimentation, sand filtration, granular activated carbon filtration, and disinfection [16], and anaerobic filter followed by UV disinfection [17]. Recently, the evolution of greywater recycling operations has been from traditional treatment technologies into more The horizontal series filters applied for treatment were four PVC cylinders with a length of 80 cm and a diameter of 15 cm, which were filled with desired filters. For optimizing the reactor configuration, various experiments have been done using diverse architectural configurations. The investigation of each combination was conducted using try and error. Finally, the reactor configuration with the highest removal efficiency was designated as the best. The used sand (S) containing four layers in filter one, with D10 = 0.6 mm, D60 = 1.2 mm, Cu = 2, and permeability = 2.7 × 10 arranged from coarse to fine. This filter was produced by recommended standards and guidance for performance, application, design, and operation and maintenance recirculating gravel filter systems by Washington state department of health [29]. The granular activated carbon (GAC) in filter two, purchased from Merck Company, Darmstadt, Germany (product number 1025141000). The zeolite (Z) in the third cylinder originated from a mine in Darab, Fars province, Iran. The peat (P) in the last filter, purchased from Biolan Oy Co., Finland, composed of raw biomaterials and applied for horticultural aims. Characterization and amounts of the applied materials in filters tabulated in Table 1.   This study operated using real GW, effluent from a student dormitory block with 250 residents. The characteristics of raw GW are represented in Table 2. A 100 L tank was used to retain the greywater for 24 h to settle the suspended solids and increasing dissolved oxygen and was also applied for flow regulation before the filters. The greywater was conveyed through a pump to the suggested treatment filter at a constant flow rate equal to 0.034 L s −1 under constant pressure conditions (constant head equal to 0.58 m) applying a flow control tool with a pre-calibrated valve. In addition, a bypass system was considered to divert excess effluent into the sewage disposal system. Afterward, the treated greywater was returned to the tank via a tube from the final filter. The efficiency of the HSF treatment system (E) was measured by analyzing the greywater samples collected prior and after the filters, as follows: where C 0 and C are indicator concentrations before and after the filters in the samples, respectively. In each experiment, the filtration rate was maintained at 2.94 m 3 day −1 and the system operated for six hours. Treated greywater samples were taken in 15, 30, 120, 180, 240, 300, and 360 minutes after the system operation. For studying the purification efficiency of the different filter materials, the performance of the system was assessed every time with two, three or all four filters. Meanwhile, the sand filter operated always on the system and every time one, two or three others were added. The water quality parameters of biochemical oxygen demand (BOD), chemical oxygen demand (COD), total dissolved solids (TDS), and turbidity were analyzed in samples according to the standard methods for the examination of water and wastewater [30]. All experiments were performed with three replications. Subsequently, based on the findings of the effluent water quality, the hierarchical clustering analysis using the Ward clustering method and squared Euclidean distances were applied to classify the studied adsorbents.

Clustering Methodology
A multivariate approach such as cluster analysis (CA) can be utilized to cluster the features or observations [31]. The hierarchical cluster analysis (HCA) is the extensively applied type of CA, which is based on the degree of similarity or dissimilarity of the features or observations, utilizing an amalgamation technique. The degree of similarity or dissimilarity can be estimated in various methods like Euclidean, squared Euclidean, and Manhattan distances. In addition, various amalgamation techniques like between-groups linkage, within-groups linkage, nearest neighbor, furthest neighbor, centroid clustering, median clustering, and Ward's method can be applied [32].
Since each cluster consists of some points, numerous amalgamation techniques like Ward's technique, average linkage, complete linkage, and simple linkage can be applied to compute the distance and similarity between different points [32][33][34]. In real-world applications, squared Euclidean distances and Ward's method are usually selected. Let C i , C j , and C ij as i th , j th clusters, and combination of C i and C j . In Ward's technique, the distance between C i and C j is computed by: where r i , r j , and r ij are the corresponding centroids of C i , C j , and C ij , respectively. It should be noted that the squared Euclidean distances of x = (x 1 , . . . , x k ) and y = (y 1 , . . . , y k ) is defined by The structure of HCA algorithm is summarized in Appendix A.

