Bioresource Technology 117 (2012) 107–116

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Evolution of process parameters and determination of kinetics for co-compostingof organic fraction of municipal solid waste with poultry manure

Ivan Petric ⇑, Azra Helic, Edisa Avdihodzic AvdicDepartment of Process Engineering, Faculty of Technology, University of Tuzla, Univerzitetska 8, 75000 Tuzla, Bosnia and Herzegovina

h i g h l i g h t s

" Seven process parameters were monitored in co-composting of OFMSW and poultry manure." The mixture of 60% OFMSW/20% manure/10% compost/10% sawdust gave the highest OM degradation." Nine kinetic models were analyzed with four statistical indicators." Satisfactory fitting of proposed kinetic model to the experimental data of OM was achieved." The number of measured variables influences kinetics more than the number of kinetic parameters.

a r t i c l e i n f o

Article history:Received 22 February 2012Received in revised form 11 April 2012Accepted 13 April 2012Available online 26 April 2012

Keywords:CompostingKinetic modelMunicipal solid wastePoultry manureNumerical methods

0960-8524/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.biortech.2012.04.046

⇑ Corresponding author. Tel.: +387 35 320 766; faxE-mail address: ivan.petric@untz.ba (I. Petric).

a b s t r a c t

This study aimed to monitor the process parameters and to determine kinetics in composting of organicfraction of municipal solid waste (OFMSW) and poultry manure. The experiments were carried out withthree different mixtures. The results showed that the mixture 60% OFMSW, 20% poultry manure, 10%mature compost and 10% sawdust provided the most appropriate conditions for composting process.Using nine kinetic models and nonlinear regression method, kinetic parameters were estimated andthe models were analyzed with four statistical indicators. Kinetic models with four measured variablesproved to be better than models with less number of measured variables. The number of measured exper-imental variables influences kinetics more than the number of kinetic parameters. Satisfactory fittings ofproposed kinetic model to the experimental data of OM were achieved. The model is more suitable fordata obtained from composting of mixtures with much higher percentage of OFMSW than percentageof poultry manure.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The land scarcity problem for waste disposal in most of the ur-ban places prompted the environmentalists searching for anappropriate processing, recycling, and disposal of municipal solidwaste (MSW) as an important and integral part of a solid manage-ment system. Composting has gained an important role in munici-pal solid waste management. Composting is an effective and safeway for reduction of the manure’s mass and volume, for destruc-tion of pathogens and stabilization of nutrients and organic matterin it (Tiquia et al., 2000). Co-composting means composting of sev-eral types of residual matters altogether such as: olive mill waste-water with solid organic wastes (Paredes et al., 2000), roseprocessing waste with organic fraction of municipal solid waste(Tosun et al., 2008), poultry manure and wheat straw (Petric andSelimbašic, 2008), municipal solid wastes and sewage sludge

ll rights reserved.

: +387 35 320 741.

(Fourti et al., 2010), physic nut deoiled cake with rice straw anddifferent animal dung (Das et al., 2011), sewage sludge, barksand green waste (Watteau and Villemin, 2011).

To find out biodegradation behavior of wastes is important foran optimized design regarding composting process parameterssuch as processing time, size of reactor or pile area and the productquality. A high degradation rate usually indicates lower capital andoperational costs for composting plants (Tosun et al., 2008). Incomposting, there are different process management options,but, for the majority, maximizing the decomposition rate is oneof the main objectives (Baptista et al., 2010).

Kinetic model can be used as model to study composting pro-cess on an industrial scale for the optimization of the process.The degradation rate of waste can be predicted using kinetic mod-els of the process indicators (temperature, organic matter content,moisture content, oxygen concentration, pH, C/N ratio, particlesize, etc.). Kinetics of the process is used to determine waste biode-gradability and generate a useful measure for the loss of organicmatter during composting. It is determined by using reliable data

108 I. Petric et al. / Bioresource Technology 117 (2012) 107–116

obtained by experimental studies under controlled conditions.Modeling of composting kinetics is necessary to design and operatecomposting facilities that comply with strict market demands andtight environmental legislation (Hamelers, 2004).

Composting kinetics was the main objective in the studies ofmany researchers. Review of the literature showed that first-orderequation is the most common form of description of the compost-ing process (Baptista, 2009). Many researchers found that the or-ganic matter degradation follows first-order kinetics in thecomposting of different wastes (Haug, 1993; Keener et al., 1993;Hamoda et al., 1998; Bari et al., 2000; Paredes et al., 2000; Külcuand Yaldiz, 2004; Baptista et al., 2010).

Külcu and Yaldiz (2004) have determined kinetics in aerobiccomposting of agricultural waste, where the compost materialswere prepared by mixing grass trimmings, tomato, pepper andeggplant wastes. The seven different first-order kinetic modelswere applied to the experimental values for modeling of decompo-sition rate. While process-monitoring parameters (CO2, tempera-ture, pH, moisture content) were used for interpretation of theprocess, degradation of organic matter was used for interpretationof the process success. The suitability of the models was comparedand evaluated using chi-square, root mean square error and mod-eling efficiency.

Hamoda et al. (1998) evaluated the reaction rate constant forfirst-order kinetics of composting of municipal solid waste withMSW compost, without expressing the reaction rate constant asa function of any experimental data. Basnayake et al. (2007) ap-plied Monod model and Michaelis–Menten kinetic analysis oncomposting of separated municipal solid waste (food waste, coco-nut fiber, yard waste, fish and meat waste and mixed waste). Ku-mar et al. (2009) performed simplified kinetic analysis forcomposting of municipal solid waste using Michaelis–Mentenequation. Baptista et al. (2010) used simulation model developedby Haug (1993) (with correction factors for temperature, oxygen,moisture and free air space) in order to test the application ofthe model for the description of the composting process in full-scale mechanical–biological treatment plants.

According to our knowledge and based on an extensive reviewof the literature, there are no studies dealing with first-order ki-netic models for co-composting of organic fraction of municipal so-lid waste with poultry manure, directly based on organic matterdegradation and experimental data like pH or electrical conductiv-ity. Therefore, the objectives of this study are the following: (1) toprovide experimental data (organic matter content, temperature,moisture content, carbon dioxide concentration, oxygen concen-tration, aeration rate, pH and electrical conductivity) for determi-nation of kinetics for co-composting of organic fraction ofmunicipal solid waste with poultry manure, (2) to estimate kineticparameters in six literature and three proposed kinetic modelsusing several statistic indicators, (3) to perform fitting of the bestkinetic model to the experimental data.

