Greenhouse gas emissions from windrow composting of organic wastes: Patterns and emissions factors

The experiment was conducted at the West Marin Composting Facility in Nicasio, California, from February through September 2016. A windrow pile (15 m × 2 m × 4 m) comprised of manure (mostly cattle, with goat, horse, chicken) and yard waste (branches, leaves, grass) was composted. The starting pile was 55% manure by volume and 91% manure by mass. The pile was managed as a commercially-produced compost pile: with weekly turning using a large-scale mechanical windrow turner (Scarab 18), periodic watering, and a 98 d duration.

Conditions inside the piles were monitored continuously using 27 automated sensors. Nine O₂ (Apogee SO-110), temperature (Campbell 107), and moisture (Campbell CS616) sensors each were placed in three transects along the length of the pile, at three heights (0.50 m, 1 m, 1.5 m), in the center of the pile (figures S1-2). Sensors were removed briefly for weekly pile turning (<60 min). Compost grab samples were collected weekly into 1 gallon Ziploc bags from three heights (0.5 m, 1 m, 1.5 m—at 2 m horizontal depth) before and after pile turning (n = 6). Samples were analyzed for gravimetric moisture content by weighing 10 g samples before and after drying at 105° C for 24 h. Compost pH was measured by suspending 3 g of sample in 12 g of water [21]. We estimated porosity by weighing a mason jar with a volume of 100 ml of collected compost, adding 100 ml distilled water, and re-weighing to determine the volume of the pore space [22] (n = 5 per sample). Note that collecting samples for porosity may result in disturbance to the sample and thus not represent exact in situ conditions. Potential net N mineralization and nitrification were determined using dark laboratory incubations; 3.5 g compost samples were extracted before and after incubation (7 d) in 75 ml of 2 M KCl (n = 3 per sample) [23]. Total C and N were measured on air-dried, sieved (2 mm) and ground samples using an elemental analyzer (Carlo Erba Elantech, Lakewood, NJ, USA).

The pile was outfitted with a greenhouse gas and wind measurement system. Four gas towers (figure S2, A–D) were placed at cardinal directions along each edge of the windrow, 1 m from the pile edge; each of these towers held four gas intakes (heights of 0.25, 1, 2, and 3.5 m). Air samples were continuously drawn through 16 Teflon tubes, delivering gas samples to a G2308 cavity ring down laser spectrometer (Picarro, Santa Clara, CA), which measured real-time CO₂, N₂O and CH₄ concentrations. The instrument was calibrated using a three-point calibration curve, using zero (pure N₂ gas), intermediate (304 ppm CO₂, 0.513 ppm N₂O, 1.05 ppm CH₄), and high (1000 ppm CO₂, 10 ppm N₂O, 10 ppm CH₄) standard concentrations. The instrument sensitivity is high for CH₄ and N₂O (raw precision <10 ppb and <25 ppb, respectively), but low for CO₂ (<20 ppm). While one intake port delivered gases to the greenhouse gas analyzer (1 min per intake), the other 15 lines were flushed with ambient air.

Two wind towers on the NW and SE corners of the windrow held four 3D sonic anemometers (Gill Windmaster Pro) each, at four heights (0.25 m, 1 m, 2 m, 3.5 m), measuring wind speed and direction at 0.1 Hz. The high frequency wind and gas concentration data were combined to measure greenhouse gas concentrations upwind and downwind of the pile, and calculate the flux of CO₂, CH₄ and N₂O from composting, using MMB [11, 24]. The flux (g m⁻² s⁻¹) from the source was approximated:

$$\begin{eqnarray}&&{\rm{Flux}}=\displaystyle \frac{1}{{L}}\,\displaystyle {\int }_{0}^{\infty }{\bar{{u}}}_{{z}}\left({\bar{{c}}}_{{z},{d}}-{\bar{{c}}}_{{z},{u}}\right)\,{\rm{d}}{z},\end{eqnarray} \tag{ 1 }$$

where L (m) is the fetch (horizontal distance that air travels over the source), ${\bar{{u}}}_{{z}}$ represents mean horizontal wind speed at height z (m s⁻¹), and ${\bar{{c}}}_{{z},{d}}$ and ${\bar{{c}}}_{{z},{u}}$ are the mean gas concentrations (g m⁻³) at height z downwind and upwind of the source. The flux describes the mass of trace gas emitted per unit time, per cross-sectional area of the source (the windrow pile).

