Prediction of methane production performances based on determination of organic components for different vegetable wastes

The rapid development of the economy has led to rapid consumption of fossil fuels, which results in extremely serious environmental problems. Biomass energy has been accepted as a way to reduce the usage of fossil fuels due to its cleanliness and renewability. In this study, vegetable wastes (VWs), an abundant kind of biomass resource, were treated by anaerobic digestion (AD) to be converted into methane. The total solids (TS), volatile solids (VS), elemental contents, and organic components of 17 kinds of typical VWs were systematically determined. The methane production performances were then measured and ranged from 120.1 mL/g VS (for pepper stem) to 377.7 mL/g VS (for bok choy). To easily and quickly predict the methane yields of VWs, a curvilinear relationship between different organic compositions (e.g., cellulose, hemicellulose, lignin, non-structural carbohydrate, protein, and VFA contents) and methane production was established and proved to be a useful tool for methane prediction. Four kinetic models (first-order model, Fitzhugh model, Cone model, modified Gompertz model) were applied to simulate the process of AD, and Cone and modified Gompertz models were shown to describe the AD process well. This study will not only provide basic data about the characteristics and methane production of 17 kinds of VWs but also contribute a method for predicting the methane yields of vegetable wastes, which is also valuable in future agro-industrial applications.


Introduction 
With the fast development of the economy, the energy demand is increasing every year, which leads to more and more depletion of fossil fuels.According to China Statistical Yearbook 2017, China, which is the second largest economic system in the world, consumed a large amount of coal, oil, and natural gas, which accounts for 86.7% of the total energy consumption in 2016 [1] .Inevitability, in the process of utilizing those fossil fuels, various kinds of soil, water, and air pollutants were generated [2][3][4] .Biomass energy, which is renewable and clean [5] , has been regarded as one of the most popular alternatives to fossil fuels [6] .
Vegetable waste (VW) is produced in the process of vegetable planting, growing, harvesting, and processing [7] .
It is an increasing biomass resource, accounting for approximately 30% of the total vegetable yield [8][9][10] .Most VWs are currently treated by landfilling and incineration, which lead to secondary pollutants in the environment [7] .
Anaerobic digestion (AD), which can efficiently convert biomass wastes to methane [11,12] , has been applied in managing many conventional agricultural residues, such as lignocellulosic materials and animal manures [13][14][15][16] .VW has a high moisture content, organic content, and high biodegradability, which make it a desired feedstock for AD [7,17] .
Prior studies have assessed the AD of VWs, and most of them used mixed VWs [10,[18][19][20] .However, AD performance of mixed VWs will be greatly influenced by the vegetable variety, location, season, and even collection method.These factors can cause fluctuations in feedstock and AD performance.It is scientifically necessary and important to investigate waste characteristics and methane production performances, which were rarely reported before.Moreover, as an important energy conversion parameter, the methane yield is directly influenced by the organic components, such as cellulose, hemicellulose, protein, and non-structural carbohydrate [12,21] .
It will be very helpful if the methane production of VWs could be quickly estimated by simply measuring the organic components.
The major objectives of this study were as follows: 1) to systematically investigate the characterization and methane production of 17 representative kinds of VWs in China; 2) to explore the correlation between biochemical components and methane yields to predict the methane production performance easily through properties analysis; and 3) to provide some basic data and useful reference for future scientific research and industrial utilization of VWs.

Feed and inoculum
Seventeen typical kinds of VW, which were massively produced or abandoned in the field, were collected from farms in Shandong province, China, where the output of vegetables is enormous.The common and Latin names, the used parts and the abbreviations are shown in Table 1.After collection, some vegetable wastes were divided into leaves and stems.All the samples were shredded to 3-5 mm using a grinder (JOYOUNG, China).They were then stored at 4°C in a refrigerator.The sludge was obtained from Donghuashan biogas plant, where only pig manure was used as feedstock.Before use, the inoculum was placed in room temperature.Then, the supernate was removed and the precipitate was use as inoculum.