2. Methods

2.1. Reactor system

A laboratory-scale composting system as shown in Fig. 1 wasused for this study. The experiment was lasted over the period of22 days using three identical specially designed reactors made ofstainless steel (volume 35 l, height 0.55 m, internal diameter0.36 m). The reactors were insulated with a layer of polyethylenefoam (10 mm of thickness). A vertical rotating axis with bladesmixing on intermittent schedule, fixed at perforated plate madeof stainless steel (with holes of 5 mm), ensures the complete mix-ing of compost mass. Mixing was performed once a day. The reac-

tors were equipped with a valve for dropping the leachate andcondensate. On the reactor lid, there were two holes, for the shaftmixer and for the thermocouple. Two stainless steel tubes (tube forgas sampling with valve and tube for discharge of exhausted gases)were welded to the reactor lid.

Each reactor was connected with an air compressor (Trudbenik,Bosnia and Herzegovina), which provided air into the reactors at acontrolled rate (0.9 l air min�1 kg�1

OM) based on published recom-mendations (Külcu and Yaldiz, 2004; Petric and Selimbašic,2008). Measurement of airflow was carried out using airflow me-ters (Valved Acrylic Flowmeter, Cole-Parmer, USA).

Thermocouple was inserted through a drilled rubber stopper,which is then inserted through a hole in the reactor lid. In all reac-tors, temperature was measured through thermocouples type T(Digi-Sense, Cole-Parmer, USA), placed in the middle of substrate.Thermocouples were connected through the acquisition moduleTemperature Data Acquisition Card Thermocouple CardAcq(Nomadics, USA) on a laptop.

At reactor outlet, the gas mixture passed through a gas washingbottle with 1 M sodium hydroxide and a gas washing bottle with0.65 M boric acid, in order to remove carbon dioxide and ammonia,respectively. The gas washing bottles were changed daily.

2.2. Materials

Organic fraction of municipal solid waste (OFMSW), poultrymanure, mature compost and sawdust were used as experimentalmaterials. Basic physical and chemical characteristics of thesecomponents are shown in Table 1.

The preparation of materials for composting was done in thelaboratory. Part of prepared waste was separated for analysis whilethe remaining waste mixed with the rest of the material forcomposting.

The organic fraction of municipal solid waste was simulatedusing the following components and mass percentage: food waste(56.5%), paper and cardboard (30.4%), and yard waste (13.1%). Foodwaste was collected from several restaurants and main city marketin Tuzla, Bosnia and Herzegovina. This type of waste was com-posed of fruit (plums, watermelons, apples + peels, grapes, pears,lemons, oranges + peels, peaches, etc.), vegetables (cabbages, pota-toes + peels, onions, eggplants, peppers, tomatoes, etc.), pie crustsand pasta. Paper used in the experiment, was consisted of officepaper and old newspapers (chopped off on the size 3–4 cm). Card-board was prepared by shredding old cardboard boxes to the sizeof 3–4 cm. Yard waste was composed of chopped grass, leavesand branches and it was collected from city parks and house gar-dens. The components of OFMSW were manually mixed in plasticboxes to achieve better homogenization of material.

The role of poultry manure was to adjust the ratio of carbon/nitrogen. Manure was collected in a large polyethylene bags fromthe poultry farm near Gracanica. After bringing it to the laboratory,part of the manure was separated for analysis while the rest wasused to prepare the mixture with organic fraction of municipal so-lid waste.

Sawdust was used as a bulking agent, in order to increase aera-tion of composting mixtures and to optimize substrate properties(moisture, porosity, C/N ratio, pH). Sawdust was collected in largepolyethylene bags from sawmills near Lukavac and Tuzla.

Mature compost was added as inoculum in order to acceleratethe start of the composting process. The compost was transportedto the laboratory in polyethylene bags from the private compostingplace near Lipnica.

Composting materials were manually mixed in plastic boxes, inorder to achieve better homogenization of mixtures. After mixing,reactors were filled with composting materials. The masses of

Fig. 1. Schematic diagram of reactor system for aerobic composting (1 – air compressor, 2 – airflow meter, 3 – reactor, 4 – thermocouple, 5 – laptop, 6 – gas-washing bottlewith solution of sodium hydroxide, 7 – gas-washing bottle with solution of boric acid, 8 – Gas Chromatograph).

Table 1Basic physical and chemical characteristics of organic fraction municipal solid waste, poultry manure, mature compost and sawdust before mixing (three replicates, meanvalue ± standard deviation).

Material TS (% w.b.) MC (% w.b.) OM (% d.b.) Ash (% d.b.) pH EC (dS m�1) C (%) N(%) C/N

OFMSW 40.17 ± 0.50 59.83 ± 0.50 91.69 ± 0.37 8.31 ± 0.37 4.98 ± 0.05 1.19 ± 0.07 50.94 ± 0.37 0.66 ± 0.03 77.18Poultry manure 28.97 ± 0.92 71.03 ± 0.92 78.89 ± 1.02 21.11 ± 1.02 8.31 ± 0.08 3.77 ± 0.06 43.83 ± 1.02 5.02 ± 0.02 8.73Compost 67.49 ± 0.67 32.51 ± 0.67 40.31 ± 0.88 59.69 ± 0.88 6.92 ± 0.07 0.35 ± 0.06 22.39 ± 0.88 1.21 ± 0.05 18.50Sawdust 89.97 ± 0.40 10.03 ± 0.40 99.9 ± 0.17 0.10 ± 0.17 5.31 ± 0.05 0.24 ± 0.04 55.50 ± 0.17 0.28 ± 0.04 198.21Mixture 1 54.44 ± 1.82 45.56 ± 1.82 87.53 ± 0.02 12.47 ± 0.02 5.17 ± 0.19 1.50 ± 0.07 48.63 ± 0.02 1.21 ± 0.32 40.19Mixture 2 41.83 ± 1.72 58.17 ± 1.72 85.61 ± 2.69 14.39 ± 2.69 6.27 ± 0.19 2.17 ± 0.36 47.56 ± 2.69 1.30 ± 0.35 36.58Mixture 3 39.62 ± 4.01 60.38 ± 4.01 83.89 ± 2.71 16.11 ± 2.71 6.37 ± 0.29 2.62 ± 0.38 46.61 ± 2.71 1.38 ± 0.27 33.77

TS – total solids; MC – moisture content; OM – organic matter content; EC – electrical conductivity; w.b. – wet basis; d.b. – dry basis.

I. Petric et al. / Bioresource Technology 117 (2012) 107–116 109

composting mixtures in the reactors 1, 2 and 3 were 7.90, 10.80and 12.65 kg, respectively.

Three mixtures for composting reactors were prepared with dif-ferent percentages of materials: reactor 1 (60% OFMSW, 20% poul-try manure, 10% mature compost, 10% sawdust), reactor 2 (50%OFMSW, 33.33% poultry manure, 8.33% mature compost, 8.33%sawdust), reactor 3 (42.86% OFMSW, 42.86% poultry manure,7.14% mature compost, 7.14% sawdust).

Basic physical and chemical characteristics of composting mix-tures in reactors are shown in Table 1.

2.3. Methods

2.3.1. Analysis and samplingThe same procedures were used for characterization of mix-

tures as well as for characterization of OFMSW, poultry manure,compost and sawdust.

