Treatment of highly turbid water in disaster conditions using coagulation-flocculation process: modeling and optimization

Over the past decades, disasters, especially floods and earthquakes, have had adverse effects on water quality such as increasing suspended solids (SS) and colloids more than the standard range (Ang et al. 2016). Among the various water quality indicators, turbidity is one of the most significant indicators for detecting freshwater characteristics (Antov et al. 2018). Typically, the drinking water-based on turbidity value can be divided into four categories as follows: low turbidity (less than 50 NTU), medium turbidity (50–100 NTU), high turbidity (100–200 NTU), and very high turbidity (more than 300 NTU) (Altaher 2012). Excessive turbidity in drinking water can also be associated with the presence of organic materials, dye, and microorganisms such as viruses, and certain bacteria, therefore, it increases the health and environmental concerns (Aboubaraka et al. 2017; Al-Husseini et al. 2018).

Based on the above reasons, the World Health Organization (WHO) proposed that acceptable levels of turbidity in drinking water must be less than 5 NTU (Al-Husseini et al. 2018). Therefore, to achieve WHO standards various chemical-physical technologies such as sequencing batch reactor (SBR) (Azimi et al. 2019), ceramic membranes (Park et al. 2020), ozonation (Hajiali & Pirumyan 2014), and coagulation (Zhang et al. 2018) have been used for removing turbidity from water. Among them, the coagulation–flocculation (CF) process is the most reliable technology to reduce any pollutants (turbidity, particulates, and organic matters) from drinking water (Zhao et al. 2012).

Previous studies confirmed that the CF process has been extensively applied in water treatment plants (WTPs) because of its superior properties such as simplicity, cost-effectiveness, high efficiency, non-toxic method, and low energy consumption (Lim et al. 2018; Ozairi et al. 2020). Typically, the CF process has been performed by using various coagulants such as iron-aluminum salts, and long-chain polymers for the removal of turbidity from raw water (Baruth 2005). Previous works indicated that aluminum salts (alum) with a chemical formula (Al₂(SO₄)3.18H₂O) has a better potential for the removal of turbidity than ferric salts. However, alum cost is higher than ferric chloride (Gobena et al. 2018; Kumari & Gupta 2020). Studies have shown that there is a direct correlation between the use of alum and neuropathological diseases such as Alzheimer's (Huang et al. 2015). Therefore, it is necessary to optimize the main factors such as coagulant dosage, mixing speed, effluent pH, temperature, and time, which have a direct effect on the CF process (Corral Bobadilla et al. 2019).

In recent years, several methods such as an artificial neural network (ANN) (Singh & Gupta 2012), adaptive neuro-fuzzy inference system (ANFIS) (Kim & Parnichkun 2017), multi-layer perceptron (MLP) (Gagnon et al. 1997), generalized regression neural network (GRNN) (Specht 1991), and response surface methodology (RSM) (Corral Bobadilla et al. 2019) have been applied for modeling and optimization of the CF process. Among the aforementioned methods, RSM as a statistical method has been widely used for analyzing, modeling, and optimizing the effect of different variables and their responses on the system (Corral Bobadilla et al. 2019). Furthermore, it can be used in various systems to obtain maximum performance with the ability to decrease the number of examinations (Singh & Kumar 2020).