Statistical analyses were performed using the open-source statistical software 'R.' Sensor readings were logged every half hour for the duration of the experiment. Two-way, repeated measures ANOVA was used to test the null hypotheses that sensor values did not change by week or position (top, middle, or bottom of pile), and to test variability of weekly physical and chemical properties of the composting material (pH, porosity, N mineralization, nitrification, C:N, moisture).

Wind speed and direction, and gas concentrations were averaged in 8 min blocks, and fluxes were calculated using equation (1), after screening for wind direction. An angle allowance of $\pm$30 degrees was used to determine upwind and downwind towers for each time block [9]. Relationships between CH₄ fluxes and environmental variables (temperature, moisture, and O₂) were examined using breakpoint analyses (R package 'strucchange'). The number of breakpoints (1, 2 or 3) with the lowest Bayesian Information Criterion (BIC) was selected; breakpoint significance was determined using a structural change test [25].

Emissions factors were estimated by calculating the area under the mean daily flux curves, for each greenhouse gas, using the R package 'pracma'; we multiplied this integral (g m⁻²) by the pile volume (m³) divided by the fetch (m) to estimate the total emissions. For CO₂ fluxes, we additionally perform a mass balance on C, using laboratory analyses of C content in the composting material over the course of the experiment. We compared our emissions factors to those in the literature, by performing a meta-analysis of published studies that measured greenhouse gas emissions from composting.

To estimate net emissions from composting, we created a lifecycle model, using a functional unit of 1 kg of waste (91% manure, 9% green waste by mass) over one year, employing a system boundary beginning at waste generation, and ending at its final application (land or landfill; figure S3). We simplified the model by including only major contributors to greenhouse gas emissions from waste management: avoided emissions, waste processing, and carbon sequestration from land application [26, 27]. To estimate avoided emissions, we assumed that manure would otherwise be managed in an anaerobic lagoon, a common practice in California [2], and yard waste would be used as alternate daily cover in a landfill [28]; emissions values were taken from Owen and Silver (2015) [2] (manure), and EPA's WARM model (yard waste) [29]. Carbon sequestered from compost application are from Ryals et al (2013) [30]. We assumed that 1 kg wet waste produced 0.5 kg compost.

Temperature varied over time and by position in the compost pile, and followed a pattern consistent with microbial degradation of organic feedstocks [31]. Temperatures rose rapidly initially, then decreased as easily-decomposed substrates were consumed (figure 1(a)). Within that pattern was a weekly oscillation, corresponding to turning events: temperature rose steeply following turning, then declined. The lowest temperatures occurred at the start (day 0), and the highest daily mean temperatures were reached on day 8 (74 ± 0.2, 71 ± 0.2, 68 ± 0.1 °C, for top, middle and bottom positions). Temperature varied significantly between weeks (p < 0.05) but not by position in the pile (table S2). Variability of temperature over time is well established in the literature [6, 32]; the composting process is understood as a series of discrete sub-processes (e.g. mesophilic, thermophilic) due to the discontinuity of shifts in temperature and in microbial communities [33]. Variability by location in the pile was observed in another study [34], but was not observed here, possibly due to the high relative lability of manure.

Zoom In Zoom Out Reset image size

Moisture affects the composting rate and end-product characteristics. Model estimates and field data suggest that wet-weight moisture contents around 50% are best for O₂ diffusion, water potential, and microbial growth rates during composting [35]. Moisture in the pile varied in response to water consumption and evaporation, and were controlled to keep conditions favorable for decomposition (figure 1(b)). Daily mean moisture varied between 36% ± 0.01% and 56% ± 0.04%. All positions showed moisture increases with pile turning and watering (figure 1(b)). Moisture differed significantly between weeks (p < 0.001), but not by position (table S2).