The moisture content and total solids were analyzed by dryoven method at 105 �C for 24 h (APHA, 1995). The organic matter(OM) content (volatile solids) and ash content were measured byburning oven at 550 �C for 6 h (APHA, 1995). The loss of OM wascalculated from the initial and final organic matter contents,according to the following equation (Haug, 1993; Külcu and Yaldiz,2004):

k ¼½OMm ð%Þ � OMp ð%Þ� � 100� �

OMm ð%Þ � ½100� OMp ð%Þ�� 100 ð1Þ

where OMm is the OM content at the beginning of the process(mas.%) and OMp is the OM content at the end of the process.

Kjeldahl nitrogen determination was performed according tothe standard procedure (APHA, 1995). The carbon content (%C)

was calculated from the ash fraction (%ash), according to the fol-lowing aquation (Haug, 1993):

%C ¼ ð100�%ashÞ � 1001:8

ð2Þ

Electrical conductivity and pH were measured in aqueous ex-tract. This aqueous extract was obtained by mechanically shakingthe samples with distilled water at a solid to water ratio of 1:10(w/v) for 1 h. The suspension was centrifuged and filtered througha Whatman no. 42 filter paper. The pH and electrical conductivitymeasurements were carried out using a PC 510 Bench pH/Conduc-tivity Meter (Oakton, Malaysia) with two separated electrodes.

Concentrations of oxygen and carbon dioxide were measured byClarus500 Gas Chromatograph (Perkin–Elmer Arnel, India),equipped with a thermal conductivity detector (TCD-R, range 2,time constant 200, autozero ON, polarity POS, filament voltageON), light gas analyzer Model 4016 and Arnel TotalChrom Work-station software. Helium was used as a carrier gas (PFlow-He initialsetpoint 34.00 ml/min, auxiliary pneumatics, flow – He setpoint40 ml/min). The gas mixture consisting of CO, CO2, CH4 and O2

(MESSER, Germany) was used for calibration of the device. GasChromatograph has two packed columns: 70 HayeSep N 60/80, 1/800 SF (maximum temperature 150 �C) and 90 Molecular Sieve13 � 45/60, 1/800 SF (maximum temperature 375 �C). Conditionsfor GC oven ramp are: initial temperature 60 �C, initial hold12 min, total run time 12 min, maximum temperature 150 �C.The samples of output gas from reactors were taken using a specialcylinder volume of 500 ml (RESTEK, United Kingdom), equippedwith two hoses connected with cylinder outputs (regulated withvalves) at top and bottom of the cylinder. The sampling was per-formed in the following way: first, the valve on the tube at reactorlid was opened. Secondly, the valve at the cylinder bottom was

Table 2Applied kinetic models in the study.

Model no. Kinetic model References

1 kT ¼ k23 � aðT�23Þ Haug (1993) modified2 kT ¼ a � expf½ðb � TÞ þ ðc � Mc

T Þ�g Külcu and Yaldiz (2004)

3 kT ¼ a � expfb � ½ðMi�cd Þ þ ð

T�fg Þ�g Ekinci et al. (2001)

4 kT ¼a

McT�ðC�bÞ � expf½ðT � cÞ � ðd � Mc

T Þ�gKülcu and Yaldiz (2004)

5 kT ¼ a � bC � expf½ðc � TÞ � ðd � McT Þ�g Külcu and Yaldiz (2004)

6 kT ¼ eðVh�aÞ � k23 � bðT�23Þ Külcu and Yaldiz (2004)

7 kT ¼ a � bO2 � exp½ðc � TÞ þ ðd � McT Þ� This study

8 kT ¼ a � exp½ðb � TÞ þ ðc � O2T Þ � ðd � EC

T Þ� This study

9 kT ¼ Oð1�aÞ2 � bðT�23Þ � pHc � ðMc

T Þd This study

kT – the reaction rate constant (day�1); T – the process temperature (�C); Mc – thedaily moisture content (% w.b.); C – the daily concentration of CO2 in compostingreactor (%); a, b, c, d, f, g – are constants; Vh – aeration rate (l air min�1 kg�1

OM); EC –electrical conductivity (dS m�1); O2 – oxygen concentration (%).

110 I. Petric et al. / Bioresource Technology 117 (2012) 107–116

opened and the valve at the cylinder top was closed. Thirdly, thecylinder hose at the bottom was connected with the tube (the tubewas welded at reactor lid), where the cylinder was filled with agaseous mixture from the reactor. Fourthly, when the cylinderwas completely filled with gaseous mixture, the valve at the cylin-der bottom was closed and the hose cylinder was disconnectedfrom the tube at reactor lid (with simultaneous closing of the valveon the tube welded at reactor lid). Between each two analyzes, thecolumn was first flushed with helium and then with a gaseoussample.

The composting material was mixed once a day. After mixing,samples (about 25 g) were taken from different places in the sub-strate (top, middle, and bottom). The analysis of the fresh sampleswas performed immediately after taking them out of the reactors.

2.3.2. Statistical analysis of experimental dataStatistical analysis (ANOVA analysis, and the least significant

difference for mean at 95%) was performed with statistical packageSTATGRAPHICS Plus 5.1 (Statistical Graphics Corporation, 1996), ondata obtained in the composting mixture at the different compost-ing times.

2.3.3. Kinetic models and statistical indicatorsDetermination of kinetics was based on experimental data ob-

tained under controlled laboratory conditions. For description oforganic matter degradation as a function of time, the first-orderkinetics was used (Haug, 1993; Hamoda et al., 1998; Külcu andYaldiz, 2004):

dðOMÞdt

¼ �kT � OM ð3Þ

where OM is the mass of biodegradable volatile solids at any time ofthe composting process in kilograms, t is time in days, kT is the reac-tion rate constant in days�1.

For modeling the decomposition rate, experimental data wereapplied to six different literature kinetic models (Haug, 1993; Ekin-ci et al., 2001; Külcu and Yaldiz, 2004), and to three new models(Table 2).

Using the nonlinear regression method with the Levenberg–Marquardt algorithm (Press et al., 1992), kinetic parameters wereestimated from experimental data obtained from the process inreactors. The suitability of the models was compared and evalu-ated using correlation coefficient (R2), adjusted correlation coeffi-cient (R2

adj), root mean square deviation (Rmsd), variance (s2) and95% confidence interval:

R2 ¼ 1�Pn

i¼1ðki;exp � ki;preÞPni¼1ðki;exp � �kexpÞ2

2

ð4Þ

R2adj ¼ 1� ð1� R2Þ � ðn� 1Þ

n� pð5Þ

Rmsd ¼1n�Xn

i¼1ðki;exp � ki;preÞ2

h i12 ð6Þ

s2 ¼Pn

i¼1 ki;exp � �kexp� �2

n� 1ð7Þ

where ki,exp is the experimental decomposition rate (day�1), ki,pre isthe predicted decomposition rate (day�1), n is the number of obser-vations, p is the number of parameters in the model, �kexp is the meanexperimental decomposition rate (day�1).

