3.1. Ultimate Experimental Specific CH4 Yield (Bo) of Single Substrates
The results from the BMP assays of the single substrates are shown in
Figure 2. Depending on the prevalent biochemical composition of the substrates, it is possible to divide the results into two groups. The first group includes the substrates with lignocellulosic nature, namely the Paunch Wastewater (PWW) and Bovine Manure (BM), which had low CH
4 production due to their high content in scarcely degradable lignocellulose (
Table 2): from the start of the BMP assay until day 12, the cumulative CH
4 yields of both substrates were almost similar (
Figure 2). However, from day 12, the increase of the PWW yield slowed down and approached its plateau, whereas the BM yield continued to rise until reaching its stable value from approximately day 25. The above behaviors are similar to those found in previous studies on digestion of bovine manure [
44] and PWW [
13], showing a relatively higher rate of degradation of PWW compared to manure.
BM resulted in a
Bo, at 30 days, of 0.206 ± 0.003 m
3 CH
4 kg
−1 VS and an
fd of 0.46 ± 0.00, which are in the range of
Bo values reported for dairy manure (0.089–0.303 m
3 CH
4 kg
−1 VS) [
44,
45] and close to the biodegradability published in previous studies (0.54) [
22]. PWW resulted in a
Bo and an
fd of 0.154 ± 0.011 m
3 CH
4 kg
−1 VS and 0.34 ± 0.01, respectively. These values are lower than those found for PWW in Australian slaughterhouses (0.309 m
3 CH
4 kg
−1 VS and 0.84) [
13]. Since the composition of ruminal content depends on how long the grass remains in the stomachs of animals [
46], the above differences can be attributed to variations in the animals handling before slaughter. According to Australian regulation, animals must stay 24 h in yards before slaughter to be checked and to ensure that they are healthy [
47]. However, in Colombian slaughterhouses, animals can be slaughtered 6 h after arrival [
48].
The second group is formed by Offal Wastewater (OWW) and Slaughter Wastewater (SWW), which, contrary to the first group, are richer in lipids and proteins (
Table 2), resulting in a relatively higher CH
4 production (
Figure 2). During the first 3 days, the CH
4 yield of OWW and SWW did not present significant differences (
p > 0.05). However, from day 4 to 10, the CH
4 yield of OWW increased at a higher rate than SWW and then slowed down from day 11 until it reached a steady-state at about day 25. On the other hand, in the case of SWW, the CH
4 yield presented an almost constant increase until about day 17, where it declined and achieved a plateau on day 25. Previous studies have shown how anaerobic digestion of wastes with high lipid concentrations result in a long lag period, due to LCFA accumulation and inhibition. For instance, Jensen et al. (2014) [
4] reported a lag period of 18 days during anaerobic digestion of lipid-rich wastewater (10 g/L). In turn, Harris et al. (2018) [
49] evidenced 7 days of lag period for anaerobic digestion of DAF (dissolved air flotation) sludge (10.5 g lipid/L). Likewise, Andriamanohiarisoamanana et al. (2017) [
17] found that the BMP curve of crude glycerol presented an atypical shape (constant increase in the first 5 days followed by a slow CH
4 production until day 15 and then an exponential behavior) due to LCFA inhibition. On the contrary, in the current study, the BMP assays of SWW and OWW started CH
4 production from the first day, their curves had a typical behavior and their lipids concentration was lower than 10 g/L. This indicates how LCFA is unlikely to be a source of inhibition during anaerobic digestion of the tested slaughterhouse wastewater streams.
