The experiments occurred between November/2016 and January/2017, carried out at Embrapa Swine and Aves (Concordia, Brazil: 27° 18’ 46’’S, 54° 59’ 16’’W). The research was carried out in two replicates of the treatments, using carcasses of different animal species for each replicate: I) accelerated composting of pig carcasses; II) accelerated composting of poultry carcasses.

The study was performed in two replicates due to the volume of the reactor in commercial size, which requires a large number of animals and substrate for composting.The study consisted in evaluating 4 treatments regarding the time in which the reactor remains at rest between the aeration periods (drum rotation). The times interval between the aeration periods studied were: Treatment 1 (T1): 1 hour; Treatment 2 (T2): 2 hours; Treatment 3 (T3): 3 hours; Treatment 4 (T4): 4 hours. Drum rotation time was 24 minutes for all treatments at 0.16 rpm. Sawdust, a by-product of the local wood industries, was used as the carbon source and bulk agent.

In the first replica, where these treatments were evaluated for accelerated composting of pig carcasses, a carcass mass:sawdust ratio of 0.63:1 was used for all treatments. In the second replicate, in which the same treatments were repeated, but with accelerated composting of poultry carcass, a carcass of poultry:sawdust ratio of 2:1 was used for the 4 treatments studied. Table 1 shows the physical-chemical characteristics of the animal carcasses and sawdust used in each of the replicates. For each replica, the characterization of the carcasses used in each treatment is presented, showing small variations among the treatments due to the heterogeneity of the carcasses. Also, the characteristics of sawdust are presented which, because it is more homogeneous, did not make it necessary to characterize it for each treatment. Finally, the biomass mixture (animal carcass + sawdust) used in each treatment is presented, the biomass characteristics being the result of the weighted average between the carcass and sawdust characteristics used. Carcasses were previously ground, homogenised and weighed, and they were manually added to the reactors mixed with sawdust, which was also weighed.

Table 1:

Replica	Material	Treat	NM (kg)	DM (%)	N (%)*	C (%)*	K (mg/kg)*	pH
I	Pig carcass	T1	188	28.41	8.99	55.0	8954.59	5.29
		T2	188	31.67	11.07	61.80	8929.95	6.25
		T3	188	30.52	10.28	59.44	9376.98	5.32
		T4	188	31.45	10.42	54.95	12411.12	5.51
	Sawdust	-	300	55.94	0.160	47.76	909.25	5.65
	Mix (Pig Carcass + Sawdust)	T1	488	45.33	2.30	49.52	4008.68	5.51
		T2	488	46.59	2.79	51.15	3999.19	5.88
		T3	488	46.14	2.60	50.58	4171.41	5.52
		T4	488	46.50	2.64	49.5	5340.30	5.59
II	Poultry Carcass	T1	300	34.82	7.80	56.22	6479.04	5.85
		T2	300	34.44	9.15	52.33	8170.73	5.78
		T3	300	35.61	9.35	55.46	7346.25	5.85
		T4	300	34.83	8.34	55.11	7499.28	5.78
	Sawdust		300	54.88	0.13	47.89	352.42	5.79
	Mix (Poultry Carcass + Sawdust)	T1	600	44.85	3.97	52.05	3415.73	5.78
		T2	600	44.66	4.64	50.11	4261.58	5.75
		T3	600	45.24	4.74	51.67	3849.34	5.78
		T4	600	44.85	4.23	51.50	3925.85	5.75

*On dry basis. NM (natural matter); DM (dry matter); N (nitrogen); C (carbon); K (potassium).

Four reactors were used for each experimental stage (replicate), and one was used for each treatment. The RDRs were made of carbon steel with a length of 1.7 m, diameter of 2.3 m and total volume of 3.6 m³. Fifty percent of the useful volume of the RDRs was used for the biomass, and the control of the rotation frequency was automated. A schematic diagram of the reactor is presented in Figure 1.

Figure 1:

The reactors were equipped with fans that worked intermittently during the experiments, ensuring air renewal inside the reactors and contributing to biomass aeration. The air flow rate (m³.h^-1) in each reactor was as follows: T1) 290.82; T2) 216.44; T3) 226.21; and T4) 262.17.