Oxygen concentrations affect microbial activity, nutrient mineralization, and greenhouse gas dynamics during composting [36]. In this experiment, O₂ concentrations followed an oscillating pattern, with the lowest daily mean values occurring near the beginning and end of the experiment, and highest daily means occurring in the middle (figure 1(c)). Similar patterns have been measured in composting bioreactors [36], while field scale data are highly variable and dependent upon feedstock and management [37]. Daily mean O₂ concentrations ranged from 0.95% to 15% over the 98 d study. The lowest daily mean values were 1.6 ± 0.06% (day 65), 2.3 ± 0.09% (day 87), and 0.95 ± 0.37% (day 0), and the highest daily mean values were 15 ± 0.29% (day 19), 15 ± 0.09% (day 48), and 14 ± 0.09% (day 34), for the bottom, middle and top positions. O₂ level varied significantly by week (p < 0.05), and by the interaction between week and position (p < 0.05) (table S2). Estimated weekly mean O₂ consumption rates ranged from 5.3 ± 0.002 mg O₂ m⁻² s⁻¹ (week 1) to 0.39 ± 0.000 03 mg O₂ m⁻² s⁻¹ (week 13). Declining O₂ consumption is consistent with slower microbial decomposition, and declining availability of labile material.

The average pH values of the composting material ranged from 7.6 ± 0.07 to 9.3 ± 0.01. pH can be an important predictor of N mineralization rates and denitrification potential. However, N mineralization tends to be most sensitive at pH 7 or lower [38, 39]. pH values did not vary significantly by position, week, or whether samples were taken pre- or post-mixing (table S3a). Porosity varied between 0.59 ± 0.02 and 0.84 ± 0.01, and varied significantly by week (p < 0.001), position (p < 0.001), and the interaction between week and position (p < 0.05) (table S3b). Porosity provides an index of physical limitations on O₂ diffusion [40, 41]; its value varied over space and time, reflecting feedstock heterogeneity.

The C:N ratio of composting material is an indicator of microbial degradability and the maturity of the composting process [6], and may predict CO₂ and N₂O emissions [20, 42]. The C:N ratio declined throughout the experiment, mostly due to lowering C concentrations. This decline is consistent with a well-functioning composting process [6]. The C:N ratio varied significantly by week (p < 0.001), and by the interaction between position and state of mixing (pre- or post-mixing) (p < 0.05) (table S4). The feedstock had initial an C:N of 26 ± 0.7 for all positions, consistent with best management practices (25–30) [6], and final values near 20 (21 ± 0.7, 21 ± 0.6, and 19 ± 0.6, for the bottom, middle, and top). Microbes preferentially utilize available forms of C in feedstocks, and emit CO₂ (as they metabolize aerobically), or CO₂ and CH₄ (anaerobic decomposition) [6].

Potential net nitrification remained near zero throughout the experiment (table S5). Daily mean values ranged between −0.20 ± 0.06 and 0.02 ± 0.01 mg N g⁻¹ compost, and weekly means varied from −0.14 ± 0.04 to 0.01 ± 0.06 mg N g⁻¹. Potential net nitrification varied significantly with the interaction between position and week, with higher net NO₃⁻ consumption in the top position during the second half of the experiment (p < 0.05). The low rates of nitrification measured are consistent with high temperature inhibition of nitrification [43, 44].

Potential N mineralization values also bracketed zero, with daily means ranging from −0.15 ± 0.06 to 0.34 ± 0.02 mg N g⁻¹ compost, and weekly means varying from −0.11 ± 0.04 and 0.22 ± 0.08 mg N g⁻¹ compost. Mineralization declined throughout the experiment; values differed significantly by week (p < 0.05) (table S5). The highest compost N mineralization rates tend to occur during the thermophilic phase [43, 44].

