Experimental and statistical modeling of the effect of process modification and wastewater characterization on greenhouse gas emissions for a dairy industry wastewater treatment plant: A minimization approach

In recent years, corrective preventions and policies have been applied to develop system performance by mitigating greenhouse gas (GHG) emissions due to raising attention on the sustainable operation of wastewater treatment plants (WWTPs) (Yapıcıoğlu 2018). The management of WWTPs has concentrated on mitigating operating costs while obtaining effluent discharge limits and minimum GHG emissions (Metcalf & Eddy 2014). The effluent quality and GHG emissions of WWTPs have been mainly influenced by operating conditions, which are hydraulic retention time (HRT) and solid retention time (SRT) for WWTPs. This topic had significant attention worldwide in the last decades. GHG emission generator points have critical importance in determining economic performance in terms of GHG emissions. According to the European (EU) Green Deal, GHG emissions should be reduced in considerable amounts, and this reduction of GHG emissions would lead to economic wealth in the world (EU 2018). WWTPs should obtain incentives to enhance and carry out climate-neutral process management to align with the Green Deal objectives. Especially, industrial WWTPs are an essential element of the EU Green Deal and circular economy (EU 2018). According to the EU Green Deal, there will be a 55% reduction in total GHG emissions by 2030. Within the scope of the Green Deal, the EU envisages reducing the GHG emission value from water treatment by 30% in the next 10 years. From this perspective, GHG emissions should be taken under control and reduced for these types of plants (Adebayo et al. 2021; Moretti et al. 2021; Pahunang et al. 2021; Udemba et al. 2021; Voss et al. 2021).

Industrial WWTPs can be regarded as one of the GHG emission resources due to high organic wastewater content and high energy consumption (Galve et al. 2021; Hashem et al. 2021; Kumar et al. 2021; Pata & Kumar 2021; Asmal et al. 2022; Raji & Packialakshmi 2022). From this perspective, this study mainly aimed to mitigate GHG emissions from a dairy industry WWTP in terms of the EU Green Deal. There are several GHG emission mitigation techniques such as carbon capture systems, innovative wastewater treatment technologies (biochar, microbial fuel (MFC) cell, and microbial electric synthesis (MES) applications), microalgal technology, carbon-neutral processes, as well as process and operational conditions’ modification (Qambrani et al. 2017). Industrial WWTPs have high first investment and operating costs. Due to this reason, process modification or optimization for available treatment processes is the best alternative for reducing GHG emissions, rather than carbon capture systems or innovative wastewater treatment processes, if the discharge standards are obtained with the available treatment systems. In this context, process modification has been fulfilled in order to obtain possible reductions in GHG emissions of a dairy WWTP, in this study. Operational conditions (HRT and SRT) of the aeration tank have been modified to design conditions with the aim of reducing GHG emissions.

This paper aimed to investigate the effect of wastewater characterization and process modification on GHG emissions from an industrial WWTP within the scope of water–energy nexus. The focus of the study is to use process modification in order to reduce GHG emissions and to determine and validate the correspondence between GHG emissions and wastewater quality. The argument of this study is that the mitigation of GHG emissions was aimed at using process modification in terms of the European Green Deal. The major argument of this paper is the validation of the minimization of GHG emissions by applying design conditions with in situ GHG monitoring. Statistical and experimental results have been theoretically discussed and modeled at the end of the study. Analysis of variance (ANOVA), correlation analyses, heat mapping, and principal component analyses (PCA) were performed to determine the correspondence between GHG emissions and wastewater characterization. From this point of view, a benchmarked assessment has been applied to define the correspondence of GHG emissions and operational conditions as well as influent and effluent quality. In this context, a new empirical model that depended on the correlation of wastewater characterization has been developed. From this point of view, GHG emissions have been figured out using this correspondence with wastewater characterization and the results have been compared with the in situ monitoring values. In this paper, chemical oxygen demand (COD), total suspended solid (TSS), fats, oil, and grease (FOG), and Total Kjeldahl Nitrogen (TKN) were considered as inputs, pollutant removal (effluent values) as desirable outputs, and GHG emissions were regarded as undesirable outputs. These key parameters have been selected for the dairy WWTP. Especially COD, FOG, TSS, and TKN are the main indicator pollutant parameters for dairy industrial plants (Metcalf & Eddy 2014).

