Sage Journals: Discover world-class research

Abstract

The biochemical methane potential (BMP) test is significant for the landfill industry as it provides a means to evaluate the gas potential, and therefore potential degradability, of both incoming and in-place municipal solid waste (MSW). However, the BMP test is not standardized making comparison of BMP results across sites problematic. For example, the BMP test duration has historically ranged from 20 days to several months with most current BMP tests lasting 60 days. However, the gas generation data can potentially be modelled for any of those durations to produce a prediction of the ultimate BMP value (BMP_ULT). Currently, the predicted BMP_ULT values of 23 long-duration (115–150 days) BMP tests were used to determine the required quantity of data (i.e. number of days) needed to produce an accurate BMP_ULT prediction. Results showed that no single test duration produced both accurate and efficient results, so a novel performance-based endpoint was proposed. The relative change in predicted BMP_ULT values with respect to time (dBMP_ULT/dt) was chosen as a potential performance-based completion metric. Results indicate that once the absolute normalized dBMP_ULT/dt value is within <2.5, <1.5 and <0.6% day⁻¹ that the predicted BMP_ULT is within 20, 10 and 5% of the true BMP_ULT, respectively. Overall, the use of performance-based metrics for determining BMP test completion will allow for the collection of partial data sets, reduced experimental times and verification of results.

Keywords

Methane production biochemical methane potential municipal solid waste BMP test completion metrics landfill landfilled MSW

Introduction

Municipal solid waste (MSW) is a highly heterogeneous material, which, unlike most geotechnical materials, is susceptible to, and will experience biological decay. During anaerobic decay the degradable organic fraction of MSW is converted to CO₂ and CH₄ which produces heat and reduces solid volume (Ivanova et al., 2008a; Kelly et al., 2006; Krause et al., 2016). An estimation of the total potential decay of MSW over time is an important factor in its classification for use in landfill management research and industrial design processes (Barlaz et al., 1989; Cho et al., 2012; Stevens, 2012). The biochemical methane potential (BMP) test is significant as it provides a means to evaluate the total anaerobic gas potential of incoming waste streams prior to landfilling (Fleming and Mathison, 2013; Owen et al., 1979; Owens and Chynoweth, 1993; Pearse et al., 2018; Raposo et al., 2012). The BMP test can also be used to estimate the remaining degradation potential for partially degraded and landfilled waste which can no longer be readily classified using other compositional methods (Kaartinen et al., 2013; Sel et al., 2016). A measure of the BMP, both prior to and during the landfilling process, is critical for important purposes such as designing a landfill gas or heat extraction system, estimating greenhouse gas emissions, and quantifying the rate and magnitude of potential settlement (Hucl, 2021; Ivanova et al., 2008a; Kelly et al., 2006; Morhart, 2022; Stevens, 2012; Usova, 2012). In addition, the BMP test can function as an index test to allow for the comparison of data between various landfills and within the literature (Krause et al., 2016; Pearse et al., 2018; Raposo et al., 2011).

The BMP test is one of the main biological index tests used for MSW (Pearse et al., 2018; Sel et al., 2016); however, this test still has major limitations. Results from BMP testing can be largely variable and inconsistent even when performed on the same test medium. For example, an interlaboratory study conducted in 2011 compared the BMP results from several different research groups all using different BMP test methodologies (Raposo et al., 2011). To control for variability, all laboratories that participated in the study performed the BMP tests on the same relatively simple and standardized lab chemicals including starch, cellulose, gelatin and mung bean. Overall, there was significant variation between the laboratories in both the measured decay rates and final BMP values for all tested materials. Clearly, if variable BMP procedures produced inconsistent results using simple standard test materials, then these same variable procedures are certain to introduce errors when comparing BMP test results for complex MSW.

A major unstandardized test parameter used for BMP analysis is the overall test duration. This variation in BMP test duration is necessary to allow for complete degradation (reaching the ultimate BMP value (BMP_ULT)) of MSW samples and can range from several months to years, even under controlled laboratory conditions (Bao, 2011; Barlaz et al., 1989; Cho et al., 2012; Ivanova et al., 2008a; Rao et al., 2000). For a MSW sample, measuring BMP_ULT experimentally becomes an exercise in impracticality and a shorter test duration must be used. The first BMP test duration used to test MSW by Owens and Chynoweth in 1993 was 75 days. Since that time, a range of other test durations have been utilized from 20 days to over 100 days, with many researchers studying MSW currently opting for a test duration of 60 days (Bayard et al., 2010; Chickering et al., 2018; Fleming and Mathison, 2013; Francois et al., 2007; Hansen et al., 2004; Kelly et al., 2006; Krause et al., 2018; Owens and Chynoweth, 1993; Pearse, 2019; Stinson and Ham, 1995; Zhu et al., 2009). However, rather than arbitrarily using the 60-day BMP value as a standardized index value, these 60 days of experimental data can be fitted to a gas prediction model to predict the BMP_ULT value (Bao, 2011; De La Cruz and Barlaz, 2010; Krause et al., 2016; Mathison, 2015; Pearse, 2019).

