Next Article in Journal
Flood Estimation and Control in a Micro-Watershed Using GIS-Based Integrated Approach
Previous Article in Journal
Waterborne Disease Risk Assessment and Mapping for a Floating Village by Combining 3D Hydraulic Simulation and Quantitative Microbial Risk Assessment
Previous Article in Special Issue
Unlocking the Potential of Wastewater Treatment: Machine Learning Based Energy Consumption Prediction
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:

Machine Learning Techniques in Dosing Coagulants and Biopolymers for Treating Leachate Generated in Landfills

HYDROLAB, Centro de Investigación, Innovación y Transferencia de Tecnología, Universidad Católica de Cuenca, Cuenca 010102, Ecuador
Grupo de Investigación Ambiente, Ciencia y Energía de la Universidad Católica de Cuenca, Cuenca 010102, Ecuador
Escuela de Arquitectura, Universidad Católica de Cuenca, Cuenca 010102, Ecuador
Author to whom correspondence should be addressed.
Water 2023, 15(24), 4200;
Submission received: 17 August 2023 / Revised: 11 September 2023 / Accepted: 14 September 2023 / Published: 5 December 2023
(This article belongs to the Special Issue Mathematical Modelling and Model Analysis for Wastewater Treatment)


The leachate discharges generated in sanitary landfills contain many pollutants that are harmful to the environment; treatments are scarce and should be carried out better. The use of coagulation–flocculation processes has been one of the most widely used, but due to the complexity of the characterization of the leachate, the dosing strategy of coagulants and biopolymers needs to be clarified. Therefore, the present study was carried out to determine the doses of coagulants and biopolymers suitable for coagulation–flocculation processes in the treatment of leachates using computational models of machine learning techniques such as artificial neural networks (ANNs); these allow for decreasing the operations of the tests of jars in the laboratory, optimizing resources. Through laboratory experimentation, there are real results of the effectiveness of applying biopolymers in leachate treatments at different concentration levels. The laboratory results were taken as input variables for the algorithms used; after the validation and calibration process, we proceeded to estimate predicted data with the computational model, obtaining predictions of optimal doses for treatment with high statistical adjustment indicators. It is verified that the applied coagulation–flocculation treatments reduce the turbidity values in the leachate and contaminants associated with suspended solids. In this way, the jar tests are optimized so that the operational costs decrease without affecting the results of adequate dosing.

1. Introduction

Leachate is a liquid from the decomposition of solid waste that is generally found in landfills where precipitation in the area has an impact, producing an infiltration that affects the soil and bodies of water [1]; it is necessary that the leachate is purified before dumping to avoid environmental problems and achieve a better operation of the landfill taking advantage of the use of biogas [2]. Within the characterization of landfill leachate, organic, inorganic, and heavy metal compounds can be found; according to the age of the landfill, the leachate can be acidic or alkaline. The leachate presents a high concentration of anions such as chloride, nitrate, and phosphates and cations such as sodium, magnesium, calcium, and iron [3]. These pollutants, upon reaching water bodies, are a clear tracer of landfill leachate. For such reason, processes are proposed to eliminate them in their generation.
Leachate discharges in water bodies is a problem that has been considered for some years. Several treatment alternatives have been proposed, often using techniques for wastewater treatment [4], such as biological processes, but these have their limitations due to the low concentration of biodegradable compounds. The BOD5/COD ratio decreases as the lifespan of the landfill progresses, and to this is added the presence of toxic compounds that inhibit the treatment of biological processes that do not yield suitable results in individuals, so composite processes that integrate aeration, coagulation–flocculation, and membrane treatments such as reverse osmosis have been tested [5,6]. The coagulation–flocculation process has been widely accepted because of the characteristics possessed by the leachates and because of its relative simplicity in both application and evaluation by measuring turbidity as a result of the process [7]. Perhaps the initial complexity of a jar test is the biggest problem for the application of this method since the characteristics of the leachate evolve with the passage of time and external characteristics, which is why the jar test must be recurrent and periodic.
Primary treatment is a chemical treatment used in leachate treatment to enhance the removal of suspended solids, organic matter, and nutrients. In the process, chemical coagulants are usually added to primary sedimentation. This process can help reduce the loading rate of solids and organic matter in biological treatment, treatment infrastructure requirements, and total cost [8]. Efficiency in a primary treatment facility depends on the type and dosage of the coagulant, pH level, temperature, and alkalinity [9]. For these reasons, it is essential to know the age of the leachate depending on the sampling location within the landfill and the initial characteristics before applying the treatments. Coagulation–flocculation as a primary treatment can be effective in removing COD, color, turbidity, and metals depending on the type of coagulant/flocculants and the contaminants to remove [10,11,12].
Environmental processes are commonly represented by nonlinear behavior, which leads to applying nonlinear complex models because analytical solutions are not capable of solving them. Mathematical models in wastewater treatments allow processes to become predictable, and at the same time, these can be optimized, saving money and time. Recently, studies have been conducted to model water and wastewater treatment processes using machine learning (ML) data-driven models, such as decision tree (DT), artificial neural network (ANN), support vector machine (SVM), random forest (RF), and K-nearest neighbor (KNN) methods [13,14,15,16,17,18,19,20,21]. This highlights the great use that can be made of these techniques to optimize processes focused on water treatment.
Machine learning models mostly used to model wastewater treatments are ANN. ANN has been widely used in environmental disciplines. This technique is an effective data drive tool that helps in modeling different environmental processes [22,23]. ANN is trained to be capable of simulating nonlinear relations between input and output variables based on experimental data [24] to yield results close to the output target variable [25]. On the other hand, SVM is categorized as a new neural network algorithm used for forecasting. It is applied for classification and regression problems as a nonlinear and kernel-based modeling approach [26]. Also, SVM can simultaneously minimize estimation residuals, minimizing the upper limit of generalization error by improving the generalizability of traditional models [26]. Likewise, KNN is one of the easiest and simplest learning algorithms. It works with regression and classification as supervised learning approaches, and the classification process is performed using the highest vote of the k closest training points [27].
Coagulation has been applied to remove dyes from different types of contaminated wastewater [28] and subsequently remove dissolved materials through a flocculation process. The coagulant properties are important in determining coagulation performance, as they can either enhance or hinder the removal of pollutants [29]. Other parameters that must also be considered include coagulant dosage volume, pH, mixing rate and duration, temperature, and settling time [30].
Given that this process necessitates control and knowledge of these variables across a broad spectrum, which is contingent on the leachate’s concentration, it is common to conduct numerous dosing trials in the jar test process. Often, these trials depend on the experience of the technician conducting the test. The laboratory process, which was employed as input data for this research, is depicted in Figure 1.
Mathematical and statistical models are presently employed to investigate the relationships between multiple independent variables and one or more response variables. These methodologies enable the creation of experimental models and the fitting of empirical models to the data acquired through the experimental design previously selected by Khuri [31].
This approach has been extended to leachate treatment, as exemplified by the research conducted by Igwegbe et al. [17]. In their study, Igwegbe assessed the biocoagulation–flocculation (BCF) of municipal solid waste leachate using Picralima nitida extract and applied both RSM (response surface methodology) and ANN (artificial neural network) models for analysis. Notably, the coagulation process is often regarded as the cornerstone of wastewater treatment plant technology [32], and the exploration of various optimization techniques for finding suitable solutions in processes is currently under evaluation [33,34].
Wastewater treatment processes are frequently complex due to the substantial volume of data they manage, inherent variability, and nonlinear behavior. In the realm of wastewater treatment, biochemical semi-mechanistic processes currently dominate research, with computational techniques emerging as valuable complements for process optimization [35]. The treatment of leachate processes arises as a critical necessity to cleanse these environmentally harmful liquids before they can contaminate natural water bodies, thereby mitigating pollution concerns. Given the high pollutant loads present in these liquids, the refinement of processes and the application of innovative techniques become imperative in achieving purification objectives.
This research experiment proposes an alternative to optimize the jar test and decrease the test application time through the use of machine learning techniques such as artificial neural networks, support vector machines, and K-nearest neighbors. For this purpose, the leachate treatment process is simulated by adding a coagulant in various concentrations. Then, statistical methods in the validation of the models were applied with the purpose of simplifying the subsequent treatment process.