Results and Discussion
The analysis of the form and geometry of the grains of GAC, Z, P, and sand via scanning electron microscopy (SEM) was represented in Figure 2. The high porosity of GAC and rough surface of Z particles are the most important factors to make them suitable adsorbents. Likewise, less efficiency of Sustainability 2020, 12, 3194 5 of 14 sand and P is expected because of their lower porosity level. Anyway, availability and low cost of peat and sand make them inevitable to use.
where , , and are the corresponding centroids of , , and , respectively. It should be noted that the squared Euclidean distances of = ( , … , ) and = ( , … , ) is defined by ∑ ( − ) . The structure of HCA algorithm is summarized in Appendix.

Results and Discussion
The analysis of the form and geometry of the grains of GAC, Z, P, and sand via scanning electron microscopy (SEM) was represented in Figure 2. The high porosity of GAC and rough surface of Z particles are the most important factors to make them suitable adsorbents. Likewise, less efficiency of sand and P is expected because of their lower porosity level. Anyway, availability and low cost of peat and sand make them inevitable to use. The performance of horizontal series filtration including various combinations of studied adsorbents to remove (%) COD, BOD, TDS, and turbidity overtime, was represented in Figure 3.  The performance of horizontal series filtration including various combinations of studied adsorbents to remove (%) COD, BOD, TDS, and turbidity overtime, was represented in Figure 3. Table 3 represents the removal percent of COD, BOD, TDS, and turbidity by various combinations of the studied adsorbents within the HSF system after 6 h. The results indicate that the most removal was carried out in a combination of all three adsorbents. However, GAC that is extensively employed in domestic wastewater treatment is expensive, while, there are lots of natural Z and P. Thus, combining the low and high-cost materials will diminish the costs of wastewater treatment and may amend the adsorption capacity. The superiority of GAC to other materials is due to the most specific surface area (SSA), the great quantity of permeability, and the most porosity. On the other, the combinations containing peat had more removing of turbidity compared with others. The main Sustainability 2020, 12, 3194 6 of 14 reason is the sponge-like structure of peat that traps turbidity agents. This system represented better results compared to the researches of [8,9] in GW treatment after 4 h.
The performance of horizontal series filtration including various combinations of studied adsorbents to remove (%) COD, BOD, TDS, and turbidity overtime, was represented in Figure 3.   We were interested in clustering the materials based on COD, BOD, TDS, and turbidity. To compare and cluster the materials, the Ward clustering method and squared Euclidean distances were utilized. Figures 4-8 summarize the results of the clustering analysis. As these figures indicate, there were significant differences between these materials. Based on the similarity of level 95% (similarity more than 95%, or distance lower than 5%), for COD and BOD (Figures 4 and 5), these materials could be clustered in three clusters including cluster 1: GAC, GAC+P, GAC+Z, Z, and Z+P; cluster 2: GAC+Z+P; cluster 3: P. According to clustering analysis, four treatments of GAC, GAC+P, GAC+Z, and Z have the same COD and BOD adsorption rate. This means that in cluster 1, Z and Z+P treatments as inexpensive materials could be used instead of commercial adsorbents such as GAC for greywater treatment. In Iran, the approximate cost of the zeolite, peat, and granular activated carbon employed for municipal wastewater treatment is about 100, 100, and 900 US$ ton − 1 , respectively, while the cost of the GAC+Z+P would be lower than 367 US$ ton −1 . The relative cost of the GAC+Z+P used Sustainability 2020, 12, 3194 7 of 14 in the current research was much lower than granular activated carbon as a commercial adsorbent. Therefore, by combining the low-cost (zeolite and peat) and high-cost (granular activated carbon) materials, not only the expenses of greywater treatment will be decreased but also the uptake capacity may enhance.  According to Table 3 and Figures 4 and 5, the maximum removal of COD and BOD from greywater (more than 90 percent) occurred by the combination of three adsorbents (GAC+Z+P), among the three clusters. After that, the Z+P combination with removal percent of about 68 was suggested because of the lower cost compared with other combinations in cluster 1. Peat adsorbent in cluster 3 performed the minimum removal of COD and BOD in comparison with other clusters. Indeed, the great potential of GAC+Z+P to eliminate COD and BOD from greywater may be related to the cation exchangeability and the hydrophilic surface of Z and P as well as the hydrophobic surface and great surface area of GAC [23]. It is found that COD and BOD are eliminated from greywater by GAC+Z+P via chemical and physical adsorption mechanisms. Chemical adsorption is predominant for zeolite and peat, whereas physical adsorption is predominant for GAC [23].
