Optimizing Methane Production from Lignocellulosic Biomass: Low-Temperature Potassium Ferrate Pretreatment via Response Surface Methodology

Abstract

Lignocellulosic biomass like pistachio shells (PSs) is a promising feedstock for anaerobic digestion (AD), but lignin recalcitrance limits biodegradability. Conventional pretreatments suffer from high energy costs or inhibitor formation; here, potassium ferrate (PF) + low-thermal pretreatment offers a green alternative. A Box–Behnken Design was employed to optimize the PF dosage, pretreatment temperature, and time, with response variables including the methane (CH₄) yield, soluble chemical oxygen demand (SCOD)/total chemical oxygen demand (TCOD) ratio, and lignin removal efficiency. The optimized conditions (0.637 mmol/g total solids PF dose, 66.76 °C, 55.84 min) achieved a CH₄ yield of 171.00 mL CH₄/g volatile solids, representing a 4.3-fold increase compared to untreated PSs. The ANOVA results showed strong links between how much lignin was removed, the ratio of SCOD to TCOD, and the amount of CH₄ produced, with the interactions between temperature and time being the most important. This study highlights the potential of PF-based pretreatment as a cost-effective and environmentally sustainable strategy to maximize CH₄ yields from lignocellulosic waste, supporting renewable energy adoption and circular economy principles. Further studies should explore scalability and economic feasibility for industrial applications.

1. Introduction

Lignocellulosic biomass, with an estimated global annual output of 220 billion tons, is a highly promising and eco-friendly feedstock for renewable energy production [1]. Its conversion to methane (CH₄) via anaerobic digestion (AD) is widely acknowledged as an efficient method for generating bioenergy while reducing greenhouse gas emissions. Additionally, biogas production from lignocellulose residues supports sustainable waste management and the circular economy [2].

Pistachio shells (PSs), derived from the fruit of Pistacia vera L., represent a significant portion of global lignocellulosic waste. While the edible portion of the pistachio is highly nutritious, over 50% of the total fruit mass comprises the outer shell, which is typically discarded. In 2023, global production of in-shell pistachios was estimated at approximately 747.31 thousand tons [3]. PSs are composed of 20–32% hemicellulose, 30–55% cellulose, and 12–38% lignin, rendering them a rich lignocellulosic feedstock [4]. Despite their abundance, PSs are underutilized and mainly relegated to low-value applications such as animal bedding or soil conditioning, which contributes to environmental pollution. Their use as a substrate for AD not only offers a sustainable waste management solution but also enhances biogas production. This strategy is especially relevant in reducing the carbon footprint of energy generation and improving the accessibility of renewable energy resources [5].

The chemical composition of lignocellulosic biomass plays a critical role in determining both the CH₄ yield and the operational stability of AD. Parameters such as plant species, growth conditions, and maturity significantly influence AD performance. However, the complex structure of lignocellulose poses substantial challenges to biodegradation [6]. Comprising 30–50% cellulose, 15–35% hemicellulose, and 10–30% lignin, lignocellulosic biomass is inherently resistant to microbial hydrolysis. While cellulose is a crystalline polymer that resists enzymatic attack, hemicellulose is more amorphous and heterogeneous in structure. Lignin, an aromatic polymer, not only resists degradation but also obstructs enzymatic access to cellulose and hemicellulose. As such, the application of physical, chemical, or biological pretreatment methods is often essential to enhance hydrolysis efficiency and CH₄ yields during AD processes [7].

Although pretreatment strategies can increase the biodegradability of lignocellulosic biomass, conventional methods face several limitations. Thermal pretreatments are typically energy-intensive and costly, while biological methods are labor-intensive and time-consuming. Chemical pretreatments, although effective, can pose environmental risks and lead to high operating costs, depending on the reagents used [6]. Therefore, the development of cost-effective and environmentally benign pretreatment strategies is imperative to fully unlock the bioenergy potential of lignocellulosic biomass.

Among the emerging approaches, advanced oxidation processes (AOPs) have recently gained attention as promising chemical pretreatments to improve biomass hydrolysis [8]. Originally developed in the 1980s for water purification, AOPs generate hydroxyl (•OH) and more recently sulfate (SO₄•⁻) radicals, which facilitate the breakdown of biomass polymers [9]. Potassium ferrate (PF) is considered a powerful and environmentally friendly oxidant due to its high redox potential, ranging between 0.72 and 2.20 V, depending on the pH [10]. In acidic conditions, the redox potential of ferrate ions (FeO₄²⁻) reaches 2.20 V, whereas it decreases to approximately 0.72 V in alkaline environments [11]. Despite its oxidative strength, the relatively high cost of PF and its potential to inhibit microbial activity at elevated doses limit its standalone application [12,13]. Consequently, coupling PF with thermal pretreatment under moderate temperatures offers a synergistic strategy to enhance lignin solubilization while mitigating adverse microbial impacts [14,15,16]. This integrated approach holds promise for improving subsequent CH₄ production through more effective biomass disintegration.