The correlation coefficients are frequently used to judgewhether the model represents correctly the data, implying that ifthe correlation coefficient is close to one then the regression modelis correct. There are, however, many examples where the correla-tion coefficient is close enough to one but the model is still notappropriate.

Just like the correlation coefficients, variance and root meansquare deviation are recommended to be used for comparing var-ious models representing the same dependent variable. A modelwith smaller variance and root mean square deviation representsthe data more accurately than a model with larger values of theseindicators.

For the regression model to be stable and statistically valid, theconfidence intervals must be much smaller (or at least smaller)than the respective parameter values (in absolute values). Anunstable model may yield very inaccurate derivative values andabsurd results for even a small range of extrapolation. For anunstable model, a small change in the data (by adding or removinga data point, for example) may lead to large changes of the param-eter values.

For determining the values of kinetic parameters, numericalsoftware package Polymath 6.0 (Shacham et al., 2004) was used.

3. Results and discussion

3.1. Changes of OM, T, MC, CO2, O2, pH and EC

Organic matter is mineralized after composting, mostly due tothe degradation of easily degradable compounds, which are uti-lized by microorganisms as carbon and nitrogen sources. Whiledegrading organic compounds, microorganisms convert 60–70%carbon to carbon dioxide and utilize remaining 30–40% into theirbodies as cellular components (Barrington et al., 2002). The rateof organic matter loss is an indicator of the overall composting rate.Watteau and Villemin (2011) used microstructure characterization(transmission electron microscopy, TEM) to monitor organic mat-ter dynamics during co-composting process of sewage sludge,green waste and barks. This method can be used in this study in or-der to specify the nature and level of biodegradation of heteroge-neous residues (both as raw matters and during the compostingprocess). TEM can show which compounds were not biodegradedduring composting. The organic matter content decreased in allreactors during composting (Fig. 2(a)) and this decrease was morepronounced during the first stages of the process. Organic matterdegradation in reactor 1 was faster than degradation in other reac-tors. The smallest loss of organic matter was achieved in reactor 3.After 22 days of composting, organic matter losses were 70.81%,63.03% and 49.75% for reactors 1, 2 and 3, respectively. These re-sults can be explained by higher carbon content (that is C/N ratio)

Fig. 2. Changes of the process parameters during composting process: (a) organic matter; (b) temperature; (c) moisture content; (d) CO2 concentration; (e) O2 concentration;(f) pH; (g) EC (note: three replicates for OM, MC, pH and EC, standard deviations are shown).

I. Petric et al. / Bioresource Technology 117 (2012) 107–116 111

in reactor 1, which provided a favorable condition for the growthand biological activity of microorganisms. There were statisticallysignificant differences between organic matter content for all reac-tors (P < 0.05).

Temperature profiles of the reactors are shown in Fig. 2(b). Afterthe initial filling of the reactors, a rapid increase in temperaturewas produced in all reactors, indicating a marked microbial activ-ity. The lag phase on temperature curve was not recorded because

112 I. Petric et al. / Bioresource Technology 117 (2012) 107–116

the original substrate was rich in microorganisms. The changes incompost temperature followed a pattern similar to a typical com-posting process. Initially, the temperature of composting mixturesrose as a consequence of the rapid breakdown of the readily avail-able organic matter and nitrogenous compounds by microorgan-isms (thermophilic phase). As the organic matter became morestabilized, microbial activity, the organic matter decompositionrate, and the temperature all decreased gradually to ambient lev-els. At the end of the process, when there is a smaller amount ofeasily degradable organic matter (reactor 3), the composting mix-ture cools down and the temperature of the decomposing materialbecomes comparable to the ambient temperature. This is not thecase with bulk temperatures in the reactor 1 (from ninth to seven-teenth day) and reactor 2 (from nineteenth to twenty-second day).It was probably the result of the secondary fermentation in theprolonged duration induced by more available organic componentsand active microbes. Reactor 3 reached the highest temperaturewhereas reactor 1 reached the lowest temperature. Temperatureof compost mixture in reactor 3 decreased immediately afterreaching the maximum. This is probably because of the smaller ini-tial moisture content and therefore the greater free air space. Be-cause of greater pores inside the compost mixture, the constantair flow in the reactor enhanced cooling of the composting mixture.Therefore, readily degradable matter was not consumed com-pletely by thermophilic microorganisms. There were statisticallysignificant differences in temperature regime between reactors 1and 2 (P < 0.05).

Moisture content is a critical parameter in the composting pro-cess. It influences the microbial activity, free air space, oxygentransfer and temperature of the process (Haug, 1993). Evolutionof moisture content in the mixtures for all reactors is shown inFig. 2(c). Moisture content varied between 62.50% and 68.52%(reactor 1), between 58.01% and 68.21% (reactor 2) and between52.92% and 61.82% (reactor 3). According to these results, moisturecontent above 60% seems to be suitable for greater degradation oforganic matter during composting process. Moisture loss duringthe composting process could be considered as an index of decom-position rate because the heat generation that accompanieddecomposition drove vaporization (Liao et al., 1997). There werestatistically important differences between moisture content val-ues in all three reactors (P < 0.05).

Evolution rates of carbon dioxide (Fig. 2(d)) for the reactorsshowed a similar behavior with temperature variation. Therefore,CO2 concentration might be used as another indicator for measur-ing the composting process besides temperature. During the pro-cess, the maximum CO2 evolution rate occurred within a fewdays of the process, probably due to easily biodegradable sub-strates. Carbon dioxide rate increased in all reactors in proportionto microorganism activity during the process. This rate was higherin reactor 3 than in other reactors only during the first day(13.89%). The maximum concentrations in reactors 1 and 2 wereachieved after second day, 10.43% and 11.26%, respectively. Afterreaching the maximum, CO2 concentration decreased in all reac-tors due to degradation of easily degradable organic compounds.Second increase of CO2 concentrations in all reactors could be asign that there was a degradation of hardly degradable organiccompounds. Statistical analysis showed that there were significantdifferences in CO2 evolution between reactors 1 and 2 (P < 0.05).

The composting process can be defined as aerobic when theoxygen content exceeds 5% (Haug, 1993). Moreover, the oxygenconcentration needs to be at least within 10–18% to prevent a de-crease in metabolic activity based on carbon dioxide evolution.Data for oxygen concentration (Fig. 2(e)) showed that oxygen lev-els did not cause anaerobic conditions during the process. The min-imum O2 concentration (6.89%) was observed in reactor 3. It wasnoticed that the profile of consumed O2 looks like a reflection of

generated CO2 in the mirror (Fig. 2(d) and (e)). As in the case ofCO2 concentration there were significant differences in O2 evolu-tion between reactors 1 and 2 (P < 0.05).