For TDS (Figure 6), we reached two clusters including cluster 1: GAC, GAC+P, and GAC+Z+P; and cluster 2: GAC+Z, Z+P, Z, and P. According to Figure 6 and Table 3, performance of GAC  According to Table 3 and Figures 4 and 5, the maximum removal of COD and BOD from greywater (more than 90 percent) occurred by the combination of three adsorbents (GAC+Z+P), among the three clusters. After that, the Z+P combination with removal percent of about 68 was suggested because of the lower cost compared with other combinations in cluster 1. Peat adsorbent in cluster 3 performed the minimum removal of COD and BOD in comparison with other clusters. Indeed, the great potential of GAC+Z+P to eliminate COD and BOD from greywater may be related to the cation exchangeability and the hydrophilic surface of Z and P as well as the hydrophobic surface and great surface area of GAC [23]. It is found that COD and BOD are eliminated from greywater by GAC+Z+P via chemical and physical adsorption mechanisms. Chemical adsorption is predominant for zeolite and peat, whereas physical adsorption is predominant for GAC [23].
For TDS (Figure 6), we reached two clusters including cluster 1: GAC, GAC+P, and GAC+Z+P; and cluster 2: GAC+Z, Z+P, Z, and P. According to Figure 6 and Table 3, performance of GAC  Table 3 and Figures 4 and 5, the maximum removal of COD and BOD from greywater (more than 90 percent) occurred by the combination of three adsorbents (GAC+Z+P), among the three clusters. After that, the Z+P combination with removal percent of about 68 was suggested because of the lower cost compared with other combinations in cluster 1. Peat adsorbent in cluster 3 performed the minimum removal of COD and BOD in comparison with other clusters. Indeed, the great potential of GAC+Z+P to eliminate COD and BOD from greywater may be related to the cation exchangeability and the hydrophilic surface of Z and P as well as the hydrophobic surface and great surface area of GAC [23]. It is found that COD and BOD are eliminated from greywater by GAC+Z+P via chemical and physical adsorption mechanisms. Chemical adsorption is predominant for zeolite and peat, whereas physical adsorption is predominant for GAC [23].  [35]. After that, among the cluster 2 materials, peat with TDS removal percent of 53.99 was suggested due to the lower cost. In the case of turbidity (Figure 7), the materials were clustered in three clusters including cluster 1: P, GAC+Z, Z, and GAC+P; cluster 2: Z+P and GAC+Z+P; and cluster 3: GAC. According to Figure 7 and Table 3, cluster 2 had the best performance to reduce the turbidity, in which GAC+Z+P removes about 93.01% of initial turbidity after 360 min. After that, the removal percent of turbidity was 88.03% by Z+P and 81.74 by P that ranked in second and third places, respectively. Indeed, the great potential of GAC+Z+P to eliminate turbidity from greywater may be related to the cation exchangeability and the hydrophilic surface of Z and P as well as the hydrophobic surface and great surface area of GAC [23]. Because of the complex surface of GAC+Z+P that includes a broad range of pore size distribution, multiple mechanisms are involved in the uptake of turbidity from greywater. Therefore, by combining the low-cost (Z and P) and high-cost (GAC) adsorbents, not only the expenses of greywater treatment will be decreased but also the uptake capacity may enhance. From a practical point of view, Z+P combination and P are suggested for turbidity removal from greywater due to the lower cost and acceptable performance. As Figure 3 indicates, the required time in the equilibrium state in all treatments was about 300 min. The removal percentage of COD, BOD, TDS, and turbidity improved with the contact time and reaches equilibrium after 300 min for GAC+Z+P. 99.45%, 99.45%, 99%, and 98% of initial COD, BOD, TDS, and turbidity concentrations, respectively, were removed by using GAC+Z+P within 300 min of circulation. Therefore, this system should be worked in closed-loop until the treated greywater becomes appropriate to employ for restricted and unrestricted irrigation of landscape as well as groundwater recharge.
