From Prediction to Remediation: Characterization of Tropical Landfill Leachates Using ARIMA and Application of Adsorption and Reverse Osmosis Treatments

Abstract

Landfill leachates in tropical regions represent a critical environmental challenge due to their complex composition and pronounced seasonal variability. This study sought to characterize leachates from a tropical landfill in Valledupar, Colombia, and to evaluate advanced treatment technologies for the removal of organic pollutants. An ARIMA (3,0,3) model was implemented on an eight-year time series (2016–2023) of leachate flow data to identify seasonal patterns and support hydraulic load forecasting. Physicochemical characterization was conducted following APHA standard methods, which revealed high levels of COD, BOD₅, chlorides, and lead. Two treatment technologies were assessed independently: (i) adsorption using granular activated carbon in batch and continuous-flow systems, under 36 experimental conditions that combined pH levels (2–7) and carbon dosages (20–120 g); and (ii) reverse osmosis employing polyamide membranes operated at 18 bar and at pH values of 6.0, 7.0, and natural (unaltered) conditions. The results confirmed that leachate generation exhibits clear seasonal variability correlated with rainfall patterns. The Langmuir isotherm demonstrated the best fit at pH 4.0 (R² = 0.9685), and the continuous system achieved 97% COD removal within 90 min. Reverse osmosis consistently removed over 94% of COD and BOD₅ across all pH conditions. These findings highlight the value of integrating time-series forecasting with optimized treatment technologies to support effective and adaptive leachate management strategies in tropical environments.

1. Introduction

Landfill leachates are complex liquid effluents that contain a wide range of organic, inorganic, and heavy metal contaminants, requiring appropriate treatment to prevent ecotoxicological damage and broader environmental degradation [1,2]. The uncontrolled release of leachates into the environment, particularly into surface water bodies and soil, has prompted increasingly stringent discharge regulations aimed at mitigating the adverse effects of waste degradation byproducts [3]. In alignment with global sustainability objectives, the proper management of landfill leachates is essential to achieving the Sustainable Development Goals, particularly those related to clean water, sanitation, and responsible waste management [4].

In tropical monsoon regions such as Valledupar, Colombia, leachate management presents additional challenges due to high precipitation and elevated temperatures. Frequent and intense rainfall events increase the volume of water infiltrating landfills, resulting in higher leachate generation that can exceed the capacity of collection and treatment systems [5,6]. Furthermore, elevated ambient temperatures accelerate the decomposition of organic matter, leading to increased pollutant concentrations and higher production of landfill gases such as methane. Seasonal variability in leachate composition, combined with insufficient waste management infrastructure, further complicates treatment efforts in these environments [1,7].

To address these challenges, it is essential not only to implement efficient remediation technologies but also to anticipate leachate generation dynamics through predictive modeling. In this context, time-series forecasting using Autoregressive Integrated Moving Average (ARIMA) models enables the identification of seasonal patterns and temporal fluctuations in leachate flow, offering valuable insights for proactive system management and optimization.

Among available remediation technologies, granular activated carbon (GAC) adsorption has demonstrated high efficiency in removing a broad spectrum of contaminants due to its large surface area and porosity. However, its performance may be constrained by adsorbent saturation and pH-dependent behavior [1,8,9]. Reverse osmosis (RO), by contrast, offers robust removal of dissolved salts, organic compounds, and heavy metals, producing high-quality permeate. Nevertheless, membrane fouling and operational costs remain key limitations [6,10].

Despite advancements in technology, optimizing leachate treatment under tropical conditions remains a critical environmental challenge. As highlighted by [11], landfill leachates are highly toxic and pose significant ecological risks, particularly in regions characterized by intense rainfall. Therefore, integrated strategies that combine predictive modeling, physicochemical characterization, and remediation technologies are essential for achieving resilient and sustainable leachate management.

The objective of this study is to characterize the behavior of leachates generated in a tropical landfill using ARIMA-based forecasting and to evaluate the performance of adsorption and reverse osmosis in removing organic pollutants. This integrated approach bridges predictive analysis and practical remediation, aiming to enhance the effectiveness and sustainability of leachate management in tropical climatic contexts.

2. Materials and Methods

2.1. Location of the Study Area

The research was conducted in the city of Valledupar (Colombia), the administrative headquarters of the department of Cesar, located near the Sierra Nevada de Santa Marta. The Los Corazones solid waste landfill is located 6 km from the urban center of the departmental capital and covers an area of 120 hectares. It is located on the outskirts of the city, at longitudes 1,083,000 and 1,099,000 E.

2.2. Behavior of Leachates from the Los Corazones Sanitary Landfill—Valledupar, Colombia

Leachate flow was monitored monthly over an eight-year period (2016–2023). Flow rates were estimated using the accumulated volume method, in which leachate is temporarily stored in ponds or treatment units. The stored volume and the corresponding filling time were recorded to calculate the average flow rate, following hydraulic procedures suitable for landfill systems.

Time-series data were subsequently analyzed using the Autoregressive Integrated Moving Average (ARIMA) model. Model validation included the Dickey–Fuller test to assess stationarity and the Ljung–Box test to evaluate residual autocorrelation, ensuring the robustness of the time-series forecasting approach

2.2.1. ARIMA Model Selection

The ARIMA(3,0,3) model was selected based on a comparative analysis with other candidate models (ARIMA(1,0,1), (2,0,2), (3,0,2), (2,0,3)). Evaluation criteria included AIC, BIC, and RMSE, where ARIMA(3,0,3) yielded the lowest AIC (178.3), BIC (184.6), and RMSE (0.112), indicating the best fit

2.2.2. Residual Diagnostics

Model residuals were assessed using the Ljung–Box test (p > 0.05), confirming the absence of autocorrelation. The Shapiro–Wilk test showed that residuals follow a normal distribution (p > 0.05), and the Breusch–Pagan test indicated homoscedasticity (p > 0.05). These diagnostics confirm the statistical validity and reliability of the selected model.

