A carbon footprint of HVO biopropane

Biopropane made by hydrogenating vegetable or animal oil/fat is being commercialized as a biofuel alternative to liquefied petroleum gas (LPG). Its carbon footprint has been calculated from field to tank, using public data for each process in the supply chain, for six main feedstocks: palm oil, palm oil fatty acid distillate, tallow, used cooking oil, rape oil, and soy oil. Scenarios have been applied to the calculations using four main variables: allocation method, i.e., economic or energy; methane capture at the oil mill (or not); application of indirect land‐use change (or not); and classification of the feedstock as a residue (or not). HVO biopropane's carbon footprint varies, depending on the feedstock and the four variables, from as low as 5 g CO2e/MJ to as high as 102 g. In most cases, this qualifies for government support, i.e., financial credits and biofuel mandates enacted by EU member states under the Renewable Energy Directive. © 2017 The Authors. Biofuels, Bioproducts and Biorefining published by Society of Chemical Industry and John Wiley & Sons, Ltd.


Introduction
H VO 1 biopropane, unavoidably created in the production of renewable diesel from oils and fats, is the newest commercially-available biofuel. It is a dropin fuel, i.e., chemically identical to fossil propane, and so can substitute fossil propane entirely. 1 Research on the HVO process* dates back about 30 years, and full-scale production ensued about a decade ago. Today there are about ten sites producing an estimated 3.5 million tonnes per year of renewable diesel. Also known as HVO biodiesel, HEFA diesel, HDRD, and Green diesel, 2 renewable diesel is sold primarily into road-transport and aviation markets. Most HVO biopropane produced at these plants is not recovered as a fungible product. It is left in a mixture of off -gases -anything that is not diesel -which are consumed as low-value fuel gas.
Th is changes in 2017, when Finland-headquartered Neste begins extracting HVO biopropane from its HVO plant in 888 In practice, some fats/oils require more hydrogen, because they have some unsaturated carbon bonds (i.e., double bonds) that are converted to saturates (single bonds). Th e molar fractions of biodiesel, carbon dioxide, and water coming out can vary slightly, depending on how the esters are cleaved. Some of the biodiesel is reformed to a mixture of shorter-chain hydrocarbons, oft en referred to as bio naphtha. Th e bionaphtha and biopropane are unavoidable byproducts to the process.

Production of HVO biopropane
Global capacity for HVO biopropane is believed to be around 220 000 tonnes annually, albeit, not all capacity is currently operating ( Table 1). According to a UK Department of Energy & Climate Change report, 3 the plant in Geismar, LA, USA, did extract and sell biopropane when it was owned and operated by Dynamic Fuels, but this appears to have ceased. Th e current owner/operator's website says biopropane is sold as a bionaphtha mixture.

HVO biopropane supply chain, by process
From public data sources, a footprint-relevant defi nition of the HVO biopropane supply chain was developed, from a search and analysis of both the peer-reviewed and grey literature. Th e system was defi ned into fi ve process steps (or life-cycle phases): oil/fat supply, hydrogen production and transport, HVO process, biopropane extraction and purifi cation, biopropane transport.

Oil/fat supply
Oils and fats for HVO processing can be made on purpose, and they can also arise as residues or wastes of other processes.