Muyibi & Alfugara (2003) used alum and Moringa oleifera seed as coagulants for the removal of turbidity from water at different levels; low turbidity (21.5–49.3 NTU), moderate turbidity (51.8–114 NTU) and high turbidity (163–494 NTU). Their results showed that the process can achieve the minimum residuals of turbidity, 1.4, 1.9, and 0.9 NTU, using Moringa oleifera, alum, and alum with Moringa oleifera, respectively (Muyibi & Alfugara 2003). Another study by Baghvand et al. (2010) reported that the highest turbidity removal by using alum and ferric chloride as coagulant was 82.9–99.0 and 92.9–99.4%, respectively (Baghvand et al. 2010). Agbovi & Wilson (2019) examined and optimized the removal efficiency of turbidity using an amphoteric chitosan-based flocculant–ferric chloride by RSM. They found that the turbidity removal under optimum conditions was 96.7%. Furthermore, the results showed that the process is highly dependent on the pH, FeCl₃ dosage, and the flocculant dosage (Agbovi & Wilson 2019). Usefi & Asadi-Ghalhari (2019) used the coagulation–flocculation process in the elimination of turbidity using rice starch and the system was optimized using the central composite design (CCD) approach. The results demonstrated that 98.4% of turbidity was removed at the optimal point. Moreover, they showed that among the four independent variables (pH, settling time, rice starch dosage, and slow mixing), pH had the most significant effect on the turbidity removal (Usefi & Asadi-Ghalhari 2019).

Therefore, due to the importance of optimizing the coagulant factors, herein, the influences of operational parameters (coagulant dose and pH, rapid mixing, slow mixing, and sedimentation time) on the removal of turbidity were investigated using the CCD-RSM method. Furthermore, this is the first study that has completely described the role of rapid mixing; slow mixing on the coagulation-flocculation process.

Commercial aluminum sulfate (Al₂ (SO₄)₃·14H₂O; 17%), with a molecular weight (M_W) of 342.15 g/mol was purchased from Goglagh Company, Iran. All reagents with high analytical grade were purchased from Merck Company, Germany.

To prepare turbid water in desired turbidities, clay soil (without any modification) from the suburb of Kermanshah city was collected and dried at 150 °C for 150 min using an oven (Memmert 854, Germany). The synthetic turbid water was prepared daily using 10 g of dried clay soil in 2 L of tap water and mixed at 30 rpm for 1 hour to complete hydration of the particles. Consequently, the desired range of turbid water (10–350 NTU) was prepared using the stock solution.

The results of the ANOVA are shown in Table 3. According to the table, it is clear that terms such as A, B, D, E, F, AB, AE, AF, BF are significant (P-value < 0.05). The quality of model fitness (Equation (3)) was investigated by the R² coefficient (Noordin et al. 2004; Almasi et al. 2017), in which R² and adjusted R² were 0.8547 and 0.7871, respectively. The high adequate precision (greater than 4) and low coefficient of variation are attributed to the high accuracy and reliability of the proposed models (Zangeneh et al. 2016; Nadarajan et al. 2018). Herein, adequate precision and CV values were 17.20 and 52.41, respectively. Based on the results, the quadratic model was more suitable for the dataset. Furthermore, the F-value of the model (12.64) confirmed the significance of the model. Based on Equation (3), the initial turbidity, settling time, pH, coagulant dosage, and rapid mixing had a positive and significant effect on turbidity removal.

The experiments were performed to better understand the impact of varied initial turbidity (10–350 NTU) and settling time (10–50 min) on the turbidity removal. The simultaneous effect of the initial turbidity and settling time on the turbidity removal is shown in Figure 2. As can be observed, when the initial turbidity increased from 10 to 350 NTU the removal efficiency of the system increased by more than 99%. On the other hand, minimum residual turbidity was attained at the initial turbidity of 350 NTU. The lower elimination of turbidity at low initial turbidity (10 NTU) may be due to the particle size reduction, which creates smaller flocs with a lower tendency for settling (Camacho et al. 2017). Furthermore, Figure 2 shows that the maximum efficiency of 99.14% was achieved at a settling time of 50 min so that when the settling time increased to 50 min, the minimum residual of turbidity (3 NTU) was attained.