Traditional modeling of AD processes relies on simplified mathematical approaches such as the modified Gompertz and logistic models to estimate CH₄ yields and determine kinetic parameters [12]. While these models provide valuable baseline insights, they often fail to capture the nonlinear and complex behavior inherent to AD systems, resulting in suboptimal parameter estimation and limited predictive power. To overcome these limitations and optimize process conditions more accurately, the response surface methodology (RSM) has emerged as a robust statistical tool for experimental design and multivariate optimization [13].

This study utilizes the RSM to optimize low-thermal (<100 °C) PF pretreatment for PSs, aiming for improved CH₄ yield, hydrolysis efficiency, and lignin removal. In this work, lignin removal was experimentally confirmed, whereas cellulose and hemicellulose contents were not directly measured. Based on previous findings [17], we hypothesize that under controlled alkaline conditions (pH adjusted using NaOH), PF oxidation selectively depolymerizes lignin and facilitates partial hemicellulose solubilization, releasing lower-molecular-weight sugars. These structural modifications likely enhance cellulose and hemicellulose accessibility to microbial hydrolysis during AD, thereby contributing to the observed methane yield improvements. To the best of our knowledge, only a limited number of studies have investigated PF-based pretreatment for lignocellulosic substrates. This study represents one of the earliest optimization attempts, specifically focusing on PSs.

2. Materials and Methods

2.1. Substrate and Inoculum

Pistachio hull waste was collected from a facility in Gaziantep, Türkiye and stored at +4 °C to prevent premature biodegradation. The shells were manually separated and sieved to obtain a uniform particle size of 1–2 mm for consistent substrate quality [14]. The inoculum was sourced from an active anaerobic digester in Trabzon, Türkiye, ensuring a viable microbial community for the AD process.

2.2. PF-Based and Thermal Pretreatment Procedure

PF (K₂FeO₄, ≥98% purity) was used as the oxidizing agent [15]. For each experimental run, 10 g of PS substrate (1–2 mm) was suspended in 100 mL of distilled water. PF was added at concentrations of 0.33, 0.66, and 0.99 mmol per gram of total solids (TSs). Right after adding PF, the pH of the mixture was changed to 10.0 using 0.1 M NaOH to help the PF oxidize quickly in the alkaline environment. The mixtures were stirred magnetically at 300 rpm (HJ-4, Ronghua Ltd., Tianjin, China) at 25 °C for 2 h to allow initial oxidation [9,16]. Following this, samples were subjected to thermal pretreatment at the designated temperatures (60, 80, or 100 °C) for the specified times (30, 60, or 90 min) according to the Box–Behnken Design (BBD) (Figure S1). Untreated PS samples served as the control. All experiments were performed in triplicate. Following pretreatment, samples were centrifuged at 4000 rpm for 10 min. The supernatant was collected for analysis of initial SCOD and dissolved lignin content. The entire pretreated slurry (solids + liquid) was used for BMP assays.

2.3. Experimental Design and RSM-Based Optimization

To optimize the pretreatment conditions and assess their effects on lignin removal, the soluble COD/total COD (SCOD/TCOD) ratio, and CH₄ yield, a three-factor BBD was employed within the framework of the RSM. The selected independent variables were the PF dosage (mmol/g TSs), 0.33, 0.66, and 0.99; pretreatment temperature (°C), 60, 80, and 100; and pretreatment time (min), 30, 60, and 90 (Table S1, where the full experimental design is presented). The PF dosage range was adapted from a previous study [17], while the pretreatment temperature was limited to 100 °C due to high operational costs and potential material loss associated with higher thermal conditions [14]. These factors were chosen based on preliminary experiments and the literature-reported effective ranges [9,18,19]. A total of fifteen experimental runs were generated by the BBD matrix, including three replicates at the center point to estimate pure error and validate the model’s adequacy. Experimental responses from each run were used to develop second-order polynomial regression models, and analysis of variance (ANOVA) was performed to assess the significance of both individual factors and their interactions. The optimization process aimed to determine the most effective combination of PF dosage, pretreatment temperature, and pretreatment time to maximize CH₄ yield and substrate solubilization while minimizing lignin content. All experiments were carried out in triplicate, and response data were analyzed using Design-Expert^® software (Version 13, Stat-Ease Inc., Minneapolis, MN, USA) for model development, 3D surface plot visualization, and multi-objective optimization.