In addition, by considering the simultaneous changes of COD, BOD, TDS, and turbidity together (total features), according to Figure 8 and Table 4, these materials could be clustered in three different clusters including cluster 1: GAC, GAC+P, GAC+Z, Z, and Z+P; cluster 2: GAC+Z+P; and cluster 3: P. According to these results, the existence of GAC in the filter components increased the filter ability to reduce the total features. However, based on the lower cost and more availability of peat and zeolite, the combinations of Z+P and Z can be suggested as the more favorable materials to improve greywater quality. This finding is in agreement with [23] that in studying the greywater treatment using single and combined adsorbents in batch experiments, concluded the best For TDS (Figure 6), we reached two clusters including cluster 1: GAC, GAC+P, and GAC+Z+P; and cluster 2: GAC+Z, Z+P, Z, and P. According to Figure 6 and Table 3, performance of GAC adsorbent (with TDS removal of 84.37 percent) is equivalent to combination of three materials (with TDS removal of 84.65 percent) in cluster 1. Some researchers have confirmed this capability of GAC in TDS reduction [35]. After that, among the cluster 2 materials, peat with TDS removal percent of 53.99 was suggested due to the lower cost. and cluster 3: P. According to these results, the existence of GAC in the filter components increased the filter ability to reduce the total features. However, based on the lower cost and more availability of peat and zeolite, the combinations of Z+P and Z can be suggested as the more favorable materials to improve greywater quality. This finding is in agreement with [23] that in studying the greywater treatment using single and combined adsorbents in batch experiments, concluded the best performance of a combination of activated carbon, zeolite, and nano zero-valent iron. In the case of turbidity (Figure 7), the materials were clustered in three clusters including cluster 1: P, GAC+Z, Z, and GAC+P; cluster 2: Z+P and GAC+Z+P; and cluster 3: GAC. According to Figure 7 and Table 3, cluster 2 had the best performance to reduce the turbidity, in which GAC+Z+P removes about 93.01% of initial turbidity after 360 min. After that, the removal percent of turbidity was 88.03% by Z+P and 81.74 by P that ranked in second and third places, respectively. Indeed, the great potential of GAC+Z+P to eliminate turbidity from greywater may be related to the cation exchangeability and the hydrophilic surface of Z and P as well as the hydrophobic surface and great surface area of GAC [23]. Because of the complex surface of GAC+Z+P that includes a broad range of pore size distribution, multiple mechanisms are involved in the uptake of turbidity from greywater. Therefore, by combining the low-cost (Z and P) and high-cost (GAC) adsorbents, not only the expenses of greywater treatment will be decreased but also the uptake capacity may enhance. From a practical point of view, Z+P combination and P are suggested for turbidity removal from greywater due to the lower cost and acceptable performance.
As Figure 3 indicates, the required time in the equilibrium state in all treatments was about 300 min. The removal percentage of COD, BOD, TDS, and turbidity improved with the contact time and reaches equilibrium after 300 min for GAC+Z+P. 99.45%, 99.45%, 99%, and 98% of initial COD, BOD, TDS, and turbidity concentrations, respectively, were removed by using GAC+Z+P within 300 min of circulation. Therefore, this system should be worked in closed-loop until the treated greywater becomes appropriate to employ for restricted and unrestricted irrigation of landscape as well as groundwater recharge.