2. Materials and Methods

2.1. Study Area

The leachate sample was taken at the “Pichacay” landfill. The Pichacay landfill has been located in the Santa Ana parish, in the town of Pichacay, 21 km from the city of Cuenca, in the province of Azuay in southern Ecuador, since 2001. Figure 2 shows the location of the landfill where the samples were collected.
This landfill operates under public administration and serves 7 of the 15 cantons of the province of Azuay: El Pan, Sevilla de Oro, Guachapala, Sigsig, Chordeleg, Gualaceo, and Cuenca, burying approximately 92.7% of the total solid waste [36].

2.2. Leachate Sample

The samples were collected directly in the North 2 phase. The leachate generated in the cells reached a pipe, which is easily accessible. About 10 gallons of leachate sample was collected to provide sufficient quantity for multiple tests. They were then taken to the HYDROLAB water quality laboratory at the Catholic University of Cuenca and stored at 4 °C prior to further testing and analysis. To choose the parameters that were analyzed in laboratories, we evaluated the typical parameters for leachate characterization as presented in studies such as Hussein [37] and Feng [38]. Table 1 shows the raw leachate characterization values. The parameters analyzed indicate an alkaline leachate with pH values above neutral (>7). The values for COD, BOD, and nutrients are not as high as those reported by other studies, such as Patel [39]. In this context, it is necessary to analyze the age of the leachate, which is a fundamental parameter in the composition of the liquid.
For the coagulation and flocculation of the leachate, we worked with hexahydrated ferric chloride (100.3%). As a control parameter of the efficiency, we measured the turbidity in NTU. This coagulant was chosen because it is industrially accepted and widely applied in water and wastewater treatment in primary treatment. Stock solutions of the coagulants were prepared and stored at 4 °C for experimental use. The concentration of prepared stock solutions of ferric chloride was 8.5 g/L. Application of stock solution is preferred compared to solid coagulant for testing since the dissolved coagulants can mix rapidly compared to the solid coagulant [40].
In addition to working with ferric chloride, a biopolymer obtained from cassava starch was also tested.

2.3. Experimental Setup

The objective was to perform a jar test that mimicked the treatment conditions, obtain the best dosing parameter, and run the machine learning turbidity as the control parameter because of its ease of implementation in the laboratory and low cost.
The rapid mixing was carried out at 70 rpm for one minute. After the time required for the quick mixing, the agitation speed was decreased to 30 rpm for 20 min. The coagulant and the biopolymer were added to guarantee the union of the small flocs formed in the coagulation. After this, a sedimentation test was carried out for 30 min.