Characterization methods
(1) Determination of total solids (TS) and volatile solids (VS) TS and VS contents of substrates and inoculum were determined using the APHA standard method [22] .The TS and VS contents of inoculum were measured as (6.00±0.02)%and (2.57±0.00)%(w/w), respectively.
(2) Elemental analysis The elemental contents of C, N, and H were detected using an organic element analyzer (Vario EL cube, Germany).The oxygen elemental content was then calculated by the equation: C + H + N + O = 99.5% (VS based) [23] .
(4) Determination of non-structural carbohydrate, protein, and VFA The non-structural carbohydrate contents were determined by DNS colorimetric method using 3,5-Dinitrosalicylic acid (DNS) reagent [25] .The protein contents were tested by Branford method using Coomassie Brilliant Blue G250 [26] .The volatile fatty acid (VFA) contents of the raw materials were measured using a gas chromatograph (GC, Agilent 7890A, USA) equipped with a flame ionization detector and DB-wax capillary column (30 m× 530 μm × 1.0 μm).
All the characterizations were measured with three parallels.

Anaerobic digestion experiments
The organic loading and the feed-to-inoculum ratio of all the anaerobic digesters were set to 5 g VS/L and 1:1 [14] .The inoculum and the substrate were firstly added into each digester.Tap water was then poured in to get a working volume of 250 mL.Nitrogen was filled into each digester to remove the oxygen and create an anaerobic condition.All the digesters were put in an incubator with the temperature of 37°C.The blank control without raw material added was also set in the same condition to detect the methane produced by inoculum.Every condition has three parallels.

Biogas analysis
The biogas production was calculated by incorporating the pressure difference measured by a 3151 WAL-BMP-Test system pressure gauge (WAL Messund Regelsysteme GmbH, Germany) with the following Equation (1) [27] : where, V b means the daily biogas production, L; Δp stands for the absolute pressure difference, kPa; V h represents the headspace volume of the anaerobic digester, L; C is the molar volume in normal temperature-pressure, 22.4 L/mol; R refers to the thermodynamic constant, 8.314 kPa/K• mol, and T is the temperature of the incubator, 310.15 K.The biogas compositions were then tested by a GC (Agilent 7890B, USA) equipped with a thermal conductivity detector and an analytical column of Agilent Hayesep Q.

Maximum theoretical methane production and biodegradability
Maximum theoretical methane production (TMP) means the maximum methane production of the substrate in theory.It was calculated based on the compositions of organic elements, such as carbon (C), hydrogen (H), oxygen (O), and nitrogen (N), as shown in Equations ( 2) and (3) [28] : 16 14 The biodegradability (BD) was calculated as the ratio of the cumulative methane production (CMP) to TMP, as shown in Equation ( 4): CMP BD TMP  (4)

Kinetic models
Many models have been applied to simulate the AD process [29][30][31][32] , among which the following four commonly used kinetic models (first-order model, Fitzhugh model, Cone model, and modified Gompertz model) were employed, shown as Equations ( 5)-( 8) [31,[33][34][35] : ) where, B represents the CMP, mL/g VS, for a anaerobic digestion time t, d; B o represents the simulated maximum methane yield, mL/g VS; k hyd stands for hydrolysis rate constant, 1/d; R max is the maximum methane production rate, mL/(g VS• d); e equals to 2.7183; is the lag phase time, d, and n stands for a dimensionless factor.