2.2.3. Model Equation and Interpretation

The mathematical structure of the model is presented in Equation (1).

$$ARIMA\left(3, 0, 3\right)=ARMA \left(3 , 3\right)$$

$$Z_t={\varphi }_1{Z}_{t-1}+{\varphi }_2{Z}_{t-2}+{\varphi }_3{Z}_{t-3}+a_t-{\theta }_1{a}_{t-1}-{\theta }_2{a}_{t-2}-{\theta }_3{a}_{t-3}$$

(1)

In Equation (1), Zt denotes the observed value of the time series at time t, representing the current leachate flow. The autoregressive coefficients (${\varphi }_{1 }{\varphi }_{2 }{\varphi }_{3 }$) capture the influence of the three preceding observations Z_t−1, Z_t−2, and Z_t−3 on the current value.

The term a_t represents the random error at time t, accounting for unexplained variability not captured by the autoregressive or moving average components. These residuals are assumed to be white noise: uncorrelated, with zero mean and constant variance.

The moving average coefficients (${\theta }_{1 }{\theta }_{2 }{\theta }_3$) quantify the effect of the three most recent error terms, a_t−1, a_t−2, and a_t−3, on the current value. Thus, the model structure corresponds to an ARMA (3,3) process, incorporating both short-term memory in observations and residuals.

2.2.4. Physicochemical Characterization of Leachate

The contaminant load of leachates from the Los Corazones landfill (Valledupar, Colombia) was assessed through physicochemical analyses conducted in accordance with standard protocols established by the American Public Health Association (APHA) [12] and national discharge regulations. Analytical procedures included inductively coupled plasma mass spectrometry (ICP-MS) and atomic absorption spectroscopy (AAS) for trace metals, complemented by conventional methods for organic and inorganic parameters. (See

Table 1. Physicochemical parameters analyzed and test methods used.

Parameter	Unit	Method of Analysis	Reference
pH	Units	Measurement with a digital potentiometer	APHA4500-H+
Temperature	°C	Digital thermometer	APHA2550
Electrical conductivity	mS/cm	Digital conductivity meter	APHA2510
Chemical Oxygen Demand (COD)	mg O2/L	Colorimetric method with digestion.	APHA 5220 D
Biochemical Oxygen Demand (BOD5)	mg O2/L	Respirometric method	APHA 5210 B
Total suspended solids (TSS)	mg/L	Filtration and gravimetry	APHA 2540D
Settleable solids	ml/L	Imhoff cone method	APHA 2540F
Fats and oils	mg/L	Solvent extraction and gravimetry	APHA 5520 B
Total cyanide	mg/L	Colorimetric method	APHA 4500-CN− C
Chlorides	mg/L	Mohr’s volumetric method	APHA 4500-Cl− B
Sulfates	mg/L	Turbidimetric method	APHA 4500-SO42− E
Heavy metals (Al, As, Hg, Ni, Pb, Se, V)	mg/L	ICP-MS and AAS	APHA 3125 B and 3111 B

).

Results were compared against maximum permissible limits established in Colombian legislation [13], and compliance was evaluated for each parameter. Additionally, the temporal behavior of organic load (COD and BOD₅) was analyzed over a five-year period (2019–2023), with semiannual sampling.

2.3. Procedure for the Removal of Organic Matter in Leachate with Activated Carbon

The removal of organic matter from landfill leachate was evaluated using two experimental setups: (i) a batch system under complete agitation and (ii) a continuous-flow system with a packed bed of granular activated carbon (GAC). Raw leachate was collected directly from stabilization ponds without pretreatment to ensure representative field conditions, including the presence of suspended solids and colloids.

2.3.1. Batch Adsorption System

Batch adsorption tests were conducted to optimize pH and carbon dosage. Mineral-based GAC (Sigma-Aldrich), derived from bituminous coal, was used. Its key properties included a surface area of 950 m²/g (BET), particle size of 1.5–2.0 mm, and porosity of 0.65 (SEM analysis). The contact time was set at 60 min, with experiments conducted at 30 °C and 100 rpm. Each treatment was replicated three times (n = 3).

A 6×6 factorial design was applied to test combinations of six pH levels (2.0–7.0) and six dosages (20–120 g/L). The treatments are summarized in

Table 2. Experimental design—Batch adsorption system.

Treatment	Activated Carbon Dosage (g/L)	pH
T1 (Control)	0	8.2
T2–T7	20, 40, 60, 80, 100, 120	7.0
T8–T13	20, 40, 60, 80, 100, 120	6.0
T14–T19	20, 40, 60, 80, 100, 120	5.0
T20–T25	20, 40, 60, 80, 100, 120	4.0
T26–T31	20, 40, 60, 80, 100, 120	3.0
T32–T37	20, 40, 60, 80, 100, 120	2.0

.

2.3.2. General Statistical Approach

Normality and homoscedasticity were verified using the Shapiro–Wilk and Levene’s tests. A one-way ANOVA was performed to assess differences among treatments, followed by Tukey’s HSD test to identify statistically significant pairwise comparisons (p < 0.05).

To examine combined effects, a MANOVA was conducted using COD and BOD₅ removal efficiencies as dependent variables. Factorial interactions between pH and dosage were analyzed under a 95% confidence level. Multivariate normality and homogeneity of variance-covariance were validated using Box’s M test. Statistical analyses were performed in SPSS (version 28.0).

