Ex-ante life cycle assessment of commercial-scale cultivated meat production in 2030

Purpose Cultivated meat (CM) is attracting increased attention as an environmentally sustainable and animal-friendly alternative to conventional meat. As the technology matures, more data are becoming available and uncertainties decline. The goal of this ex-ante life cycle assessment (LCA) was to provide an outlook of the environmental performance of commercial-scale CM production in 2030 and to compare this to conventional animal production in 2030, using recent and often primary data, combined with scenario analysis. Methods This comparative attributional ex-ante LCA used the ReCiPe Midpoint impact assessment method. System boundaries were cradle-to-gate, and the functional unit was 1 kg of meat. Data were collected from over 15 companies active in CM production and its supply chain. Source data include lab-scale primary data from five CM producers, full-scale primary data from processes in comparable manufacturing fields, data from computational models, and data from published literature. Important data have been cross-checked with additional experts. Scenarios were used to represent the variation in data and to assess the influence of important choices such as energy mix. Ambitious benchmarks were made for conventional beef, pork, and chicken production systems, which include efficient intensive European animal agriculture and incorporate potential improvements for 2030. Results and discussion CM is almost three times more efficient in turning crops into meat than chicken, the most efficient animal, and therefore agricultural land use is low. Nitrogen-related and air pollution emissions of CM are also lower because of this efficiency and because CM is produced in a contained system without manure. CM production is energy-intensive, and therefore the energy mix used for production and in its supply chain is important. Using renewable energy, the carbon footprint is lower than beef and pork and comparable to the ambitious benchmark of chicken. Greenhouse gas profiles are different, being mostly CO_2 for CM and more CH_4 and N_2O for conventional meats. Climate hotspots are energy used for maintaining temperature in reactors and for biotechnological production of culture medium ingredients. Conclusions CM has the potential to have a lower environmental impact than ambitious conventional meat benchmarks, for most environmental indicators, most clearly agricultural land use, air pollution, and nitrogen-related emissions. The carbon footprint is substantially lower than that of beef. How it compares to chicken and pork depends on energy mixes. While CM production and its upstream supply chain are energy-intensive, using renewable energy can ensure that it is a sustainable alternative to all conventional meats. Recommendations CM producers should optimize energy efficiency and source additional renewable energy, leverage supply chain collaborations to ensure sustainable feedstocks, and search for the environmental optimum of culture medium through combining low-impact ingredients and high-performance medium formulation. Governments should consider this emerging industry’s increased renewable energy demand and the sustainability potential of freed-up agricultural land. Consumers should consider CM not as an extra option on the menu, but as a substitute to higher-impact products.


Appendix A -Overview of previous Life Cycle Assessments about cultivated meat
A few unique studies have been done to date to assess the (projected) environmental impacts of CM production: • Tuomisto and Teixeira de Mattos (2011) assess a hypothetical large-scale production system that uses cyanobacteria hydrolysate as main input for culture media.
Three different countries of production are considered.
• Tuomisto et al. (2014) assess a few hypothetical large-scale production systems with cyanobacteria, wheat and corn hydrolysate as main input for culture media.
Worst and best case scenarios are considered.
• Mattick et al. (2015) assess a hypothetical mature production process with detailed modeling based on Chinese Hamster Ovary (CHO) cell cultures. Hydrolysate as well as defined basal media are included.
• Tuomisto et al. (2022) assess a bench-scale production process for CM in hollow fiber bioreactors using experimental data from mouse myoblast C2C12 cells.
Various scenarios are included, among which for different cell metabolism, different media formulations and different energy sources.
Building on these unique studies, a couple of additional (scenario) analyses have been done (Smetana et al. 2015(Smetana et al. , 2018Lynch and Pierrehumbert 2019). These have also been included as far as possible in this appendix.

Appendix B -Complete impact assessment results for CM and conventional meats
The impact assessment results for the CM baseline model in the three energy scenarios is provided in

Appendix C -Acknowledgement of organizations that contributed with data and/or cross checks of data ranges
In Table C.1, the organizations that agreed to be acknowledged are summed up. A few organizations did not want to be acknowledged or did not respond, for various reasons.