2.4. Biochemical Methane (CH₄) Potential (BMP) Assay

BMP tests were performed in 500 mL anaerobic glass bottles. Each bottle contained the pretreated slurry equivalent to 10 g PSs, anaerobic inoculum added at an inoculum-to-substrate ratio of 2.0 on a volatile solids (VSs) basis, and distilled water to achieve a total working volume of 400 mL [20,21]. The initial pH was adjusted to 7.0 ± 0.1 using 0.1 M NaOH or HCl. The headspace was flushed with nitrogen gas for 2 min, and the bottles were sealed with rubber stoppers and aluminum caps [22]. The bottles were incubated at 39 °C for 30 days. CH₄ production was monitored daily via water displacement. Manual mixing was performed twice daily [15]. Liquid samples were periodically withdrawn, filtered (0.45 μm), and analyzed for pH and SCOD [16]. Each condition was tested in triplicate, alongside inoculum-only blanks and untreated PS controls. The net CH₄ yield (mL CH₄/g VSs) was calculated by subtracting the blank CH₄ production [23].

2.5. Analytical Methods

TSs, VSs, SCOD, and TCOD were analyzed according to APHA standard methods [23]. C, H, N, and S contents were determined using a Costech NA 2500 analyzer (Costech Analytical Technologies, Valencia, CA, USA), while O was calculated by difference. pH was monitored during AD using a calibrated pH meter (FE28-S, METTLER-TOLEDO, Greifensee, Switzerland) [24]. The lignin content of PSs before and after pretreatment was analyzed via a fiber analyzer (ANKOM A2000i, New York, NY, USA) [25]. For SEM analysis, pretreated and untreated PS samples were oven-dried (60 °C), mounted on stubs, sputter-coated with gold, and imaged using a Hitachi S-3000N microscope at 500× and 1000× magnifications [26]. CH₄ yield is expressed as net mL CH₄ produced per gram of VSs added.

3. Results and Discussion

3.1. Physicochemical Characterization of Pistachio Shells and Inoculum

Table 1. Characterization of inoculum and substrate.

Parameters	Pistachio Shell	Inoculum
%TSs a	92.91 ± 1.21	15.44 ± 2.10
%VSs a	86.92 ± 1.97	5.45 ± 1.209
TCOD (mg O2/L)	69,244 ± 545	8111 ± 335
SCOD (mg O2/L)	n.a.	1798 ± 147
Lignin b	34.32 ± 2.19	n.a.
Cellulose b	23.11 ± 1.79	n.a.
Hemicellulose b	21.55 ± 1.57	n.a.
pH	n.a.	6.58 ± 0.21
%C b	49.12 ± 1.8	4.52 ± 0.33
%H b	5.48 ± 0.19	n.a.
%O	45.17	n.a.
%N b	0.13 ± 0.03	0.40 ± 0.04
%S b	0.1 ± 0.02	n.a.

Note: ^a as total weight of the sample; ^b as TSs of the sample; n.a. = not available. SCOD = soluble chemical oxygen demand; TCOD = total chemical oxygen demand. Values are means ± standard error from triplicate experiments.

presents the physicochemical properties of PSs and anaerobic inoculum. The TS and VS contents of PSs were 92.9% and 86.9% (w/w), respectively, indicating a high organic load. The chemical composition of lignocellulosic biomass can vary due to environmental and botanical factors, which influence its bioenergy potential [26]. Compared to woody biomass, lignocellulose materials such as PSs exhibit greater compositional diversity and resemble fibrous plants. According to the literature, PSs typically contain 30–55% cellulose, 20–32% hemicellulose, and 12–38% lignin [27]. In this study, PSs were found to contain 24.2% cellulose, 22.1% hemicellulose, and 37.5% lignin. This high lignocellulose content (over 75%) suggests their potential for CH₄ production via AD, though pretreatment is necessary to enhance biodegradability [28].