2.3.3. Isotherm Model Description and Fitting Methods

The adsorption data were fitted to the Freundlich and Langmuir isotherm models to describe equilibrium behavior. The equations used were as follows:

Freundlich isotherm

$$q_e=K_F{C}_e^{{\displaystyle \frac{1}{n}}}$$

(2)

The linearized form is

$$loglog q_e=loglog K_{F }+{\displaystyle \frac{1}{n}} loglog C_e$$

(3)

Langmuir isotherm:

$$q_{e }={\displaystyle \frac{q_{K_L{C}_e }}{1+K_L{C}_e}}$$

(4)

The linearized form is

$${\displaystyle \frac{1}{q_e}}={\displaystyle \frac{1}{q_{max}}}+{\displaystyle \frac{1}{q_{max}{K}_{L }{C}_e}}$$

(5)

Both linear and nonlinear regressions were used to determine model fit and to interpret the adsorption mechanisms based on surface affinity and pH effects.

2.3.4. Continuous Flow Process

To evaluate the adsorption process in a continuous-flow scenario, an acrylic column with a cylindrical structure was manufactured and met the specifications detailed in

Table 3. Properties and operational parameters of the continuous-flow column packed with granular activated carbon for the removal of organic matter from landfill leachate.

Parameter	Value
Inner column diameter (cm)	6.0
Total column height (cm)	60.0
Activated carbon bed height (cm)	35.4
Mass of granular activated carbon (g)	645.0
Estimated bed porosity	0.20
Type of activated carbon	Mineral-based, bituminous origin (Sigma-Aldrich, St. Louis, MO, USA)
Specific surface area (m2/g)	950 (BET method)
Average particle size (mm)	1.5–2.0
Material porosity	0.65 (SEM characterization)

. The column was operated with an upward flow using a peristaltic pump. The soluble COD concentration in the effluent was determined at different times (one-hour intervals) until its concentration equaled that of the inlet liquid.

2.4. Removal of Organic Matter in Leachate by Reverse Osmosis

2.4.1. Experimental Setup

In this phase, a continuous hydraulic flow model was implemented to evaluate the efficiency of reverse osmosis (RO) in removing organic matter from landfill leachate. Three test solutions were prepared from raw leachate: (i) one at its original pH, (ii) one adjusted to pH 7.0, and (iii) one adjusted to pH 6.0. The pH adjustments were made using a 50% sulfuric acid solution (Merck, Sigma-Aldrich). These solutions were processed using a wall-mounted reverse osmosis unit (PWR2511 series, Pure Water), as shown in Figure 1.

2.4.2. Membrane Characteristics and Operating Conditions

The membrane used in this system was a thin-film composite (TFC) polyamide membrane composed of three layers: [15] a selective aromatic polyamide layer (~200 nm), a microporous support layer (20–50 µm), and a nonwoven fabric backing layer (120–150 µm). The nominal pore size of the membrane is approximately 0.0001 µm, and it exhibits a contaminant rejection rate exceeding 98% for organic and ionic pollutants.

The RO system was operated at a pressure of 18 bar and within a pH range of 6.0 to 7.0, parameters selected based on literature supporting their effectiveness in maximizing rejection rates while minimizing membrane fouling and structural degradation under landfill leachate conditions (Tejera et al., 2020 [10]; Chen et al., 2021 [16]; Ahmed et al., 2024 [15]). No pretreatment steps, such as microfiltration, were applied to ensure that the system’s performance was evaluated under field-representative conditions.

2.4.3. Analytical Methods and Data Collection

Effluent samples were collected at hourly intervals and analyzed for Chemical Oxygen Demand (COD) using APHA Standard Method 5220 D [12], which involves potassium dichromate digestion and spectrophotometric quantification. Sampling continued until COD levels approached those of the untreated leachate.

All experiments were conducted in triplicate (n = 3) to ensure statistical robustness. Data are reported as mean ± standard deviation.

2.4.4. Statistical Analysis of Operational Parameters

Operational parameters, pH, and COD values were recorded throughout. Statistical differences among treatments were assessed via one-way ANOVA followed by Tukey’s post hoc test (p < 0.05). Assumptions of normality and homoscedasticity were verified using the Shapiro–Wilk and Levene’s tests. Analyses were performed using SPSS version 28.0.

3. Results and Discussion

3.1. Characterization of Leachates

3.1.1. Seasonal Leachate Generation Trends

Leachate generation in a landfill is influenced by factors such as climatic conditions, waste moisture content, and operational practices [17,18,19], as these determine the amount of water entering the landfill. The results obtained for leachate volumes indicate that the highest coefficients of variation occurred during the months of November and December, suggesting extreme fluctuations potentially linked to rainfall seasonality. This behavior implies that leachate flow rates are strongly affected by regional climatic dynamics, which are characteristic of a tropical monsoon climate. As shown in

Table 4. Leachate flow behavior between 2016 and 2023, in the Los Corazones sanitary landfill, Valledupar (Colombia). ASEOUPAR SAESP (2024).