Standardization issues relating to the BMP test are well known, and at a 2015 conference in Leysin Switzerland, many best practices and guidelines for BMP test parameters were discussed and established (Holliger et al., 2016). One of these guidelines was to determine BMP test duration as a function of gas production rather than a fixed duration, yet many researchers studying MSW still use a set test duration of 60 days or greater (Bayard et al., 2018; Krause et al., 2018; Pearse, 2019). One possible explanation is the large proportion of lignocellulosic matter contained in MSW, and the slow anaerobic decay rate of this material group (Barlaz et al., 1989; De La Cruz and Barlaz, 2010; Holliger et al., 2016).

Clearly, aspects of the BMP test can be improved to provide better results for the assessment of MSW gas production potential. The goal of this study was to determine a methodology to assess BMP real-time data for the prediction of BMP_ULT, thus creating a more standardized method for use by researchers and industry to determine MSW gas production. To determine the validity of a single BMP test duration (such as 60 days), a method of determining data requirements was developed using 23 exhumed MSW samples from a regional landfill in Saskatchewan, Canada. Each of these samples was BMP tested, and the resulting data were modelled using a first-order decay gas prediction model. The fitted gas prediction model was then iteratively truncated to assess how the predicted BMP_ULT changes with variable quantities of input data. Following this assessment, a novel performance-based method of determining test duration is proposed along with a range of completion metrics based on the desired accuracy of the BMP_ULT prediction. Using this methodology, a more accurate assessment and verification of the BMP_ULT can be made that is based on test performance and not reliant on a pre-determined test duration.

Methods

Description of MSW

Landfilled MSW samples were collected from the Loraas Ltd landfill located north of Saskatoon, SK. The Loraas landfill is a medium-sized landfill, which mainly accepts industrial, commercial and institutional (ICI) waste; however, in the last 10–15 years, the landfill has accepted increasing quantities of MSW from surrounding communities. A sonic drill rig was used to collect eight continuous core samples (15 cm in diameter), which allowed for visual classification and approximate dating of the waste over the landfill depth. The extracted cores spanned the entire depth of the landfill and ranged from 23.6 to 25.4 m in length. Based on placement records and data markers found in the cores, the waste age at the time of extraction ranged from 2 to 31 years. For a more detailed description of the waste’s age, visual description and landfill layout, refer to Hucl (2021).

Each of the eight MSW cores was separated into three discrete samples based on depth increments of approximately 10 m to produce a total of 24 samples for laboratory analyses. These 24 MSW samples were sub-sampled to reduce sample volume and processed with a hammer mill (to a particle size <10 mm) to create a uniform sample for laboratory testing. Please refer to Casavant et al. (2019) for a complete overview of the subsampling and sample preparation procedures. Note that one of the MSW samples experienced repeated inhibition during BMP testing and was therefore removed from this study.

BMP test procedures

The BMP test methodology utilized both automatic and manual BMP tests based on the methods used by Mathison (2015), which is a modified version of the initial BMP test procedures from Owens and Chynoweth (1993). Automatic and manual BMP tests were considered to increase sample throughput given the automatic BMP tests were limited to 15 channels at one time.

The Automatic Methane Potential Test System II from BPC Instruments (Lund, Sweden) was used to perform the automatic BMP tests. A processed MSW sample of 20–50 g was slurried with distilled deaired water and 50–150 mL of anaerobic sludge from the local wastewater treatment facility. The test vessels were purged with N₂ gas for 2 minutes after assembly, to ensure initial anaerobic conditions and were then placed into a water bath to maintain a temperature of 37°C. The test duration varied from 115 to 155 days to ensure sufficient gas data was collected for an accurate prediction of the BMP_ULT. One blank test vial containing only deaired water and inoculum was also tested for every five BMP test vessels. The blank gas production data was subtracted from the MSW sample gas production data to remove the effects of inoculum from the BMP results.

The manual BMP test method only differed from the automatic method with respect to the size of test vessels and the method of measuring produced gas. The manual BMP tests were performed in 2 L flasks, with 50–200 g of MSW and 400–1500 mL of inoculum used per test vessel. To prevent over-pressuring of the vessels, the produced gas was collected in a Tedlar bag attached to the lid of each vial. The collected gas was periodically measured using a GEM 3000 probe to obtain gas composition, and a water displacement column was used to determine the total gas volume. All gas measurements are recorded in units of standard temperature and pressure and have been corrected for water vapour pressure.

To verify consistent results between the manual and automatic test methodologies, a single MSW sample was tested in triplicate using both methods. The automatic method produced a range of BMP_ULT values of 45.3 to 55.5 L_CH4 kg_VS⁻¹, with a mean of 48.5 L_CH4 kg_VS⁻¹. The manual method had a range of BMP_ULT values between 50.0 and 51.3 L_CH4 kg_VS⁻¹, with a mean value of 50.7 L_CH4 kg_VS⁻¹. Given the small differences in both the ranges and mean values of BMP_ULT, the results from both methods were assumed to be consistent with each other.