3.2. ANOVA and Model Fitting

This study employed the RSM to investigate the effects of three critical process parameters—PF dose, production temperature, and pretreatment temperature—on three key response variables: the CH₄ yield, the SCOD/TCOD ratio, and lignin removal. The experimental design facilitated systematic exploration of parameter interactions and their effects on process performance, with the ultimate goal of optimizing bioprocess conditions for enhanced bioenergy production. The ANOVA results for the CH₄ yield, SCOD/TCOD ratio, and lignin removal responses from the BBD design are given in

Table 2. ANOVA for RSM models.

Source	Sum of Squares	df	Mean Square	F-Value	p-Value		Adjusted R2	Prediction R2	R2	MSE
CH4 yields (model)	23,272.03	9	2585.7	10.30	0.0028	Significant	0.8395	−0.1238	0.9297	103.44
A—PF	4919.33	1	4919.3	19.59	0.0031
B—pretreatment temperate	2457.01	1	2457.0	9.78	0.0167
C—pretreatment time	7174.82	1	7174.8	28.57	0.0011
AB	136.89	1	136.9	0.55	0.4844
AC	403.61	1	403.6	1.61	0.2454
BC	2430.49	1	2430.5	9.68	0.0171
A2	1777.68	1	1777.7	7.08	0.0324
B2	566.81	1	566.8	2.26	0.1767
C2	2867.7	1	2867.7	11.42	0.0118
Residual	1758.02	7	251.5
Lack of fit	1758.02	3	586.0
Pure error	0	4
Cor. total	25,030.05	16
SCOD/TCOD (model)	0.0234	9	0.0020	12.59	0.0015	Significant	0.8670	0.1882	0.9418	8.49 × 10−5
A—PF	0.0045	1	0.0040	21.89	0.0023
B—pretreatment temperate	0.0028	1	0.0030	13.65	0.0077
C—pretreatment time	0.0091	1	0.0090	44.21	0.0003
AB	0.0004	1	0.0004	1.84	0.2165
AC	0.0009	1	0.0009	4.51	0.0713
BC	0.0026	1	0.0026	12.37	0.0098
A2	0.0019	1	0.0019	9.42	0.0181
B2	0.0001	1	0.0001	0.3309	0.5831
C2	0.0009	1	0.0009	4.58	0.0696
Residual	0.0014	7	0.0002
Lack of fit	0.0012	3	0.0004	8.06	0.0359
Pure error	0.0002	4	0.0001
Cor. total	0.0248	16
Lignin removal (model)	283.49	7	40.5	13.63	0.0004	Significant	0.8467	0.4231	0.9138	1.57
A—PF	55.12	1	55.12	18.55	0.002
B—pretreatment temperate	9.03	1	9.03	3.04	0.1153
C—pretreatment time	117.81	1	117.81	39.64	0.0001
AC	23.04	1	23.04	7.75	0.0213
BC	19.8	1	19.8	6.66	0.0296
A2	26.36	1	26.36	8.87	0.0155
C2	29.06	1	29.06	9.78	0.0122
Residual	26.75	9	2.97
Lack of fit	23.22	5	4.64	5.27	0.0663
Pure error	3.53	4	0.882
Cor. total	310.24	16

.

The second-order model developed for the CH₄ yield was found to be statistically significant (F = 10.30, p = 0.0028 < 0.05). The model can explain approximately 93% of the variance in the CH₄ yield, according to the high coefficient of determination (R = 0.9297). The adjusted R² value (0.8395) also supports the strong generalizability of the model. However, the negative predicted R² (−0.1238) suggests that the model may perform poorly on unseen data, indicating a potential overfitting issue. The predicted R² value for the CH₄ yield was negative (−0.1238); however, this does not indicate a low explanatory power of the model. Indeed, the actual R² (0.9297) and adjusted R² (0.8395) values were found to be quite high. A negative predicted R² is particularly common in multivariate biological systems where a limited number of experiments are used and indicates that the model’s external predictive power may be limited. This limitation was addressed through validation experiments, and close agreement was obtained between the model predictions and experimental results. The adequacy of the developed RSM models was further confirmed by the normal probability plots of residuals and the predicted vs. actual plots (Figure S1), which indicated that the models were statistically valid and that the residuals were approximately normally distributed.

Among the statistically significant factors in the model are the PF dose (A), pretreatment temperature (B), and pretreatment time (C), with the latter having the most substantial effect. The interaction between temperature and time (B × C) was also found to be significant, highlighting the necessity of evaluating these parameters in conjunction. The significance of the quadratic terms A² and C² further implies a nonlinear relationship between these factors and the CH₄ yield. Although the lack-of-fit test is not directly reported, the overall model fit appears to be robust.