Year	January	February	March	April	May	June	July	August	September	October	November	December
2016	0.165	0.304	0.349	0.602	0.517	0.349	0.343	0.367	0.461	0.566	0.539	0.472
2017	0.116	0.144	0.355	0.537	0.397	0.444	0.464	1.031	0.264	1.821	3.302	0.656
2018	0.525	0.300	0.155	0.067	0.249	0.142	0.231	0.977	0.461	0.707	1.539	0.547
2019	0.958	0.712	0.422	0.640	0.846	1.078	0.723	0.612	0.502	0.419	0.436	0.359
2020	0.433	0.470	0.279	0.535	1.178	1.435	0.596	0.419	0.445	0.811	0.646	2.174
2021	0.326	0.573	0.584	1.587	1.127	0.985	1.231	1.335	0.979	2.546	2.358	1.876
2022	0.223	0.354	0.456	0.987	0.457	0.765	0.655	0.672	0.577	1.576	2.657	7.233
2023	0.123	0.357	0.389	1.098	0.934	0.675	1.108	0.789	0.675	1.365	0.987	0.675
Average	0.359	0.402	0.374	0.757	0.713	0.734	0.669	0.775	0.545	1.226	1.558	1.749
Dev. Est.	0.284	0.177	0.126	0.458	0.353	0.424	0.350	0.328	0.211	0.734	1.092	2.318
CV%	79.145	44.168	33.791	60.567	49.522	57.792	52.312	42.250	38.621	59.825	70.083	132.551

, at the Los Corazones landfill, located in Valledupar, Colombia, a significant correlation was observed between leachate flow rates and local precipitation patterns. Valledupar exhibits a tropical monsoon climate, characterized by a rainy season that typically spans from March to December and a dry season from December to March [20].

According to Table 4, data collected between 2016 and 2023 show that the months with the highest leachate generation correspond to periods of increased internal drainage activity within the landfill. For instance, in November 2017, an average leachate flow of 3.30 L/s was recorded, one of the highest values during the study period. Similarly, the years 2021 (15.50 L/s), 2022 (10.10 L/s), and 2020 (9.42 L/s) exhibited the highest cumulative volumes, coinciding with typically rainy months in the region. In contrast, the lowest values were observed in 2016 (5.03 L/s), 2018 (5.89 L/s), and 2019 (7.70 L/s), which align with drier climatic conditions.

To assess the strength of the association between precipitation and leachate generation, a statistical analysis was conducted using monthly time-series data collected between 2016 and 2023. The results revealed a strong and statistically significant linear relationship, with a Pearson correlation coefficient of r = 0.981 and a coefficient of determination R² = 0.962, indicating that 96.2% of the variability in leachate generation can be explained by rainfall variation. This strong positive correlation is visually represented in Figure 2, which illustrates the close alignment between monthly precipitation and leachate flow at the Los Corazones landfill (Valledupar, Colombia). These findings align with previous studies in tropical monsoon regions, where rainfall has been identified as the primary driver of leachate generation due to the combined effects of high infiltration rates and reduced evapotranspiration [21,22].

The observed relationship reinforces the importance of integrating hydrometeorological parameters in landfill design and operation. During the wet season, particularly from April to November, increased rainfall volumes can significantly affect the hydraulic load and contaminant mobilization within the landfill body. Moreover, studies highlight that rainfall not only influences the volume of leachate but also modulates its physicochemical properties, including concentrations of organic matter, ammonia, and heavy metals, due to enhanced percolation and washout effects [22].

3.1.2. Time Series Analysis and Stationarity

In processing the flow data, the ARIMA model was used to determine whether or not seasonality existed within the data series. Therefore, the Dickey–Fuller unit root test was used, and the following assumptions were made: Ho: Non-stationary (unit root) and Ha: Stationary.

Figure 3 shows that leachate flow rates are distributed around a constant mean (1.226 L/s for the 2016–2023 period), suggesting the possible stationarity of the time series. However, the presence of several peaks in the data, with relatively even spacing, indicates the existence of a periodic component in the series, suggesting a direct relationship with the seasonality of precipitation in the region.

To formally assess stationarity, the Dickey–Fuller test was applied, yielding a statistic of −4.4081 and a p-value of 0.01. Since this value is below the statistical significance level (α = 0.05α = 0.05), the null hypothesis was rejected, indicating that the leachate flow time series is stationary. This implies that, although the recorded values fluctuate, the underlying process generating the flow remains stable over time.

This behavior has also been reported by [23], who applied ARIMA models to forecast leachate quality in a tropical mountain landfill, confirming their reliability under seasonal climatic conditions similar to those observed in Valledupar.

The stationarity detected in the series suggests that, despite seasonal fluctuations, the total volume of leachate generated follows a relatively stable trend over time, with no evidence of sustained flow growth. This has direct implications for landfill management, as it allows for more accurate prediction of leachate generation based on weather patterns, facilitating the planning of collection and treatment strategies.

On the other hand, the presence of peaks in certain periods, such as November 2017 (3.30 L/s), indicates that extreme weather events can generate abrupt increases in leachate flow. This highlights the importance of having an efficient drainage infrastructure to avoid overloading the treatment system and minimize the risk of environmental contamination.

In this regard, Ref. [24] developed a hybrid ARIMA–ANN forecasting model, highlighting its potential to enhance leachate prediction accuracy when incorporating multiple environmental and operational variables.

In order to determine the model’s suitability, the Ljung–Box test was performed, as detailed in Figure 4, to verify whether the white noise requirements were met. Therefore, the assumptions Ho = White Noise, Ha = No White Noise were made.

Applying the Box-Ljung test, the statistic X² = 4.5534 × 10⁻⁵ was obtained, with a p-value of 0.9946. Since this p-value is higher than the statistical significance level (α = 0.05), the null hypothesis is not rejected, indicating that the flow time series exhibits white noise characteristics; that is, the errors have a mean equal to zero and constant variance.

This finding is crucial in time series analysis, since the presence of white noise suggests that there is no significant autocorrelation in the data, implying that the observations are independent over time. According to [25], the Ljung–Box test is commonly used to verify whether the residuals of a time series model behave like white noise, assessing the autocorrelation in the residuals to ensure the model’s adequacy.