Ammonia is another potential cause of inhibition, which results from substrates with high protein content. In this regard, the BMP assay with SWW presented a final NH
3 concentration of 21.12 ± 0.25 mg L
−1, which is higher than the measured inhibition coefficient K
I50-
NH3 of the inoculum (18.53 ± 0.34 mg L
−1). Various studies investigated ammonia inhibition effects on BMP assays and reported experimental curves that were qualitatively similar to the present study. For instance, Nielsen and Angelidaki (2008) [
50] evaluated the anaerobic digestion in BMP assays of cattle manure, with different initial total-N concentrations. The ammonia inhibition was evidenced in the slope of the cumulative CH
4 curves, which decreased with increasing initial nitrogen. In particular, samples with a total-N concentration of 3.0 g L
−1 and 3.5 g L
−1 achieved the same ultimate CH
4 yield. However, the former sample reached 80% of its ultimate CH
4 yield at 13 days while the latter reached 80% at 21 days; this result also highlights how ammonia inhibition follows a threshold behavior [
35]. Similarly, Cuetos et al. (2017) [
51] investigated the effect of active carbon addition in the anaerobic digestion of poultry blood (which is similar to the slaughter wastewater of this study). The experiments with lower activated carbon contents resulted in NH
3 inhibition and a significantly lower rate at the beginning of the BMP curve (specifically, during the first 13 days). The aforementioned analysis and studies confirm the likelihood of NH
3 accumulation and inhibition during the mono-digestion of SWW.
SWW and OWW BMP assays resulted in a
Bo of 0.505 ± 0.008 and 0.425 ± 0.015 m
3 CH
4 kg
−1 VS, respectively. Although OWW has the highest lipids content, it presented lower
Bo than SWW due to the concomitant presence of lignocellulosic material (
Table 2). The
Bo of SWW was close to the values of 0.500 and 0.570 m
3 CH
4 kg
−1 VS reported in the studies of Jensen et al. (2014; 2015) [
4,
52], while the
fd resulted in a value of 0.80 ± 0.01, which is close to the results of a similar BMP study investigating blood biodegradability (
fd of 0.77) [
12]. On the other hand, the
Bo of OWW is lower when compared to studies investigating similar substrates. For instance, Jensen et al. (2014) [
4] found a
Bo between 0.721 and 0.931 m
3 CH
4 kg
−1 VS for an offal wastewater stream. Nevertheless, this wastewater also contained the waste stream from the cleaning of red viscera, resulting in a higher lipid concentration (up to 11.64 kg m
−3) compared to the OWW stream in the current study, thus explaining the relatively higher
Bo. Regarding the
fd from OWW (0.63 ± 0.02), to the best of the author’s knowledge, there is no available comparison in the literature.
3.2. Experimental Ultimate Specific CH4 Yield of AcoD
Figure 3 shows the composition (lipids, proteins and carbohydrates) and the ultimate experimental yield
Bo of the different AcoD mixtures evaluated (the BMP curves are depicted in
Supplementary Data Figure S1). On the whole, for both binary and multicomponent mixtures, the
Bo increased directly with the proportion of lipids and decreased with the proportion of carbohydrates. Therefore, the highest
Bo corresponds to the binary mixtures of SWW and OWW (S33:O67 and S67:O33) and the ternary mixtures where SWW and OWW were present simultaneously (S33:O33:B34 and S33:O33:P34).
The ternary and quaternary mixtures had significantly higher
Bo (
p < 0.05) than binary mixtures with a similar biochemical composition. For instance, the combinations with the mixing ratio of S33:B67 and S33:P33:B34 have almost the same composition (~11%VS lipids, ~31%VS protein and ~58%VS carbohydrates); however, the latter mixture showed a
Bo 40% higher than the former. Likewise, the ternary mixture O33:P33:B34 exhibited a
Bo 10% higher than binary mixtures O33:P67 and O33:B67, despite having similar compositions (~15%VS lipids, ~20%VS protein and ~65%VS carbohydrates). When comparing the ternary mixtures with the highest
Bo (mixtures S33:O33:B34 and S33:O33:P34) to the binary mixture with the highest
Bo (S67:O33), the ternary mixtures have similar
Bo (4–14% difference), while having 33% fewer proteins and 25% fewer lipids than the binary mixture. The above evidence a higher synergy between macromolecules on CH
4 production in multicomponent mixtures than in binary mixtures. This result is in agreement with the study by Astals et al. (2014) [
12], who suggested that in addition to macro-composition, the structure of the substrates also affects their interaction. In this sense, there are differences in carbohydrates structure between PWW and BM and the kind of proteins between SWW and OWW.