2.4. Application of Artificial Neural Network (ANN)

Linear and nonlinear algorithms were used to analyze the best prediction method. Within the linear algorithms, linear regression models (LMs), generalized linear models (GLMs), and penalized linear models (GLMnet) were tested. For the nonlinear algorithms, we worked with support vector machines (SVMs), classification and regression trees (CARTs), and K-nearest neighbors (KNNs).
The laboratory test data were organized for use and processed in the R Studio program. R Studio was used to validate the dataset to confirm the final model’s accuracy; a percentage was allocated for training data (80%) and validation data (20%). For a better evaluation of the dataset, we used the 10-fold function for cross-validation, which allowed us to evaluate the linear and nonlinear regression algorithms that would work for this case; the algorithms were evaluated using the MAE, RMSE, and R2 metrics.
The set of algorithms (linear and nonlinear) used detailed the interactions that must be fulfilled for the ANN process so that the best process would be presented and thus replicate the neural network to predict the estimated data. Finally, laboratory tests were repeated to validate the predicted results.

3. Results

3.1. Relation of Water Quality Parameters

Turbidity, pH, and suspended solids parameters were measured in the initial tests and related to the dosage of ferric chloride and organic polymer.
The control parameter for treatment efficiency was turbidity, which is related to the suspended solids in the leachate. This has made it possible to evaluate the removal efficiency according to the dosage applied to have a basis for analysis to test the neural networks.
The turbidity measurement was performed for sufficient data at various scales to apply the proposed algorithms. In addition, the COD point measurements were entered into the experiment to verify the efficiency of the treatment.
The initial and final turbidity measurements show nonlinear changes, meaning they do not offer a consistent behavior since the initial turbidity values were 1000 NTU, the highest value read by the turbidity meter, and the lowest turbidity value obtained was 2.39 NTU. The removal percentages provide an idea of how effective the process was. The following are some examples of the removal percentages in the leachate experiment (Table 2). The equipment used for the determination is HACH model 2100Q.
According to Table 2, it is evident that the lowest efficiency is achieved at the lowest coagulant dosage (0.5 mL), the best removal percentage (99.01%) is performed with 9 mL, and then the removal begins to decrease, indicating a nonlinear relationship, if not with an adequate dosage zone.
To verify an adequate treatment of leachate contaminants, the presence of organic matter measured as COD has been related to turbidities, whose measurements will be used later to validate the algorithms.
Figure 3 shows the results of measuring these parameters over time with the most effective dosage (9 mL). Reducing the turbidity for this test from 164 to 27 NTU with an efficiency of 83.5% and decreasing the COD from 2727 to 801 mg/L with an efficiency of 70% thus validated the relationship that exists between the decrease in suspended solids and COD with a similar tendency although with lower efficiencies.

3.2. Relationship of the Parameters with the Dose of Coagulant

A matrix of correlations between the variables considered for the jar test trials is presented below (Figure 4). This matrix is calculated to use the input or output variables for the ANN subsequently.
One of the highest correlations observed is between turbidity and coagulant dosage, which is why this parameter is used to test the neural network. All the relationships are in numerical form; for a better analysis, they are detailed in Table 3.
According to Table 3, the parameters whose intersection has a number closer to a positive or negative one are the ones that have the best relationship between them, both directly proportional (+) and indirectly (−). It can be seen that the best correlation is the turbidity with the dosage of organic polymer (−0.814), which indicates that the higher the dosage of the polymer, the lower the turbidity in the leachate.
The relationship between turbidity and dosage of the coagulant is relatively high compared with the other parameters and, like the previous one, is negative (−0.444), indicating a decrease in turbidity according to the dosage of ferric chloride.

3.3. Results Related to Observed Data, Simulated Data, and ANN Algorithms

As mentioned in the methodology section, linear and nonlinear algorithms have been tested to train a neural network to obtain optimal dosing of the coagulants to obtain the best turbidity parameter, optimizing the leachate treatment jar test. Simulations have been run for each proposed network to verify the best fit.
The input hyperparameters (Table 3) have been adjusted to various settings within the algorithms, resulting in predictions of input turbidity, which were then compared with the actual measurements to determine the error. Examples of these results are provided in Table 4.
Table 5 presents the metrics used to validate the best-fit algorithm. Mean absolute error (MAE), root mean square error (RMSE), and root mean square error (R-squared) were tested. The purpose of using three indicators is to determine which algorithm best fits the scenario when two metrics have comparable values. R-squared indicates the proportion of variation in the response variable y that is accounted for by the independent variable X in the linear regression model. A higher R-squared value indicates that the linear regression model explains a larger amount of the variability.
The nonlinear CART and KNN algorithms appear to have similar results and show slightly bad errors, and these algorithms have the best fit to the data in their R2 measures.
SVM has the lowest MAE and RMSE metrics and a higher R-squared, indicating that it is the best-fitting algorithm, so it will be used to predict data. Figure 5 shows the observed and simulated data using the SVM application.
As already indicated by the previous metrics analyzed, the best fit is for the SVM algorithm; it can be seen in the figure above that there is also a specific zone in the dosing of the coagulants where the fit is the best. This gives a fundamental criterion to optimize the test since it allows us to know in which dosages the turbidity (and associated contaminants) will be the lowest.
A new experiment is proposed to validate the use of the artificial neural network, working with dosages that yielded the best adjustments, according to Figure 4. A working range from a dosage of 0.1 mL to 22 mL is presented, classified in three experiments with several proposed trials. The fields of the experiments are 0.5–13 mL, 3–17 mL, and 8 to 22 mL. Figure 6 shows the turbidity results for each of the ranges proposed.
In the jar test trials carried out about the optimum dose, multiple results were achieved about the chloride dosage and the final turbidity. Figure 4 shows the results obtained in the experimentation with final turbidity ranging from 5 to 3 NTU. They were the best experiments and demonstrated that the optimal doses varied in different milliliters for the coagulation to be effective and obtain a suitable leachate quality.