The volatile solids (VS)-to-total solids (TS) ratio of the MSW samples varied from 0.19 to 0.59 with a mean of 0.38 (±0.11), whereas the VS/TS ratio of the inoculum remained fairly constant (24 and 30⁻¹g TS/L). Together, this resulted in the inoculum-to-substrate ratio (ISR) of the tests varying from 0.06 to 0.31 with a mean of 0.12 (±0.06). The ISRs presented here are lower than the range of 2–4 recommended by Holliger et al. (2016) but are comparable to ISRs noted in literature for other BMP tests performed on large MSW samples with a high lignocellulosic content (Mathison, 2015; Pearse, 2019). As an additional means of test verification, a number of samples were spiked with 1⁻¹M sodium acetate to act as a control, after BMP test completion had been verified. The gas generated by the sodium acetate was compared against the theoretical stoichiometric value, and it was found that gas production ranged from 89 to 112% of the theoretical potential with a mean of 100% and a standard deviation of 9%.

Model convergence analysis algorithm and statistical analyses

Experimental and modelled BMP_ULT

Several different methods of predicting BMP_ULT values from ongoing BMP tests have been used in the literature. In general, BMP_ULT can be predicted in one of three ways: (1) by constructing a constituent model of the waste; (2) by simulating the life cycles of the bacterial groups associated with anaerobic decay (i.e. Monod model or modified Gompertz equation) or (3) by empirically curve fitting a simple mathematical model such as a first-order decay model to the set of experimental data (Beaven, 2008; De La Cruz and Barlaz, 2010; Ivanova et al., 2008b; Krause et al., 2016; Li et al., 2013; Mathison, 2015; Pearse, 2019). A first-order decay gas prediction model (equation 1) was chosen in this study based on: the simplicity of the model, the ability to easily account for lag time, and the excellent fit of the model to the experimental data (see ‘Results and discussion’ section).

BMP (t) = {BMP}_{ult} (1 - e^{- k (t - t lag)})

(1)

where BMP(t) (L_CH4 kg_VS⁻¹) is the cumulative BMP at time t (days); BMP_ULT (L_CH4 kg_VS⁻¹) is the BMP_ULT determined via the model; k is the decay rate (days⁻¹) and t_lag (days) is lag time. It is worth noting that since the samples used were exhumed landfill samples, the value of BMP_ULT is a measure of the remaining methane potential of the samples. If the methane generation trends of a BMP test do not align well with a first-order decay model or any of the other previously mentioned gas prediction models, then some major inhibitory effect has taken place and the data from that test should not be used (Holliger et al., 2016; Koch et al., 2019).

Curve fitting was performed using the scipy.optimize.curve_fit function in Python. This function utilizes a method of least squares to optimize the input parameters, BMP_ULT and decay rate (k) while minimizing the residual sum of squares between the experimental data, and the output model. A summary of the algorithm used to identify t_lag and apply the curve fitting function can be found in Mathison (2015).

Model convergence analysis

If a gas prediction model is accurately predicting the BMP_ULT of an ongoing test, then the addition of further experimental data will not significantly change the model’s output, which is referred to here as model convergence. Rather than attempting to set a single BMP test duration, the ability to verify model convergence becomes the new goal in defining BMP test duration. The intention of a model convergence analysis is to obtain a set of predicted BMP_ULT values while gradually reducing the quantity of modelled input data (Figure 1). The first step is to model the predicted BMP_ULT value using the entire experimental data set which is then assumed to be the ‘true’ BMP_ULT value. This assumption was deemed appropriate as all BMP tests performed for this study had test durations ranging from 115 to 155 days, which was far in excess of the 60 day test duration typically found in the literature. Next, the experimental dataset was truncated by one data point, and the gas prediction model was then fitted to the new truncated dataset. The model output along with the quantity of data (days) modelled was then stored. Finally, the previous two steps would be repeated, with the experimental data set continuing to be further truncated and the reduced model output stored. The cycle of data truncation and analysis continues until the curve fitting function becomes unstable, at which point the data sets containing the iterative BMP_ULT predictions and the corresponding quantities of modelled data (days) were output for further analysis.

Figure 1.

Model convergence analysis algorithm which outlines the process of iteratively truncating and modelling the results of the reduced BMP dataset.

Convergence tolerances

A measure of BMP test completion is required to compare data requirements between replicates or different samples undergoing model convergence analysis. Convergence tolerance refers to the quantity of input data required for the output of the gas prediction model to converge within a certain percent tolerance of the final model output. The convergence tolerances used here are 5, 10, 15 and 20%. These convergence tolerances were arbitrarily picked as a sensible range for comparison purposes. A convergence tolerance of 5% was intended to represent a tolerance appropriate for research, a convergence tolerance of 20% represents an initial approximation of BMP results and the convergence tolerances of 10 and 15% were meant to provide a range for better interpretation of results and trends.