Characteristics of VWs
As seen from Table 2, all the selected VWs had extremely different characteristics.Relatively low TS contents of 3.93%± 0.05% to 18.74%±0.14%(w/w) and VS contents of 2.56%±0.05% to 14.36%±0.20%(w/w) were obtained, which were close to other reported mixed VWs [7,18] and indicating the high moisture contents of VWs.In particular, leaf vegetable wastes (e.g., lettuce, spinach, bok choy, Chinese cabbage, crown daisy) had a moisture content of more than 90% [10,20] .In addition, it was found that the leaves of a single species had higher TS contents and VS contents than stems of the same species in this study, except for pepper, whose VS content of stem (13.67%±0.42%,w/w) was slightly higher than that of leaf (12.36%±0.01%,w/w).Generally high values of VS/TS ranging from 65.23% to 84.20% were calculated for all raw materials, implying high organic matter contents, which were desirable for AD [36] .The C/N ratios ranged from 5.05 to 16.46, most of which were slightly lower than the ideal range of C/N (from 15 to 30) [15,37] .

Methane production performance and biodegradability
As seen from Figure 1a-1d, the CMP of all the VWs were in the range of 120.1±1.8 mL/g VS to 377.7±12.1 mL/g VS.The majority of the substrates showed a CMP of more than 150.0 mL/g VS, and nearly half of them had methane yields higher than 200.0 mL/g VS.Among the substrates, leaf vegetable wastes performed especially well, and the highest CMP was produced by bok choy with a value of 377.7±12.1 mL/g VS, which might be attributed to the low TS contents.Generally, the stems of one single species had significantly lower CMP than the leaves of the same species for most VWs.The hardness of stems could be the main reason as they were difficult to dissolve in water.However, the CMP of celery stem was obviously higher than celery leaf, which might be because of the much higher non-structural carbohydrate content of celery stem (3.52%±0.05%)than leaf (0.81%±0.02%) [12] .The lowest CMP was produced by pepper stem (120.1±1.8 mL/g VS) due to its high lignin content of 15.62%±0.78%,which might influence the methane production [32] .
BD means the proportion of organic matters of substrates that are converted during the AD process [13] .A higher BD always indicates a higher conversion efficiency during AD [28] .The BD of studied VWs ranged from 29.25% to 64.04% as shown in Figure 2. It was found that those substrates (e.g., spinach, bok choy, crown daisy, celery stem, and potato leaf) that had the high CMP of more than 250 mL/g VS (296.3±19.5, 377.7±12.1,275.6±4.3,275.2±7.0, and 256.1±11.5 mL/g VS, respectively) also showed relatively high BD which exceeded 50% (60.93%,64.04%, 54.14%, 65.56%, and 51.78%, respectively).The lowest BD was found in pepper stem which also performed the lowest CMP, probably due to the very low nitrogen content (2.88%) which might limit the growth of anaerobic microorganisms [34] , and high lignin content (15.62%±0.78%)which was hard to be utilized in AD [37] .Some other VWs, such as Chinese cabbage, zucchini leaf, and eggplant leaf, also showed relatively low BD (37.33%, 32.98%, and 34.82%, respectively), which could be partly because of the low C/N (5.83, 8.48, and 8.58, respectively) that were extremely less than the optimal C/N ratio for AD [15] .
In conclusion, the high CMP and BD for the majority of the selected VWs verified that VWs would be suitable feedstock for methane production during AD.It is worthwhile to employ AD to manage the huge amount of VWs that were left in the field in the future.

Correlation between organic components and CMP
Organic components were completely different in all studied VWs, which resulted in the significant difference of methane production performances.To assess the correlation between organic components and methane yields, a non-linear regression relationship between different organic matters and the CMP was established.It was found that the function forms for all the organic matters were obviously distinct, e.g., for cellulose and hemicellulose, the forms were linear function together with logarithmic function; for protein, the form was simple linear function; and for lignin, non-structural carbohydrate, and VFA, the forms were linear relation together with quadratic linear relation.Such distinction was shown as the following equation (Equation ( 9)), in which all the coefficients were calculated using least squares method with a software (Eviews 9.0).
The simulated curvilinear correlation (R 2 ) was 0.800.In addition, as shown in Figure 2, the values of CMP′ were very close to CMP, which indicated the reliability of this equation for methane production prediction.The formula contributed by this research would not only provide useful reference for future biomethane production tests as they can predict the CMP by easily estimating the organic components but also offer a co-digestion design policy to select the mixing ratio of various VWs according to the contents of different organic compositions.