Senthil Kumar et al. (2016) reported that the Moringa oleifera seeds could eliminate turbidity from underground water. The results completely demonstrated that by increasing the initial turbidity from 50 to 135 NTU, the removal efficiency increased from 54.67 to 74.28%. It indicates that at higher turbidity concentrations more interaction between coagulants and colloidal particles takes place (Senthil Kumar et al. 2016). Ramavandi (2014) used the Plantago ovata as a coagulant to remove turbidity, in which the turbidity varied from 50 to 300 NTU. The study showed that at the first stage of the process with an increase in initial turbidity from 50 to 250 NTU, the turbidity removal efficiency decreased from 99 to 95%, further increase in initial turbidity from 250 to 300 NTU was the cause of increase in the removal efficiency of the process (Ramavandi 2014). Daryabeigi Zand & Hoveidi (2015) compared the efficiency of two commercial coagulants (aluminum sulfate and poly-aluminum chloride) under similar conditions (dosage of 10–20 mg/L, pH 4–8) for treatment of high turbidity (10–1,000 NTU). The results confirmed that initial turbidity had a positive and significant effect on turbidity removal so that when turbidity increased to 500 and 1,000 NTU, the turbidity removal efficiency also remained high (Daryabeigi Zand & Hoveidi 2015).

Altaher et al. (2016) represented the turbidity removal of alum with a coagulant aid. The results showed that with increasing settling time, the amount of residual turbidity decreased, indicating that settling time plays a very important role in the CF process (Altaher et al. 2016). Similar results by Sasikala & Muthuraman (2017), in which the natural coagulants were used to remove turbidity from surface water, explicitly declared that by increasing settling time in the coagulation process, the turbidity removal efficiency increased (Sasikala & Muthuraman 2017). Also, Mohammed & Shakir (2018) reported that the removal of residual turbidity increased when settling time increased to 20 min.

Generally, one of the most challenging parameters in the CF processes is pH, and the only reliable way to determine the appropriate range of this parameter is by performing laboratory-scale tests (Orooji et al. 2016). On the other hand, changing the pH value influences the substance hydrolysis load, and therefore pH optimization is very important to achieve suitable efficiency in CF processes (Al-Husseini et al. 2018). In this section, the simultaneous effects of the desired pH (5–9) and coagulant dosage (50–250 mg/L) are discussed. Based on Figure 3, it can be observed that the increase in the pH from 5 to 9 could increase the turbidity removal to 99.14%. In this study, we used alum as a coagulant, and the best removal efficiency occurred at pH 9. It is better to state that the pH range in the alum is between 5.5 and 8.5 (Altaher 2012).

The coagulant dosage can be deemed as one of the operational parameters for determining the optimal conditions of the process (Bazrafshan et al. 2015). Moreover, it is necessary to adjust the coagulant dosage because increasing the coagulant dosage may lead to an increase in operating costs, sludge production, shorten filter life, and decrease alkalinity (Ødegaard et al. 2010). According to Figure 3, it is clear that when the coagulant dosage was boosted to 250 mg/L, the turbidity removal declined, and the best condition was observed at a low coagulant dosage (50 mg/L). Several studies have shown the effect of pH and coagulant dosage on the CF process. Mohammadi-Moghaddam et al. (2015) studied the feasibility of polyaluminum ferric chloride for the treatment of highly turbid water (250 and 500 NTU) and pH value varied from 5 to 11. Their results indicated that increasing pH (more than 9) could result in decreasing the efficiency of the system, therefore, they chose a pH range of 7–8.5 to obtain the minimum residual turbidities (≤0.6 NTU) (Mohammadi-Moghaddam et al. 2015). Hussain et al. (2019) applied pine cone extract as a natural coagulant for the treatment of turbid water, and the pH value was varied from 2 to 12. Their results showed that by increasing the pH from 2 to 7, the removal efficiency of the process decreased from 75% to almost 45%, but with a further increase in pH from 7 to 12, the removal efficiency increased significantly (Hussain et al. 2019).