Main process design parameters values, sources and data quality
The main process design parameters are provided in  Table D.6. Some quantitative data on model inputs is confidential and therefore not included, for example the estimated energy use for recombinant protein production. Production of meat per production run 3,080 kg Calculated, based on production line presented in Specht (2020) and projectspecific parameters

-No
Number of production runs needed for producing 10 kton 3,243 Calculated (annual production/productio n in one production run)

Data collection procedure for culture medium usage
Data collection for the culture medium scenarios took place over the course of 2019 -2022. There were two main rounds of data collection (the first in 2019 -2020 and the second in 2022), which were followed by additional communication with individual parties for verification or better understanding of data.

First round of data collection
In the first round of data collection, the following specific questions were asked regarding production of meat from a specific amount of medium. In all cases we explicitly asked for answers regarding the current situation (in that case 2019), and the anticipated situation 10 years into the future.
• What is the estimated output of wet cellular mass relative to the volumetric capacity of your largest proliferation vessel (kg/L)?
o What is the %dry mass in this output?
o What total volume of medium per batch is used to create this mass?
• What culture medium do you expect will be used in the future/are you aiming for (e.g. basal medium, hydrolysate, yeast extract, premix based on a variety of feedstocks)?
o What do you expect to be the quality grade requirements (pharma, food or feed grade)?
o What do you expect to be the main feedstocks and production processes for medium production?
o If you expect it to be a combination of ingredients from different sources, please list the percentages of these ingredients (e.g. xx% basal medium, xx% yeast extract and xx% foodgrade single ingredients) • Do you expect your feed conversion efficiency to go up with increase in production scale? o Please list any metabolites (e.g. lactate, ammonia) you are currently measuring during your bioprocess.
• Does your medium contain any growth factors or recombinant proteins (including insulin, transferrin, etc)? Please list on a level of aggregation that you feel comfortable with (total --> compound specific).
In the end, it is important that we gain insight into the total quantity/volumes and costs per kg/ton final product. A higher level of detail means a more accurate analysis.
o Are any of these produced in-house? Please list on a level of aggregation you are comfortable with.
o If so, what are the cost savings per quantity of protein?
o If you are able to say, are any growth factors used species-specific? If so, has there been any efficiency gain in a measured parameter (e.g. proliferation, differentiation, etc) using the species-specific protein? Please list any information here.
Most parties filled in most of the questions, to varying degrees of accuracy. In addition, a data sheet was included and (partially) filled in by multiple parties, which resulted in additional insights in media composition and additives.
The parties indicated they currently used DMEM/F12, in cases supplemented with various elements (recombinant proteins, amino acids, salts, growth factors). Based on interviews with (industry) experts, it is assumed that DMEM/F12 is not used in the future, but instead the medium components are delivered to the factory as a dry powdered mix (DPM). All parties indicated serum-free media would be the standard in the future and most parties indicated they were aiming to include hydrolysates.
In order to calculate the total amount of ingredients in the DPM, the primary data from companies was translated to its individual ingredients. Three medium scenarios were defined based on the range of data received (see CE Delft, 2021). The mid-medium scenario was compared to the scenarios presented in Specht (2020) and the amount of medium was found to be between the low-and average media use scenarios from Specht (2020).
The report was sent for checking to the parties involved before publishing, which did not result in any changes.

Second round of data collection
For the second round of data collection, the same parties from the first round were contacted to provide information for an update of the medium scenarios. Two reasons for this were that a few years had gone by (delivering additional insights for medium use, also at slightly larger scales than bench-top), and additional analysis showed that the low medium scenario from CE Delft (2021) resulted in cell mass with a relatively low dry matter (dm) content, which could not be generalized across different cell types and was too low to be comparable to conventional meat.
The approach for data collection was different in the second round. A table with three scenarios was presented to the parties (Table D.7). These scenarios were largely similar to those presented in CE Delft (2021), but the low medium scenario was adapted to match the 'enhanced catabolism cell types'-medium scenario in Humbird (2021), corrected for a lower dm content (20% instead of 30%).
General and specific questions were asked to the parties. The specific questions relate to the data provided by individual parties and are therefore excluded here due to confidentiality.