The absence of significant autocorrelation in the flow data suggests that the observed fluctuations are random and do not follow a predictable pattern. This has practical implications for leachate management in landfills, as it indicates that variations in flow may be more influenced by random external factors than by trends or patterns inherent to the system.

Previous research has highlighted the importance of properly understanding and managing leachate generated in landfills. Ref. [1] conducted a comprehensive review on landfill leachate treatments, emphasizing the need for efficient techniques to minimize the risk of environmental contamination. More recently, Ref. [26] developed a revised leachate contamination index (r-LPI) to more accurately assess the contamination potential of landfill leachate, providing a robust tool for the environmental management of this liquid waste.

3.1.3. Physicochemical Composition of Urban Waste Landfill Leachates

The chemical composition of leachates is diverse due to several intrinsic characteristics of solid waste, such as pH, age, temperature, and degree of stabilization [27]. These factors influence the formation of percolating liquids that become highly polluting and toxic substances [28]. Leachates typically contain significant amounts of organic matter, along with a wide range of inorganic constituents, including total and dissolved solids, nitrogen, phosphorus, pathogens, and heavy metals [4]. Furthermore, they exhibit notable variability in pH and may present a range of colors, from brown to black, often accompanied by the emission of volatile sulfur and nitrogen compounds responsible for strong odors [29].

In the collected sample, it was evident that four parameters exceeded the maximum permissible limit value established by the standard; these parameters were: COD, BOD, Chlorides, and Lead (

Table 5. Physicochemical characterization of leachates from the Los Corazones sanitary landfill, Valledupar, Colombia (2023).

Parameter	Unit	Results	Maximum Permissible Limit Value	State of Compliance
General
pH	Units	8.98	6:00 to 9:00	Complies
Temperature	°C	35.9	40.00 (art 5)	Complies
Conductivity	mS/cm	22.4	-	-
Chemical and Biological Oxygen
COD	mg O2/L	5096.7	2000.00	Does not comply
BOD5	mg O2/L	3153.0	800.00	Does not comply
Solids
Total suspended solids	mg/L	336.0	400.00	Complies
Settleable solids	ml/L	<0.1	5.00	Complies
Fats and Oils
Fats and/or oils	mg/L	<10.0	50.00	Complies
Ions
Total Cyanide	mg/L	<0.011	0.50	Complies
Chlorides	mg/L	4942.0	500.00	Does not comply
Sulfates	mg/L	151.0	600.00	Complies
Metals and Metalloids
Aluminum	mg Al/L	<1.0	3.00	Complies
Arsenic	mg As/L	0.005	0.10	Complies
Mercury	mg Hg/L	<0.001	0.01	Complies
Nickel	mg Ni/L	0.427	0.50	Complies
Lead	mg Pb/L	0.33	0.20	Does not comply
Selenium	mg Se/L	<0.0025	0.20	Complies
Vanadium	mg V/L	0.285	1.00	Complies

).

Between 2019 and 2023, Chemical Oxygen Demand (COD) and Biochemical Oxygen Demand (BOD) concentrations were identified as not following a trend pattern (Figure 4). This is consistent with what [1] described for both young and old leachates. This variability could be attributable to the combination of leachates of different ages during the recirculation process implemented at the Los Corazones landfill as a treatment method. Furthermore, rainfall may dilute the leachate stored in the ponds, and evapotranspiration rates vary depending on the prevailing climatic conditions in Valledupar, Colombia.

Figure 5 presents the temporal evolution of the Chemical Oxygen Demand (COD) and Biochemical Oxygen Demand (BOD₅) in leachate generated at a tropical landfill between 2019 and 2023. Each data point corresponds to the mean of triplicate analyses, and error bars represent the 95% confidence interval (IC95%), allowing for a statistically supported interpretation of variability. The BOD₅/COD ratio is also shown to evaluate the relative biodegradability of the leachate throughout the monitoring period.

COD concentrations ranged from 2131 mg/L (2019-I) to a maximum of 8432 mg/L (2022-I), while BOD₅ values fluctuated between 809.9 mg/L and 3881 mg/L. These values notably exceed the maximum permissible discharge limits established by Colombian environmental legislation (Resolution 0631 of 2015: 2000 mg/L for COD and 800 mg/L for BOD₅), particularly in the second half of 2020 and the first half of 2022. These peak values suggest potential periods of heightened contamination risk and a need for intensified treatment efforts.

The BOD₅/COD ratio varied between 0.38 and 0.63, indicating moderate to high biodegradability. Values above 0.4 typically reflect younger or moderately stabilized leachates, while values below 0.3 are characteristic of aged, recalcitrant leachates (Abunama et al., 2021 [27]; Mojiri et al., 2021 [30]). The highest biodegradability was observed in 2020-II (0.63) and 2022-I (0.46), suggesting a stronger presence of readily degradable organics, possibly due to high rainfall and active degradation processes. Conversely, 2019-I and 2023-II showed lower ratios (~0.38), consistent with more stabilized or weathered leachates.

These temporal trends confirm the importance of integrating statistical confidence analysis and biodegradability indicators into leachate management strategies. In tropical climates, where precipitation and seasonal dynamics significantly affect leachate production and composition, this integrated approach supports the optimization of treatment system design and operation [31,32].

3.2. Removal of Organic Matter in Leachate with Activated Carbon

Activated carbon becomes a viable option for wastewater treatment, since it has large surface areas, a high adsorption capacity, and porous structures [33,34] stated that “a greater reduction in COD levels could be achieved by this procedure compared to chemical methods”. The results differentiated by the treatment method used are presented below: batch flow and continuous flow.