3.4. Verification of the SVM

To verify that the SVM works and predicts the estimated values, we work with different data between the ranges already obtained in the laboratory tests. These observed data are changed to calculate predicted values, i.e., the chloride dose is estimated as a function of the input variables.
Table 6 shows the input data to the neural network according to the ranges previously obtained. By changing the observed data, we estimate predicted values; this means that the chloride dose is calculated as a function of the input variables.
The dosage of starch, turbidity, solids, pH, and dilution are the input variables taken by the algorithm to estimate the doses of ferric chloride necessary for the treatment; with the input data, the neural network is run again with the algorithm that obtained the best adjustment, which was SVM. It also shows the estimated chloride data for the different variables. An external cross-validation of the data obtained through the neural network and through experimentation has been carried out, comparing the values between them. To achieve this, we have modified the algorithm to use the experimental dosage as the input parameter, allowing us to estimate turbidity. Dosage test ranges were made from simulated data (4–17 mL). These data are presented in Table 7.
The analysis of turbidity values for validation reveals variations in error magnitudes. While the SVM algorithm provides suitable adjustment results, it is important to acknowledge that in certain applications, additional validation using different parameters may be necessary.
The results display predicted turbidity data as a function of chloride dosage since a significant portion of the generated data focused on these parameters in an initial process optimization test. While it has been established that there is a correlation between the removal of suspended solids and the elimination of the suspended COD fraction, this research has identified a gap in pinpointing a specific factor or mechanism that predicts this behavior.

4. Discussion

Landfill leachate has a characterization depending on the age of the leachate and the parameters that characterize it; this is already an indicator of the type of treatment to which it can be subjected. According to the parameters obtained in the characterization before the experiment, it can be assumed that it is medium-aged leachate [41]. The parameters of COD, BOD, BOD/CBD ratio, pH, and suspended solids indicate that the leachate may have an age of about 5 to 10 years, according to what Teng [42] reports.
Machine learning techniques have been used to study the problem of leachates. From determining the COD load, they can generate [43] or optimize various treatment objectives for leachate generated at a landfill site [44].
Although the efficiency of the study is tested by measuring turbidity because it is a quick and inexpensive parameter since several data are needed to validate the algorithm, this parameter indicates a high efficiency in the leachate treatment. In the study conducted by Cheng [45], an efficiency higher than 58% is achieved in COD removal by applying polyaluminium ferric chloride. In the research developed by Lee [46], the efficiency in the removal of suspended solids is tested by measuring turbidity. In that investigation, the turbidity value decreased with increasing coagulant dosage. The turbidity value decreased steadily at 5 g/L dosage, and 10 g/L dosage approaches the turbidity value (25 NTU). These results indicate that the optimum dosage required to precipitate the fine soil particles is 10 g/L. In this research, the optimal ferric chloride dose was 0.1–22 mL with a concentration of 40 g/L, a relatively higher value. Still, the matter can be justified when working with leachates with high concentrations of solids.
The study presented by Patel [40] tests several coagulants, including ferric chloride, for leachate treatment, showing significant decreases in several pollutants, such as total organic carbon and COD.
COD removal has been related to turbidity reduction, reaching COD efficiencies of 70%, similar to research conducted by Masouleh et al. [47], who achieved 80.8% removal with the application of UV and peroxymonosulfate and the research undertaken by El Mrabet [48], who reached efficiencies of 70% with a persulfate/iron(II)/UV-A irradiation process.
The application of artificial intelligence techniques to optimize leachate treatment processes has been used in recent years. Biglarijoo et al. [49] introduced ferric chloride as a suitable catalyst compared to iron sulfate using the AHP method and applied neural networks together with genetic algorithms to introduce optimal models and conditions. Azadi et al. [50] applied an artificial neural network (ANN) to model the temporal variations of landfill leachate COD in the temporal variations of COD of landfill leachate in the photocatalytic treatment process. Different ANN structures were developed, trained, validated, and tested using data from 150 experiments. The optimal ANN structure was determined based on three performance measures: MAPE, NRMSE, and R.
Some authors have applied coagulation–flocculation as a primary treatment and used machine learning (ML) algorithms to estimate some efficiencies of the treatment. Beshrati [16] used Alyssum mucilage and polyaluminium chloride as coagulants on oily saline wastewater and applied ANN and Adaptive Neuro Fuzzy Inference System (ANFIS) to predict COD removal efficiency based on input parameters (pH, coagulant dose, and contact time), obtaining high accuracy in both models for predicting COD removal. Adaobi [17] used Picralima nitida extract (PNE) on landfill leachate and applied response surface methodology (RMS) and ANN using coagulation–flocculation variables like pH, PNE dosage, and time as inputs to estimate the COD reduction process, obtaining superior accuracy on ANN. Also, Igwegbe et al. [17] used Luffa cylindrical as a biocoagulant in the coagulation–flocculation process to treat dye-polluted wastewater and applied RSM and ANN to predict color, total suspended particles, and COD removal based on biocoagulant dosage, pH, and stirring time as input variables in the model, obtaining better results in ANN. The mentioned authors had better results with ANN models; however, in our case, SVM yielded better results.
Some authors have also predicted turbidity removal efficiencies by applying machine learning. Kusuma [18] tested coagulation–flocculation on synthetic turbid wastewater and predicted turbidity removal efficiency by applying RSM and ANN with input variables like initial turbidity, coagulant dosage, mixing time, and mixing speed, showing that ANN performed better than RSM. Adeogun [19] also used RSM and ANN to predict COD, total dissolved solids, turbidity, and energy consumption based on input variables like pH, time, and current density, obtaining better results for ANN. Likewise, Ezemagu [18] applied RSM and ANN to predict turbidity removal efficiency using dosage time and temperature as input variables. The authors mentioned above showed suitable efficiency in ANN models to predict turbidity removal efficiency; however, in our case, the best model was SVM.