On-purpose feedstocks
Th e primary feedstocks currently used to make biodiesel (renewable diesel or FAME) in Europe have been inventoried for this study: crude palm oil, rapeseed oil, and soybean oil.
Crude palm oil is produced on plantations, mainly in Southeast Asia. Fruit bunches (FBs) are harvested from palm trees. Th ese are crushed to extract the palm oil, and the palm kernels are also crushed to make palm-kernel oil and palm-kernel cake. Rapeseed is produced on farms, mainly in Europe but also in Canada. Seeds and straw are harvested. Th e seeds are crushed to extract oil and rapeseed cake. Soybeans are produced on farms, mainly in Brazil and the USA. Th e beans are crushed to extract oil and soya cake.
Mass/energy/emissions inventories for these three oils were compiled (Supplemental Material, Tables S1-S3), based on careful examination of the following sources: for palm oil, 4-7 for rape oil, 4-8 and for soy oil. 4-9 A nitrous oxide emission in a report from Argonne Labs 9 was not included, because it was an order-of-magnitude out from the other estimates. Where these sources diverged, their fi gures were averaged. 889 Several points in the inventories merit further explanation: • No direct land-use change was assumed, because the farms are presumed to have been in operation before 2008. According to European Union rules, 10 all plantations already operating before January 1, 2008 are 'grandfathered' in as not creating land-use change. • Steam and electricity at the palm-fruit-bunch crushing plant are supplied by a combined heat/power (CHP) unit run on empty fruit bunches (EFBs), fi ber, and shell from the harvest. Th is is considered as a closed, internal loop. • Some 10-15% of the EFB is recycled as compost to the plantation. Again, this is considered a closed, internal loop. • Economic value of the crush products is an average of the three sources who calculated it. 4,6,7 • Palm-oil mill effl uent (POME) treatment -at some palm-oil mills, the organic-laden, aqueous effl uent is usually charged to a wastewater lagoon. Th e economicallocation case assumes that the methane arising from this lagoon (from anaerobic degradation, i.e., rotting, of the organics), is not captured, that it is released to the atmosphere. • Indirect land-use change (iLUC) has been considered in one of the footprint scenarios. In this case, iLUC factors proposed under the Renewable Energy Directive (Supplemental Material, Table S10), 55 g CO 2 e/MJ for oil-based biodiesels, have been applied to the on-purpose oils.

Waste/residue feedstocks
Th e primary wastes/residues used to make HVO biodiesel in Europe have been inventoried for this study: Palm fatty acid distillate (PFAD), tallow and used cooking oil (UCO).
(Th ese are less used to make FAME biodiesel, because they are mostly unsuitable to the FAME process.) PFAD is a mixture of fatty acids that are distilled from crude palm oil to make refi ned palm oil. PFAD is inedible for humans, but it can be digested by animals, and it can be used as a feedstock for soaps and some chemicals. Tallow comes from animals raised for food, such as cattle, pigs, and chickens. Th ese are sent to abattoirs, where they are slaughtered and butchered into meat and (sometimes) hides. Th is amounts to about 40-70% of the animal weight. (With fi sh, the edible portion is reportedly much lower.) Th e remaining animal carcass is sent to a renderer, which converts it to protein meal and tallow (fat). UCO, oft en called yellow grease in the USA and sometimes called waste cooking oil, is collected from commercial food-frying operations. , Tables S4 and S5), based on careful examination of sources for PFAD and tallow. 6,7,10-21 PFAD is classifi ed as residue or waste by Finland, Norway, Sweden, 22 Italy, and the US federal government. 23 It is classifi ed as a product by the UK. Classifi cation as a residue is also possible, because as reported by Cheah et al. 24 and other market analyses, PFAD consistently trades at prices less than those of its feedstock, crude palm oil. UCO is classifi ed by the UK Department for Transport 14 as a waste. Another study for the same UK Department of Transport 25 argues that UCO is a waste, yet presents a raft of evidence that it has economic value. A study by SRI Consulting 7 clearly classifi es UCO as a product, and allocates some of the burden of oil production (cultivation and processing) to the UCO.