Physicochemical analyses of the material were performed at the Laboratory of Physicochemical Analysis of Embrapa Swine and Poultry. Animal carcass (pig and poultry) samples were collected during carcass grinding for carcass characterisation. Ten subsamples were collected and mixed, and four samples of approximately 1 kg were collected from the mixture and analysed. The samples were frozen and freeze-dried using a JJ Científica LJI-030® freeze drier. All biomass samples collected from the RDR during the experiment were collected at the end of drum rotation to ensure proper mixing of the material. Ten subsamples were collected from each reactor from different points inside the reactor, mixed and homogenised, and one sample was then collected from the mixture and sent to the laboratory.

The following physicochemical parameters were analysed in the laboratory: pH, dry matter (DM), carbon (C), nitrogen (N), nitrogen nitrate (N-NO₃ ^-), nitrogen nitrite (N-NO₂ ^-), nitrogen and potassium (K). pH was determined by potentiometry. DM was determined by heating to 105°C for 18 hours. K was determined according to the Association of Official Analytical Chemists (AOAC, 1995). C and N concentrations were determined using a Thermo-Scientific^TM Flash 2000 CHNS/O elemental analyser. O N-NO₃ ^-and N-NO₂ ^- were determined according to the official procedure APHA 4500- as described by the American Public Health Association (APHA, 2012). All analyses were performed in duplicate, and only errors lower than 5% were accepted.

Biomass temperature during composting was determined by placing four iButton thermometers in each reactor, which were mixed with the biomass, and the thermometers recorded the temperature every 30 min. Because it was not possible to determine the position of the iButtons inside the RDRs, daily temperatures were estimated as the average maximum temperatures recorded daily by the iButtons.

RDR outlet air was collected continuously from the air outlet tube at intervals of approximately 5 minutes for each treatment. External air samples were collected at a single point of the external environment (inlet air). The following gases were evaluated in the inlet and outlet air: CO₂, CH₄, NH₃ and N₂O. Automated gas monitoring was performed using a multipoint sampler INNOVA 1309 coupled to a photoacoustic analyser INNOVA 1412 (Lumasense Technologies, Denmark).

Gas emissions during composting were determined as the difference between inlet and outlet gas concentration (Angnes et al., 2013) using the following Equation 1:

E i m = Q × ( C i , o u t m − C i , i n m ) (1)

E i m = ( m g . m − 3 )

C i , o u t m ( m g . m − 3 )

C i , i n m ( m g . m − 3 )

Biomass composition and mean daily maximum temperature were analysed using repeated measures analysis, considering the effects of block (pig and poultry carcasses), treatment, composting time, interaction between treatments, and 16 types of variance and covariance matrix structures, using PROC MIXED in Statistical Analysis System^© (SAS, 2012). The structure used for analysis was selected based on the lowest Akaike Information Criterion (AIC). Restricted maximum likelihood estimation was used. When significant differences were detected by the F test, a protected t-test was used to compare the treatments (p≤0.05).

A total of 27 linear and non-linear models were fitted to the experimental data to evaluate the effect of temperature on CO₂-C and NH₃-N emission. The analyses were performed using the GENMOD and NLMIXED procedures in SAS (2008). The best models were selected based on the AIC. Coefficients of determination (R²) were calculated for all models. All models are valid for a biomass temperature range from 19 °C to 72 °C.

For estimated gas emissions using mass balance, a variance analysis of the model was performed considering block (pig and poultry carcasses) and treatment effects.

The results presented in the present study are the mean adjusted for least squares between the two replicates for each of the treatments.

Temperature curves characteristic of composting processes, presenting mesophilic, thermophilic, cooling and maturation or curing phases, were observed (Luo et al., 2014; Singh; Kalamdhad, 2013). All treatments reached the thermophilic phase (>50 °C) on day 3, indicating rapid microbial activity (Kalamdhad; Kazmi, 2009).

During composting, temperature is recommended to remain within the thermophilic range long enough to eliminate all pathogens present in the degrading material. In the present experiment, the temperature remained within the thermophilic range for 4, 8, 11 and 11 days for treatments T1, T2, T3 and T4, respectively. Therefore, higher turning frequencies resulted in less time in the thermophilic range because more frequent aeration speeds up the composting process (Rodríguez et al., 2012) and promotes heat dissipation.

The maximum biomass temperatures were 61.05 ºC (day 5), 62.71 ºC (day 6), 69.83 ºC (day 7) and 69.81 ºC (day 7) for T1, T2, T3 and T4, respectively. These values were higher than previously reported maximum temperatures of 58 °C (Kalamdhad; Kazmi, 2009) and 60 °C (Singh; Kalamdhad, 2013). This difference may have been due to the small volume of the RDRs used in the previous studies (250 L and 550 L, respectively), which promoted energy dissipation.