5. Conclusions

Suspended solids present in the leachate, which were measured and correlated with the turbidity parameter, were effectively eliminated through the application of both organic and inorganic coagulants. Impressive removal efficiencies of 98% for suspended solids and 70% for organic matter were attained. These controlled parameters, derived from numerous repetitions, have been utilized as inputs for algorithms aimed at optimizing the process.
In the context of the proposed assays, the support vector machine (SVM) emerges as the optimal technique, exhibiting the most fitting indicators. This achievement was observed across dosages ranging from 0.1 to 22 mL. Subsequently, these established efficiencies were successfully validated through experimental data generated by the algorithm.
This research offers additional applicability to novel techniques aimed at optimizing a crucial process, such as jar testing, as an initial step in leachate treatment. These algorithms can effectively utilize and learn from the expanded dataset generated by establishing an initial dosing range. This knowledge can then be applied to assess and optimize other parameters essential for the effective treatment of this highly polluting waste.
The effectiveness of the ML will depend a lot on the definition of the input variables that directly influence the prediction of the data on the quantity and quality of data that are available for the training and validation set for the respective dose prediction, and on the algorithms to be executed since these will determine that the results are presented with the lowest error metrics and are more accurately predicted results.

Author Contributions

Conceptualization, C.M. and M.Q.; methodology, C.M.; software, D.H.; validation, C.M. and D.H.; formal analysis, D.H.; investigation, C.M.; writing—original draft preparation, C.M.; writing—review and editing, C.M. All authors have read and agreed to the published version of the manuscript.


Research funds for laboratory development CIITT PICCIIT 19-26.

Data Availability Statement

Data available for analysis on the HYDROLAB website (, accessed on 13 September 2023).

Conflicts of Interest

The authors declare no conflict of interest.