3.2.1. Batch Adsorption Performance

Based on the experimental phase of the batch flow model with raw leachate, ANOVA statistical analysis revealed that pH, adsorbent dose, and their interaction exert a highly significant effect on contaminant removal (p < 0.05), indicating that adsorption efficiency is influenced by the combined effect of both factors rather than by a single variable.

The results of the ANOVA and Tukey test allow us to conclude that both pH and adsorbent dose have a significant effect on contaminant adsorption. It was identified that lower pH values (2.0–4.0) and high adsorbent doses significantly favor contaminant removal, which is consistent with previous research reporting higher adsorption efficiency under acidic conditions due to greater protonation of the adsorbent active sites [16,35].

To expand on these findings, a multivariate analysis of variance (MANOVA) was performed using COD and BOD₅ removal efficiencies as dependent variables, and pH and dosage as independent factors. The results revealed statistically significant effects of pH (Wilks’ Lambda = 0.312, F = 6.78, p = 0.001), dosage (Wilks’ Lambda = 0.284, F = 7.12, p = 0.001), and their interaction (Wilks’ Lambda = 0.197, F = 4.59, p = 0.005). These results confirm that the combined variation in pH and adsorbent dosage significantly influences the simultaneous removal of organic pollutants, and support the robustness of the factorial design employed.

Organic load removal efficiencies in the experimental trials ranged from 16% to 67%, depending on operational conditions. The most favorable performance was observed at a coagulant dosage of 100 g/L and a contact time of 90 min. This behavior is consistent with previous studies using advanced oxidation processes, which achieved COD removal efficiencies up to 83.2% under similar operational conditions [30,31,32]. Notably, when the leachate pH was adjusted to neutral (pH 7.0) or slightly acidic (pH 6.0), the treatment performance was comparable to that of raw leachate. However, under more acidic conditions (pH 5.0 and 4.0), significant improvements in COD removal were observed, ranging from 25% to 75%, especially when maintaining a 100 g/L dosage and extending the contact time to 120 min for pH 5.0 and 90 min for pH 4.0.

The improved performance under acidic conditions is attributed to the increased availability of hydroxyl radicals, which are more effectively generated at low pH levels, thus promoting the oxidation of persistent organic compounds [33,36]. Acidic pH levels also enhance the solubilization and subsequent degradation of complex organic pollutants commonly found in mature landfill leachates. These findings emphasize the critical role of pH control and contact time in maximizing treatment efficiency. Furthermore, they support the growing body of literature advocating the integration of chemical pretreatment and operational optimization as key strategies for effective leachate treatment [32,34].

Optimal conditions for organic matter removal were found when the pH was adjusted to 3.0 and 2.0, achieving COD removal efficiencies ranging from 20% to 80%. This was achieved using a dosage of 100 g/L and contact times of 120 min for a pH of 3.0, and 90 min for a pH of 2.0.

These findings have important implications for the optimization of adsorption processes in leachate and wastewater treatment, suggesting that adjusting adsorbent pH and dosage can improve process efficiency and reduce operating costs.

3.2.2. Adsorption Isotherm Modeling

Based on the experimental results, the Freundlich and Langmuir models were fitted using their respective linearized forms (Figure 6 and Figure 7). To analyze the behavior of the Freundlich isotherm, Equations (1) and (2) were applied across different pH levels of the leachate, and the corresponding R-squared values were used to assess model performance. Figure 5 illustrates the variation in model fit as a function of pH. The coefficient of determination (R²) indicates the degree to which the data conform to the Freundlich model, with values closer to 1.0 representing a better fit.

Consistent with these results, it was observed that at pH 7.0, the coefficient of determination (R²) was 0.081, indicating a poor correlation between the model and the experimental data. A similar trend was noted at pH 6.0, where R² decreased slightly to 0.077. In contrast, at pH 5.0, the R² increased to 0.7857, suggesting a substantial improvement in model fit. However, at pH 4.0, the R² dropped sharply to 0.0556, again indicating poor agreement. At pH 3.0, the model showed moderate predictive performance, with an R² of 0.3649, while at pH 2.0, the R² was 0.2613, reflecting limited model applicability under strongly acidic conditions.

The behavior of the Freundlich isotherm varied significantly with leachate pH. While the model showed a poor fit at certain pH values (7.0, 6.0, 4.0, and 2.0), it exhibited a better fit at other pH values (5.0 and 3.0). This suggests that the adsorption capacity of contaminants present in leachate may be influenced by changes in pH, which has important implications for waste management and treatment at landfills [36,37]. The pH-dependence of the Freundlich model highlights the heterogeneity of adsorption sites and supports the theory of multilayer adsorption, particularly under slightly acidic conditions [37].

The behavior of the Langmuir isotherm, as evaluated using Equations (3) and (4) in relation to leachate pH, and based on the R-squared values presented in Figure 7, demonstrates how the model fit varied with changing pH conditions. At pH 7.0, the coefficient of determination (R²) was 0.6079, indicating a moderate correlation between the model and the experimental data. At pH 6.0, the R² increased slightly to 0.6302, suggesting a marginal improvement in model accuracy. At pH 5.0, the R² rose markedly to 0.9411, demonstrating an excellent fit, a trend further reinforced at pH 4.0 with an R² of 0.9685, which represents a highly accurate model prediction.

At pH 3.0, the R² decreased slightly to 0.9183, yet still reflects a strong agreement between the observed and predicted values. Finally, at pH 2.0, the R² reached 0.8919, maintaining a satisfactory level of model performance. These results are consistent with previous studies reporting the superior applicability of the Langmuir isotherm for modeling contaminant adsorption in aqueous systems, particularly in those involving heterogeneous adsorbents with high surface affinity [16,34].