Comparing the temperature profile among the four treatments, the temperature in T1 was significantly different from T3 and T4 between days 7 and 14 (p<0.05) and from T2 between days 8 and 14. No significant differences in temperature were observed between T2 and T3 (p>0.05), and a significant difference between T2 and T4 was only observed on day 14. No significant differences were observed between T3 and T4 (p>0.05) during the entire experiment.

The pH behaviour was also characteristic of composting processes (Figure 2b) (Fernández et al., 2010). The pH increased to alkaline values in the beginning of composting, indicating that the acids produced in the beginning of the degradation process were quickly consumed (Jiang et al., 2015), and then decreased to values close to neutral, which is characteristic of the biological process (Fernández et al., 2010). Also, the decrease of the pH at the end of the composting can be related to the nitrification process, caused by a release of H⁺ during this related process and the low capacity of buffering of the biomass (Cáceres et al., 2016; Cáceres et al., 2018).

Figure 2:

The pH of the mix was initially 5.7, 5.82, 5.66 and 5.66 for T1, T2, T3 and T4, respectively, reaching maximum values of 8.45 (T1), 8.36 (T2), 8.32 (T3) and 8.30 (T4) followed by final values of 7.15 (T1), 7.25 (T2), 6.91 (T3) and 7.11 (T4) at the end of the composting process. Significant differences between treatments were only observed between T1 and the remaining treatments on day 9 (p<0.05).

CO₂-C emissions during composting results from biochemical reactions by the anaerobic microbiota. T1 presented higher daily CO₂-C emissions than the remaining treatments in the beginning of the composting process, reaching values higher than 17500 mg.kg_carcass ^-1.d^-1 (Figure 3a). This result may be explained by the higher turning frequency, which promoted microbial activity and the resulting high gas production. For the remaining treatments, CO₂-C emissions also peaked in the beginning of the composting process (T2, 9881.06 mg.kg_carcass ^-1.d^-1; T3, 13033.96 mg.kg_carcass ^-1.d^-1; and T4, 12023.51 mg.kg_carcass ^-1.d^-1) and then decreased until the end of the experiment. This result was in agreement with previous reports that CO₂ emissions are highest during the first stages of composting (Ahn et al., 2011; Arriaga et al., 2017; El Kader et al., 2007; Wang et al., 2014) due to high temperatures, which results in higher emissions during the thermophilic phase (Wang et al., 2014).

Figure 3:

Total CO₂-C emissions were not significantly affected by the treatments (0.0968, 0.0847, 0.0853 and 0.1005 kg.kg_carcass ^-1 for T1, T2, T3 and T4, respectively). Although T1 presented the highest daily emissions in the beginning of the composting process, the remaining treatments presented longer thermophilic phases, resulting in similar total losses for all treatments.

Comparing the treatments throughout the experiment, aeration frequency significantly affected CO₂-C emissions, resulting in significant differences between T1 and the remaining treatments on day 5 (p<0.05). On days 6, 9, 11, 12 and 13, significant differences were only observed between T1 and treatments T2 and T4. Significant differences in CO₂-C emissions between T2 and T4 were only observed on day 5. Zeng et al. (2018) observed no significant differences (p>0.05) in CO₂-C and CH₄-C emissions among treatments with different aeration intervals, which may have been due to the fact that the aeration intervals tested by these authors were shorter (10, 30 and 50 minutes as well as a treatment with continuous aeration) than in the present study.

Although CH₄ is characteristic of anaerobic processes, it can also be generated during composting due to formation of small anaerobic zones, resulting from high O₂ demand during biodegradation (Hazarika et al., 2017). The highest CH₄-C emissions for T1 (215.08 mg.kg_carcass ^-1.d^-1), T2 (91.58 mg.kg_carcass ^-1.d^-1) and T3 (143.97 mg.kg_carcass ^-1.d^-1) was observed on day 4, and the highest CH₄-C emissions for T4 (110.44 mg.kg_carcass ^-1.d^-1) was observed on day 5. Significant differences in daily CH₄-C emissions (p<0.05) between T1 and the remaining treatments were only observed on day 6.

The highest CH₄-C emissions were observed during the thermophilic phase when microbial activity was higher, resulting in higher oxygen consumption, thereby promoting the creation of anaerobic zones. This phenomenon was also observed in a previous study (Chowdhury; Neergaard; Jensen, 2014b).