The ANOVA results for the SCOD/TCOD ratio were also found to be highly significant (F = 12.59, p = 0.0015 < 0.05). The model exhibits a high explanatory power with an R² of 0.9418, indicating that 94% of the variance is accounted for. The adjusted R² (0.8670) further confirms this finding. However, the predicted R² value (0.1882) remains relatively low, suggesting limited predictive capability for new data. All linear terms—PF dose, temperature, and time—were statistically significant. The B × C interaction was found to be significant (p = 0.0098), while the A × C interaction was near the threshold of significance. The quadratic term A² was significant, and C² exhibited a borderline level of significance, indicating a potential nonlinear effect. The lack of fit was statistically significant (p = 0.0359 < 0.05), implying that unexplained variation remains in the model and more complex modeling may be required.

The model developed for lignin removal yielded the highest F-value (13.63) and was the most statistically robust among the three responses (p = 0.0004 < 0.05). It explained 91.38% of the variance (R² = 0.9138), and its adjusted R² value (0.8467) indicates strong reliability. Although the predicted R² (0.4231) was positive, its relatively lower value compared to the adjusted R² suggests that the model’s predictive power is more limited than its explanatory power. Among the linear factors, the PF dose and pretreatment time were statistically significant, while temperature was not. Both the AC and BC interaction terms were important, showing that the variables work together to affect lignin removal. The significance of the quadratic terms A² and C² further indicates nonlinear relationships between these factors and lignin degradation. The lack of fit was not statistically significant (p = 0.0663 > 0.05), suggesting a satisfactory fit of the model to the experimental data and that additional terms were not necessary.

The matrix plot presented in Figure 1 was created to statistically evaluate the pairwise relationships between the modeled response variables and to visually reveal the structural patterns of these relationships. Diagonal panels show histograms confirming near-normal distributions, while off-diagonal panels present pairwise scatter plots with fitted regression lines and 95% confidence intervals. Strong positive correlations are observed, especially between the CH₄ yield and SCOD/TCOD (R² = 0.942) and between the CH₄ yield and lignin removal (R² = 0.914), indicating predictive consistency and interdependence within the BBD-based RSM model. These results align well with the ANOVA outputs (p < 0.05, high R², low residuals), confirming the model’s statistical validity. Overall, the plot suggests that enhanced lignin degradation and increased solubilized organic content positively affect CH₄ production. The combined effects of thermal and chemical pretreatment improve biomass accessibility and bioconversion efficiency, emphasizing the synergistic influence of pretreatment factors on biogas yield.

3.3. Response Surface Plot

In Figure 2, the three-dimensional response surface graphs reveal a statistically significant interaction between the PF dosage and pretreatment temperature on the CH₄ yield. The dome-shaped surface profile indicates the presence of a clear optimum point, under which CH₄ production is maximized. While increasing the PF concentration initially enhances the CH₄ yield, this improvement plateaus beyond a certain threshold. Similarly, low pretreatment temperatures limit the CH₄ output due to insufficient hydrolysis, whereas an optimal temperature range results in peak production. The contour lines and color gradients further demonstrate the sensitivity of the CH₄ yield to these two factors, effectively delineating the critical optimization zones [18].

In the SCOD/TCOD response surface, both the PF dosage and temperature contribute positively to the increase in soluble organic matter. Elevated PF concentrations enhance the chemical degradability of the substrate, promoting the release of soluble organic compounds. Simultaneously, higher temperatures facilitate cell wall disruption and accelerate hydrolysis rates. However, temperatures approaching 100 °C lead to a plateau in the SCOD/TCOD, suggesting the potential formation of inhibitory compounds or saturation of reaction kinetics [29]. These results underscore the need for careful optimization of both PF dosage and temperature in pretreatment design.

Regarding the lignin removal model, increasing the PF dosage directly improves lignin solubilization, likely due to the strong oxidative potential of PF and its enhanced penetration into the lignin matrix. Elevated temperatures further weaken phenolic–polymer linkages within lignin structures, facilitating their conversion to more soluble forms [29]. The contour intensity on the response surface indicates that maximum lignin removal occurs under combined high-PF and -temperature conditions. Enhanced lignin removal is assumed to increase substrate accessibility to microbial hydrolysis, thereby positively impacting CH₄ production [30]. However, excessive chemical dosing and elevated temperatures may lead to higher energy consumption and costs; thus, economic sustainability should also be considered in process optimization.

