The implications of facility design and enabling policies on the economics of dry anaerobic digestion 1

10 Diverting organic waste from landfills provides significant emissions benefits in addition to increasing 11 landfill capacity and creating value-added energy and compost products. Dry anaerobic digestion (AD) is 12 particularly attractive for the organic fraction of municipal solid waste because of its high-solids 13 composition and minimal water requirements. This study utilizes empirical data from operational 14 facilities in California in order to explore the key drivers of dry AD facility profitability, impacts of 15 market forces, and the efficacy of policy incentives. The study finds that dry AD facilities can achieve 16 meaningful economies of scale with organic waste intake amounts larger than 75,000 tonnes per year. 17 Materials handling costs, inclusive the disposal of inorganic residuals from contaminated waste streams 18 and post-digester mass (digestate) management, are both the largest and the most uncertain facility cost. 19 Facilities that utilize the biogas for vehicle fueling and earn associated fuel credits collect revenues that 20 are 4-6x greater than those of facilities generating and selling electricity and 10-12x greater than facilities 21

. A set of facility operational modeling equations, described in Section 2.1, determine the mass 116 and energy flows of the waste as it moves through sorting and digestion, and the generated biogas as it 117 moves from creation (in the digester) to the final energy delivery. Key assumptions for this analysis are 118 shown in Table 1. The model assesses three pathways for biogas utilization: (1) on-site combined heat 119 and power (CHP) generation using biogas ("Electricity scenario"), (2) upgrading biogas to pure methane, 120 referred to as renewable natural gas (RNG) or biomethane, and compressing for on-site vehicle fueling 121 ("RNG Vehicle Fuel scenario"), or (3) upgrading biogas to RNG for natural gas pipeline injection with 122 unspecified end-use ("RNG Non-Vehicle scenario"). Cost and revenues are calculated for each step in the 123 model, as denoted by the red and green dollar symbols in Figure 1 and described in Section 2.2. Capital, 124 operational, and labor costs are calculated for the sorting and digestion facility, the gas upgrading or 125 energy conversion equipment, and the energy delivery infrastructure. Trucking and disposal costs for the 126 inorganic residuals and post-digester digestate are also modeled. Revenues for a given energy product are 127 calculated based on current California market and policy incentive landscapes, while the revenue 128 associated with incoming waste into the facility (i.e., the tipping fee) is not modeled directly but is instead 129 an output of the model. Key cost parameters are shown in Table 2.   Table 1 shows the modeled assumptions, including ranges representing uncertainty, for key operational 149 parameters, and Section 2 of the Supplementary Information provides more detail on parameters not 150 described here. The facility modeled in this study accepts source-separated organics (SSO) from 151 municipal residential and commercial sources and municipal yard waste. The level of contamination in 152 SSO was assumed to be 40%, as reported by currently operating facilities; the facility employs manual 153 sorting to remove 85% of the contamination, then mixes the SSO with yard waste in a 3-to-1 ratio (see 154 SI). We assumed an average biogas yield of 85 standard cubic meters of biogas per wet metric tonne of 155 waste (scm per tonne) and a range of 65 to 105 scm per tonne for the combined SSO and yard waste in the 156 digester. Our base performance and bounded assumptions are consistent with assumptions used in recent 157 literature (Tominac et al., 2021) as well as empirical data on dry AD system biogas yields: the SSFSC and 158 ZWEDC facilities in California reported 94-100 and 62-78 scm per tonne at, respectively, while  159 Angelonidi and Smith (2015) report 78-90 scm per tonne for facilities in Europe that process mixed food 160 and green waste and 50-106 scm per tonne for facilities with dry continuous systems. These biogas yield 161 ranges depend on various digester design parameters such as feedstock moisture content, operating 162 temperature, and retention time. We also assumed the generated biogas has a methane content of 55% 163 (range 50-60%) by volume based on empirical facility data from ZWEDC and others ( The residuals (i.e., inorganic contamination) separated from the waste once it reaches the AD facility 170 must be landfilled; we estimated the cost of disposing residuals based on typical landfill tipping fees in 171 California (see Table 2). California tipping fees are similar to the average tipping fees across the U.S. in 172 terms of both the average and the range of values (CalRecycle, 2015; Environmental Research &  173 Education Foundation, 2018). We assumed the cost of transporting the residuals to the landfill are 174 negligible, as the total tonnage of inorganic contamination hauled out is small compared to other inbound 175 and outbound tonnages, and digestion facilities are often co-located with landfills or other waste transfer 176 operations (e.g., as is the case for the observed facilities). A small portion of residuals will be recyclable, 177 but the monetary value of these materials, if any, is negligible based on our observations. 178 Digester capital costs were modeled using a standard exponential equation (see Table 2). The capital cost 179 equation for the base scenario was calibrated to the ZWEDC facility capital cost of $43.5M (see SI for 180 details, including comparison to other reported facility costs). A scaling factor of 0.7 was assumed; 181 though wet AD facilities exhibit a scaling factor of 0.6 (Sanscartier et al., 2012), previously reported costs 182 of dry batch AD facilities appear to scale more linearly with facility capacity (Angelonidi and Smith, 183 2015). This is likely due to the fact that dry AD facilities, particularly those with batch processes, operate 184 multiple identical reactors in parallel and therefore do not exhibit the same economies of scale as simply 185 increasing the size of a wet AD tank. We assumed annual operating costs (not including labor) of 5% of 186 capital cost based on ZWEDC facility data. Gas upgrading costs (for removal of CO 2 ) for the RNG 187 energy utilization scenarios take a linear form, based on combined capital and operational costs from 188 European data collected from 2007-2008 (Ong et al., 2014). Labor was separated into three categories 189 (i.e., overhead, operations, and sorting) that scale differently with facility size and operational parameters 190 (see SI for assumptions). Annual employment costs, including employer-paid benefits, were based on 191 California-specific wage data (see Table 2; see SI for details). 192