Methane fluxes varied over time, with most emissions occurring over short periods early in the composting process. Daily mean CH₄ fluxes ranged from 17 μg CH₄ m⁻² s⁻¹ to 4.4 mg CH₄ m⁻² s⁻¹. The highest CH₄ emissions occurred during the first three weeks of composting, during which 75% of total CH₄ was emitted. Half of the total CH₄ emissions occurred in the first 10 d, and 36% occurred in the first week (figure 2(a)). Most fluxes measured were small, with few days contributing the majority of CH₄ emissions. Integrating the daily mean flux curve, the total CH₄ emitted was 1.7 ± 0.32 g CH₄ kg⁻¹ wet feedstock² . This emission factor falls between the median (1.4 g CH₄ kg⁻¹), and mean (2.2 g CH₄ kg⁻¹) literature values found for windrow composting (figure 3, table S1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Carbon dioxide fluxes were also variable. Daily mean CO₂ fluxes ranged over two orders of magnitude (0.3–30 mg CO₂ m⁻² s⁻¹) and were an indicator of the C lability of the feedstock. Similar to CH₄, most CO₂ emissions occurred early, with almost 50% emitted in the first three weeks (figure 2).

Because CO₂ emissions from biological systems are considered to rapidly cycle, these biogenic emissions are usually considered to have no net global warming effect [26, 45–48]. Due to a low sensitivity of the analyzer to CO₂, the net CO₂ fluxes measured (19 ± 3.7 g CO₂ kg⁻¹ wet feedstock³ ) likely present an underestimation of this flux. Utilizing laboratory measurements of C content of the composting material over the course of the experiment to perform a C balance on the composting system, we found an upper limit on these emissions to be 100 g CO₂ kg⁻¹. The upper and lower estimates are within the range of the CO₂ emissions factors found in the literature, as is the high proportion of C loss via CO₂ versus CH₄ (96% to 4% for the upper emissions factor in this study; other studies found even higher proportions lost as CO₂) [49, 50]. From a greenhouse gas mitigation perspective, it is preferable that the C emitted from composting be emitted as CO₂ rather than as CH₄.

Nitrous oxide (N₂O) concentration differences remained below the instrument detection limit (25 ppb + 0.05%), likely reflecting an inability to detect a small signal using micrometeorological methods. The equivalent fluxes of this detection limit (45 μg m⁻² s⁻¹) were lower than those observed in the literature (table S1; figure S4). Low emissions may be explained by the high C:N measured, above the threshold (25:1) associated with N₂O emissions [20] for the first four weeks (and remaining above 20 throughout), during which decomposition is occurring most rapidly. Additionally, as NO₃⁻ is a precursor to N₂O, the low NO₃⁻ concentrations and nitrification rates measured throughout support the lack of N₂O production. Periodic anaerobic conditions and high temperatures likely inhibited nitrification.

Methane was the major contributor to non-biogenic greenhouse gas emissions from composting, and the majority of that CH₄ was emitted early in the composting process. The timing of emissions suggests that efforts to further decrease direct emissions from composting should focus on increasing O₂ availability in the early stages, when decomposition and C fluxes (as CH₄ and CO₂) are at their highest.

We identified 17 studies with 53 estimates of total emissions (emissions factors) from composting [5, 7, 10, 36, 49–61]; the studies spanned different composting processes and methods for emission measurement, feedstocks, and time scales (figure 3; table S1). Most studies used static chamber measurements, which yield periodic measurements of greenhouse gas concentrations at the surface of the composting pile. In the spatially heterogeneous and dynamic composting pile, using static chambers can underestimate fluxes. The measurements reported in the literature varied in duration, from 21 to 150 d. Most studies measured the composting of food and/or green waste; three studies [50, 52, 58] considered manure. The current study used methods to continuously measure whole-pile emissions, exceeded 90 d in duration, and used a high-emitting feedstock (manure); thus, we would expect emissions measured to exceed most of those in the literature. That the resulting emissions factor is near the literature's median value suggests that management may play a larger role than feedstock in predicting greenhouse gas fluxes.

Environmental variables in the composting pile—O₂, moisture, and temperature—exhibited threshold effects on CH₄ emissions (figures S5–S7). The breakpoints (thresholds) for CH₄ emissions occurred at O₂ concentrations of 3.3%, 5.7%, and 6.0% (p < 0.005); these represent O₂ levels where the modeled relationship between O₂ and CH₄ change [25], with higher emissions occurring below, and lower emissions occurring above those thresholds (figure S5).