Although T1 was the treatment with highest number of rotations, it presented the highest total CH₄-C emissions (1270 mg.kg_carcass ^-1) because aeration promotes CH₄-C release from the biomass (Zeng et al., 2018). The total CH₄-C emissions for the remaining treatments were 732.7 mg.kg_carcass ^-1 for T2, 739.84 mg.kg_carcass ^-1 for T3, and 720.64 mg.kg_carcass ^-1 for T4, but they were not significantly different from T1.

From the total C mass measured as gas emissions (CO₂-C + CH₄-C), 98.7% (T1), 99.1% (T2), 99.1% (T3) and 99.2% (T4) resulted from CO₂-C emissions. These levels may be considered adequate because they indicate CH₄-C emissions of 1-4% of the initial C (Arriaga et al., 2017).

The N present in the organic matter mainly originates from proteins, cellular compounds and DNA, and these are biodegraded during the first days of composting and transformed into NH₄ ⁺, which may be converted to and volatilised as NH₃ (Cáceres; Malinska; Marfà, 2018) or converted to nitrite and nitrate (Wang et al., 2018).

NH₃-N emissions peaked on day 5 (Figure 4a), and it was highest at T1 (2049.62 mg.kg_carcass ^-1.d^-1). Total N emissions was also highest at T1 (8493.86 mg.kg_carcass ^-1), which may be explained by the higher turning frequency, promoting NH₃ volatilisation. Similarly, a previous study has shown that NH₃ loss is also highest for the compost pile with the highest number of turnings (Parkinson et al., 2004). Another study has also observed higher NH₃ emissions for a treatment with aeration compared to a treatment without aeration (Jiang et al., 2015). The high NH₃ emissions observed for all treatments during the first days of composting may be attributed to the high NH₄ ⁺ availability, temperature and pH (Angnes et al., 2013; Chowdhury; Neergaard; Jensen, 2014b; Luo et al., 2014; Jiang et al., 2015), especially during the thermophilic phase because NH₃ is not soluble in water at high temperatures (Wang et al., 2018).

Figure 4:

Although nitrogen losses due to N₂O-N emissions were lower, special attention should be given to this gas because it can potentially contribute considerably to the greenhouse effect. N₂O is mainly produced during nitrification/denitrification, originating from intermediate reactions. The highest N₂O emissions were observed on day 5 for T1, T2 and T4, and the highest N₂O emissions were observed on day 4 for T3 (Figure 4b). T1 presented the highest emissions (17.23 mg.kg_carcass ^-1.d^-1). Maximum emissions for the remaining treatments were 9.45, 11.3 and 11.36 mg.kg_carcass ^-1.d^-1 for T2, T3 and T4, respectively. Similarly to NH₃-N volatilisation, N₂O-N emissions practically zeroed after day 15. NO₃ ^- was not observed in the biomass during the experiment, indicating that nitrification was the main pathway of N₂O emission. This result has been previously observed (Chowdhury; Neergaard; Jensen, 2014a; Chowdhury; Neergaard; Jensen, 2014b; Zhu-Barker et al., 2017) because turning/aeration promotes nitrification and consequently N₂O emissions (Arriaga et al., 2017).

Significant differences in NH₃ emissions between T1 and the remaining treatments were only observed on day 6 (T2, T3 and T4; p<0.05), and significant differences in N₂O emissions between T1 and the remaining treatments were only observed on day 10 (T2 and T4; p<0.05). No significant differences were observed among T2, T3 and T4 (p>0.05). Zeng et al. (2018) tested different aeration intervals and observed no significant differences in N₂O and NH₃ emissions among treatments (p>0.05).

Because temperature is a parameter that is easy to control during composting, mathematical models were built to estimate gas emissions relative to the temperature of the material during biodegradation. Models were built for CO₂-C and NH₃-N emissions because these were the main gas losses of C and N, respectively, measured in the present study.

CO₂-C emissions relative to temperature was better fitted by a non-linear model, showing the same partitioning temperature for all treatments and different coefficients for each treatment (Table 2). The partitioning temperature was estimated at 31.27 °C. At this temperature, the model predicted constant CO₂-C emissions of 299.4 mg.kg_carcass ^-1. Above this temperature, CO₂-C emissions may be estimated using a specific linear equation for each treatment.