Since the beginning of the process, the pH tended to move to-wards first neutral and then alkaline (Fig. 2(f)), when formed acidswere converted to carbon dioxide by microbial action. Commondrop in pH value was not observed in any mixture because initialmixtures were quite acidic (5.17, 6.27 and 6.37 in reactor 1, 2,and 3, respectively). The pH of the mixtures increased significantlyin reactors 1 and 3 for the first 3 days and in reactor 2 for the first5 days. The rapid increase in pH during the thermophilic phase canbe attributed to the production of ammonia associated with pro-tein degradation in the samples and to the decomposition of organ-ic acids (Liu et al., 2011), or bioconversion of organic nitrogen intofree ammonia (Altieri et al., 2011). The maximum values of pHwere 7.99 (seventh day), 8.61 (seventh day) and 8.75 (eighthday) for the reactors 1, 2 and 3, respectively. After reaching themaximum, pH slightly decreased in all three reactors. The final val-ues of pH were 7.65, 8.02 and 8.11 for the reactors 1, 2 and 3,respectively. It was observed that there were significant differ-ences in pH between reactor 1 and reactors 2 and 3 (P < 0.05).

Evolution of electrical conductivity in the mixtures for all reac-tors is shown in Fig. 2(g). The EC value reflected the degree of salin-ity in the compost, indicating its possible phytotoxic/phyto-inhibitory effects on the growth of plant if applied to soil (Huanget al., 2004). During the process, the electrical conductivity de-creased in all three reactors (from 1.501 to 0.774 dS m�1 in reactor1, from 2.169 to 1.398 dS m�1 in reactor 2 and from 2.627 to1.830 dS m�1 in reactor 3). The reason for decrease of EC is proba-bly due to low concentration of soluble salts. It was observed thatthere were significant differences in EC between all three reactors(P < 0.05).

According to Petiot and de Guardia (2004), the volume of labo-ratory reactor used in this study (with provided external insula-tion) should allow the self heating of the substrate and thereforeeven to simulate full-scale composting. Moreover, Fig. 2(b) showsthree of four phases in aerobic composting process: mesophilic(initial) phase, thermophilic phase and cooling phase. Maturationphase has not been shown because it was not the aim of this study(only active, hight-rate composting in a reactor). Compost maturitywas not examined in this study since the experiments were termi-nated after the compost temperature reached the ambient level.

Taking into account the changes of physico-chemical propertiesof the three mixtures, it seems that the mixture in reactor 1 (60%OFMSW, 20% poultry manure, 10% mature compost and 10% saw-dust) provided the most appropriate conditions for the activephase of composting process (mesophilic phase, thermophilicphase and cooling phase). As already mentioned, the rate of organicmatter loss is an indicator of the overall composting rate. Thegreatest loss of OM was achieved in reactor 1. The highest temper-ature is achieved in reactor 3, but with a sudden drop to ambienttemperature immediately after reaching the maximum. On theother hand, the peak temperature in reactor 1 was lower than inreactor 3, but after reaching its peak the temperature in this reac-tor was maintained for longer time. Initial moisture content wasthe highest in the reactor 1 and it seems to be suitable for greaterdegradation of organic matter during composting process compar-ing to the initial moisture contents in reactors 2 and 3. New in-creases of CO2 concentrations and decreases of O2 concentrationsin reactor 1 in the middle of the process are probably indicatorsof degradation of hardly degradable organic compounds. The finalvalues of pH (7.65) and EC (0.774 dS m�1) for mixture in reactor 1were the lowest among the three mixtures and they are in therange of the recommended values (Woods End Research Labora-tory, 2000; Watson, 2003). It seems that more poultry manure inthe initial mixture accelerated the process in the beginning but

Table 3Estimated kinetic parameters in the kinetic models with statistical analysis for reactor 1.

Model no. Kinetic models Kinetic parameters 95% Confidence R2R2

adjRmsd Variance

1 kT ¼ k23 � aðT�23Þ a = 0.0043 3 � 10�4 0.9962 0.9960 0.00296 2 � 10�4

2 kT ¼ a � exp ðb � TÞ þ c � McT

� �� �� �a = 1 � 10�5 0.0012 0.9974 0.9972 0.0024 2 � 10�4

b = �0.2118 1.3607c = 5.2777 11.0445

3 kT ¼ a � exp b � Mi�cd

� þ T�f

g

� h in oa = 0.0284 0.02071 0.9973 0.9965 0.0025 2 � 10�4

b = �14.8434 1.9942c = 39.5455 0.8456d = 17.5232 0.5181f = 54.5527 0.7913g = 16.7311 0.4323

4 kT ¼a

McT� C�bð Þ � exp ðT � cÞ � d � Mc

T

� �� �� � a = 0.0108 2 � 10�5 0.5727 0.5053 0.0314 0.0273b = �3.4750 0.0225c = 0.1465 4.995 � 10�5

d = �2.6421 0.00065 kT ¼ a � bC � exp ðc � TÞ � d � Mc

T

� �� �� � a = 2 � 10�6 4 � 10�7 0.9975 0.9972 0.0024 1 � 10�4

b = 0.7535 0.1142c = �0.0370 0.0082d = �4.7038 0.0620

6 kT ¼ eðVh�aÞ � k23 � bðT�23Þ a = �3.3005 0.0371 0.9973 0.9972 0.0025 2 � 10�4

b = 0.4118 0.05647 kT ¼ a � bO2 � exp ðc � TÞ þ d � MC

T

� h ia = 0.0384 1 � 10�7 0.9996 0.9995 0.0009 3 � 10�5

b = 0.8206 3 � 10�6

c = �0.0032 1 � 10�7

d = 1.1536 1 � 10�6

8 kT ¼ a � exp ðb � TÞ þ c � O2T

� � d � EC

T

� �n oh ia = 3.0338 9 � 10�5 0.9996 0.9995 0.0010 3 � 10�5

b = �0.0949 1 � 10�6

c = �5.4244 0.0010d = �17.9943 0.0004

9 kT ¼ Oa2 � b

ðT�23Þ � pHc � McT

� �d a = �0.4802 0.0006 0.9999 0.9998 0.0005 8 � 10�6

b = 0.9139 0.0076c = 0.0212 0.0025d = �2.9434 0.0390

Table 4Estimated kinetic parameters in the kinetic models with statistical analysis for reactor 2.