The effects of AcoD on the reduction of initial lignocellulosic material composition and final NH
3 concentration (see
Supplementary Data Table S1 for NH
3 calculation details) are shown in
Table 6, taking biodegradability (
fd) as an indicator. In the case of BM and PWW, the co-digestion with OWW and SWW in binary or multicomponent mixtures allowed to achieve mixtures with relatively lower lignocellulosic content; this reduced the recalcitrant character of the mixture and as a consequence increased the biodegradability
fd above the values of both BM and PWW mono-digestion (0.46 and 0.34, respectively). On the contrary, the binaries AcoD between BM and PWW presented a high lignocellulosic composition, which resulted in an
fd around 0.44. Previous studies have demonstrated that the AcoD with lignocellulosic residues is an alternative to enhance the C/N ratio of animal manure; however, this requires pretreatment [
53].
In the case of OWW, all its mixtures presented higher
fd than its mono-digestion (0.63), since fatty wastes are suitable co-substrates to lignocellulosic and protein wastes [
12]. In turn, SWW showed the highest degradability of individual substrates (0.80) due to its content of soluble proteins in the blood (e.g., albumin and globulin), which are hydrolyzed fast and then converted to CH
4 while producing NH
3. In the case of SWW, AcoD offers the opportunity to reduce the risk of ammonia inhibition, through mixtures with substrates with lower protein content. For instance, the addition of PWW to SWW in binary mixtures allowed to reduce the inhibition risk by NH
3 and achieved an
fd around 0.7. The ternary mixture with a mixing ratio S33:O33:P34 exhibited an
fd (0.83) higher than SWW mono-digestion, which is consistent with its balanced composition of carbohydrates, lipids and proteins (
Figure 3).
On the other hand, important inhibition risk occurred during binary AcoD mixtures between BM and SWW, as indicated by the final NH
3 concentration being higher than K
I50-NH3, which led to a significantly lower
fd (
p < 0.05) than the other AcoD mixtures of SWW. A similar result was presented by Andriamanohiarisoamanana et al. (2017) [
17], who investigated the AcoD of meat and bone meal and manure in BMP assays. This study showed how the increase of meat and bone meal content from 10% to 66%VS caused inhibition by NH
3 and, as a consequence, the conversion rate of meat and bone meal to CH
4 was reduced. In the current study, the inhibitory effects between SWW and BM were mitigated in ternary and quaternary mixtures by dilution with OWW and PWW. Similarly, previous studies have highlighted lignocellulosic as a suitable co-substrate for anaerobic digestion of blood. For instance, López et al. (2006) [
54] evaluated the AcoD of ruminal content and blood in batch digesters. The results showed an organic matter degradation from 55 to 70% when ruminal content/blood ratio (on a TS basis) varied between 2 and 8; the authors highlighted how during AcoD blood generates extra buffer capacity and brings micronutrients to the system. Cuetos et al. (2013) [
55] conducted batch experiments on AcoD of poultry blood with maize residues. When maize concentration increased from 15% to 70% (VS basis), the CH
4 production raised from 0.130 to 0.188 m
3 kg
−1 VS. Similarly, also in CSRT digesters, the AcoD of blood and organic fraction of municipal solid waste has been implemented in order to achieve stable operations, with a CH
4 yield between 0.200 and 0.289 m
3 kg
−1 VS [
56].
Because of the aforementioned drawbacks, the mixtures between BM and PWW and between BM and SWW can lead to low values of biodegradability and instabilities, respectively, in the digestion process (see bold/italic values in
Table 6). Hence, these mixtures were excluded from the following sections to focus on the seemingly synergistic mixtures.