When all three models are evaluated collectively, lignin removal, the SCOD/TCOD ratio, and the CH₄ yield are found to represent complementary aspects of lignocellulosic biomass digestion. A holistic analysis of the response surfaces reveals meaningful positive correlations among these variables. In particular, high PF dosages and elevated temperatures facilitate more efficient delignification, enhancing microbial access to biodegradable components such as cellulose and hemicellulose [31]. The breakdown of the lignin structure increases substrate solubility, which directly results in elevated SCOD/TCOD ratios [17]. This rise in soluble organics supports microbial activity, ultimately improving the CH₄ yield. Accordingly, improvements in pretreatment conditions not only enhance CH₄ production but also boost overall process efficiency.

In conclusion, the findings demonstrate that careful optimization of the pretreatment parameters—especially PF dosage and temperature—enables effective removal of recalcitrant biomass components. This, in turn, increases the fraction of biomass available for microbial digestion and maximizes energy recovery within the specified dosage range. The multivariate analysis conducted using the RSM provides a systematic, data-driven pathway to identify optimal conditions, offering a robust framework for process enhancement.

3.4. Model Equations

The response surface models for the CH₄ yield, the SCOD/TCOD, and lignin removal—developed based on the BBD—have revealed both the individual and interactive effects of the three main factors in the system (A: PF dosage; B: pretreatment temperature; C: pretreatment time). In particular, the inclusion of two-factor interaction terms (AB, AC, BC) in the model equations is of great significance for understanding the multivariate nature of the system, as well as the synergy or antagonism among the factors (Equations (1)–(3)) [32].

CH₄ yields = 191.70 + 24.80 × A + 17.53 × B + 29.95 × C + 5.85 × AB − 10.05 × AC + 24.65 × BC + −20.55 × A² − 11.6 × B² − 26.1 × C²

(1)

SCOD/TCOD = 0.1822 + 0.0237 × A + 0.0187 × B + 0.0338 × C + 0.0098 × AB − 0.0152 × AC + 0.0253 × BC − 0.0215 × A² + 0.0040 × B² − 0.0150 × C²

(2)

Lignin removal = 19.91 + 2.62 × A + 1.06 × B+ 3.84 × C − 2.4 × AC + 2.23 × BC − 2.5 × A² − 2.62 × C²

(3)

The AB interaction (PF dose × pretreatment temperature) exhibits a positive coefficient across all three response variables (CH₄ yield: +5.85; SCOD/TCOD: +0.0097; lignin removal: +0.75). This suggests that a simultaneous increase in the polyphenol dosage and pretreatment temperature slightly enhances system performance. However, the magnitude of this interaction is relatively limited, and according to the ANOVA results, it was not found to be statistically significant (p > 0.05). Therefore, the AB interaction can be considered a secondary factor with minimal influence on the system.

In contrast, the AC interaction (PF dose × pretreatment time) shows a notably negative effect in the model (CH₄ yield: −13.05; SCOD/TCOD: −0.0140; lignin removal: −2.45). This indicates that the combined application of high polyphenol doses and extended treatment durations may negatively impact system performance. It is likely that under such conditions, radical generation increases uncontrollably, leading to indiscriminate degradation of organic matter rather than selective lignin oxidation. According to the ANOVA results, this interaction was statistically significant for both the CH₄ yield and lignin removal, identifying it as a critical parameter that must be carefully managed within the system. The strongest interaction effect is attributed to the BC interaction (pretreatment temperature × pretreatment time), which shows a significantly positive influence on all response variables (CH₄ yield: +20.65; SCOD/TCOD: +0.0253; lignin removal: +2.60) and is statistically significant according to the ANOVA (p < 0.05). These findings indicate that co-optimization of temperature and duration can effectively enhance lignin breakdown and promote the release of soluble organic compounds, thereby contributing to increased CH₄ production. As such, the BC interaction holds primary importance in system optimization and necessitates joint consideration of these process parameters.

In conclusion, the derived RSM model equations demonstrate that not only individual factors but also inter-factor interactions play a decisive role in process performance. Avoiding antagonistic combinations such as the AC interaction and harnessing synergistic effects like the BC interaction are critical for achieving both selective lignin degradation and a high CH₄ yield. Therefore, interaction terms within the model should be evaluated not only from a statistical standpoint but also in terms of reaction mechanisms and overall system stability.