General questions:
Our questions are regarding estimated culture medium usage for the year 2030. The table below shows the scenarios that we developed based on received data. Please note that the unit is in g/kg meat cells (excluding scaffold, so 100% meat cells).
• Regarding amino acids and sugars (note: these values are higher than used in the current report, because the current numbers could not be generalized across different cell types and therefore we used literature data to define the lower range): o Is the amount of amino acids what you would expect?
o Is the amount of sugars what you would expect?
o If for both, lower amounts than the 'low' scenario are possible, are you able you substantiate with data or calculations?
• Do you expect albumin will be used?
• Does anything seem off in the values/ranges from your perspective? Some important conclusions that came from this second round of data collection were: • Amino acids quantities were within range of expectations • Sugars quantities were slightly below expectations, and were expected to be higher than amino acids quantities (even when accounting for the fact that when hydrolysates are used, higher quantities of amino acids will be needed than when only single amino acids are used) • Some experimental data was provided to substantiate lower medium use than the low medium scenario, but this could not be generalized across cell types and it was uncertain whether this could also be achieved at larger scales than lab-scale • Albumin was unlikely to be included in future medium formulations for most parties, and already excluded currently for some parties • Concentrations of solids were too high, resulting in a medium too viscous for cell culture, and therefore additional water would be needed (although scenarios were reported that indicated certain fed-batch feeding strategies could reduce the amount of water in the medium needed down to around 20L) • Total minimum amount of solids needed is 600 -800 g/kg meat cells The additional information was used to determine new low, mid and high medium scenarios (see main paper, Table 1).
A mass balance check was performed for the three medium scenarios and for the upper and lower bound of expected protein content in the final product, to check whether more carbon was supplied than produced and what the carbon consumption efficiency would be in the different scenarios (Table D.8). Average carbon contents of amino acids, glucose and protein of 41%, 40% and 51% respectively were used for the calculations.
This verified that in theory, ample carbon is supplied to produce the required cell mass. The amounts of ingredients needed for 1 kg of meat cells was also compared to Humbird (2021) and Tuomisto et al. (2022) (Table D.9). In general, the consumption of amino acids and glucose will be dependent on a cell's metabolism, which is known to vary between species and cell type (O'Neill et al. 2022). Despite this variability, both the amount of amino acids and of glucose in this study fall within the range of the compared studies, which model data from CHO and C2C12 cells, respectively. The amino acid consumption in the baseline (mid medium) scenario in this study and in the scenarios including hydrolysates in Humbird (2021) are slightly higher than in the optimized scenarios of Tuomisto et al. (2022). This could be explained by the fact that Tuomisto et al. (2022) use defined media made with single amino acids, and no hydrolysates (DMEM/F12 + FBS or Essential 8). In these defined media, the ratio of amino acids:glucose is almost 1:3. When hydrolysates are used as a (partial) source of amino acids, current evidence from both CM producers and literature shows that relatively more amino acids are needed, because the media composition is not defined and not optimal (Table   D.9). Glucose consumption in the baseline (mid medium) scenario in this study falls between the optimized scenarios of Tuomisto et al. (2022) and the enhanced catabolism scenario in Humbird (2021) (which is a more probable cell type for commercial cell cultures than wild-type).

Data collection and assumptions for energy consumption
Data on energy consumption were retrieved from engineering consultants in the field working for CM companies to develop scaled-up facility designs. Energy consumption for this study relies on modeling of similar processes and scales in SuperPro designer software. See Figure 2 (main paper) for the production line characteristics. The primary assumption was that metabolic heat generated was similar to other cell/protein production processes. Further assumptions are that biomass production is proportional to oxygen uptake rate (OUR) and that the cells are the final product. System characteristics used for the model, such as cell densities, average OUR and aeration rate (vvm), are provided in Table D.1. Table D.10 provides the model energy consumption results for one production line, including relevant assumptions and data modifications. Total energy consumption of the facility is a multiplication of the energy consumption of one production line times the amount of production lines at the facility (3,243, see Table D.1). Mbtu/hr). This is based on calculations for metabolic heat production of the cells at the stated cell densities and oxygen uptake rates (OUR). Original numbers provided were for cells with a volume of 5,000 um, for which peak heat rate was calculated to be 260 kW continuously (0.9 MBtu/hr). For cells with a volume of 3,500 um, this is estimated to be 70% of that (3,500/5,000), as metabolic heat production is linearly related to (molar) mass at constant OUR (Altwasser et al. 2017 In addition, HVAC power consumption was included. This is based on an average power consumption of 0.3 kW/m 2 facility size (number from primary data collection). 77% of this is electricity and 23% is heat/steam (Tschudi et al., 2001). Facility size is estimated at 3000 m 2 , or 3 times the floorspace occupied by equipment.
Total annual energy consumption is reported in Table D.3.
Average (macro)nutritional compositions are provided in Table E.1. For CM, a few parties were able to provide dry matter content, but information about macronutrients was not provided. Therefore deliberately broad ranges based on various conventional meat products are assumed and reported here in the row for CM.