3.3. Kinetic Model Selection
The goodness of fit of the Gompertz and first-order models, and the respective estimated kinetic parameters, are summarized in
Table 7. The best model was selected based on two statistical criteria: the normalized root mean square error (NRMSE) and the regression coefficient (R
2). NRMSE is the standard deviation of the prediction errors (residuals). Thus, NRMSE is a measure of how far the experimental points are from the simulated curves. R
2 provides a further measure of how well the model can reproduce the experimental data. For all mixtures, the Gompertz model resulted in a better fit of the experimental data compared to the first-order model. In particular, the ranges of NRMSE and R
2 were 0.011–0.044 and 0.992–0.999, respectively, in the modified Gompertz model and 0.037–0.134 and 0.918–0.988, respectively, in the first-order model. The confidence interval of the estimated parameters for Gompertz (reported as standard error, and shown in
Supplementary Data Table S2), is in all cases below 3% for the simulated ultimate yield
P and below 4% for the maximum specific CH
4 production rate
Rmax. For the lag-phase λ, the average error is 17%, with the highest value of 70% in the case S33:P33:B34, due to the smallest estimated value of the lag-phase (0.152 days). Given the better goodness of fit and the acceptable parameter identifiability, the Gompertz kinetics was selected for the following model-based analysis of the AcoD synergy (
Section 3.4).
Figure 4 shows a selection of six AcoD BMP experimental data, together with the fitted Gompertz and first-order model; the complete set of curves is shown in
Supplementary Data Figure S2.
Figure 4a–c show three experiments which resulted in the smallest differences in the goodness of fit between the two models, with all cases achieving high values of the regression coefficient (R
2 > 0.98). These experiments correspond to the AcoD mixtures S33:P67; S33:P33:B34 and O33:P33:B34; it can be noted how they all have relevant content of the lignocellulosic substrates manure (BM) and paunch (PWW). In these cases, hydrolysis is significantly the rate-limiting step in the CH
4 production [
53]. For first-order models, the hydrolysis rate coefficient of these mixtures resulted in the range 0.06–0.12 d
−1, which is similar to the value of 0.1 d
−1 reported for paunch content by Jensen et al. (2016) [
13].
On the other hand,
Figure 4d–f shows the three experiments that presented the greatest deviation from the first-order model, namely, S33:O67, O67:P33 and O67:B33. It can be noted how these cases have a relevant content of lipid-rich offal wastewater (OWW). The lipid content from these mixtures caused an initial low CH
4 production, which is reflected in a significant value of the lag-phase (
λ) between 2 and 3 days. After the lag-phase the CH
4 production occurred at a relatively high rate (
Rmax between 0.036 and 0.044 m
3 CH
4 kg
−1 VS d
−1), which is comparable to the other mixtures. Similar behavior is reported by Astals et al. (2014) [
12] in the anaerobic digestion of olive oil; the authors attributed the behavior to an initial LCFA absorption onto the surface of the microorganisms, which is followed rapidly by conversion to CH
4.
In general, ternary and quaternary AcoD mixtures had lower
λ values (range: 0.152–1.466 days; average 0.95 days) compared to binary mixtures (range: 0.281–2.982 days; average: 1.61 days) (
Table 7). The λ range obtained in the current research is lower than values reported in previous research on slaughterhouse wastewater anaerobic digestions, with the work of Jensen et al. (2014) [
4] reporting values of up to 18 days for lipid-rich streams. There is limited information on
Rmax in the anaerobic digestion of slaughterhouse wastewater. Hernández-Fydrych et al. (2019) [
57] analyzed the CH
4 production kinetics of pretreated combined slaughterhouse wastewater by BMP assays. The authors fitted a Gompertz model and calculated a
Rmax of 0.0125 and 0.0140 m
3 CH
4 kg
−1 VS d
−1 for autoclaving and mechanical pretreatment, respectively. These values are lower than those found in this study (0.022–0.044 m
3 CH
4 kg
−1 VS d
−1). Therefore, the possibility of controlling the mixture ratios of slaughterhouse wastewater streams in anaerobic co-digestion can have kinetics advantages, when compared to the digestion of the wastewaters’ individual streams or combined as a whole.