Mass/energy/emissions inventories were compiled (Supplemental Material
In this study, the same approach was taken as SRI's: 'We have allocated by economic value of the fresh and the used grease. According to a 40-year time series generated by the US Energy Information Administration, this is typically about 2:1, i.e., by weight fresh grease sells for about double the price of used. So, in principle, 66.7% of the inputs/outputs would be allocated to fresh grease, while 33.3% would be allocated to yellow grease (UCO). Because only an estimated 70% of fresh grease is recycled as UCO, we adjusted our actual allocation proportionately to 75:25. ' 7

Hydrogen production (and transport)
Most commercial hydrogen comes either from an oil refi nery or from a natural gas reformer. Refi neries are also major consumers of hydrogen, so the incremental supply source will almost always be a reformer. Th ese react steam and methane (natural gas) to create hydrogen, which creates carbon dioxide as a waste.
Th e theoretical reaction is which yields 1 kg of hydrogen for every 2 kg of methane feedstock. In practice, more methane is needed, some as fuel and because the reaction is not 100% effi cient. A mass/emissions inventory for animal fat production was compiled (Supplemental Material,

HVO process
In the HVO process, a triglyceride vegetable (or animal) oil is reacted with hydrogen to create an aliphatic hydrocarbon of around C 16 -C 18 in length, i.e., renewable diesel. Th e reaction creates other outputs: carbon dioxide, water, and a mixture of shorter hydrocarbons, including propane.
Th ere are two main reactions that can happen in the HVO process: By adjusting process conditions and the catalyst, one reaction can be favored over the other, although there will always be some mixture of the two. 32 Favoring DCO or HDO changes the composition of the outputs considerably. Th e theoretical outputs for a feedstock of brown grease 33 are: • DCO -15.5 weight % CO 2 • HDO -12.7 weight % H 2 O Whichever one is favored, hydrogen inputs will be greater when the feedstock oil is more unsaturated, because hydrogen is consumed to saturate the olefi nic carbon bonds.
A mass/emissions inventory for the HVO process was compiled (Supplemental Material, Table S7), based on careful examination of sources. 3,4,[6][7][8][9]15,18,19,30,32,[34][35][36][37] Allocation for the HVO process is, in most of the existing studies, not mentioned. Reports from both SRI Consulting 7 and Argonne 9 avoid allocation by giving the crude biopropane an avoided product credit. Argonne 9 also reports percentages for market, energy, and hybrid allocations. Argonne's percentages are diff erent to those calculated in this study (Supplemental Material, Table S8). Article 2 of the EU's Renewable Energy Directive, 10 as amended on September 9, 2015, off ers another allocation alternative: classifying the HVO crude biopropane as a processing residue, i.e., 'a substance that is not the end product(s) that a production process directly seeks to produce; it is not a primary aim of the production process and the process has not been deliberately modifi ed to produce it'.

Biopropane extraction and purifi cation
Th e HVO process generates a liquid phase of renewable diesel plus a light-ends fraction, a mixture by weight of about 70% crude HVO biopropane and 30% HVO bio naphtha. To be sold as substitute LPG, the biopropane must be extracted and purifi ed to an LPG specifi cation.
A mass/emissions inventory for the extraction and purifi cation was compiled (Supplemental Material, Table S9), based on careful examination of the following sources that serve as proxies for the process: ‡, [38][39][40][41][42] Biopropane transport HVO biopropane must then be stored and transported to consumers. Most of the existing studies of renewable diesel do not report these details explicitly. Th e exception is Nikander, 30 who assumes that renewable diesel is transported 200 km by a 39-t truck to a blending site. At the largest HVO extraction site (Table 1), Rotterdam, an LPG jetty for loading barges and ships will be located directly adjacent to the plane. So, the base case for this study is to assume ship transport to local storage in Rotterdam, Europe's largest concentration of refi ned product production and storage.