In addition, by considering the simultaneous changes of COD, BOD, TDS, and turbidity together (total features), according to Figure 8 and Table 4, these materials could be clustered in three different clusters including cluster 1: GAC, GAC+P, GAC+Z, Z, and Z+P; cluster 2: GAC+Z+P; and cluster 3: P. According to these results, the existence of GAC in the filter components increased the filter ability to reduce the total features. However, based on the lower cost and more availability of peat and zeolite, the combinations of Z+P and Z can be suggested as the more favorable materials to improve greywater quality. This finding is in agreement with [23] that in studying the greywater treatment using single and combined adsorbents in batch experiments, concluded the best performance of a combination of activated carbon, zeolite, and nano zero-valent iron. The treated greywater from HSF using GAC+Z+P is proper for employing in restricted and unrestricted irrigation of landscape as well as groundwater recharge according to Iranian standards [23], where the residuals COD, BOD, TDS, and turbidity in treated greywater are 25.29 mg L −1 , 15.85 mg L −1 , 429.80 mg L −1 , and 12.93 NTU, respectively. These results are in agreement with [36] that used the greywater treated by a horizontal sub-surface flow constructed wetland reactor to irrigate the landscape in small communities of Morocco. In comparison between raw greywater and tap water, those irrigated by treated greywater appear to grow faster, because of the richness of this water by nutrients. [37] analyzed and verified the samples experimentally to indicate the improvement of physical, chemical, and biological factors to assure the quality of the treated greywater from an anaerobic filter used in irrigation in Iraq. [38] used a bio-remediation way of greywater recycling in India and concluded that the reuse of the recycled effluents could be prescribed for domestic use, irrigation, and landscaping. [21] indicated that user opinions towards greywater treatment and reuse were only favorable towards non-potable purposes, chiefly because of perceived pollution or lack of trust in the level of treatment suggested by the treatment system. [39] proposed the treated greywater achieved by gravity governed filtration method and disinfection for domestic, agriculture, and also for aquifers artificial recharge to prevent saltwater intrusion along the coastal aquifers in Kuwait. Table 4 summarizes the rate of similarity between the materials for Ward's method based on COD, BOD, TDS, turbidity, and total features, respectively. The results of Table 4 confirmed the aforementioned output of dendrograms. For example, for COD removal from greywater, the similarities of various combinations were GAC and Z 96%, GAC and P 75%, GAC and GAC+P and GAC+Z 99%, GAC and Z+P 96%, GAC and GAC+Z+P 92%, Z and P 75%, Z and GAC+P and GAC+Z 96%, Z and Z+P 99%, Z and GAC+Z+P 92%, and similarity between P, GAC+P, GAC+Z, Z+P, and GAC+Z+P 92% was 75%. The greater the similarity between the two combinations, the more similar their performance in GW treatment.
An economic analysis of the horizontal series filtration was performed using the rate of return (ROR) method as described in the previous researches [39][40][41]. Briefly, with regard to the comparison between the ROR and the minimum attractive rate of return (MARR), a project can be rejected or accepted. The project is accepted economically when the ROR ≥ MARR and rejected economically when ROR < MARR. To compare two or more projects, extra investment analysis is conducted. To assess the performance of GAC, P, Z, and GAC+Z+P in an industrial horizontal series filter for greywater treatment, the economic analysis based on the Iranian market (Table 5) was carried out. The initial investment, annual cost, annual incomes, salvage value, useful life, and MARR for each project are presented in Table 5. From the results presented in this  By solving the equations 3 to 6, the RORs for GAC, P, Z, and GAC+Z+P were obtained as 18%, 12%, 12%, and 40%, respectively. The industrial horizontal series filter using P and Z as adsorbents are not economical since their RORs are lower than 15%, whereas the industrial horizontal series filter using GAC and GAC+Z+P as adsorbents are economical since their RORs are higher than 15%. Between GAC and GAC+Z+P adsorbents in an industrial horizontal series filter, GAC+Z+P is selected since its ROR (40%) is higher than GAC (18%). In addition, this triple combined adsorbents have a lower initial investment. The volume of greywater, which can be treated by GAC, P, Z, and GAC+Z+P in an industrial horizontal series filter is 8000, 5600, 6600, and 8760 m 3 year −1 , respectively. The weight of adsorbents which were used in this system is depended on regeneration efficiency. Thus, more attention should be performed to study the reusability of the GAC, P, Z, and GAC+Z+P in an industrial horizontal series filter. The weight of the GAC, P, Z, and GAC+Z+P to treat this volume of greywater can be approximated 15,480, 124,488, 22,320, and 54,000 kg year −1 , respectively. The performance of