This study is unique and novel in that the effect of design conditions on GHG emissions for an industrial WWTP in terms of the EU Green Deal was investigated using various statistical approaches and in situ GHG emission monitoring. Also, the other originality of this work is that process modification was investigated as a GHG emission minimization technique. There is a gap in the literature on this topic. So, this study could be a guide for similar industrial WWTPs, which have highly organic content in terms of the reduction of GHG emissions using process modification. This study has dealt with the investigation of the optimal operation of an industrial WWTP in order to mitigate GHG emissions. In this context, process modification could be a GHG emission minimization technique (Sweetapple et al. 2014; Barbu et al. 2017). For the extended aeration (EA) process, process modification has been fulfilled. First, process conditions were modified in terms of HRT and SRT, and then, CO₂, CH₄, and N₂O emissions were monitored in this plant.

For WWTPs, optimal operating conditions could be defined using optimization methods coupled with an estimative mathematical model of the WWTP (Kim et al. 2015). A few studies have investigated optimization solutions that carried out the analysis or operational design of WWTPs. Several researchers have concentrated on model calibration, simulation, and energy efficiency in WWTPs (Kim et al. 2015; Kyung et al. 2015; Yapıcıoğlu & Demir 2021). Kim et al. (2015) studied the optimization of operating conditions in terms of GHG emissions. Another study is related to Kyung et al. (2015), in which they used model optimization. Yapıcıoğlu & Demir (2021) investigated the reduction in GHG emissions, and the effects of design and operational conditions on GHG using Monte Carlo simulation. In this study, ANOVA, correlation analyses, heat mapping, and PCA were simultaneously performed to determine the correspondence between wastewater characterization and GHG emissions. Also, this study is unique in that in situ GHG emissions were measured under design parameters and operational conditions for a full-scale industrial WWTP. The validation of the proposed model has been ensured by in situ monitoring of GHG emissions for design and operational conditions. Also, the effect of wastewater characterization was investigated on GHG emissions. Data interpretation has been done according to correlation analysis. GHG emissions have been also estimated based on this correlation resulting from statistical analysis, which defines the relationship of GHG emissions and wastewater characteristics. In this context, a specific new calculation model has been developed in this paper. Also, this study is original in that model validation has been ensured with in situ GHG emission monitoring in the plant. Zhou et al. (2022) have used a normalizing method for GHG emission and wastewater characterization in a municipal plant. This study confirms statistical analysis with in situ monitoring of GHG emissions using the closed chamber method. Huang et al. (2023) have used the correspondence between GHG emissions and pollutant parameters in a municipal WWTP. They focused on pollutant removal to determine the correspondence with GHG emissions. On the contrary, statistical analyses and in situ GHG emission monitoring have been applied in this study.

Process modification could be a GHG emission mitigation technique (Rodríguez-Caballero et al. 2014; Sweetapple et al. 2014; Barbu et al. 2017). For the EA process, process modification was applied. First, process conditions were modified according to Table 2. The inputs and desirable outputs of the modified process condition for the aeration tank are given in Table 1. Many researchers proposed the operation of WWTPs under design conditions for minimum energy consumption (Castellet-Viciano et al. 2018). The design conditions are 36 h of HRT and 20 days of SRT. From this perspective, the operating conditions have been adjusted to the design conditions. A limitation has been defined for HRT and SRT in this specific range to define GHG emissions not to disrupt the effluent quality and discharge standards in this industrial WWTP. As well, a jar test has been performed before monitoring to detect the optimum operating conditions. GHG emissions have been measured for both design and operation and modified conditions in the plant. Table 1 presents the inputs and desirable outputs of the design conditions.