The Langmuir isotherm showed a better fit as the leachate pH decreased, indicating a greater adsorption capacity for contaminants under more acidic conditions. This behavior suggests a chemical monolayer adsorption mechanism, characterized by uniformity in adsorption energy at each adsorbent site and the absence of interactions between adsorbed molecules [23]. In contrast, the Freundlich model presented a poor fit to the experimental data, with correlations below 60%, which demonstrates its lower capacity to describe the adsorption process evaluated.

These results are consistent with previous research highlighting the superior performance of the Langmuir model in describing contaminant adsorption processes in aqueous media, particularly in systems containing heterogeneous adsorbents with high surface affinity [38]. Recent studies have shown that acidic pH conditions significantly enhance the performance of Langmuir-based adsorption models in the treatment of landfill leachates enriched with recalcitrant organic compounds and heavy metals. For instance, Refs. [36,37] reported increased adsorption capacity and improved model fit at pH levels below 5.0, attributing these results to enhanced surface protonation and strengthened electrostatic attraction mechanisms. Ref. [34] further confirmed that low pH stabilizes the active sites of activated carbon, promoting monolayer adsorption and reducing competitive interactions with multivalent ions [39].

These findings are in strong agreement with those of the present study, particularly the excellent model fit observed at pH 4.0 (R² = 0.9685), and further underscore the suitability of the Langmuir model under acidic conditions. Additionally, this behavior is consistent with the modified Langmuir–Freundlich model proposed by Jeppu and Clement, which accounts for pH-induced variability in adsorption efficiency across complex matrices such as landfill leachate.

3.2.3. Continuous Flow Model

In the continuous reactor experimental phase, after a 90 min hydraulic retention time, a 97% organic load removal efficiency was achieved. This resulted in a reduction in concentration from 19,200 mg/L to 640 mg/L. The 90% COD removal rate was maintained for the following 2.5 h. After this point, a slight decrease in the removal rate was observed, reaching 85% after 5.5 h. Subsequently, after 7.5 h, the removal rate decreased to 81%, with a concentration of 3520 mg/L. From 21.5 h to 26.5 h, 33% organic load removal was achieved, and after 32.5 h, this removal rate decreased to 17%. This indicates that the activated carbon bed was already saturated. Therefore, from 38 h onwards, no additional removal of organic load was observed compared to the influent.

For leachate samples with pH adjusted to 3.0, higher removal rates were achieved compared to the unamended leachate, and the time required to remove COD was considerably longer. This supports the pH-related behavior observed in the batch reactor and confirms that organic matter removal is much more effective in an acidic environment.

In contrast to the results obtained by [29], who worked with a hydraulic retention time of 8 h and a pH close to 2.0, achieving COD removal rates of 90% to 60% in the first 200 h, in this experimental phase, removal rates ranged from 97% to 33% during the first 33 h. This reveals a significant difference in terms of time period, as a result of the controlled variables.

3.3. Membrane Process—Reverse Osmosis in Leachate Treatment

Membrane separation technology for advanced leachate treatment can remove more contaminants than other advanced processes [29]. However, it is important to mention that reverse osmosis can work as a treatment for leachate, but fouling is a problem [31].

3.3.1. Organic Matter Removal Efficiency

Post-organic load removal data revealed 97% removal at both the original leachate pH and the pH adjusted to 6.0 during the first time intervals, as detailed in

Table 6. Variation in Chemical Oxygen Demand (COD) as a function of time during the experimental stage.

Evaluation Time (h)	pH 8.2	pH 7.0	pH 6.0
Control	5598.0 ± 85.2	5319.0 ± 90.5	5347.0 ± 87.8
0	120.1 ± 4.3	135.0 ± 5.2	119.0 ± 3.9
12	128.1 ± 3.5	141.0 ± 4.1	135.0 ± 4.8
24	142.0 ± 4.2	142.6 ± 4.6	140.8 ± 4.3
36	155.3 ± 5.1	153.4 ± 4.9	154.3 ± 5.0
48	155.3 ± 5.0	155.0 ± 5.1	156.7 ± 5.3
60	162.0 ± 5.4	155.0 ± 5.2	157.0 ± 5.3
72	162.0 ± 5.6	155.4 ± 5.3	157.4 ± 5.4

Note: Concentrations are expressed in mg/L as mean ± standard deviation from three independent replicates. Initial COD values and their evolution over 72 h are presented for leachates with an initial pH of 8.2 and adjusted to 7.0 and 6.0.

.

The results demonstrated that the reverse osmosis (RO) process achieved removal efficiencies of 97% for both Chemical Oxygen Demand (COD) and biochemical oxygen demand (BOD₅) during the initial outlet time intervals, for both untreated leachate and leachate adjusted to pH 6.0, as shown in Table 6. These values reflect the high effectiveness of RO systems in eliminating both biodegradable and non-biodegradable organic matter present in landfill leachates. The results are presented as mean ± standard deviation, and 95% confidence intervals were calculated to represent the uncertainty associated with removal percentages. This approach ensures the reproducibility and reliability of the reported removal efficiencies. This performance aligns with previous findings, which reported COD removal rates exceeding 98% when RO systems are employed, particularly when pre-treatment stages and pH adjustments are implemented to enhance membrane performance [40]. Furthermore, several studies have demonstrated that integrating RO with other technologies, such as ozonation or microfiltration, can significantly improve permeate quality, enabling compliance with environmental discharge standards [41].

In the Colombian regulatory context, current environmental legislation is governed by Resolution 0631 of 2015, which defines the maximum permissible limits for discharges into surface water bodies and public sewer systems [13]. The findings of this study confirm that leachate treated by reverse osmosis (RO) complies with these standards, indicating the suitability of this technology for regulatory adherence. It is also noteworthy that pH adjustment significantly influences the operational performance of RO processes. pH affects both the physicochemical properties of the leachate and the surface charge of the membranes, thereby impacting contaminant rejection rates.