Appendix F -Conventional meats ambitious benchmarks theoretical background
Ambitious benchmarks were created for conventional meats in order to lead to highly robust conclusions regarding sustainability claims of CM. A summary of the improvements modeled for conventional meats is provided in Table F.1. The improvements are discussed per topic below.
Food additives such as 3NOP have been developed that claim a reduction of methane emissions from enteric fermentation. Measured reductions vary widely and range from a negligible to over 90% (Klop 2016;Dijkstra et al. 2018;Honan et al. 2021;van Gastelen et al. 2022). In this study, a reduction of 15% was modeled, while adding 1.5 grams of enzymes per day to feed regimes. Sustainable energy was used for production of these enzymes.
Soy products contribute significantly to the carbon footprint and biodiversity impacts of pork and chicken (and to a lesser extent beef). This is for a large part due to land use change (LUC). Certification of LUC-free soy is available in the market, and while the efficiency of certifications vary, improvements in these schemes are made. Therefore it was assumed that LUC associated with soy will be zero in 2030 (both in m2a and in kg CO2), for both conventional and cultivated meats.
Ammonia (NH3) contributes substantially to multiple environmental indicators and this is most significantly so in cattle (beef and dairy) production. Ammonia emissions can be reduced through additional outdoor grazing.
Outdoor grazing was increased in the models by adding 50% of the difference between the current Dutch average and the required amount for organic production, in hours of outdoor grazing per year. This results in 5.4% reduction in ammonia emissions (calculations based on Hoving et al. 2014).
Energy used for ambitious benchmarks of conventional meat production was assumed to be sustainable, both at the farm and at feed production plants.
These changes result in a carbon footprint that is 15%, 26% and 53% lower for the 2030 benchmark for beef, pork and chicken respectively. Sustainable energy (electricity and heat) at farm and in feed compound production and soybean production. Chicken Agri-footprint: Chicken meat, fresh, at slaughterhouse/NL Economic − No LUC or associated GHG emissions related to soy in feed. − Sustainable energy (electricity and heat) at farm and in feed compound production and soybean production.

A. Cell culture medium
See paragraph 2.3.3. in the main text.
Results are provided in Table G.1.

B. Cell density
Maximum cell density during proliferation stages is a parameter for which many companies are optimizing. This is highly dependent on the adopted product system and bioreactor types. In this study, we have adopted a median maximum cell density expected to be achieved by CM companies in commercial-scale production. In the stirred-tank reactor (STR) system that we model, it may be feasible to increase cell densities by a factor of 1.4 (to 7.1*10 7 cells/ml, the higher density scenario), but not much more, according to experts in the field. At the baseline cell volumes, this density is close to the theoretical maximum cell volume fraction of 0.25 (Humbird 2021). It is plausible that certain large-scale production systems will not manage to operate at the cell densities that we model in the baseline scenario, and therefore model lower densities by a factor of 10 (10*10 6 cells/ml). In other reactor systems, cell densities up to 10 9 cells/ml are currently already feasible, albeit not yet at very large scales (Allan et al. 2019). However, limitations regarding metabolite formation and oxygen availability can be expected in those situations.
Higher cell densities would mean that more meat cells can be grown in a reactor of the same volume. It affects the energy demand for heating and cooling, as cultures with higher cell densities need less heating and more cooling, and cultures with lower cell densities need more heating and less cooling (per unit of volume). It is assumed that the total cooling load needed remains unchanged. Lower cell densities mean that more water has to be added to the growing medium to fill up the reactors, and more reactors are needed to produce the same amount of CM, with subsequently more energy and water used for cleaning. Changes in cell density during proliferation stages have no effect on assumed cell density during differentiation and maturation.
Results are provided in Table G.2.