A sensitivity analysis has been done for all datasets and these parameters. An uncertainty analysis has been performed before statistical analysis. Mean values have been considered and selected in statistical analysis. ANOVA, correlation analyses, heat mapping, and PCA were performed to determine the correspondence between GHG emission and wastewater characterization. R ‘pheatmap’ (Kolde 2019; İsmail et al. 2022), ‘Hmisc’ (Harrell 2021), ‘factoextra’ (Kassambara & Mundt 2020) and ‘FactoMineR’ (Le et al. 2008) packages and minitab for all analyses and charts package program (Minitab Inc. 2017) were used. The data were statistically analyzed using Levene's test for variance equality assumption and the Shapiro–Wilk test was applied for normality assumption (p > 0.05). For this purpose, two-way ANOVA in repeated measures and Tukey's HSD multiple comparison test were used to determine the differences between the groups and analyzed to determine whether there was a difference between the groups. The Pearson correlation was used to determine the relationships between variables. In addition, multivariate PCA was performed to examine the dimensions of the data. The data are presented as n, mean, and standard deviation. All analyses were performed at p < 0.05 significance level.

In order to understand the effect of independent variables on the dependent variable, linear multiple regression analysis was applied. The outputs have been divided into GHG emissions and pollutant removal, and the relevant variables for the EA process, DAF, and GHG emissions (GHG) are included in the analysis separately. In this way, it is thought that the causal properties of the variables will be examined in more detail. In this paper, COD, TSS, FOG, and TKN were considered as inputs, pollutant removal (effluent values) was considered as the desirable output, and GHG emissions were regarded as undesirable outputs. A mass balance that contains carbon and nitrogen content and removal has been performed before the assays.

PCA is a multivariate statistical method that is widely used for pattern identification and variable reduction (Şahin & İşler 2021). The purpose of PCA is to minimize the dimensionality of the data while preserving the variation of the original dataset. PCA decomposes the highly correlated variables of a dataset into smaller unrelated variables known as principal components. They are linear combinations of weighted original variables. In general, the first few principal components are responsible for almost all the variability in the data, and the rest is explained by the later components. The first principal component tends to account for most of the variability of the data. The first step in using PCA to select variables is to look for PCAs with eigenvalues greater than 1 that show PCs that explain the most variation in the dataset.

According to the analyses results, process modification has a serious impact on GHG emissions. CO₂ emissions were the lowest direct GHG emission in December. CO₂ emission was not observed at the DAF tank. Table 2 shows the GHG emission monitoring results and operating conditions related to the available process conditions. The highest GHG emission was observed in the EA process in September as N₂O emissions. It could be originated from wastewater nitrogen content and the denitrification process. Caniani et al. (2015) similarly reported that N₂O emission had the highest value for the membrane processes for a WWTP. According to the results, GHG emissions were closely correlated with process conditions and operational parameters of the WWTP. Also, Rodríguez-Caballero et al. (2014) researched the process conditions on GHG emissions for municipal wastewater treatment. They similarly found that process conditions have a considerable impact on GHG emissions. Also, similar results have been obtained in terms of seasonal variations with this study. They similarly reported the lowest GHG emissions in winter and the highest GHG emissions in summer.

According to the results, lower direct GHG emissions were measured at design conditions for all types of GHG emissions. From these results, it could be said that if the plant is operated at design conditions, lower GHG emissions would be released. Also, process modification has a serious impact on GHG emissions. The results revealed that N₂O emissions had the highest value among GHG emissions due to its global warming potential. The highest GHG emission was related to N₂O emissions in September at operational conditions (HRT = 26 h, SRT = 15 days). The lowest GHG emission is CO₂ emission in June with a value of 300 mg/L at design conditions (HRT = 36 h, SRT = 20 days). From this perspective, it could be said that if the plant is operated at design conditions, lower GHG emissions could be monitored. If the aeration tank is operated at 36 h of HRT and 20 days of SRT at design conditions, an average reduction of 45.4, 45.3, and 45.2% in CO₂, CH₄, and N₂O emissions have been ensured, respectively. An average reduction of approximately 46% in overall GHG emissions has been reported in the study. A limitation has been designed for GHG emission monitoring points because the assumption contains that GHG emissions could be originated from organic material removal. If chemical and energy consumption has been considered, then a larger GHG emission has been calculated. Inorganic material removal has been applied in the screens and settling tanks. These units are not monitoring points, so it could be said that they have no effect on this correlation.