3.5. Optimization of Pretreatment Conditions

The experimental results obtained in this study were evaluated within the framework of multi-response optimization based on the RSM, using Design-Expert^® software to optimize three key performance indicators: the CH₄ yield, the SCOD/TCOD ratio, and lignin removal. A multi-criteria strategy was adopted during the optimization process, aligning with both environmental and process efficiency goals. While the independent variables—PF dose (A), pretreatment temperature (B), and processing time (C)—were constrained within specific ranges, the aim was to maximize the CH₄ yield, the SCOD/TCOD ratio, and lignin removal. In addition, the contour plots for the optimum values of the responses are provided in Figure S2. This approach targeted enhanced CH₄ production, along with maximized biomass solubilization and lignin degradation. Throughout the process, the PF dosage and temperature levels were limited in accordance with economic and operational constraints.

Validation experiments revealed deviations between the predicted and actual values. The CH₄ yield showed acceptable agreement (predicted: 161.6 mL/g VSs; experimental: 148.7 mL/g VSs; 5.8% deviation). Deviations for the SCOD/TCOD (predicted: 0.163; experimental: 0.149; 9.4%) and lignin removal (predicted: 18.41%; experimental: 16.28%; 14.1%) were higher. This aligns with the lower predictive R² values for these responses (Table 2) and suggests the models capture the main trends but have limitations in precisely predicting solubilization and delignification under the optimized point. In conclusion, the developed RSM model demonstrates strong predictive power for the simultaneous control and optimization of both individual and multiple response variables. The high desirability score and low error margins indicate that the model can serve as an effective tool for process control and the development of environmentally sustainable bioconversion systems.

Although there are no prior studies in the literature that directly optimize PF pretreatment conditions for biogas production from lignocellulosic PS waste, our previous work involving similar lignocellulosic feedstocks, such as hazelnut shells, is available [17]. In that study, PF application alone resulted in limited lignin solubilization but achieved CH₄ yields of approximately 175 mL/g VSs. In the present study, under comparable conditions—specifically Run 5 (PF dose: 0.99 g/g TSs; processing time: 60 min)—a CH₄ yield of 170.7 mL/g VSs was obtained, which is largely consistent with the previous findings (Figure S1). Moreover, our earlier research showed that increasing the PF dose from 0.75 to 1.0 mmol/g TSs did not yield a significant improvement in CH₄ production, a trend that has also been confirmed in this study. Here, the optimum PF dose identified through multi-response optimization was 0.637 mmol/g TSs—substantially lower than the maximum values reported in the previous literature. These results support the notion that increasing the PF dosage does not necessarily translate to performance gains and suggest that a more environmentally sustainable approach is feasible through the use of lower chemical inputs.

3.6. Process Mechanisms, Structural Changes, and Implications

The highest CH₄ yield achieved in this study was 234.7 mL CH₄/g VSs, obtained under high-intensity pretreatment conditions (0.66 mmol/g TSs PF, 100 °C, 90 min). This represents a 5.8-fold increase compared to untreated PSs, which yielded approximately 40 mL CH₄/g VSs. Notably, this maximum yield condition is distinct from the multi-objective optimized condition (0.637 mmol/g TSs, 66.76 °C, 55.84 min; 171 mL CH₄/g VSs), which prioritized balanced performance across multiple response variables while minimizing resource inputs. The synergistic mechanism of the combined PF–thermal pretreatment effectively disrupts the lignocellulosic structure through complementary pathways (Figure 3). Thermal action at moderate temperatures (60–100 °C) disrupts hydrogen bonding between lignin and hemicellulose, leading to the solubilization of hemicellulose into pentose sugars such as xylose and arabinose [33]. This weakens the lignin matrix and increases biomass porosity. Simultaneously, PF oxidation involves FeO₄²⁻ radicals selectively cleaving phenolic and ether bonds in lignin, such as β-O-4 linkages, generating soluble fragments [34]. This process also reduces the crystallinity of cellulose, thereby enhancing enzymatic susceptibility. The synergy between these processes is particularly evident at elevated temperatures, which accelerate PF decomposition and radical generation [35]. While alkaline conditions (pH 10) favor PF reactivity, localized pH changes during pretreatment were not measured, highlighting the need for further investigation [17]. Together, these processes break down structural barriers, significantly improving microbial access to cellulose and hemicellulose during AD.

A previous study [17] demonstrated that specific PF dosages selectively reduced lignin in lignocellulosic biomass and facilitated the release of lower-molecular-weight sugars from hemicellulose, both of which enhanced microbial accessibility during AD. Consistent with these findings, the higher methane yields observed in the present study for PF-pretreated lignocellulose, compared to the untreated control, can be reasonably attributed to such structural degradations.