3.4. Evaluation of Synergy Effects
Figure 5 represents the synergistic effects of AcoD based on CH
4 yield (ϕy), lag-phase (ϕλ) and CH
4 production rate (ϕR). The predictive BMP curves along with the modified Gompertz plots are depicted in
Supplementary Data Figure S3. All mixtures resulted in an experimental CH
4 yield higher than the expected (ϕy > 0). This result agrees with the evaluation presented in
Table 6 and reaffirms the AcoD ability to reduce the inhibition risk by NH
3 and to improve the biodegradability of slaughterhouse wastewater and manure. Regarding the kinetic synergy, antagonistic effects were observed in some mixtures (left side of
Figure 5). Four AcoD mixtures resulted in a negative synergy with respect to the lag-phase (ϕλ < 0); these mixtures were characterized by a relatively high lipid proportion (23–34%VS), which slowed down the production of CH
4 during the first 2 or 3 days (
Table 7). This observation is in agreement with the study on AcoD of dairy manure, meat, bone meal and crude glycerol carried out by Andriamanohiarisoamanana et al. (2017) [
17], where an increase of glycerol proportion from 13%VS to 37%VS doubled
λ. Additionally, antagonistic effects for
Rmax (ϕR < 0) were presented in four AcoD experiments.
Comparing the binary and multicomponent AcoD, greater synergy was observed in the latter. The binary mixtures exhibited synergistic factors between 4.2% and 38.0% for ϕy, between 3.4% and 81.5% for ϕλ and 5.6% and 29.5% for ϕR. Meanwhile, the ternary and quaternary mixtures showed synergistic factors between 14.5% and 41.9% for ϕy, between 31.1% and 87.9% for ϕλ and 2.1% and 73.9% for ϕR. This highlights the advantage of multi-component AcoD over binary ones, both in the final CH
4 yield and in the kinetics of production. Similar findings were found by Ara et al. (2015) [
18] during AcoD of organic fraction of municipal solid waste, primary sludge and thickened waste activated sludge; the ternary mixtures exhibited CH
4 yields between 12 and 27% higher than binary mixtures. Additionally, Castro-Molano et al. (2018) [
39] observed higher ϕy factors in ternary mixtures (25–167%) than binary mixtures (5–68%) when chicken manure was co-digested with industrial wastes.
The results showed seven mixtures in which all three synergistic factors were positive (ϕy > 0, ϕλ > 0 and ϕR > 0); these mixtures were considered fully synergistic and depicted on the right side of
Figure 5. However, the synergistic effects in the AcoD with the mixing ratio of S67:O33 were relatively small, with values below 10%; these small values of synergy are generally considered not significant in AcoD studies [
23]. Furthermore, the binary mixtures with significant synergy presented the BM or PWW as main substrates. This analysis suggests that when wastes with potential high CH
4 yield (e.g., SWW and OWW) are combined with the wastes with lower potential (e.g., BM and PWW), strong positive interactions are generated; on the other hand, weaker interactions occur when mixing wastes with similar characteristic (e.g., SWW with OWW and BM with PWW). Similar evidence can be found in the literature, such as in a study by Astals et al. (2014) [
12], where the AcoD of DAF sludge and blood did not present significant synergy in CH
4 production; however, when DAF sludge was blended with paunch waste, the resulting CH
4 yield was 15% higher than expected. Likewise, Pagés-Diaz et al. (2014) [
21] found antagonist effects in CH
4 production rate and no significant interaction in CH
4 yield when manure was co-digested with various crops (green fruit, vegetable residues and straw). Nevertheless, the AcoD of manure with slaughterhouse wastes presented significant synergy in both the production rate and yield of CH
4.