Model no. Kinetic models Kinetic parameters 95% Confidence R2R2

adjRmsd Variance

1 kT ¼ k23 � aðT�23Þ a = 0.8461 0.7057 0.0229 �0.0236 0.0485 0.05932 kT ¼ a � exp ðb � TÞ þ c � Mc

T

� �� �� �a = 1 � 10�11 2 � 10�11 0.7604 0.7364 0.0240 0.0153b = �0.0288 0.0817c = 8.8441 0.7175

3 kT ¼ a � exp b � Mi�cd

� þ T�f

g

� h in oa = 0.0298 0.0447 0.0229 �0.2645 0.0485 0.0733b = �2.6033 3.4145c = 40.9930 10.0646d = 17.4630 6.5218f = 55.2803 8.9703g = 15.5768 4.5613

4 kT ¼a

McT�ðC�bÞ � exp ðT � cÞ � ðd � Mc

T � �� � a = 0.0351 1 � 10�5 0.9969 0.9964 0.0027 2 � 10�4

b = �707.9814 3.9236c = 0.1647 1 � 10�5

d = �2.4848 0.00015 kT ¼ a � bC � exp ðc � TÞ � d � Mc

T

� �� �� � a = 5 � 10�6 3 � 10�7 0.9995 0.9995 0.0010 3 � 10�5

b = 0.0328 0.0262c = �0.3354 0.0027d = �1.5050 0.0244

6 kT ¼ eðVh�aÞ � k23 � bðT�23Þ a = �1.4635 2.1347 0.0229 �0.0236 0.0485 0.0593b = 0.8461 0.8124

7 kT ¼ a � bO2 � exp ðc � TÞ þ d � MCT

� h ia = 0.0002 0.0001 0.9997 0.9996 0.0009 2 � 10�5

b = 0.8254 0.0079c = 0.0718 0.0095d = 2.5377 0.1735

8 kT ¼ a � exp ðb � TÞ þ c � O2T

� � ðd � EC

T Þh i

a = 0.1424 5 � 10�7 0.9995 0.9994 0.0011 3 � 10�5

b = �0.0385 1 � 10�7

c = �4.8567 0.0001d = �34.4951 4 � 10�5

9 kT ¼ Oa2 � b

ðT�23Þ � pHc � ðMcT Þ

d a = �0.3039 0.0050 0.9999 0.9998 0.0005 8 � 10�6

b = 1.1102 0.0472c = �3.6132 0.9538d = 4.4450 1.9852

I. Petric et al. / Bioresource Technology 117 (2012) 107–116 113

Table 5Estimated kinetic parameters in the kinetic models with statistical analysis for reactor 3.

Model no. Kinetic models Kinetic parameters 95% confidence R2R2

adjRmsd Variance

1 kT ¼ k23 � aðT�23Þ a = 0.3280 6 � 10�4 0.8455 0.8382 0.0197 0.00982 kT ¼ a � exp ðb � TÞ þ ðc � Mc

T � �� �

a = 2 � 10�7 2 � 10�9 0.6093 0.5703 0.0314 0.0260b = �0.1070 4 � 10�4

c = 5.0400 0.00473 kT ¼ a � exp b � Mi�c

d

� þ T�f

g

� h in oa = 7 � 10�4 6 � 10�5 0.9969 0.9960 0.0028 2 � 10�4

b = �9.0780 0.0980c = 19.8383 0.0349d = 4.0913 0.0032f = 48.0504 0.0114g = 2.2631 0.0016

4 kT ¼a

McT�ðC�bÞ � exp ðT � cÞ � ðd � Mc

T � �� � a = 0.0015 2 � 10�6 0.6301 0.5717 0.0305 0.0259

b = �15.8499 0.0582c = 0.1633 4.126 � 10�5

d = �3.7136 5.651 � 10�4

5 kT ¼ a � bC � exp ðc � TÞ � d � McT

� �� �� � a = 2 � 10�7 9 � 10�9 0.9973 0.9969 0.0026 2 � 10�4

b = 0.0140 0.0096c = 0.4534 0.0026d = �2.0010 0.0246

6 kT ¼ eðVh�aÞ � k23 � bðT�23Þ a = �7.3673 1.7261 0.9969 0.9967 0.0028 2 � 10�4

b = 0.0147 0.03267 kT ¼ a � bO2 � exp ðc � TÞ þ d � MC

T

� h ia = 2.6519 0.0001 0.9994 0.9993 0.0012 4 � 10�5

b = 0.7886 3 � 10�5

c = �0.0384 2 � 10�6

d = 0.0449 1 � 10�5

8 kT ¼ a � exp ðb � TÞ þ c � O2T

� � d � EC

T

� �h ia = 0.1811 6 � 10�7 0.9996 0.9995 0.0010 3 � 10�5

b = �0.0253 1 � 10�7

c = �5.5840 0.0001d = �22.5055 2 � 10�5

9 kT ¼ Oa2 � b

ðT�23Þ � pHc � McT

� �d a = �0.5047 2 � 10�7 0.9999 0.9999 0.0005 8 � 10�6

b = 0.9689 8 � 10�7

c = �0.5890 8 � 10�7

d = �2.3560 2 � 10�6

114 I. Petric et al. / Bioresource Technology 117 (2012) 107–116

at the same it decelerated the process in the latter phases. Besidesthe changes of physico-chemical properties of the three mixtures,the choice of mixture which is the most suitable for successfulcomposting was also influenced by two important factors: theduration of this study and the applied airflow rate. The studywas limited on the active phase of composting process (22 days).Applied airflow rate (0.9 l air min�1 kg�1

OM) was constant duringthe study. In the future study, it would be valuable and interestedto investigate the changes of physico-chemical properties in thematuration phase as well as to optimize (adjust) the airflow rateaccording to extent of OM degradation and microbial activity.

Fig. 3. Fitting of the model to the experimental data (R1 – reactor 1, R2 – reactor 2,R3 – reactor 3).

3.2. Application of the six literature kinetic models to the experimentaldata

Calculated values of kinetic parameters for existing modelsfrom literature and statistical analyses are presented in Tables 3–5.

Statistical analyses showed that the highest values of correla-tion coefficient and adjusted correlation coefficient, and the lowestvalues of root mean square deviation and variance were found inmodel 5 (Külcu and Yaldiz, 2004). This model also showed verygood 95% confidence intervals, which were smaller or at leastsmaller than the respective parameter values (in absolute values).These findings are valid for all three reactors.

Other models had worse statistical results than model 5 eitherin one, two or all three reactors. Statistical analyses showed thatthe worst results in all three reactors were found in model 2. Itshould be noted that model 5 has four kinetic parameters andthree measured variables (temperature, moisture content, CO2

concentration). On the other hand, model 2 has three kineticparameters and two measured variables (temperature, moisturecontent).

Successful convergence of a nonlinear regression model is oftendependent upon the use of good initial guesses for the modelparameters. Good initial parameter guesses can be typically ob-tained from a linearization of the nonlinear equation so that linearregression or multiple linear regression can be applied. The initialestimates are required for all model parameters. Models 1, 2, 5 and6 are firstly solved with multiple linear regression and then withnonlinear regression.

Due to the complexity, models 3 and 4 can be solved by usingnonlinear regression only, taking the values from the literature(Külcu and Yaldiz, 2004) as the initial guesses.

For some kinetic parameters in models 2, 3 and 6, confidenceintervals showed higher values than the values of kinetic parame-ters. Nonlinear regression is very sensitive to the initial guessesthat are given for the individual kinetic parameters. For largernumber of kinetic parameters in the model it is necessary to give

Table 6The maximum and mean differences between experimental and model data for the organic matter content.