  1. Jegan Durai, N.; Gopalakrishna, G.V.T.; Padmanaban, V.C.; Selvaraju, N. Oxidative removal of stabilized landfill leachate by Fenton’s process: Process modeling, optimization & analysis of degraded products. RSC Adv. 2020, 10, 3916–3925. [Google Scholar] [CrossRef]
  2. Simmons, N. Pfas Concentrations of Landfill Leachates in Victoria, Australia-Implications for Discharge of Leachate To Sewer. In Proceedings of the 17th International Waste Management and Landfill Symposium, Santa Margherita di Pula, Italy, 30 September–4 October 2019. [Google Scholar]
  3. Mishra, S.; Tiwary, D.; Ohri, A. Leachate characterisation and evaluation of leachate pollution potential of urban municipal landfill sites. Int. J. Environ. Waste Manag. 2018, 21, 217–230. [Google Scholar] [CrossRef]
  4. Müller, G.T.; Giacobbo, A.; dos Santos Chiaramonte, E.A.; Rodrigues, M.A.S.; Meneguzzi, A.; Bernardes, A.M. The effect of sanitary landfill leachate aging on the biological treatment and assessment of photoelectrooxidation as a pre-treatment process. Waste Manag. 2015, 36, 177–183. [Google Scholar] [CrossRef]
  5. Wiszniowski, J.; Robert, D.; Surmacz-Gorska, J.; Miksch, K.; Weber, J.V. Landfill leachate treatment methods: A review. Environ. Chem. Lett. 2006, 4, 51–61. [Google Scholar] [CrossRef]
  6. Kuusik, A.; Pachel, K.; Kuusik, A.; Loigu, E. Landfill runoff water and landfill leachate discharge and treatment. In Proceedings of the 9th International Conference on Environmental Engineering, Vilnius, Lithuania, 22–24 May 2014. [Google Scholar] [CrossRef]
  7. de Oliveira, M.S.; da Silva, L.F.; Barbosa, A.D.; Romualdo, L.L.; Sadoyama, G.; Andrade, L.S. Landfill Leachate Treatment by Combining Coagulation and Advanced Electrochemical Oxidation Techniques. ChemElectroChem 2019, 6, 1427–1433. [Google Scholar] [CrossRef]
  8. Chagnon, F.J.F.; Harleman, D.R.F. Chemically Enhanced Primary Treatment of Wastewater. In Water Encyclopedia; John Wiley & Sons, Ltd.: Hoboken, NJ, USA, 2005; pp. 659–660. ISBN 9780471478447. [Google Scholar]
  9. Jiang, J.Q. The role of coagulation in water treatment. Curr. Opin. Chem. Eng. 2015, 8, 36–44. [Google Scholar] [CrossRef]
  10. Marañón, E.; Castrillón, L.; Fernández-Nava, Y.; Fernández-Méndez, A.; Fernández-Sánchez, A. Coagulation-flocculation as a pretreatment process at a landfill leachate nitrification-denitrification plant. J. Hazard. Mater. 2008, 156, 538–544. [Google Scholar] [CrossRef]
  11. Boumechhour, F.; Rabah, K.; Lamine, C.; Said, B.M. Treatment of landfill leachate using Fenton process and coagulation/flocculation. Water Environ. J. 2013, 27, 114–119. [Google Scholar] [CrossRef]
  12. Verma, M.; Naresh Kumar, R. Can coagulation–flocculation be an effective pre-treatment option for landfill leachate and municipal wastewater co-treatment? Perspect. Sci. 2016, 8, 492–494. [Google Scholar] [CrossRef]
  13. Ahmadi Azqhandi, M.H.; Ghaedi, M.; Yousefi, F.; Jamshidi, M. Application of random forest, radial basis function neural networks and central composite design for modeling and/or optimization of the ultrasonic assisted adsorption of brilliant green on ZnS-NP-AC. J. Colloid Interface Sci. 2017, 505, 278–292. [Google Scholar] [CrossRef] [PubMed]
  14. Byliński, H.; Sobecki, A.; Gebicki, J. The use of artificial neural networks and decision trees to predict the degree of odor nuisance of post-digestion sludge in the sewage treatment plant process. Sustainability 2019, 11, 4407. [Google Scholar] [CrossRef]
  15. Salem, M.; EL-Sayed Gabr, M.; Mossad, M.; Mahanna, H. Random Forest modelling and evaluation of the performance of a full-scale subsurface constructed wetland plant in Egypt. Ain Shams Eng. J. 2022, 13, 101778. [Google Scholar] [CrossRef]
  16. Fard, M.B.; Hamidi, D.; Yetilmezsoy, K.; Alavi, J.; Hosseinpour, F. Utilization of Alyssum mucilage as a natural coagulant in oily-saline wastewater treatment. J. Water Process Eng. 2021, 40, 101763. [Google Scholar] [CrossRef]
  17. Igwegbe, C.A.; Onukwuli, O.D.; Ighalo, J.O.; Menkiti, M.C. Bio-coagulation-flocculation (BCF) of municipal solid waste leachate using Picralima nitida extract: RSM and ANN modelling. Curr. Res. Green Sustain. Chem. 2021, 4, 100078. [Google Scholar] [CrossRef]
  18. Kusuma, H.S.; Amenaghawon, A.N.; Darmokoesoemo, H.; Neolaka, Y.A.B.; Widyaningrum, B.A.; Anyalewechi, C.L.; Orukpe, P.I. Evaluation of extract of Ipomoea batatas leaves as a green coagulant–flocculant for turbid water treatment: Parametric modelling and optimization using response surface methodology and artificial neural networks. Environ. Technol. Innov. 2021, 24, 102005. [Google Scholar] [CrossRef]
  19. Ejimofor, M.I.; Ezemagu, I.G.; Menkiti, M.C. RSM and ANN-GA modeling of colloidal particles removal from paint wastewater via coagulation method using modified Aguleri montmorillonite clay. Curr. Res. Green Sustain. Chem. 2021, 4, 100164. [Google Scholar] [CrossRef]
  20. Adeogun, A.I.; Bhagawati, P.B.; Shivayogimath, C.B. Pollutants removals and energy consumption in electrochemical cell for pulping processes wastewater treatment: Artificial neural network, response surface methodology and kinetic studies. J. Environ. Manag. 2021, 281, 111897. [Google Scholar] [CrossRef]
  21. Adesina, O.A.; Taiwo, A.E.; Akindele, O.; Igbafe, A. Process parametric studies for decolouration of dye from local ‘tie and dye’ industrial effluent using Moringa oleifera seed. S. Afr. J. Chem. Eng. 2021, 37, 23–30. [Google Scholar] [CrossRef]