Table 2:

Gas	AIC	R2	Proposed Model
CO2-C	2455.25	0.75	y ^ C O 2 − C = { 299.4 ; i f   t ≤ 31.27 430.98. t − 13178.2 ; i f   t > 31.27   a n d   T r e a t = T 1 325.98. t − 9894.61 ; i f   t > 31.27   a n d   T r e a t = T 2 309.30. t − 9372.99 ; i f   t > 31.27   a n d   T r e a t = T 2 296.97. t − 8987.41 ; i f   t > 31.27   a n d   T r e a t = T 4
NH3-N	1757.1	0.78	y ^ N H 3 − N = { 48.03 ; i f   t ≤ 28.79 5.09. t − 125,40 ; i f 28 .79<t ≤ 50 .93 200.05. t − 10055.2 ; i f   t > 50.93   a n d   T r e a t = T 1 50.01. t − 2413.33 ; i f   t > 50.93   a n d   T r e a t = T 2 30.04. t − 1396.16 ; i f   t > 50.93   a n d   T r e a t = T 3 29.79. t − 1384.56 ; i f   t > 50.93   a n d   T r e a t = T 4

y ^ C O 2 − C : Estimated CO₂-C emissions (mg.kg_carcass ^-1.d^-1);

y ^ N H 3 − N : Estimated NH₃-N emissions (mg.kg_carcass ^-1.d^-1); t: biomass temperature (°C); Treat: Treatment.

NH₃-N emissions were also best fitted by a non-linear model with two partitioning temperatures (28.79 and 50.93 °C). NH₃-N emissions were constant under 28.79 °C (48.03 mg.kg_carcass ^-1). At temperatures greater than 50.93 °C, there was a specific model for each treatment. Between these two temperatures, the emissions were best fitted by the same linear model for all treatments.

The behaviour of the CO₂-C and NH₃-N gas emissions models is presented in Figure 5. As previously discussed, T1 presented the most pronounced differences in gas emissions compared to the other treatments during the experiments. In the presented models, T1 stood out from the remaining treatments due to its higher slope.

Figure 5:

Carbone (C) mass balance was determined to validate the composting process and determine the experimental error in gas emissions measurements. C mass balance was determined by calculating C gas emissions (GE) and C mass loss (ML). ML is the difference between the C biomass at the beginning and at the end of the composting process, and GE is the C mass calculated as the sum of CH₄-C and CO₂-C. Although N gas emissions (NH₃-N and N₂O-N) were also evaluated, N mass balance was not determined because N₂ emissions, resulting from denitrification and responsible for a part of the N loss, was not determined, which would impair the determination of errors using the mass balance.

The error between ML and GE was highest for T2 (25.82%) and lowest for T3 and T4 (≤2.5%) (Table 3). The highest error found was similar to that found for all treatments in a composting study by Arriaga et al. (2017), which was explained by heterogeneity between subsamples, a problem that was also observed in the present study. The overall mean ML for all treatments was 0.091 kg.kg_carcass ^-1, and the mean GE was 0.097 kg.kg_carcass ^-1 with a mean error of 14.09%. This error can be considered adequate and may due to possible experimental errors, especially to likely variations in the speed of air traversing the reactor caused by different distributions of the material inside the reactors following each turning, thereby resulting in ML variations throughout the day. Linear regression between ML and GE resulted in the following equation: GE = 1.059 × ML (R² = 0.77).

Table 3:

Treatment	ML (kg.kgcarcass -1)	GE (kg.kgcarcass -1)	C Error (%)	K Error (%)
1	0.084	0.098	19.76	19.62
2	0.081	0.085	25.82	14.48
3	0.099	0.86	2.49	18.54
4	0.099	0.101	2.50	12.28

ML: difference between C biomass at the beginning and at the end of the composting process; GE: C mass (CH₄-C + CO₂-C) measured by gas emission; C Error: (GE-ML)/GE; K Error: difference (in percentage) between K mass at the beginning and at the end of the composting process.

Although mass balance for non-volatile elements has not been determined in some composting studies (Arriaga et al., 2017; Mulbry; Ahn, 2014; Zhu-Barker et al., 2017), this parameter is important to check for possible sampling and subsampling errors in the collection of material for physicochemical analyses. In the present study, K mass balance was calculated and found to be lower than 20% for all treatments, which was considered satisfactory due to the heterogeneity of the composting material, especially due to the difficulty in characterising poultry and pig carcasses. A mean error of 14.48% was observed for K, a non-volatile element, for all treatments. This value was similar to those found in other studies of gas emissions during composting (Angnes et al., 2013; El Kader et al., 2007).