Reactor Maximum difference (%) Mean difference (%)

0–5 days 6–14 days 15–22 days 0–22 days 0–5 days 6–14 days 15–22 days 0–22 days

1 1.31 3.91 2.04 3.91 0.69 1.76 1.25 1.302 1.66 3.37 4.07 4.07 0.66 1.38 2.55 1.603 1.64 3.06 3.93 3.93 0.50 1.49 2.99 1.76

I. Petric et al. / Bioresource Technology 117 (2012) 107–116 115

adequate initial guesses, which would decrease possibility ofobtaining bad statistical results.

Another reason for bad statistical results is the heterogeneity ofcompost material, which implies the difficulty of obtaining ‘‘ideal’’experimental data for statistical analyses. This problem could bereduced with the number of replicates higher than three (used inthis study). Moreover, TEM characterization (Watteau and Ville-min, 2011) can highlight two fractions of organic matter, on easilyavailable and a more recalcitrant one, and also a remaining micro-bial activity.

3.3. Application of the three proposed kinetic models to theexperimental data

In addition to six literature models, three new models havebeen proposed in order to predict more accurate data by thesemodels (Tables 3–5). It should be mentioned that fittings of morethan three new kinetic models to the experimental data were at-tempted (results not shown), but for these models the fittingsdid not converge to a single solution.

According to the statistical results of all proposed models,model 9 best described the data obtained by experiment. Amongall nine kinetic models, model 9 has the best value of all statisticalindicators. Based on comparison of statistical indicators for litera-ture model 5 and new model 9, it can be concluded that model 9 ismore suitable than model 5. Moreover, the other two proposedmodels 7 and 8 have better statistical indicators than literaturemodel 5. From these results, it can be observed that models withmore parameters and more measured experimental variables willgive better fitting than models with fewer parameters and fewermeasured experimental variables. Furthermore, it seems that thenumber of measured experimental variables represents moreinfluential factor than the number of kinetic parameters. Thisobservation was confirmed by the fact that models 5 and 9 havethe same number of kinetic parameters (four), but model 9 hasfour measured variables (temperature, moisture content, pH, O2

concentration) in comparison to the three measured variables inmodel 5 (temperature, moisture content, CO2 concentration).

Fittings of kinetic model 9 to the experimental data of organicmatter content in reactors 1, 2 and 3 are shown in Fig. 3. It has beennoticed that there is relatively good agreement between the resultsobtained by experiment and model in all three reactors. Some dis-crepancies were noticed immediately after thermophilic phase andduring the cooling phase which are related to the degradation ofhardly biodegradable compounds. It should bear in mind that com-posting material in this study represents a highly heterogeneousmixture of several different substrates with different reactionrates. In order to quantify the above mentioned discrepancies,the maximum and the mean difference between experiment andmodel for the organic matter content were found and presentedin Table 6. The maximum differences in the three reactors rangedfrom 3.91% to 4.07%, whereas the values of mean differences wereat interval from 1.30% to 1.76%. Model 9 showed the best fit to datafor reactor 1, which had the highest percentage of OFMSW in theinitial mixture. The maximum and mean differences were the leastduring the first 5 days of the process in all reactors. The maximumvalues of both differences were observed between 15th and 22nd

days for reactors 2 and 3, and between 6th and 14th days for reac-tor 1. There were two possible reasons for observed discrepancies,both related to a highly heterogeneous nature of composting mate-rial used in this study. The first reason for discrepancies lies in thereliability of experimental data. Organic matter content is the mainmeasured variable in this study. Therefore, measurement of organ-ic matter content should be performed with greater accuracy thanthe measurements of other variables. Some inconsistencies in theexperimental data of organic matter content were noticed, suchas increases of OM content on the 14th day in reactor 1 (Fig. 3),on the 9th day in reactor 2 (Fig. 3) and on the 6th day in reactor3 (Fig. 3). These strange phenomena were most likely due to sam-pling errors. Possible solution to this problem may be in perform-ing more experiments and more sampling, both with different andunder same experimental conditions in reactors (reproducibility ofexperiments). At the same time, due attention must be also givento the other measured variables (O2 concentration, CO2 concentra-tion, MC, pH, EC, temperature). The second reason for discrepanciesis related to the first-order kinetics and to the proposed kineticmodels. It is obvious that the applicability of first-order kineticson simulation of organic matter degradation showed some limita-tions in this study. Keener et al. (1993) showed applicability oftheir first-order kinetic model over a short time period (approxi-mately 3 days) in the composting of chicken manure. They empha-sized that after 3 days the rate constant had to be updated in orderto reflect the changes in material composition. In the case of yardwaste, the model is applicable over a much longer period (Marruget al., 1993). Mason (2006) stated that limited evidence exists forthe applicability of a first-order model to substrate degradation.Alternatively, a double exponential approach, incorporating sepa-rate terms for rapidly and slowly degradable substrates (Haug,1993) could be used for future investigation. This approach has alimitation in the fact that there is often no sharp distinction amongrapid and slow fractions, particularly in composting systems wherelong detention time are sometimes required for complete stabiliza-tion (Haug, 1993). Nevertheless, the latter approach was applied insome studies (Haug, 1993; Kim et al., 2000; Tosun et al., 2008;Mason, 2009). On the other side, kinetic models (Briški et al., 2003;Petric and Selimbašic, 2008) used calculated values of reaction or-ders higher than 1, and with these models relatively good agree-ment between model and experimental data was achieved for awhole period of composting process. Therefore, one of the possibledirections for future work would be to study the nth-order kinetics.Also, more than four measured variables should be included insome future kinetic models. At the same time, there is a possiblerisk of creating some complex models with great number of kineticparameters, multiple solutions and bad statistical indicators.

Although presented kinetic approach has some limitations, thedeveloped kinetic models could be used as a tool for satisfactoryprediction of organic matter content and for further research ofcomplex relationships between different measured variables.

4. Conclusions

The results showed that the mixture 60% OFMSW, 20% poultrymanure, 10% mature compost and 10% sawdust provided the mostappropriate conditions for composting process.

116 I. Petric et al. / Bioresource Technology 117 (2012) 107–116

The most suitable model was the proposed model:

kT ¼ Oa2 � b

ðT�23Þ � pHc � Mc

T

�d

Kinetic models with four measured variables proved to be bet-ter than models with less number of measured variables. The num-ber of measured experimental variables influences kinetics morethan the number of kinetic parameters.

Satisfactory fittings of the kinetic model to the experimentaldata were achieved. The model is more suitable for data of mix-tures with much higher percentage of OFMSW than percentageof poultry manure.

Acknowledgements

The research conducted and presented within the study was apart of Research Project ‘‘Possibilities for the application of com-posting process of municipal solid waste with different additivesin a reactor system’’, financially supported by the Federal Ministryof Education and Science of Bosnia and Herzegovina.