  22. Dai, H.; Tian, Y.; Dai, B.; Skiena, S.; Song, L. Syntax-directed variational autoencoder for structured data. arXiv 2018, arXiv:1802.08786. [Google Scholar]
  23. Montáns, F.J.; Chinesta, F.; Gómez-Bombarelli, R.; Kutz, J.N. Data-driven modeling and learning in science and engineering. Comptes Rendus Mécanique 2019, 347, 845–855. [Google Scholar] [CrossRef]
  24. Abunama, T.; Othman, F.; Ansari, M.; El-Shafie, A. Leachate generation rate modeling using artificial intelligence algorithms aided by input optimization method for an MSW landfill. Environ. Sci. Pollut. Res. 2019, 26, 3368–3381. [Google Scholar] [CrossRef] [PubMed]
  25. Abdallah, M.; Abu Talib, M.; Feroz, S.; Nasir, Q.; Abdalla, H.; Mahfood, B. Artificial intelligence applications in solid waste management: A systematic research review. Waste Manag. 2020, 109, 231–246. [Google Scholar] [CrossRef] [PubMed]
  26. Foroughi, M.; Rahmani, A.R.; Asgari, G.; Nematollahi, D.; Yetilmezsoy, K.; Samarghandi, M.R. Optimization and Modeling of Tetracycline Removal from Wastewater by Three-Dimensional Electrochemical System: Application of Response Surface Methodology and Least Squares Support Vector Machine. Environ. Model. Assess. 2020, 25, 327–341. [Google Scholar] [CrossRef]
  27. Dalhat Mu’azu, N.; Olusanya Olatunji, S. K-nearest neighbor based computational intelligence and RSM predictive models for extraction of Cadmium from contaminated soil. Ain Shams Eng. J. 2023, 14, 101944. [Google Scholar] [CrossRef]
  28. Shak, K.P.Y.; Wu, T.Y. Optimized use of alum together with unmodified Cassia obtusifolia seed gum as a coagulant aid in treatment of palm oil mill effluent under natural pH of wastewater. Ind. Crops Prod. 2015, 76, 1169–1178. [Google Scholar] [CrossRef]
  29. Sun, Y.; Zhu, C.; Zheng, H.; Sun, W.; Xu, Y.; Xiao, X.; You, Z.; Liu, C. Characterization and coagulation behavior of polymeric aluminum ferric silicate for high-concentration oily wastewater treatment. Chem. Eng. Res. Des. 2017, 119, 23–32. [Google Scholar] [CrossRef]
  30. Tang, Y.; Hu, X.; Cai, J.; Xi, Z.; Yang, H. An enhanced coagulation using a starch-based coagulant assisted by polysilicic acid in treating simulated and real surface water. Chemosphere 2020, 259, 127464. [Google Scholar] [CrossRef] [PubMed]
  31. Khuri, A.I.; Mukhopadhyay, S. Response surface methodology. WIREs Comput. Stat. 2010, 2, 128–149. [Google Scholar] [CrossRef]
  32. El-taweel, R.M.; Mohamed, N.; Alrefaey, K.A.; Husien, S.; Abdel-Aziz, A.B.; Salim, A.I.; Mostafa, N.G.; Said, L.A.; Fahim, I.S.; Radwan, A.G. A review of coagulation explaining its definition, mechanism, coagulant types, and optimization models; RSM, and ANN. Curr. Res. Green Sustain. Chem. 2023, 6, 100358. [Google Scholar] [CrossRef]
  33. Singh, N.K.; Yadav, M.; Singh, V.; Padhiyar, H.; Kumar, V.; Bhatia, S.K.; Show, P.L. Artificial intelligence and machine learning-based monitoring and design of biological wastewater treatment systems. Bioresour. Technol. 2023, 369, 128486. [Google Scholar] [CrossRef]
  34. Bellamoli, F.; Di Iorio, M.; Vian, M.; Melgani, F. Machine learning methods for anomaly classification in wastewater treatment plants. J. Environ. Manag. 2023, 344, 118594. [Google Scholar] [CrossRef] [PubMed]
  35. Alvi, M.; Batstone, D.; Mbamba, C.K.; Keymer, P.; French, T.; Ward, A.; Dwyer, J.; Cardell-Oliver, R. Deep learning in wastewater treatment: A critical review Maira. Water Res. 2023, in press. [Google Scholar] [CrossRef] [PubMed]
  36. Cobos Mora, S.L.; Solano Peláez, J.L. Sanitary landfill site selection using multi-criteria decision analysis and analytical hierarchy process: A case study in Azuay province, Ecuador. Waste Manag. Res. 2020, 38, 1129–1141. [Google Scholar] [CrossRef]
  37. Hussein, M.; Yoneda, K.; Zaki, Z.M.; Othman, N.A.; Amir, A. Leachate characterizations and pollution indices of active and closed unlined landfills in Malaysia. Environ. Nanotechnol. Monit. Manag. 2019, 12, 100232. [Google Scholar] [CrossRef]
  38. Feng, H.; Mao, W.; Li, Y.; Wang, X.; Chen, S. Characterization of dissolved organic matter during the O3-based advanced oxidation of mature landfill leachate with and without biological pre-treatment and operating cost analysis. Chemosphere 2021, 271, 129810. [Google Scholar] [CrossRef] [PubMed]
  39. Benson, C.H. Characteristics of Gas and Leachate at an Elevated Temperature Landfill. In Geotechnical Frontiers 2017; ASCE: Reston, VA, USA, 2017; pp. 313–322. [Google Scholar]
  40. Patel, H.V.; Brazil, B.; Lou, H.H.; Jha, M.K.; Luster-Teasley, S.; Zhao, R. Evaluation of the effects of chemically enhanced primary treatment on landfill leachate and sewage co-treatment in publicly owned treatment works. J. Water Process Eng. 2021, 42, 102116. [Google Scholar] [CrossRef]
  41. Miao, L.; Yang, G.; Tao, T.; Peng, Y. Recent advances in nitrogen removal from landfill leachate using biological treatments—A review. J. Environ. Manag. 2019, 235, 178–185. [Google Scholar] [CrossRef]
  42. Teng, C.; Zhou, K.; Peng, C.; Chen, W. Characterization and treatment of landfill leachate: A review. Water Res. 2021, 203, 117525. [Google Scholar] [CrossRef]
  43. Azadi, S.; Amiri, H.; Rakhshandehroo, G.R. Evaluating the ability of artificial neural network and PCA-M5P models in predicting leachate COD load in landfills. Waste Manag. 2016, 55, 220–230. [Google Scholar] [CrossRef]