Performance-based BMP_ULT determination

If a single standard BMP test duration has the potential to be either insufficient or excessive for some fraction of MSW samples, then a different metric is required to determine BMP test completion. Firstly, this test completion metric should be based on individual test performance to allow for verification of individual tests. Secondly, the completion metric should be based on data present at the time of test completion. If excessive quantities of data past the point of completion are required to verify test completion, then the performance-based metric becomes inherently redundant.

As BMP data is gathered, the predicted BMP_ULT value for a test will asymptotically approach the true BMP_ULT value and the relative change in predicted BMP_ULT with respect to time will approach 0. The rate of change in predicted BMP_ULT with respect to time has the potential to be a performance-based metric which can be assessed for each individual test. Additionally, this potential metric can be normalized using the current predicted BMP_ULT value at the time of assessment. For these reasons, the rate of change in predicted BMP_ULT with respect to time will be assessed as a potential candidate for a performance-based BMP test completion metric.

To assess the relative change in predicted BMP_ULT values, the first-order derivative of the normalized predicted BMP_ULT curves can be calculated with respect to time (dBMP_ULT/dt). The first-order derivative was calculated using a backwards difference approximation and assessed at convergence tolerances of 5, 10 and 20%. Finally, the absolute value of dBMP_ULT/dt was used to account for both the over and under.

Results and discussion

The experimental results and discussion section includes four subsections. The first subsection shows the typical use of a first-order model to predict BMP_ULT using 1 of the 23 samples as an example. Secondly, a numerical investigation of the modelled BMP results was undertaken to determine the test duration requirements. Thirdly, a set of literature BMP data from a separate site was tested using the same analytical methods to verify the conclusions drawn from that numerical investigation. Finally, a final analysis of the modelled BMP results was performed to determine a preliminary range of performance-based experimental endpoints.

BMP results and first-order decay modelling

Figure 2(a) shows the cumulative BMP results for 1 of the 23 MSW samples tested in triplicate (BH 18-02 Mid 1) with the remaining 22 MSW samples tested presented in the supporting information (Supplemental Figure S1). Although all three curves in Figure 2(a) are replicates of the same sample, there is a variable lag time between 20 and 60 days which leads to significant differences in the required quantity of data to predict BMP_ULT. MSW is known to be a highly heterogeneous media and, even after processing, some degree of inconsistency will remain even in a large sample (20–50 kg). Sub-sampling an imperfect test media will lead to composition differences between replicates and therefore their corresponding test behaviours are expected to have inherent variability. The variable t_lag could also be indictive of a low ISR (Holliger et al., 2016), although it has been noted in the literature that BMP tests with low ISRs (<0.4) performed on MSW rich in lignocellulosic material still produce a valid measure of BMP_ULT (Mathison, 2015; Pearse, 2019). This particular sample was chosen to highlight that the methods developed herein could be demonstrated with a realistic MSW sample instead of an ideal sample with more consistent triplicate results. Despite the differences in t_lag, all three replicates achieved similar final gas generation values ranging from 121 to 135 L_CH4 kg_VS⁻¹. Unfortunately, replicate 3 did experience early inhibition (85 days), but the experimental data were still utilized, as the prediction of BMP_ULT was comparable to the longer duration replicates (130 days).

Figure 2.

(a) Cumulative BMP curves for an individual example MSW sample tested in triplicate including BMP_ULT; and (b) first-order decay gas production model fitted to the experimental data of the second duplicate (BH 18-02 Mid 1 (2)), the green line in Figure 2(a).

Figure 2(b) shows the fit of a first-order decay model applied to the second replicate in Figure 2(a). The model predictions were in good agreement for all MSW samples with R² values ranging from 0.93 to 0.99 (Supplemental Figure S2). In Figure 2, the experimentally measured BMP_ULT value (129.9 L_CH4 kg_VS⁻¹) was in close agreement with the predicted value (132.3 L_CH4 kg_VS⁻¹). This close agreement shows that the experimentally measured BMP_ULT is approaching the true material property, which verifies the assumption that the monitoring duration of 130 days was sufficient to predict the remaining BMP_ULT of the sample. This result was expected as most BMP tests performed with MSW will not exceed a test duration of 60 days unless there is a specific experimental or analytical reason (Bayard et al., 2018; Francois et al., 2007; Hansen et al., 2004; Kelly et al., 2006; Zhu et al., 2009).

Data requirements to model BMP_ULT

Figure 3(a) shows the results of a model convergence analysis when performed on the same example sample shown in Figure 2(b). The predicted BMP_ULT value is significantly overestimated at 60 days for this sample, and an extended monitoring duration of 80–90 days would sustainably improve BMP_ULT predictions. Figure 3(b) shows a normalized model convergence analysis performed with all three replicates presented in Figure 2(a). These curves were normalized so that 100% represents the predicted BMP_ULT value when the entire experimental dataset was modelled. Despite, all three replicates shown in Figure 3(b) having similar experimental and predicted BMP_ULT values the quantity of experimental data required for model convergence varies greatly between these replicates which appears to mainly be a symptom of the variance in t_lag. Similarly, the raw and normalized model convergence analysis for the remaining samples are shown in Supplemental Figures S3 and S4.