The six mixtures with significant full synergy correspond to the combinations: S33:P67; O33:P67; O33:B67; S33:O33:P34; S33:P33:B34 and S25:O25:P25:O25. These AcoD presented a lipids composition relatively lower (11–23%VS) than the rest of the mixtures (19–34%VS), while the carbohydrates and proteins did not show noticeable differences. Thus, it seems that the lipid concentration is the one that most influences the AcoD of slaughterhouses wastewater streams and BM, since a high concentration can improve CH
4 yield; however, it negatively affects the kinetics. The aforementioned fully synergistic mixtures could improve the anaerobic digestion performance of slaughterhouse wastewater streams and manure in tubular digesters. In this sense, the current results are a starting point for a second stage of investigation where the synergistic mixtures will be tested in semi-continuous laboratory trials. This will allow to determine the effect of operational variables HRT and OLR and compare the synergistic effects achieved in the batch test with the synergy in semi-continuous processes, using the same model-based analysis described in this paper. The semi-continuous operation may result in the adaptation of the microbial community to inhibitors, therefore changing the absolute value of the synergistic effects while maintaining a similar qualitative evaluation of the synergy as achieved through batch tests [
58].
3.5. Energy and Economic Feasibility
Table 8 shows a summary of the energy and economic study for the implementation of anaerobic digestion of the slaughterhouse wastewater streams and BM in mono-digestion and AcoD scenarios (see
Supplementary Data from Tables S3–S8 for complete data). Mixtures present 27% more energy potential than single substrates as a consequence of the synergistic effect on methane yield (ϕy). Likewise, the anaerobic digestion of the mixtures would need almost 30 m
3 less digester volume compared to anaerobic digestion of the single substrates. This is due to the synergistic effects on kinetics, which reduce the estimated HRT on average by 3 days.
According to the energy potentials, the treatment of slaughterhouse wastewater streams and BM through anaerobic digestion would allow an energy saving between 0.91 and 1.21 US$ m−3 of waste in the mono-digestion scenario and between 1.16 and 1.53 US$ m−3 of waste in the AcoD scenario. These values added with the saving related to the avoided costs of current waste treatment (1.30 US$ m−3 of waste) result in an economic benefit from 2.21 to 2.51 US$ m−3 of waste and from 2.46 to 2.83 US$ m−3 of waste for mono-digestion and AcoD scenarios, respectively. The economic assessment shows that the CH4 transformation into electric energy leads to higher NPV and IRR compared to the transformation into thermal energy. This is due to the low price of natural gas (0.026 US$ kWh−1) compared to electricity (0.114 US$ kWh−1). However, in both cases (electrical and thermal generation), the PBP is lower than the equipment lifetime (10 years), NPV is positive and IRR is higher than the discount rate (10%). These results confirm the energetic and economic feasibility of anaerobic digestion of slaughterhouse wastewater streams and manure. Moreover, the economic parameters (PBP, NPV and IRR) are better in the AcoD scenario than the mono-digestion scenario. This demonstrates that the synergistic effects of the mixtures also translate into economic advantages.
In developing countries, most slaughterhouses are located in small towns and supply only the local demand for meat (rural population mainly) [
7]. Therefore, these slaughterhouses have low income, which limits their investment capacity in technology. In this sense, the tubular digester is a suitable alternative for waste treatment, given its low capital cost (compared to other kind of reactors), its simplicity of operation and lack of energy requirements for its operation [
8]. Additionally, this type of waste management and renewable energy projects can access green financing. For instance, the Latin American banking sector has been developing a series of green products to finance projects that mitigate global warming [
59]. Regarding Colombia, the country will issue green bonds in 2021 directed to finance sustainable and environmentally friendly projects [
60].