C. Production run time
The production run time depends on the doubling time (for proliferation stages), on the desired maturity of cells in the final product (for differentiation and maturation), and the size of the largest proliferation vessel. Based on primary data we have decided to vary the total production run time by -25% and +25%. This does not necessarily cover all data received, but we feel that this does provide a solid basis for companies to interpret the results in comparison to their own specific process design.
For a shorter proliferation time, the cell doubling time is reduced from 30 to 22,5 days, and for a longer proliferation time increased from 30 to 37,5 days.
Shorter residence time in the reactors reduces overall energy demand, lowers medium demand during differentiation and maturation (we assume a linear relation), and results in a smaller number of reactors needed to produce the same amount of CM. Longer residence time affects these aspects reversely. We assume medium use in proliferation stages is not affected, as the same quantity of cellular biomass has to be produced, thus demanding the same amount of ingredients.
Results are provided in Table G.3.

D. Cell volume
Average cell volume differs per species type and cell type. For example, fat cells are much larger than muscle cells, and within the different types of muscle cells there is large variation. Also, small animals tend to have smaller cells than large animals. As the companies involved in this study produce a range of species and cell types, we used an average cell volume for the baseline scenario and determined smaller cell volume (500 μm3) and larger cell volume (5,000 μm3) based on primary data and literature.
The effects of smaller and larger cell volume are similar to that of lower and higher cell densities, respectively, as total cell mass harvested from one unit of volume decreases accordingly.
Results are provided in Table G.4.

E. Amount of harvests from semi-continuous process (single batch up to 10 harvests)
In the semi-continuous production process modeled in the baseline scenario, three harvests are made from the largest proliferation vessel, in one production run (see paragraph 2.3.1). In this sensitivity analysis, the effect of having less or more harvests from the largest proliferation vessel during one production run is assessed: one (batch production), five or ten harvests. This influences the number of production runs needed and the time per production run. With an increasing number of harvests, the production runs last longer, but the number of production runs decreases such that the combined operating time of the bioreactors in the facility decreases. The number of bioreactors needed also decreases. Energy demand increases-or decreases accordingly too, mainly because stand-by cooling load, aeration, and agitation are continuous and therefore in part related to combined operating time. There is no effect on culture medium use, with the exception that for the single batch scenario more water is added to the reactor to ensure adequate working volume. With an increasing number of harvests, the risk of contamination and therefore spoiled batches increases, but this was out of scope in this study.
Results are provided in Table G.5.

F. Smart cooling
Cooling energy is expected to be a major driver for energy demand in CM production. In this study, it is the dominant driver for facility energy use. In the baseline scenario, an active cooling system (using a refrigeration cycle) is modeled. This is the most conservative approach. More passive cooling (for example using cooling water and an air fin cooler) is also an option for many locations. By optimizing the cooling system to geographical location and ambient temperatures, a smart combination of active and passive cooling can lead to a reduced electricity demand for cooling. The compressor is by far the main driver of energy use in the cooling system and therefore implementing more passive cooling can drastically reduce the electricity demand. For this sensitivity analysis, the electricity demand for cooling energy is set at 50% of the baseline scenario, representative of a location where a refrigeration cycle is needed for half of the year at maximum.
Results are provided in Table G.6.

Appendix H -Ex-ante LCA additional literature review
Ex-ante LCA does not predict the future, but rather explores potential scenarios. Particularly useful are methodologies in LCA that use scenarios to assess future impacts associated with large-scale implementation of lab-or pilot-scale technologies (Cucurachi et al. 2018). The application of scenarios in ex-ante technology assessments allows for testing of policy interventions, investments, design changes and changes in socio-, techno-and economic systems. The results can also be used to estimate the validity of sustainability claims of future production and gain insights and understanding of specific design choices (van der Giesen et al. 2020).
Three topics in particular need extra attention when conducting ex-ante LCA (Cucurachi et al. 2018). The first is determining the functionality and the functional unit (FU). Novel technologies may aim to replace existing technologies based on a perceived main function, but rarely have an identical set of functionalities. The functional unit was explicitly defined broadly in this study, as many different high-protein products are compared to each other and the CM model is a general average of different products. Secondly, the lack of data that is representative of future (large-scale) production is inherent in ex-ante assessment since future systems do not exist yet and may influence the incumbent system. In this study, scenarios and conservative estimates for the baseline CM model, and ambitious benchmarks as well as global average footprints were shown for conventional meats, to treat some of this uncertainty. Thirdly, the lack of knowledge on the environmental impact of novel substances can cause skewed results, where the novel technology performs very well simply because characterization factors for its emissions are not available. As CM is in principle identical to conventional meats, and its production takes place in highly controlled environments, the skewing of results is assumed to be minimal.