Figure 3.

(a) Model convergence analysis showing changing predictions of BMP_ULT with changing quantities of data for BH 18-02 Mid 1 (2) (as for Figure 2(b)); and (b) Normalized model convergence analysis of the example MSW triplicates normalized around the final predicted BMP_ULT value.

Figure 4(a) shows the example MSW triplicates with convergence tolerances of 5, 10, 15 and 20%. A duration of about 60 days could be considered to be sufficient if the allowable convergence tolerance is relatively loose (20%). If a tighter convergence tolerance such as 5 or 10%, is required then a longer test duration of 90–100 days will be necessary. Figure 4(a) also shows that there is a large amount of variation in the quantity of data required between replicates. These differences in data requirements could be due to a number of slight compositional or biological differences between these replicates. MSW is a notoriously heterogeneous medium leading to difficulty in producing a uniform laboratory sample for testing, while still maintaining a sample that is representative of the larger waste mass that the sample was extracted from Casavant et al. (2019) and Pearse et al. (2018). Thus, despite the diligent processing of the MSW samples, variability in data requirements between replicates is to be expected. This variability in data requirements is also highlighted in Supplemental Figure S5, which contains the model convergence analysis for the other 22 samples tested.

Figure 4.

(a) Quantity of data required for model convergence of the example MSW sample from Figure 2(b) at various convergence tolerances (each bar represents a replicate); and (b) quantity of data (days) required for model convergence at a tolerance of 10% for three different MSW samples.

Figure 4(b) was produced to compare model convergence across the replicates of three different samples (top 1, mid 1 and bot 1) from the same continuous core run (BH 18-02) at a single convergence tolerance (10%). As expected based on the previous discussion of Figure 4(a), there is a substantial difference in the required quantity of data between these separate MSW samples. The top 1 sample requires approximately 45 days of data to converge at 10% tolerance, whereas the mid 1 and bot 1 samples require 80–100 days of data to converge. Figure 4(a) and (b) demonstrates the large variation in the quantity of data required for model convergence both between different samples and within the replicates of the same samples.

Figure 5(a) was developed to better assess the variability in the required quantity of experimental data across all 23 MSW samples tested. The cumulative model convergence (CMC) curves in Figure 5 show the percentage of tests that have converged for a given quantity of input data for a given tolerance. Figure 5(a) shows the required quantity of data for a single set of samples can vary from 20 to 100+ days. The large range in data requirements is partially due to the variable composition of waste deposited on site, as both the potential and rate of gas production are functions of the waste composition. Clearly, any single arbitrary test duration such as 60 days is not a sufficient standard for BMP test duration. Figure 5(a) shows that 60 days of data is insufficient as just under 60% of the BMP tests converge even at the loosest tolerance of 20%. The CMC curves indicate that a longer test duration of 80–100 days would allow for most samples to converge within a 10% tolerance. Conversely, some samples converge within only 30–40 days of data and in these cases, performing a BMP test for a duration of 100+ days would be highly inefficient, as no new meaningful data would be gathered for these samples for the last 60+ days of the test. Overall, these data show that using any single duration for the assessment of a complete BMP testing regime is problematic.

Figure 5.

(a) CMC curves over time (days) for 10, 15 and 20% tolerances using experimental data; and (b) CMC curves over time (days) for 10, 15 and 20% tolerances using literature data.

To avoid the issues associated with a single set BMP test duration, test completion should be based on model convergence rather than a set time duration. For example, a sample may exhibit a short t_lag and enter an early stationary gas production phase (20–30 days). In this case, approximately 40–50 days of data should be sufficient to verify model convergence and therefore the conclusion of the test. As a result, the time and resources required to monitor a batch of BMP tests would lessen throughout the testing period, as tests reach model convergence and can be concluded. Additionally, the use of model convergence to determine BMP test duration would allow for the allocation of additional monitoring time for samples which are typically discarded due to a long t_lag. The additional monitoring time would ensure sufficient data could be collected and the predicted BMP_ULT could still be confirmed as valid.

Verification of model convergence analysis

To determine the applicability of the model convergence analysis methodology for other BMP data, results presented in Mathison (2015) were used for comparative purposes. Model convergence analysis was performed on 16 literature samples (all tested in at least three replicates) and a second set of CMC curves were created (Figure 5(b)). The data from Mathison (2015) was considered useful for the verification of model convergence analysis as the source, and therefore composition, of MSW samples was different from the source presented herein, and the dataset contained all the cumulative gas measurements over the entire test. When comparing the CMC curves in Figure 5(a) and (b), the largest discrepancy is the percentage of converged models at any given test duration. At any chosen test duration and tolerance, there is approximately 20–30% more convergence for the literature data (Figure 5(b)) compared to the experimental data (Figure 5(a)).