There are limited studies on this topic in the literature. Kim et al. (2015) studied the optimization of operating conditions in terms of GHG emissions. Similarly, they reported a 31% reduction, while reducing the operating costs by nearly 11%. They recommended an integrated performance index including GHG emissions, operating costs, and effluent quality. The novelty of this paper is that the impact of design conditions within the scope of GHG emissions has been investigated and the possible GHG emission mitigation that was calculated depended on GHG emission measurement. In this study, approximately 46% reduction has been achieved by applying design conditions in the plant. If a WWTP is operated at design conditions, more reduction in GHG emissions has been obtained rather than the optimization of operating conditions. Also, Rodríguez-Caballero et al. (2014) investigated process conditions on GHG emissions for municipal wastewater treatment. They reported that similar process conditions have a considerable impact on CO₂, CH₄, and N₂O emissions. They reported that, similarly, various GHG emissions result in various operational conditions. Molinos-Senante et al. (2014) investigated the economic and environmental performances of WWTPs in terms of the reduction of GHG emissions. They estimated potential future reductions in GHG emissions using the economic approach in the scope of energy consumption. In this study, the focus has been on process modification. In the literature, some studies recommended energy efficiency and energy consumption reduction for carbon-neutral operations in WWTPs. Badeti et al. (2021) revealed that the simulation shows that 33% of energy could be saved, and N₂O and CO₂ emissions could also be minimized by 98 and 25%, respectively, with energy efficiency.

ANOVA, correlation analyses, heat mapping, and PCA were performed to define the correspondence between GHG emissions and wastewater characterization. In this section, all statistical analyses are shown in detail.

When the ANOVA results for inputs (pollutant parameters) and desirable outputs (pollutant removal and effluent value) were examined, it was found that, statistically, the DAF process had the highest mean and the EA process had the lowest mean in terms of COD, TSS, TKN, and FOG variables (p < 0.001). Table S1 in the Supplementary material shows the results of the ANOVA statistics. In addition, since there were no GHG emission values for FOG, only the difference between EA and DAF processes was examined and it was observed that DAF was significantly higher than EA for this parameter. On the other hand, when undesirable output (GHG) CO₂, CH₄, N₂O emission (mg/L) parameters were examined, it was reported that the EA process for CO₂ was higher in terms of statistical averages; it was observed that the EA process had the highest value. In addition, while data for DAF for HRT and SRT were not available, data for GHG were not included in the analysis because they were obtained the same in all months.

Regression analysis results overlapped with the results of ANOVA statistics. Table SM3 shows the effects and ratios of independent variables on the dependent variables of CO₂, CH₄, and N₂O. The CO₂ emission model was not significant, while DAF_PR was significant for CH₄ emissions. It was observed that the N₂O model was predicted significantly in both. When the regression coefficients in CH₄ models for the DAF process were examined, it was estimated that TKN was above DAF_P, while DAF_PR was estimated as FOG. It could be estimated that FOG removal has been applied at the DAF tank. Nitrogen removal could be obtained in the lowest value at the DAF tank. Explanation coefficients for CH₄ (R²) are DAF_P 44.42% and DAF_PR 72.92%. In other words, while DAF_PR of the independent variables explains 72.92% of the change in CH₄, DAF_P explains 44.42% of the change in CH₄. In another definition, it means that lower CH₄ emissions are released from the DAF tank rather than the EA process. Similarly, according to the ANOVA statistical results, the highest GHG emission has been reported at the EA tank for all conditions.