Kinetic analysis
Kinetic parameters of individual VW during the AD process are important for understanding AD features [38] , but there were rare literatures systematically reported for studied VWs.In this study, four widely used models (first-order, Fitzhugh, Cone, and modified Gompertz models) were chosen to simulate the AD process [31,34] , and all the parameters are shown in Table 3.

Table 3 Kinetic parameters simulated by the first-order, Fitzhugh, Cone, and modified Gompertz models
First-order Fitzhugh Cone Modified Gompertz In the first-order and Fitzhugh models, the values of R 2 were generally low, and the B o values were much higher than the CMP.Furthermore, it was found that the experimental results of bok choy could not be simulated by these two models, indicating that first-order and Fitzhugh models might not be appropriate to simulate the AD process for some VWs.
The R 2 values obtained by Cone and modified Gompertz models were in the range of 0.991 to 1.000 and 0.985 to 0.999, respectively, which were relatively high.Simulated B o values ranged from 131.6 mL/g VS to 395.0 mL/g VS and from 112.1 mL/g VS to 337.5 mL/g VS for Cone model and modified Gompertz model, respectively, both of which were close to the experimental range of CMP (120.1±1.8 mL/g VS to 377.7± 12.1 mL/g VS).These results suggested that the Cone and modified Gompertz models could simulate the AD process well.
The parameters obtained in these two models (Cone and modified Gompertz models) provided useful information for describing the AD process of VWs.In the Cone model, k hyd stands for the hydrolysis rate constant [34] .A higher k hyd value always implies an easier hydrolysis of the substrate and the k hyd values of studied VWs were ranged from 0.09 d -1 to 0.29 d -1 , which were generally higher than Salvinia molesta, rice straw and switchgrass (0.04-0.09 d -1 ) [31,34] , indicating the high degradability of VWs in AD process.In the modified Gompertz model, the kinetic parameter R max was the methane production rate and the R max values were in the range of 8.6 mL/(g VS• d) to 60.4 mL/(g VS d).A higher value normally means a higher methane production [34] , except for several substrates such as bok choy with high methane yield (377.7±12.1 mL/g VS) simulated a relatively low R max value of 27.4 mL/(g VS d).The value represents the lag phase time, which indicated the adaptation time of microorganisms for the substrates in AD [39] .In this study, the majority of the substrates was found having relatively low λ values range (from 0.3 d to 1.8 d), suggesting the fast degradation of VWs in AD [36] .Compared to these substrates, the λ value of bok choy was much higher (4.3 d), which might partly due to the lowest VS/TS ratio (65.23%) among all the studied VWs.It also indicated that big difference in organic components might influence the AD process [21,40] .

Conclusions
In this study, 17 kinds of representative VWs were systematically and comprehensively analyzed.
The results showed that all the substrates had different characteristics.The methane yields of all VWs were in the range of 120.1 mL/g VS to 377.7 mL/g VS, and 8 out of 17 had a significantly high methane production over 200.0 mL/g VS.The AD process of VWs was well simulated by the Cone and modified Gompertz models.A correlation between methane yields and individual organic components, which could be used as a quick and reliable tool for methane production prediction in the future, was established.In conclusion, the results obtained in this research provided fundamental knowledge of common vegetable wastes characterizations, their AD performances, and connection between biochemical components and methane yields, which could not only fill in the blanks in scientific research but also facilitate the future utilization of vegetable waste resources.

Figure 1
Figure 1 Cumulative methane productions (CMP) of 17 kinds of different VWs

Figure 2
Figure 2 Methane productions and BD of 17 kinds of VWs