SEM analysis revealed that significant morphological changes occurred depending on the pretreatment intensity applied, and these differences can be clearly observed in Figure 4. Untreated PSs showed a thick, smooth surface with very few pores, consistent with their high lignin content (34.3%), which inhibits microbial accessibility. In contrast, the maximum yield condition (0.66 mmol/g TSs PF, 100 °C, 90 min) resulted in severe structural disruption, including deep cracks, fragmented surfaces, and significantly increased porosity, aligning with aggressive lignin degradation and explaining the 5.8-fold increase in the CH₄ yield. The optimized condition (0.637 mmol/g TSs PF, 66.76 °C, 55.84 min) exhibited moderate structural changes, including surface erosion, micro-cracks, and increased porosity. Despite the lower severity of the treatment, these changes were sufficient to enhance bioavailability and achieve a 4.3-fold increase in the CH₄ yield [17].

Key implications of this study include dose efficiency, energy trade-offs, synergy quantification, and the structure–accessibility relationship. Effective lignin removal (16–19%) and CH₄ yield enhancement (4.3–5.8×) were achieved at low-to-moderate PF doses (0.64–0.66 mmol/g TSs), reducing chemical consumption compared to higher doses (0.99 mmol/g TSs). The 100 °C/90 min condition maximized the CH₄ yield (234.7 mL CH₄/g VSs) but required higher energy input, whereas the optimized condition (66.76 °C/55.84 min) retained over 70% of the maximum yield with approximately 50% lower thermal energy input. The combined PF–thermal pretreatment demonstrated a 40–60% improvement in the CH₄ yield compared to standalone PF or thermal treatments, confirming the synergistic effect. SEM analysis validated the direct correlation between morphological disruption (e.g., cracks and porosity) and the enhanced bioavailability of lignocellulosic biomass for microbial hydrolysis, reinforcing the effectiveness of the pretreatment mechanism.

4. Limitations

In this study, the effects of PF and thermal pretreatment on lignocellulosic biomass were optimized under laboratory-scale conditions, yielding promising results in terms of CH₄ production. However, these findings are valid only under controlled experimental settings. In real-world applications, fluctuations in the PF supply chain and operational costs associated with process scale-up may limit economic feasibility. Moreover, this study focuses on a single type of biomass (e.g., PSs) and maintains a limited reactor scale. This restricts the generalizability of the findings to other biomass sources and process configurations. The recovery, reuse, and environmental impacts of PF were beyond the scope of this investigation. Additionally, molecular-level changes in lignin structure and impacts on microbial communities were not thoroughly analyzed.

In conclusion, the potential of PF-based pretreatment to enhance CH₄ yield has been clearly demonstrated; however, the industrial applicability of this system depends on the net energy balance between energy inputs and outputs, as well as on the cost-effectiveness of PF. Future studies should aim to validate oxidative pretreatment systems in larger-scale biogas production facilities, considering reactor design, process integration, and carbon footprint within the context of energy–economic integrity. Overall, all models exhibited strong statistical correlations in explaining the experimental data. In particular, the PF dosage and pretreatment duration play a critical role across all three response variables. Nevertheless, the relatively low R² values indicate limitations in the models’ applicability to new datasets. Therefore, prospective studies should be supported by hybrid modeling strategies (e.g., artificial neural networks or genetic algorithm-assisted RSM), with additional validation particularly required during scale-up processes.

5. Conclusions

This study demonstrates the efficacy of combining PF with low-temperature thermal pretreatment to enhance CH₄ production from PSs. The RSM successfully optimized the process, identifying conditions (0.637 mmol/g TSs PF, 66.76 °C, 55.84 min) that achieved a CH₄ yield of 171 mL CH₄/g VSs—a 4.3-fold increase over untreated PSs—while balancing lignin removal and solubilization. Notably, a specific run (0.66 mmol/g TSs PF, 100 °C, 90 min) achieved a higher yield (234.7 mL CH₄/g VSs, 5.8-fold increase), highlighting the trade-off between maximizing the output and minimizing resource consumption (chemicals, energy). The synergistic pretreatment disrupted the lignocellulosic structure, enhancing microbial accessibility. While promising, the models showed limitations in predicting solubilization/delignification, and validation deviations indicate the need for cautious extrapolation. Future work should focus on scaling, economic/LCA analysis, PF recovery, and molecular-level characterization of pretreatment effects.