In the derivative models (Equations (3)–(5)), GHG^I (kg CO₂e/d) represents the estimated CO₂, CH_4, and N₂O emissions. Q is the wastewater flow rate, and it also means the volume of wastewater (m³/d). The other variable is the pollutant removal values. Correlation analysis showed that the COD removal (COD_R) value closely corresponded with CO₂ and CH₄ emissions, and nitrogen (TKN) removal (TKN_R) (kg/m³) nearly corresponded to N₂O emissions. In the calculation tools, r₁, r₂, and r₃ are the Pearson correlation coefficients resulting from statistical analysis and the values of GWP are the global warming potentials of each related GHGs. While considering negative coefficients, TSS_P (input) values have a negative impact on CO₂ and CH₄ emissions. It could be said that low TSS inputs could lead to higher CO₂ and CH₄ emissions. It could vary according to the wastewater content and organic load of the plant, although TSS_PR (effluent) values have a significant correlation with CO₂ emission (r = 0.843). It could have originated from the degradation of carbonaceous materials to carbon dioxide. FOG is an important wastewater parameter for dairy industry plants. While considering FOG effluent values, it could be said that a moderate correlation is available between CO₂ emissions (r = 0.571). Even so, FOG (effluent) has a weak correlation with CH₄ and N₂O emissions with values of 0.111 and 0.172, respectively. SRT and HRT are the main operational parameters of the EA process. Also, these operational parameters could have a great impact on GHG emissions. The results revealed that HRT strongly corresponds with CO₂ and N₂O emissions with values of 0.825 and 0.703, respectively. Also, SRT has a strong correlation with CO₂ emissions (r = 0.689).

In the literature, several studies have been focused on municipal WWTPs. Zhou et al. (2022) analyzed GHG and contaminant parameters resulting from municipal WWTPs applying the normalizing method. Similar to this study, they found that CH₄ and N₂O emissions have been highly corresponded with the COD and TKN parameters, respectively. They have only confirmed using the normalizing approach and calculated GHG emissions from organic pollutant removal. In this study, direct GHG emissions have been monitored using a flux chamber and analyzed with gas chromatography, whereas statistical analysis has been validated with the in situ experimental GHG monitoring results. Huang et al. (2023) have similarly used the correspondence between GHG emissions and wastewater quality parameters for a municipal WWTP. They reported that COD removal has a major significant effect on GHG emissions. In a study, Yapıcıoğlu (2021) figured out GHG emissions using pollutant removal. Apart from these studies, this study has validated the scientific argument with in situ experimental analysis of GHG emissions.

In this study, statistical and experimental modeling of the effect of process modification and wastewater characteristics on GHG emissions has been done for an industrial WWTP. This study confirms that process modification could be performed to reduce GHG emissions for the aeration process in WWTPs. If the aeration tank is operated at 36 h of HRT and 20 days of SRT at design conditions, an average reduction of 45.4, 45.3, and 45.2% in CO₂, CH₄, and N₂O emissions have been ensured, respectively. It is obvious that process modification has a significant effect on GHG emissions. According to correlation analysis, a moderately significant positive correlation was found between the COD effluent value and CO₂ and CH₄ values (r = 0.909 and r = 0.937). N₂O emissions were closely correlated with TKN input values (r = 0.876). From this point of view, a new GHG emission estimation tool has been developed based on this correlation. The results revealed that the estimated results have been overlapping with the in situ monitoring values. The estimated results of GHG emissions based on wastewater characterization have converged to the monitoring results by an average of 82%. The estimated results of CO₂, CH₄, and N₂O emissions based on statistical analysis have converged to the GHG emission monitoring results by an average of 77, 83, and 84%, respectively. This study focused on direct emission due to the treatment process to reveal the effect of process conditions, so these main two treatment units have been selected. Chemical and energy consumption of the plant has led to indirect GHG emissions. It is ignored because the focus is on the main treatment process. This study has added a scientific contribution to the literature on the reduction of GHG emissions from industrial WWTPs using process modification. Also, the validation of the recommended statistical model has been obtained with in situ experimental GHG emission monitoring. More research should be developed for the reduction of GHG emissions from industrial wastewater treatment plants.

This research received no external funding.

P.Y. carried out GHG emission monitoring and prepared the manuscript and H.Y. verified the analysis work. M.I.Y. compiled the manuscript with proper corrections.

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

The authors declare there is no conflict.