194
The AD process and subsequent short-term in-vessel composting cause a reduction in solid waste mass 195 (30%, according to ZWEDC tonnage reports) due to the transformation of solid mass into biogas as well 196 as loss of moisture. We modeled an arrangement where the facility pays to haul and dispose of digestate 197 at a third-party facility. For example, ZWEDC sends digestate to be composted off-site and other known 198 California AD facilities (i.e., SSFSC and MRWMD) also send their digestate for off-site management. 199 Base digestate trucking costs and tipping fees are based on ZWEDC's actual costs. The low-cost scenario 200 assumes that facilities are co-located with compost operations, so trucking costs are negligible and tipping 201 fees were modeled as the California statewide median for yard waste at compost facilities. This is a 202 reasonable lower bound given that digestate from facilities processing mixed organic waste will be more 203 heterogeneous and contaminated than yard waste, so fees would likely be higher to reflect the 204 management challenges posed by moisture, odor, and inorganic contamination (Cotton, 2019). 205

Electricity Generation 206
Assumptions regarding the combined heat and power (CHP) equipment efficiency and costs are described 207 in the SI. There are various arrangements under which dry AD facilities can sell electricity in California 208 for a premium (see compensation assumptions in Table 2). The highest price typically achievable is 209 through the state's Bioenergy Market Adjusting Tariff (BioMAT) program, which for municipal waste 210 digesters is currently $127 per MWh (CPUC, 2018). This price is multiplied by a seasonal-and time-of-211 day-varying factor that represents the value of the electricity to the grid, but we assumed facilities do not The RNG Non-Vehicle scenario assesses the economics of injecting RNG into the natural gas grid for 249 generic, untracked usage. In this case, the facility must invest in a pipeline interconnection station and 250 will receive revenues in line with wholesale natural gas prices. Sales were assumed to occur at wholesale 251 natural gas prices, which are typically $0.003-0.006 per MJ; in 2017 the California average selling price 252 was $0.003 per MJ (U.S. Energy Information Administration, 2019). There is currently no option for 253 monetizing the environmental benefits of pipeline-injected, end-use-agnostic RNG at a state or federal 254 level. In late 2018, the California legislature passed a bill authorizing the state utilities commission to 255 develop a biomethane procurement program, which in theory will raise the market value of pipeline-256 injected biomethane (California Senate, 2018). However, the scope and timeline of this bill is unknown 257 and therefore we did not include any above-market revenues for biomethane in this scenario. 258

Results and Discussion 259
Our results are reported on the basis of the levelized cost of disposal (LCOD), which can be thought of as 260 analogous to the often-used levelized cost of energy (Ayres et al., 2004). LCOD represents the per-tonne 261 tipping fee the facility would need to receive to achieve a net present value of zero (i.e., the facility earns 262 a rate of return equal to the discount rate). Figure 2

Competitiveness with landfills 316
Landfilling is the most prevalent OFMSW disposal alternative to AD, as compost facilities do not 317 commonly handle OFMSW due to contamination levels and concerns about odors and pests (Cotton,318 2019). Hence, the LCOD results in this study can be contextualized through a comparison with landfill 319 tipping fees adjusted to represent a lifetime average comparable to LCOD (denoted as LCOD-equivalent; 320 see SI for adjustment calculations). Tipping fees across California vary from $0 to 184 per tonne of waste 321 (LCOD-equivalent); zero values arise when landfills are county-owned and paid for through non-tipping-322 fee revenue mechanisms such as property taxes (see Figure S2)  landfill tipping fee equivalent, assuming waste intake greater than 100,000 tonnes per year ensures that 336 the Electricity and RNG Vehicle Fuel scenarios will be economically attractive. In more populated areas, 337 reaching this scale will be possible (even if waste is hauled no more than 20 km), while smaller cities and 338 rural areas will need to aggregate their wastes at centralized facilities . 339