The difference between these two sets of CMC curves may be attributed to a few factors. The first factor is the difference in the measurement frequency between the two datasets. The literature BMP tests were entirely performed using the manual BMP test methodology while half of the BMP tests performed for this study used the manual method and half used the automatic method. The manual BMP test method required significantly more gas to produce an accurate measurement when compared to the automatic method (300 mL of gas vs 10 mL CH₄) resulting in a lower measurement frequency for the manual method. When analysing model convergence, analysis occurs at every data point; therefore, a decrease in data frequency results in a lower resolution for the corresponding model convergence analysis.

The second factor contributing to the differences between the two sets of CMC curves was a difference in the BMP test rejection protocol. For the literature data, all samples which experienced long-term inhibition (regardless of future viability) were rejected. Despite initial inhibition, these samples still had the potential to generate a complete and valid set of BMP data but would have required a longer test duration. In contrast, in the current study BMP tests were rejected if gas generation quantities were insufficient. These differences in test rejection would cause the average time requirements for an entire BMP dataset to be greater if the rejection protocols from this experiment were applied rather than the rejection protocols used for the literature. Therefore, if the same rejection protocols were used in both cases, it would be expected that the CMC curves between both data sets would be closer.

The final factor contributing to the differences in the CMC curves are the difference in sample composition. The literature samples and the samples presented here came from two separate landfills, which accepted different types of waste. The literature samples are almost entirely MSW, while the samples presented herein are a mixture of ICI waste and MSW waste. As the quantity and rate of anaerobic gas production are both factors of material composition, it is expected that MSW samples from two separate landfills with different waste placement histories and therefore different waste compositions will produce variable results.

Initial determination of a performance-based BMP endpoint

To allow for the comparison of results between both individual samples and entire BMP datasets, dBMP_ULT/dt values were taken at specific comparison points. The comparison points used were the points at which an individual test reached a model convergence of 20, 10 and 5% (Figure 6). Figure 6 shows the dBMP_ULT/dt curves with the normalized dBMP_ULT/dt values at the comparison points for the sample in Figure 3(a). The absolute normalized dBMP_ULT/dt values were calculated for all valid replicates tested for all 23 MSW samples shown in Supplemental Figure S6 (excluding outliers), and the mean and standard deviation are presented in Table 1. Additionally, the BMP data taken from the literature were also processed to obtain absolute normalized dBMP_ULT/dt values for comparison.

Figure 6.

Predicted BMP_ULT curves from a model convergence analysis and the rate of change of the predicted BMP_ULT (dBMP_ULT/dt) curves both normalized around the predicted BMP_ULT at a convergence tolerance of 5, 10 and 20% for BH 18-02 Mid 1 (2) (as for Figure 2(b)).

Table 1.

Summary of the potential performance-based biomethane potential completion metrics values (dBMP_ULT/dt) derived from both experimental and literature data sources and the statistical comparison between the two.

Model convergence tolerance (%)	Experimental		Literature		Statistics
Model convergence tolerance (%)	Mean (% day⁻¹)	SD (% day⁻¹)	Mean (% day⁻¹)	SD (% day⁻¹)	t	p
20	2.4	1.8	3	1.9	1.21	0.23
10	1.2	0.9	1.6	1.4	1.37	0.17
5	0.6	0.1	0.7	0.7	0.79	0.43

SD: standard deviation.

The values in Table 1 provide a promising initial range of dBMP_ULT/dt values for use as performance-based completion metrics. For an approximate estimate of BMP_ULT (model convergence tolerance 20%), the results in Table 1 seem to indicate that an absolute normalized dBMP_ULT/dt value <2.5% day⁻¹ would suffice. For applications requiring more accurate results, a dBMP_ULT/dt <1.5% day⁻¹ or <0.6% day⁻¹ would be sufficient for a convergence tolerance within 10 and 5% respectively. Interestingly, even though current experimental BMP data and literature BMP data had significantly different CMC curves, the values of dBMP_ULT/dt for both data sets had remarkably similar mean values at all comparison points. A t test was performed to compare the dBMP_ULT/dt derived with the experimental data against the same values found using the literature data. The t-test produced p values of 0.23, 0.17 and 0.43 for the comparison points of 20, 10 and 5%, respectively, which confirms that there is no significant difference between dBMP_ULT/dt values between both datasets for any of the comparison points. Additionally, the MSW based completion metrics found here are in close agreement with the more general BMP completion metric of <1% of accumulated methane production per day for 3 consecutive days presented by Holliger et al., (2016). These results indicate the potential of a flexible dBMP_ULT/dt-based completion metric, which could apply to BMP tests performed with different methodologies and by different researchers at a specified confidence.