In low oxygen environments, CH₄ emission is controlled by the balance between methanogenesis and methane oxidation. Methanogenesis is favored at low O₂ concentrations. Methane oxidation rates vary as a function of CH₄ production, O₂, moisture, pH, temperature [62–67], and feedstock chemistry [15, 68] and are limited by diffusive transport [62]. While porosity, a physical control on O₂ diffusion into the pile, declined over time in our experiment, so did CH₄ emissions, suggesting that other factors, including declining temperatures and C availability, played a larger role in controlling emissions than did constraints on O₂ diffusion. As expected, we observed the highest CH₄ emissions at lower O₂ concentrations, coupled with higher moisture, temperatures, and C:N.

Methane fluxes were found to increase more rapidly with temperatures above 65^o C (p < 0.1) (figure S6). The moisture threshold was found to be 55% (p < 0.0001); CH₄ emissions increased more rapidly with moisture as levels exceeded 55% (figure S7). Moisture between 40% and 60% is recommended for composting [6]; higher levels limit O₂ diffusion into the pile [62], while lower levels preclude microbial activity. The patterns of greenhouse gas emissions from composting can also be sensitive to key biogeochemical characteristics, particularly the relative availability of C and N in the decomposing feedstock. However, here we found no clear predictive (linear or polynomial) relationships between CH₄ emissions and other chemical or physical characteristics.

Carbon concentrations play an important role in understanding the emissions patterns; the biggest fluxes occur early in composting, and a disproportionate fraction of total emissions occur during the first three weeks, when the feedstock is fresh, and the organic C and N are likely more available to microbes. As composting progresses, the greenhouse gas fluxes decrease as the C and N become more complexed. During the beginning weeks of composting, as microbial degradation speeds up, temperatures and O₂ consumption were at their highest; these conditions favor methanogenesis.

Total direct greenhouse gas emissions from composting were 57 g CO₂-e kg⁻¹ (75–150 g CO₂-e kg⁻¹ including biogenic CO₂). Direct emissions are one important, and the most uncertain, element of the lifecycle greenhouse gas emissions from composting. Two other major C fluxes include the avoided emissions—emissions that would have occurred under 'business as usual' waste management—and enhanced C sequestration from land application of compost. Using a system boundary that includes these three processes (figure S3), avoided emissions from the landfill (for green waste) and anaerobic lagoons (for manure) outweigh the direct emissions from composting (table 1). The net lifecycle emissions from composting manure and green waste and applying the compost to grasslands are negative (−690 g CO₂-e kg⁻¹ excluding biogenic CO₂; between −590 and −670 g CO₂-e kg⁻¹ including the directly measured biogenic CO₂ fluxes)⁴ , meaning that emissions are avoided on net. If all of California's dairy manure that is currently managed via anaerobic lagoon (7.4 MMT yr⁻¹) [69] were instead composted, approximately 5MMT CO₂-e yr⁻¹ would be avoided; this is equivalent to 15% of California's agricultural greenhouse gas emissions.

The biggest greenhouse gas benefit from composting is the elimination of anaerobic storage. Emissions avoided from anaerobic storage were greater than the direct emissions from composting. Though annual C sequestration from compost application is smaller in magnitude than direct and avoided emissions, its effects are potentially long-lasting. Avoided and direct emissions are a one-time phenomenon, while increased C sequestration from a single compost application can persist for several years [70]. Thus, while we modeled net C implications over one year (table 1), a longer time frame would increase the relative benefits of compost application.

Our results highlight the potential to effectively measure and minimize greenhouse gas emissions from commercial-scale composting. Managing composting piles to minimize methanogenesis—by maintaining sufficient O₂ concentrations through aeration and bulking, and focusing on the first three weeks of decomposition when temperatures and decomposition rates are high—could potentially reduce direct greenhouse gas emissions. Low emissions from composting contribute to the climate change mitigation benefit of diverting high-emitting waste streams to compost and land applying for C sequestration.