Figure 4. Economic competitiveness of AD facilities with landfills across a range of capacities (x-axis) and energy scenarios
Although outside of the scope of this analysis, it is worth noting that source-separation and/or pre-344 processing required to reduce contamination levels to levels acceptable for dry AD can result in additional 345 costs. These costs may be in the form of separate bins and collection routes to facilitate increased source-346 separation or in processing at materials recovery facilities that are designed for high organics recovery 347 (e.g., using de-packaging machines). Some of the costs and benefits of diverting organic waste from

Uncertainty 357
The LCOD can vary by more than $60 per tonne within the range of cost and performance scenarios we 358 analyzed (see Figure 5). Unit costs drive the majority of the uncertainty, while the variations in facility 359 performance do not have as large an impact on LCOD. In all cases, the cost of managing inorganic 360 residuals and solid digestate is single largest source of uncertainty, accounting for a -$25 to $42 per tonne 361 variation in LCOD. If residuals and digestate costs were held constant at the Base Cost values, the overall 362 LCOD uncertainty range for 100,000 tonne-per-year facilities would shrink by 65%, 73%, and 46% for 363 the Electricity, RNG Non-Vehicle, and RNG Vehicle Fuel scenarios, respectively (see Figure S3 for 364 comparison). In the High Performance scenarios, costs increase due to the need to process (i.e., combust 365 or upgrade) additional biogas. In the Electricity and RNG Vehicle scenarios, the revenues from increased 366 energy production outweigh this cost, but in the RNG Non-Vehicle-High Performance scenario the 367 additional market gas revenues are smaller than the additional gas upgrading costs, resulting in a net 368 increase in LCOD. 369

375
Although a wide range of costs are covered in our uncertainty analysis, facilities outside of California 376 may have costs outside of that range, particularly labor costs in much of the U.S. would likely be lower 377 than our modeled range. (Bureau of Labor Statistics, 2019). Digestate management costs may also be 378 lower in areas with inexpensive land, where odor is not a concern, and with different regulations 379 governing the year-round composting or land application. It should also be noted that, while the LCFS is 380 specific to California (a number of other states do have variations of this policy in place), facilities outside 381 of California are eligible to earn LCFS credits if they sell into the California market. 382