Conclusions

To determine the quantity of BMP data (days) required to accurately predict BMP_ULT using a model (model convergence), a method of testing the model convergence was developed. Model convergence analysis performed on 23 exhumed MSW samples revealed large variations in the required quantity of data between both different samples and even within replicates of the same sample. The CMC curves produced using current and literature data showed significant differences in the percentage of model convergence at any given duration. Despite these differences, both sets of CMC curves showed that choosing any single BMP test duration will result in one of two issues: (1) if the test duration is insufficient, a percentage of samples will produce an unreliable prediction of BMP_ULT; or (2) if a conservative test duration is used, the overall test will be inefficient as an excessive amount of time and effort will be expended on most samples. To address these issues, the relative change in modelled BMP_ULT with respect to time (dBMP_ULT/dt) was chosen as a potential performance-based completion metric for individual samples. Results indicate that once the absolute normalized dBMP_ULT/dt value is within <2.5, <1.5 and <0.6% day⁻¹ that the predicted BMP_ULT is within 20, 10 and 5% of the experimentally determined true BMP_ULT, respectively. Despite the current versus literature CMC curves having significantly different time requirements, there was no significant difference in the dBMP_ULT/dt values produced with either data set. The use of performance-based metrics for determining BMP test completion will require verification by future researchers as the dBMP_ULT/dt values given here are preliminary and may vary slightly with test methodology or MSW composition. Still, switching to performance-based test completion will provide benefits to future researchers as it allows for the collection of partial data sets, gradual reductions in time and resource requirements throughout BMP testing and would allow for a quantifiable measure of BMP test completion.

Supplemental Material

sj-docx-1-wmr-10.1177_0734242X241227373 – Supplemental material for A method for evaluating and verifying biochemical methane potential test completion performed with landfilled municipal solid waste

Supplemental material, sj-docx-1-wmr-10.1177_0734242X241227373 for A method for evaluating and verifying biochemical methane potential test completion performed with landfilled municipal solid waste by Tyler JP Casavant, Kerry McPhedran and Ian R Fleming in Waste Management & Research

Footnotes

Acknowledgements

Authors would like to thank Loraas Disposal and NSERC for providing the funding necessary for this project, Derek Stevens and Chris Tendler at Loraas for providing guidance during the fieldwork and the lab staff at the Saskatoon Wastewater Treatment Plant for providing all of the anaerobic inoculants.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Tyler JP Casavant

Kerry McPhedran

Supplemental material

Supplemental material for this article is available online.

References

Bao

(2011) Evaluating and predicting the ultimate biochemical methane potential for solid waste from short-term monitoring data. M.Sc. Thesis, North Carolina State University, Raleigh, NC, USA.

Barlaz

Ham

Schaefer

(1989) Mass-balance analysis of anaerobically decomposed refuse. Journal of Environmental Engineering 115: 1088–1102.

Bayard

Benbelkacem

Gourdon

, et al. (2018) Characterization of selected municipal solid waste components to estimate their biodegradability. Journal of Environmental Engineering 216: 4–12.

Bayard

de Araújo Morais

Ducom

, et al. (2010) Assessment of the effectiveness of an industrial unit of mechanical-biological treatment of municipal solid waste. Journal of Hazardous Materials 175: 23–32.

Beaven

(2008) Review of responses to a landfill modelling challenge. Proceedings of Institution of Civil Engineers: Waste and Resource Management 161: 155–166.

Casavant

Fleming

McPhedran

, et al. (2019) Characterisation of landfilled waste using continuous core samples. In: 18th Global Joint Seminar on Geo-Environmental Engineering, 30–31 May 2019. Montreal: Concordia University.

Chickering

Krause

Townsend

(2018) Determination of as-discarded methane potential in residential and commercial municipal solid waste. Waste Management 76: 82–89.

Cho

Moon

Kim

(2012) Effect of quantity and composition of waste on the prediction of annual methane potential from landfills. Bioresource Technology 109: 86–92.

De La Cruz

Barlaz

(2010) Estimation of waste component-specific landfill decay rates using laboratory-scale decomposition data. Environmental Science & Technology 44: 4722–4728.

10.

Fleming

Mathison

(2013) Alternate method to determine the biochemical methane potential of solid wastes. In: International Conference on Solid Waste Technology & Management. Philadelphia.

11.

Francois

Feuillade

Matejka

, et al. (2007) Leachate recirculation effects on waste degradation: Study on columns. Waste Management 27: 1259–1272.

12.

Hansen

Schmidt

Angelidaki

, et al. (2004) Method for determination of methane potentials of solid organic waste. Waste Management 24: 393–400.

13.

Holliger

Alves

Andrade

, et al. (2016). Towards a standardization of biomethane potential tests. Water Science and Technology 74: 2515–2522.

14.

Hucl

(2021) Derivation of first-order decay heat generation function and prediction of energy potential for municipal solid waste at a cold climate landfill. M.Sc. Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.

15.

Ivanova

Richards

Smallman

(2008a). The long-term settlement of landfill waste. Proceedings of Institution of Civil Engineers: Waste and Resource Management 161: 121–133.