Huang et al. (2016) investigated the turbidity removal using a titanium salt family as a coagulant. The results showed that by increasing the coagulant dosage, the efficiency of turbidity removal increased (Huang et al. 2016). Chen et al. (2015) revealed that when the dosage of coagulant increased, the turbidity removal efficiency was also increased (Chen et al. 2015). Chekli et al. (2017) used different coagulants for the removal of turbidity and their results demonstrated that by increasing the dosage of coagulants, the removal efficiency of the CF process also increased to 95% (Chekli et al. 2017). Dehkordi et al. (2017) evaluated the effectiveness of three coagulants including polyaluminum chloride, alum, and ferric chloride, with a coagulant aid to remove turbidity in slopes ranging from 50 to over 20,000 NTUs. The results demonstrated that the removal efficiency of the process increased as coagulant dosage increased (Dehkordi et al. 2017).

Rapid mixing is an important parameter that has been investigated in CF processes in which chemical coagulants are combined with freshwater to promote particle destabilization (Ramphal & Sibiya 2014). It is better to mention that rapid mixing in a shorter time increases the remaining turbidity and creates larger flocs (BinAhmed et al. 2015). On the other hand, a high value of G, which is called the rapid mixing phase in the CF processes, usually facilitates the suitable dispersion of the added coagulant to the suspension and also improves contact between the particles, the soluble materials, and the coagulant molecules (Sheng et al. 2006).

The primary purpose of mixing is to keep particles in the suspended state, additionally, slow mixing can provide a velocity gradient for collisions between particles larger than 1 μm (Zhang et al. 2013). Furthermore, the slow mixing speed must be sufficient to maintain suspending particles without floc breakage (Rossini et al. 1999). Based on the obtained results (Figure 4), it was concluded that when rapid mixing increased to 280 rpm, the efficiency of turbidity removal improved, and maximum efficiency was 99.14%. Besides, as observed, by increasing the slow mixing from 30 to 50 rpm, the efficiency of turbidity removal increased slightly. Rossini et al. (1999) investigated the effect of the rapid mixing parameter in the CF process, and the results showed that the high value of the velocity gradient in the rapid mixing could result in creating lower residual turbidity (Rossini et al. 1999). Kan et al. (2002) tested the effects of varied rapid mixing intensity (25, 80, 200, 350, and 600 s^–1) by PACl. The results confirmed that when rapid mixing intensity increased, the turbidity removal decreased (Kan et al. 2002). The results of a study by Zhang et al. (2013) verified that when the slow-mixing duration is within a certain range, such as t < 15 min at G = 15 or 38 s⁻¹, the residual turbidity declined with slow-mixing duration. However, when increasing the slow-mixing duration such as t > 15 min at G = 15 or 38 s⁻¹, the residual turbidity increases, even when the G value is as low as 4 s^–1 (Zhang et al. 2013).

The empirical results were optimized by Design-Expert software to produce an overlay plot. The amount of residual turbidity and alum dosage on the turbidity removal as responses were optimized (Figure 5). As can be seen, the optimal condition of the experiments was obtained under rapid mixing of 191 rpm, slow mixing of 43 rpm, pH 5.5, and settling time of 36 min, and high removal efficiency was between 90 and 100%.

In this study, a high concentration of turbidity (10–350 NTU) was removed by alum in batch experiments using the CF process. To optimize the impact of independent variables on the CF system, CCD based RSM was used. The results of the research revealed that the initial concentration of turbidity had the most significant role in the system so that 99.14% turbidity was eliminated at 350 NTU. Furthermore, under the optimum conditions of rapid mixing (191 rpm), slow mixing (43 rpm), pH (5.5), and also settling time (36 min), the removal of turbidity varied between 90 and 100%. This study highlighted that the CF process is the most reliable chemical method for removing highly turbid water.

The authors hereby express their gratitude to the Water and Wastewater Company of Kermanshah Province for financial support (Grant number: 129/95/21868), and Kermanshah University of Medical Sciences for supplying laboratory facilities.

All relevant data are included in the paper or its Supplementary Information.