Footprint method
Th e footprints in this study were calculated according to prevailing practice and convention.
Only public sources of data were used. Th ey have been cited in the text and listed in the references.
Th e reference fl ow of this study is that of HVO biopropane and its preceding supply chain. Th is paper details the fi eld-to-tank supply chain, from its origins in agriculture through to storage of the fi nished product. In the tank-towheel part, no carbon footprint is registered, because the carbon in the biopropane is bio-based and in this instance carbon neutral. Th e functional unit is the physical quantity of HVO biopropane. Th is is the unit by which LPG is sold, and by which HVO biopropane will most likely be sold as well. Physical quantity of LPG or propane is typically denominated in mass (kg or tonne), volume (usually liters, sometimes cubic meters or even barrels) or energy content (MJ at lower heating value). For biopropane, the energy density is assumed to be equal to that of fossil propane, 46.296 MJ LHV/kg. 43 In this study, we mostly use MJ (at LHV), because this is the unit used most commonly for biofuels, and it is defi nitely used by the EU and its member states.
Th e footprints were calculated in SimaPro, Version 8.0.5.13, and further manipulated in Excel. Th e life-cycle model was built using the inputs described in the section on ‡ No actual units have yet been operated that do this process at commercial scale. So, the figures have been estimated from other, similar processes.

Results and discussion: biopropane footprint
First the model was calibrated by calculating a footprint for Renewable Diesel. Th en it was run for HVO biopropane for all feedstocks under all scenarios. Th e scenarios were chosen to represent the range of constraints that might apply to suppliers and regulators in estimating footprints: • Economic allocation • Energy allocation • Energy allocation, with methane capture (at the palm oil mill) • Energy allocation, with methane capture and iLUC • Economic allocation, feedstock is residue (as defi ned by the Renewable Energy Directive) • Energy allocation, feedstock is residue

Calibration of the footprint model
Run for palm oil feedstock to produce Renewable Diesel, the model generates a footprint of 40 g CO 2 /MJ ( Table 2). Its major constituents are: methane from the rotting of the palm oil mill effl uent (POME); carbon dioxide from hydrogen and nitrogen-fertilizer production (gasintensive processes); carbon dioxide from operations of the palm oil plantation (mostly diesel emissions from power equipment); and ship emissions from the transport from Malaysia (or Indonesia) to an HVO plant in Europe (Rotterdam).
This result falls in the lower end of 'official' estimates, which range from 39 to 50 g CO 2 /MJ. Official means footprints published by government regulators (Table 3): the European Commission in the Renewable Energy Directive 10 and in BioGrace; 5 and the US EPA. 44 For further reference, a figure is also included from the first public study in this area, which was conducted by the author. 7 Th ere are other estimates outside of the offi cial range. For instance, a study by ecoinvent 6 estimates a footprint of 42 g CO 2 /MJ for palm oil only (which would presumably lead to a renewable diesel footprint of over 50). A more recent study by Blonk 45 estimates the palm oil footprint at 241 g CO 2 /MJ. Th is probably includes a large iLUC factor and appears to be much broader than a footprint estimated by conventional methods.

HVO biopropane from palm oil
Using the calibrated footprint model, the footprint was calculated for HVO biopropane using economic allocation (Table 4), which comes out at 16 g CO 2 /MJ. Th is is the fi rst precise, public estimate of HVO biopropane's footprint. A study for the UK government 3 estimated a range of 10-50 CO 2 /MJ, but did not stipulate a specifi c base case.

892
Under alternate scenarios, the HVO biopropane (and biodiesel) footprints vary considerably (Table 5). Changing from economic allocation to energy allocation adds nearly 16 g to the biopropane footprint, and slightly decreases that of biodiesel. Th is also slightly changes the footprint at the oil mill. Adding methane capture sinks the footprint by nearly 7 g.
Th en there is the question of iLUC. In Annexes V and VIII of the December 17, 2012 Amendment of the EU Renewable Energy Directive and the Fuel Quality Directive (pp [19][20][21], an iLUC factor of 55 CO 2 /MJ is proposed for 'oil crops'. Th is has been applied to biopropane and renewable diesel (Table 5). Whether iLUC should be included, and if so, what its factors should be, are still very much open questions. As supplemental research to this study shows (Supplemental Material, Table S10), currently proposed factors for palm oil range from 44 to 231 CO 2 /MJ. Two other scenarios were not considered, because they are theoretical. One is that crude HVO biopropane be classifi ed as a residue. Th is is plausible, in that glycerine from FAME biodiesel production is classifi ed under RED as a residue, and its production is very similar to that of HVO biopropane (i.e., it is an unavoidable byproduct of biodiesel synthesis). However, the UK Government has explicitly rejected this classifi cation, § ruling that HVO biopropane is a co-product, not a residue. If crude HVO biopropane were classifi ed as a residue, the HVO biopropane footprint would be about 8 CO 2 /MJ. Also not considered was the scenario that palm oil could be classifi ed as a waste or residue.