16.

Ivanova

Richards

Smallman

(2008b) Assessment of the anaerobic biodegradation potential of MSW. Proceedings of Institution of Civil Engineers: Waste and Resource Management 161: 167–180.

17.

Kaartinen

Sormunen

Rintala

(2013) Case study on sampling, processing and characterization of landfilled municipal solid waste in the view of landfill mining. Journal of Cleaner Production 55: 56–66.

18.

Kelly

Shearer

Kim

, et al. (2006) Relationships between analytical methods utilized as tools in the evaluation of landfill waste stability. Waste Management 26: 1349–1356.

19.

Koch

Hafner

Weinrich

, et al. (2019) Identification of critical problems in biochemical methane potential (BMP) tests from methane production curves. Frontiers in Environmental Science 7: 1–8.

20.

Krause

Chickering

Townsend

, et al. (2016) Critical review of the methane generation potential of municipal solid waste. Critical Reviews in Environmental Science and Technology 46: 1117–1182.

21.

Krause

Chickering

Townsend

, et al. (2018) Effects of temperature and particle size on the biochemical methane potential of municipal solid waste components. Waste Management 71: 25–30.

22.

Feng

Zhang

, et al. (2013) Influence of inoculum source and pre-incubation on bio-methane potential of chicken manure and corn stover. Applied Biochemistry and Biotechnology 171: 117–127.

23.

Mathison

(2015) Alternative methods for determining the biochemical methane potential of municipal solid wastes. M.Sc. Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.

24.

Morhart

(2022) Evaluation of low-grade geothermal energy recovery from a cold climate municipal solid waste landfill. M.Sc. Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.

25.

Owen

Stuckey

Healy

, et al. (1979) Bioassay for monitoring biochemical methane potential and anaerobic toxicity. Water Research 13: 485–492.

26.

Owens

Chynoweth

(1993) Biochemical methane potential of municipal solid waste and biosolids. Water Science and Technology 2: 1–14.

27.

Pearse

(2019) Biochemical methane potential of landfilled municipal solid waste using a non-slurry. M.Sc. Thesis, University of Calgary, Calgary, AB, Canada.

28.

Pearse

Hettiaratchi

Kumar

(2018) Towards developing a representative biochemical methane potential (BMP) assay for landfilled municipal solid waste – A review. Bioresource Technology 254: 312–324.

29.

Rao

Singh

, et al. (2000) Bioenergy conversion studies of the organic fraction of MSW: Assessment of ultimate bioenergy production potential of municipal garbage. Applied Energy 66: 75–87.

30.

Raposo

De La Rubia

Fernández-Cegrí

, et al. (2012) Anaerobic digestion of solid organic substrates in batch mode: An overview relating to methane yields and experimental procedures. Renewable and Sustainable Energy Reviews 16: 861–877.

31.

Raposo

Fernández-Cegrí

de la Rubia

, et al. (2011) Biochemical methane potential (BMP) of solid organic substrates: Evaluation of anaerobic biodegradability using data from an international interlaboratory study. Journal of Chemical Technology & Biotechnology 86: 1088–1098.

32.

Sel

Çakmakcı

Özkaya

, et al. (2016) Case study on prediction of remaining methane potential of landfilled municipal solid waste by statistical analysis of waste composition data. Waste Management 56: 310–317.

33.

Stevens

(2012) An empirical analysis of gas well design and pumping tests for retrofitting landfill gas collection. MSc Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.

34.

Stinson

Ham

(1995) Effect of lignin on the anaerobic decomposition of cellulose as determined through the use of a biochemical methane potential method. Environmental Science & Technology 29: 2305–2310.

35.

Usova

(2012) Numerical Analysis of Gas Flow within a Municipal Solid Waste Landfill. Saskatoon, SK: University of Saskatchewan.

36.

Zhu

Gikas

Zhang

, et al. (2009) Characteristics and biogas production potential of municipal solid wastes pretreated with a rotary drum reactor. Bioresource Technology 100: 1122–1129.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

14.63 MB

A method for evaluating and verifying biochemical methane potential test completion performed with landfilled municipal solid waste

Abstract

Keywords

Introduction

Methods

Description of MSW

BMP test procedures

Model convergence analysis algorithm and statistical analyses

Experimental and modelled BMPULT

Model convergence analysis

Convergence tolerances

Performance-based BMPULT determination

Results and discussion

BMP results and first-order decay modelling

Data requirements to model BMPULT

Verification of model convergence analysis

Initial determination of a performance-based BMP endpoint

Conclusions

Supplemental Material

sj-docx-1-wmr-10.1177_0734242X241227373 – Supplemental material for A method for evaluating and verifying biochemical methane potential test completion performed with landfilled municipal solid waste

Footnotes

Acknowledgements

Declaration of conflicting interests

Funding

ORCID iDs

Supplemental material

References

Supplementary Material

Experimental and modelled BMP_ULT

Performance-based BMP_ULT determination

Data requirements to model BMP_ULT