The authors thank Nidret Ibric, Damir Fejzic, Adnan Skejic andJasmin Pandurovic for their help in the laboratory.

References

Altieri, R., Esposito, A., Nair, T., 2011. Novel static method for bioremediation ofolive mill waste. Int. Biodeterior. Biodegrad. 65 (6), 786–789.

APHA (American Public Health Association), 1995. Standard Methods for theExamination of Water and Wastewater. APHA, Washington, DC.

Baptista, M., 2009. Modelling of the Kinetics of Municipal Solid Waste Compostingin Full-scale Mechanical–Biological Treatment Plants. Ph.D. Thesis. NewUniversity of Lisbon, Lisbon.

Baptista, M., Antunes, F., Souteiro Gonçalves, M., Morvan, B., Silveira, A., 2010.Composting kinetics in full-scale mechanical–biological treatment plants.Waste Manage. 30 (10), 1908–1921.

Bari, Q.H., Koenig, A., Tao, G.H., 2000. Kinetic analysis of forced aeration composting– I. Reaction rates and temperature. Waste Manage. Res. 18, 303–312.

Barrington, S., Choinière, D., Trigui, M., Knight, W., 2002. Effect of carbon source oncompost nitrogen and carbon losses. Bioresour. Technol. 83 (3), 189–194.

Basnayake, B.F.A., Menikpura, S.N.M., Jayakody, K.P.K., Chandrasena, A.S.H., 2007.Development of a protocol for organic waste characterization. In: EleventhInternational Waste Management and Landfill Symposium.

Briški, F., Gomzi, Z., Horgas, N., Vukovic, M., 2003. Aerobic composting of tobaccosolid waste. Acta Chim. Slov. 50, 715–729.

Das, M., Uppal, H.S., Singh, R., Beri, S., Mohan, K.S., Gupta, V.C., Adholeya, A., 2011.Co-composting of physic nut (Jatropha curcas) deoiled cake with rice straw anddifferent animal dung. Bioresour. Technol. 102, 6541–6546.

Ekinci, K., Keener, H.M., Michael, F.C., Elwell, D.L., 2001. Effects of temperature andinitial moisture content on the composting rate of short paper fiber and broilerlitter. In: ASAE Annual Meeting, California.

Fourti, O., Jedidi, N., Hassen, A., 2010. Humic substances change during the co-composting process of municipal solid wastes and sewage sludge. World J.Microbiol. Biotechnol. 26, 2117–2122.

Hamelers, H.V.M., 2004. Modeling composting kinetics: a review of approaches.Rev. Environ. Sci. Biotechnol. 3, 331–342.

Hamoda, M.F., Qdais, H.A.A., Newham, J., 1998. Evaluation of municipal solid wastecomposting. Resour. Conserv. Rec. 23, 209–223.

Haug, R.T., 1993. The Practical Handbook of Compost Engineering. Lewis Publishers,Boca Ratan, FL.

Huang, G.F., Wong, J.W.C., Wu, Q.T., Nagar, B.B., 2004. Effect of C/N on composting ofpig manure with sawdust. Waste Manage. 24, 805–813.

Keener, H.M., Marugg, C., Hansen, R.C., Hoitink, A.A.J., 1993. Optimizing theefficiency of the composting process. Science and Engineering of Composting:Design, Environmental, Microbiological and Utilization Aspects. RenaissancePublications, Worthing, Ohio, USA, pp. 54–94.

Kim, D.S., Kim, J.O., Lee, J.J., 2000. Aerobic composting performance and simulationof mixed sludges. Bioprocess Eng. 22 (6), 533–537.

Külcu, R., Yaldiz, O., 2004. Determination of aeration rate and kinetics of compostingsome agricultural wastes. Bioresour. Technol. 93 (1), 49–57.

Kumar, S., Sakhale, A., Mukherjee, S., 2009. Simplified kinetic analysis forcomposting of municipal solid waste. Pract. Periodical Hazard. ToxicRadioactive Waste Manage. 13 (3), 179–186.

Liao, P.H., Jones, L., Lau, A.K., Walkemeyer, S., Egan, B., Holbek, N., 1997. Compostingof fish wastes in a full-scale in vessel system. Bioresour. Technol. 59 (2–3), 163–168.

Liu, D., Zhang, R., Wu, H., Xu, D., Thang, Z., Yu, G., Xu, Z., Shen, Q., 2011. Changes inbiochemical and microbiological parameters during the period of rapidcomposting of dairy with rice chaff. Bioresour. Technol. 102 (19), 9040–9049.

Marrug, C., Grebus, M., Hansen, R.C., Keener, H.M., Hoitink, H.A.J., 1993. A kineticmodel of the yard waste composting process. Compost Sci. Util. 1 (1), 38–51.

Mason, I.G., 2006. Mathematical modeling of the composting process: a review.Waste Manage. 26 (1), 3–21.

Mason, I.G., 2009. Predicting biodegradable volatile solids degradation profiles inthe composting process. Waste Manage. 29 (2), 559–569.

Paredes, C., Roig, A., Bernal, M.P., Sanchez-Monedero, M.A., Cegarra, J., 2000.Evolution of organic matter and nitrogen during co-composting of olive millwastewater with solid organic wastes. Biol. Fertil. Soils 32, 222–227.

Petiot, C., de Guardia, A., 2004. Composting in a laboratory reactor: a review.Compost Sci. Util. 12 (1), 69–79.

Petric, I., Selimbašic, V., 2008. Development and validation of mathematical modelfor aerobic composting process. Chem. Eng. J. 139 (2), 304–317.

Press, W.H., Flannery, P.B., Teukolsky, S.A., Vetterling, W.T., 1992. NumericalRecipes, second ed. Cambridge University Press, Cambridge.

Shacham, M., Cutlip, M.B., Elly, M., 2004. POLYMATH, Educational Version 6.0. TheCACHE Corporation, USA.

Statistical Graphics Corporation, 1996. STATGRAPHICS, Version 2.1, Rockville,Maryland, USA.

Tiquia, S.M., Richard, T.L., Honeyman, M.S., 2000. Effects of windrow turning andseasonal temperatures on composting of hog manure from hoop structures.Environ. Technol. 21, 1037–1046.

Tosun, I., Gönüllü, M.T., Arslankaya, E., Günay, A., 2008. Co-composting kinetics ofrose processing waste with OFMSW. Bioresour. Technol. 99 (14), 6143–6149.

Watteau, F., Villemin, G., 2011. Characterization of organic matter microstructuredynamics during co-composting of sewage-sludge, barks and green waste.Biresour. Technol. 102, 9313–9317.

Watson, M.E., 2003. Testing Compost. Ohio State University Fact Sheet ANR-15-03.<http://ohioline.osu.edu/anr-fact/pdf/0015.pdf> (accessed 05.04.12).

Woods End Research Laboratory, 2000. Interpretation of waste and compost tests. J.Woods End. Res. Lab. 1 (4), 1–6.