Energy prices and policy impacts 383
The wholesale market value of energy outputs from AD facilities, absent any policy incentives, is a minor 384 source of revenue relative to what is required to offset facility costs, as highlighted in the RNG Non-385 Vehicle scenario where gas pipeline revenues offset 1-7% of facility costs. If the facility can earn retail 386 energy prices instead of wholesale prices, as is the case for the RNG Vehicle Fuel scenario, these energy 387 revenues can offset 4-37% of costs. However, participation in the vehicle fuel market is predicated upon 388 the ability to find reliable customers, ideally a fleet of medium-or heavy-duty vehicles, which may be 389 difficult for facility owners who do not own related businesses such as a waste hauling fleet, or facility 390 locations that are not close to a major road network. Adjacent investment decisions such as whether to 391 convert a trucking fleet from diesel to CNG are outside the scope of the model, but it is worthwhile to 392 note that the availability of CNG customers can factor into a facility's decision to pursue this option. In 393 the Electricity scenario, policies such as renewable feed-in tariffs or power purchase agreements valued 394 above wholesale prices can increase energy revenues, though the total value relative to the costs is at most 395 25%. The combination of the facility's ability to capture retail revenues paired with substantial state and 396 federal incentive programs make the RNG Vehicle Fuel scenario the lowest-cost option from an LCOD 397 perspective. However, these increased revenues are uncertain, varying by a factor of 5 across the cost 398 scenarios modeled, because fuel credits trade on an open market and therefore future prices will fluctuate. 399 The less money a facility earns through energy sales and related incentives, the more they must earn 400 through tipping fees in order to be financially viable. The relative importance of these two revenue 401 streams may impact the way the facility is built and operated. Figure 6 shows the share of total revenues 402 that comes from energy sales and energy-related incentives in each scenario assuming an operational 403 facility earns a tipping fee required to break-even (i.e., commensurate with the LCOD). Energy is 404 responsible for at most 7% of revenues in the RNG Non-Vehicle scenario, while the Electricity scenario 405 earns 15-25% of revenues from energy at the high end, depending on facility size. In these cases, the 406 facility would be less motivated to invest in improving energy generation processes such as optimizing 407 gas yield or reducing flaring, as their money would primarily come from waste intake. Conversely, if 408 energy is the dominant revenue source (up to 83% in the RNG Vehicle Fuel scenario), the facility may be 409 motivated to maximize energy output and become selective in the waste they accept in an effort to 410 generate as much gas as possible. This could limit the diversion opportunities for more contaminated or 411 difficult-to-handle streams such as mixed municipal solid waste. Energy and waste policy planners should 412 carefully consider the prices being offered from various sources and what it will mean for both waste 413 disposal costs as well as drivers for facility operation and investment. Additionally, regulations could be 414 put in place on the way facilities operate (e.g., minimum retention times, best practices for percolate 415 circulation) and acceptable waste streams in order to ensure that facilities being supported by public 416 policies and money are operating in a way that maximizes their social benefits. 417 New energy-related value streams could be considered to ensure that AD facilities are incentivized to 418 maximize their energy and emissions benefits. For example, electricity-generating facilities offer a 419 dispatchable form of renewable energy and therefore could be incentivized to follow specific dispatch 420 schedules that help meet peak demand and ramping needs of the grid. The current BioMAT program in 421 California accomplishes this to a limited extent through time-of-day modifying factors. The same grid 422 benefits could be achieved by injecting RNG into natural gas pipelines for use at off-site facilities, but this 423 scenario is currently the least economically viable. New monetization mechanisms for RNG used in non-424 vehicle fuel purposes could open up opportunities to reduce the carbon footprint of existing natural gas 425 power plants and decarbonize industrial processes that are too difficult or expensive to electrify, either 426 through co-location with an AD facility for direct biogas combustion and/or heat recovery from CHP 427 units or through direct sales via the natural gas grid. These added revenues would also lower the LCOD of 428 AD facilities and make them economically viable at lower tipping fees. 429  Economies of scale are important to the overall LCOD. The largest facilities we modeled (i.e., 300,000 442 tonnes per year) generally had LCOD values that were 55-70% those of the smallest facilities we modeled 443 (i.e., 25,000 tonnes per year). However, larger facilities may face barriers not explicitly considered in this 444 study such as difficulty obtaining feedstocks from local sources, odor and emissions management issues, 445 and resistance from neighboring communities. Materials handling costs, namely the disposal of inorganic 446 residuals that come into the facility and the management of post-AD digestate by third-party composters, 447 vary considerably and have the potential to be well over half of the total per-tonne costs incurred by a 448 facility. However, some of the uncertainty can be mitigated through materials contracts and holistic waste 449 management policy support. 450 RNG for use as a vehicle fuel is currently the most lucrative energy utilization pathway for AD facilities 451 due to existing U.S. Federal and state policy incentives. Upgrading the biogas to RNG for non-vehicle 452 fuel uses is not economically attractive, as the upgrading costs alone outweigh wholesale natural gas 453 revenues and no economic incentives currently in place to support the production of RNG for non-vehicle 454 applications. Lastly, electricity generation is an economically viable pathway in cases where the 455 alternative landfill tipping fees are high, and may be attractive for facilities that do not operate truck fleets 456 capable of utilizing RNG and cannot easily connect to natural gas pipelines. Advanced electricity dispatch 457 strategies and monetization of thermal energy from combined heat and power units could help make this 458 scenario more attractive. 459 Limitations of the study include significant focus on the physical, financial, and operational 460 characteristics of a specific facility in California and use of deterministic inputs as opposed to an 461 operational framework that determines facility size, operations and costs to achieve a specific financial or 462 operational objective (e.g., least cost, specific return on investment). Dry AD technology is still quite 463 nascent in the U.S. and we have limited empirical data to draw from for inputs and assumptions. While 464 this study used bounding assumptions to represent uncertainty, public data on dry AD facility financial, 465 operating, and production characteristics would generate more precise results. Opportunities for future 466 technoeconomic analysis of dry AD facilities include additional consideration of byproduct management 467 pathways such as novel composting methods, land application of raw material, or pyrolysis of digestate, 468 as well as any associated revenues from the sale of finished compost or biochar (a byproduct of 469 pyrolysis). Future studies could also quantify the benefits of various levels of RNG-and thermal energy 470 recovery-related economic incentives and opportunities for co-location with other industrial facilities that 471 can utilize a range of energy byproducts. As more dry AD facilities are built and more data becomes 472 available, research should further explore the impact of facility design parameters such as digester 473 residence times, gas storage capacity, and feedstock composition on the costs and benefits, as well as the 474 societal impacts, of dry AD. 475

Acknowledgements 476
The research for this paper was financially supported by the California Energy Commission under 477 agreement number EPC-14-044 and EPC-14-030. We thank Greg Ryan, John Pena, Amelin Norzamini, 478 Osvaldo Cordero, and Prab Sethi for their input. This work was also part of the DOE Joint BioEnergy