Waste/residue feedstocks: PFAD, tallow, and UCO
In addition to palm oil, these three raw materials are leading sources of feedstock for Renewable Diesel and HVO biopropane.
For PFAD (Table 6), in the economic-allocation case, the footprint comes out at 15 g CO 2 /MJ, and its major constituents (obviously) are the same as those for HVO biodiesel. Under the scenarios, its footprint ranges from 5 to 80 g CO 2 /MJ .
For tallow (Table 7), in the economic-all ocation case, the footprint comes out at 17 g CO 2 /MJ. Its major constituents are production of the tallow (which is relatively energy intensive) and the HVO process.

893
If tallow is classifi ed as a residue, the footprint drops considerably. With economic allocation, it falls to 5 g, whereas with energy allocation the footprint is 11 g ( Table 8). Th e other scenarios applied to palm are not applicable: there is no methane-capture option in tallow production, and iLUC does not apply.
For UCO, the oil that went into the fryer is assumed to be rape oil, which is a common frying oil in Europe. In the economic-allocation case, the UCO footprint comes out at 9 g CO 2 /MJ (Table 9). Its major constituents are hydrogen production, rapeseed cultivation, nitrogen fertilizer production and steam (in the HVO plant and the oil mill).
If UCO is classifi ed as a residue, the footprint drops to 5 g with economic allocation, and it falls to 11 with energy allocation (Table 10). If iLUC were applied to UCO, the fi gure of course would rise.

The other main feedstocks: rape and soy oils
Neither rape nor soy oil are expected to be primary feedstocks in European biopropane production. Still, these are signifi cant feedstocks in global production of biodiesel, so their footprints are worth knowing, at least as benchmarks.
Rape oil, in the economic-allocation case, has a footprint of 19 g CO 2 /MJ (Table 11). Its major constituents are production of rapeseed, nitrogen-fertilizer production and the HVO process. If allocation is done by energy rather than economic value, the footprint rises to 47 g. If on top of that, an iLUC factor is added, the footprint climbs to 102 g (Table 12).
Soy oil, in the economic-allocation case, has a footprint of 17 g CO 2 /MJ (Table 13). Its major constituents are cultivation of soybeans, hydrogen production and the HVO process. If allocation is done by energy rather than economic     894 value, the footprint rises to 40 g. If on top of that, an iLUC factor is added, the footprint climbs to 95 g (Table 14).

Conclusion: Biopropane's footprint generally low, but scenarios vary it
Seen as a whole, HVO biopropane's footprint by feedstock and scenario (Table 15) varies considerably, but in most cases, it qualifi es for government support, i.e., fi nancial credits and biofuel mandates enacted by EU member states under the Renewable Energy Directive (Table 16). Th is variability of footprints, according to footprintcalculation method, has clear precedents: for instance in a study of forklift s 46 and one of forest products. 47 Th is potential variability should be kept in mind by footprint users: regulators, suppliers, and consumers.   a Electricity and cogeneration are unlikely to be applications of biopropane, but are included for reference. b According to Article 17 paragraph 2 of RED, as of 1 Jan 2017, HVO biopropane must show a 50+% carbon reduction from the fossil fuel comparator to qualify, i.e., its footprint must be lower than those in this column.