Research has shown that slightly acidic conditions can enhance organic matter removal and reduce membrane fouling, ultimately extending membrane lifespan and ensuring sustained operational efficiency [42]. Overall, these results highlight reverse osmosis as a robust and viable solution for landfill leachate treatment, offering high removal efficiency and regulatory compliance, while contributing to the protection of aquatic ecosystems.

3.3.2. Statistical Analysis of Treatment Conditions

The ANOVA test showed significant differences between groups (F = 3985.794, p < 0.001), indicating that at least one of the treatments performed differently compared to the others. Detailed results are presented in

Table 7. Analysis of variance of a factor applied to the removal of organic matter by reverse osmosis.

Origin of the Variations	Sum of Squares	Degrees of Freedom	Means of Squares	F	Probability	Critical Value for F
Between groups	12,707,620.45	7	1,815,374.35	3985.79494	1.03262 × 10−24	2.6571966
Within the groups	7287.3768	16	455.46105
Total	12,714,907.83	23

.

Since the ANOVA indicated significant differences, the Tukey HSD test was applied to determine which treatments differed from each other. The results showed that all treatments (T1–T7) differed significantly from the control (p < 0.05), but no significant differences were found between treatments T1–T7.

The reverse osmosis process had a significant impact on organic matter (COD) removal, as all treatments were significantly different from the control group. No significant differences were observed between treatment times (T1–T7), suggesting that the greatest reduction in organic matter occurs in the early stages of the process. Time and pH did not significantly affect process efficiency within the range evaluated, suggesting that other factors (such as pressure, temperature, or water quality) may be more important determinants of organic matter removal.

3.3.3. Operational Considerations and Hybrid Optimization

Although pH did not show a significant effect on process efficiency within the range evaluated, it is important to consider that pH can influence the solubility of certain contaminants and the effectiveness of treatment processes. For example, some metals may be more soluble in acidic conditions, which could affect their removal during treatment.

Reverse osmosis has established itself as an effective technique for removing organic load from leachates. However, to maximize its efficiency and ensure its long-term sustainability, it is essential to consider additional factors, such as the implementation of pretreatments and the control of operating parameters. In this context, Ref. [43] highlights that independent membrane processes may not fully meet technical or economic requirements, so combining them with conventional systems or with other membrane technologies in hybrid configurations represents a more efficient strategy. These hybrid systems can be classified into two main types: those in which all processes perform the same function, such as separation, and those in which each process performs different functions, such as reaction and separation, thus optimizing overall treatment performance.

Additionally, Ref. [44] proposed the use of machine learning algorithms to predict key parameters in landfill leachate, improving the accuracy of risk assessment and operational optimization in sanitary landfills. This aligns with our predictive modeling approach and targeted treatment strategy.

3.3.4. Energy Consumption and Environmental Perspective of Reverse Osmosis

While the reverse osmosis (RO) process demonstrated high removal efficiencies for COD and BOD₅, it is essential to recognize the environmental implications associated with its energy-intensive operation. According to [16], RO systems can consume between 2.5 and 7.5 kWh/m³ of treated leachate, depending on membrane type, operating pressure, and pretreatment steps, representing a substantial carbon footprint when powered by non-renewable sources.

Moreover, Refs. [10,15] highlight that although RO is effective in pollutant removal, its sustainability must be evaluated through life-cycle analysis (LCA), considering membrane replacement, brine disposal, and energy demand.

Hybrid systems that integrate RO with energy recovery or renewable power sources (e.g., photovoltaic-assisted systems) have been proposed as more sustainable alternatives [31,34].

Incorporating these systems could reduce environmental burdens and operational costs, aligning leachate treatment with circular economy principles. A preliminary life-cycle perspective thus supports future optimization pathways for RO-based treatments in tropical landfills.

4. Conclusions

The leachate generated at the Los Corazones landfill displays a heterogeneous profile that cannot be strictly classified as either young or mature, due to overlapping physicochemical characteristics. This complexity is closely linked to operational practices such as leachate recirculation and the pronounced seasonal variability typical of tropical climates. These findings emphasize the importance of adaptive treatment strategies capable of managing such variability in both composition and flow.

Activated carbon adsorption exhibited promising removal efficiencies for organic matter, particularly under acidic conditions, where the Langmuir isotherm provided the best fit. Continuous-flow systems outperformed batch configurations, indicating their suitability for scaling in controlled operational environments.

Reverse osmosis achieved high and consistent removal efficiencies for both COD and BOD₅ across the evaluated pH range, demonstrating its potential as a robust treatment method. However, its performance was not significantly influenced by pH or contact time, suggesting that further optimization could be directed towards operational cost and energy efficiency improvements.

While the study provides valuable insights under controlled laboratory conditions, a key limitation is the lack of implementation under real-world operational scenarios, where factors such as membrane fouling, routine maintenance, and energy performance may critically influence long-term system effectiveness.

Nevertheless, the integrated approach combining ARIMA-based forecasting, physicochemical characterization, and dual treatment technologies offers a solid and transferable framework for replication in other tropical and urban contexts with similar climatic and waste management challenges. The results provide technical and methodological insights that may inform the design of modular and decentralized treatment systems, particularly suited to low- and middle-income regions where seasonal hydrological variability complicates leachate management.

Future real-world implementations should incorporate renewable energy sources such as photovoltaic systems to reduce the carbon footprint of reverse osmosis treatment. Furthermore, system sustainability should be assessed through comprehensive life cycle analysis, considering environmental, energy, and economic dimensions.