  44. Sabour, M.R.; Amiri, A. Comparative study of ANN and RSM for simultaneous optimization of multiple targets in Fenton treatment of landfill leachate. Waste Manag. 2017, 65, 54–62. [Google Scholar] [CrossRef] [PubMed]
  45. Cheng, Y.; Xu, L.; Liu, C. Red mud-based polyaluminium ferric chloride flocculant: Preparation, characterisation, and flocculation performance. Environ. Technol. Innov. 2022, 27, 102509. [Google Scholar] [CrossRef]
  46. Lee, H.K.; Jun, B.M.; Kim, S.I.; Song, J.S.; Kim, T.J.; Park, S.; Chang, S. Simultaneous removal of suspended fine soil particles, strontium and cesium from soil washing effluent using inorganic flocculants. Environ. Technol. Innov. 2022, 27, 102467. [Google Scholar] [CrossRef]
  47. Masouleh, S.Y.; Mozaffarian, M.; Dabir, B.; Fallah Ramezani, S. COD and ammonia removal from landfill leachate by UV/PMS/Fe2+ process: ANN/RSM modeling and optimization. Process Saf. Environ. Prot. 2022, 159, 716–726. [Google Scholar] [CrossRef]
  48. El Mrabet, I.; Benzina, M.; Zaitan, H. Optimization of persulfate/iron(II)/UV-A irradiation process for the treatment of landfill leachate from Fez City (Morocco). SN Appl. Sci. 2020, 2, 1042. [Google Scholar] [CrossRef]
  49. Biglarijoo, N.; Mirbagheri, S.A.; Bagheri, M.; Ehteshami, M. Assessment of effective parameters in landfill leachate treatment and optimization of the process using neural network, genetic algorithm and response surface methodology. Process Saf. Environ. Prot. 2017, 106, 89–103. [Google Scholar] [CrossRef]
  50. Azadi, S.; Karimi-Jashni, A.; Javadpour, S. Modeling and optimization of photocatalytic treatment of landfill leachate using tungsten-doped TiO2 nano-photocatalysts: Application of artificial neural network and genetic algorithm. Process Saf. Environ. Prot. 2018, 117, 267–277. [Google Scholar] [CrossRef]
Figure 1. Diagram of the jar test execution process for leachate treatment.
Figure 1. Diagram of the jar test execution process for leachate treatment.
Water 15 04200 g001
Figure 2. Location of the landfill where samples were obtained for treatment tests. Source: Google Earth, 2020.
Figure 2. Location of the landfill where samples were obtained for treatment tests. Source: Google Earth, 2020.
Water 15 04200 g002
Figure 3. Relation between COD decrease and turbidity.
Figure 3. Relation between COD decrease and turbidity.
Water 15 04200 g003
Figure 4. Correlation matrix of water quality variables measured in leachates with coagulant dosage.
Figure 4. Correlation matrix of water quality variables measured in leachates with coagulant dosage.
Water 15 04200 g004
Figure 5. Analysis of simulated data by application of the SVM algorithm.
Figure 5. Analysis of simulated data by application of the SVM algorithm.
Water 15 04200 g005
Figure 6. Tests carried out with the dosages chosen in the best fit of the artificial neural network.
Figure 6. Tests carried out with the dosages chosen in the best fit of the artificial neural network.
Water 15 04200 g006
Table 1. Characteristics of raw leachate samples.
Table 1. Characteristics of raw leachate samples.
Hexane soluble compounds (mg/L)23.2 ± 3.2
Alkalinity (mg/L)1251 ± 22.1
Cadmium (mg/L)0.012 ± 0.002
Chemical oxygen demand (COD) (mg/L)2014 ± 8.8
Biological oxygen demand (BOD5) (mg/L)1300 ± 9.3
Total phosphorus (mg/L)2.8 ± 1.3
Total nitrogen Kjedahl (mg/L)90 ± 13.5
Suspended solids (mg/L)398 ± 5.3
Total solids (mg/L)2388 ± 3.2
pH 7.1
Temperature (°C)17.7 ± 1.8
Table 2. Percent removal of solids measured as turbidity.
Table 2. Percent removal of solids measured as turbidity.
Initial Turbidity (NTU)Final Turbidity (NTU) Percentage of Removal (%)Ferric Chloride Dosage (mL)
Table 3. Numerical ratio of ANN parameters.
Table 3. Numerical ratio of ANN parameters.
Table 4. Comparison of observed and simulated results to identify the best algorithm.
Table 4. Comparison of observed and simulated results to identify the best algorithm.
Turbidity (NTU)Observed DataSimulated Data
Table 5. Metrics of adjustment to identify the best algorithm.
Table 5. Metrics of adjustment to identify the best algorithm.
Table 6. Input variables for chloride dose estimation.
Table 6. Input variables for chloride dose estimation.
Test Starch Dosage (mL)Ferric Chloride Dosage (mL)Turbidity (NTU)Solids (mg/L)pHDilution
Table 7. External cross-validation of neural network data and experimentation.
Table 7. External cross-validation of neural network data and experimentation.
Ferric Chloride Dosage (mL) Experimental ResultsTurbidity (NTU) Estimated by SVMFerric Chloride Dosage (mL) Estimated by SVMTurbidity (NTU) Experimental ResultsRelative Error (%) (Turbidity)
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Matovelle, C.; Quinteros, M.; Heras, D. Machine Learning Techniques in Dosing Coagulants and Biopolymers for Treating Leachate Generated in Landfills. Water 2023, 15, 4200.

AMA Style

Matovelle C, Quinteros M, Heras D. Machine Learning Techniques in Dosing Coagulants and Biopolymers for Treating Leachate Generated in Landfills. Water. 2023; 15(24):4200.

Chicago/Turabian Style

Matovelle, Carlos, María Quinteros, and Diego Heras. 2023. "Machine Learning Techniques in Dosing Coagulants and Biopolymers for Treating Leachate Generated in Landfills" Water 15, no. 24: 4200.

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop