Comparison of Biofuel Life Cycle Analysis Tools

Bioenergy plays an important role in the decarbonization of the transportation sector. Its medium and long-term benefits depend on the reduction of greenhouse gases (GHG) emissions brought forth by the conversion of renewable feedstocks, which can be quantitatively determined through a Life Cycle Analysis (LCA) methodology. In this context, identifying the main differences and commonalities in methodological structures, calculation procedures and assumptions of different LCA models are desired to demonstrate the possibility of obtaining homogeneous results for similar production chains. This report presents the main results of the study carried out for Phase 2 of the project entitled “Comparison of Biofuel Life Cycle Assessment Tools”, which is included in the activities of Task 39 (Commercializing Liquid Biofuels from Biomass) of the International Energy Agency Bioenergy Technology Collaboration Program (IEA Bioenergy).

The scope of this study is restricted to biofuels employed for transportation by road (biodiesel or FAME, Fatty Acid Methyl Esters) and air (biojet fuel or HVO/HEFA, Hydrotreated Vegetable Oil/Hydroprocessed Esters and Fatty Acids) produced from either soybean oil, palm oil and used cooking oil (UCO). Five models were considered in the study:

 BioGrace (European Community): available in https://biograce.net/home;

 GHGenius (Canada): available in https://www.ghgenius.ca/index.php/downloads;

 GREET (United States of America): available in: https://greet.es.anl.gov/index.php?content=greetdotnet;

 New EC (European Community): available in http://data.jrc.ec.europa.eu/dataset/jrc-alf-bio-biofuels_jrc_annexv_com2016- 767_v1_july17;

 Virtual Sugarcane Biorefinery - VSB (Brazil): not available to external users.

While four models are publicly available and serve regulatory purposes (BioGrace/GHGenius/GREET/New EC), the VSB was initially developed by CTBE/CNPEM to assess the sugarcane production chain, having further expanded its scope to several other feedstocks and conversion pathways within a biorefinery context. The BioGrace model, although released in 2015, still uses 2008 input data; the new JRC dataset from 2017 (and used in this study under the designation “New EC”) is freely available online. The results presented in this report are limited to the GHG emissions determined by each model with the default conditions to which they were developed using both cradle-to-gate and cradle-to-pump boundaries. The cradle-to-gate approach considers the emissions of biofuel production from the feedstock production up to the gate of the biofuel producing unit, while the cradle-to-pump analysis includes additional impacts of biofuel distribution to fuel pumps.

BioGrace estimates the highest emissions in soybean and UCO pathways. GHGenius also estimates the lowest GHG emissions for three pathways: UCO biofuels and soybean FAME. VSB follows with HVO/HEFA from soybean and palm and GREET has the lowest emissions only for palm FAME. Palm biofuels can also be assessed in two variants: with or without the capture of CH4 from palm oil mill effluent. This option is only taken into account by BioGrace and New EC, both of which clearly demonstrate the impact in carrying out this additional operation associated to palm oil extraction.

Regarding the comparison of models and pathways, the most discrepant results for FAME and HVO/HEFA production from the three assessed feedstocks (soybean, palm and UCO) were due to several reasons:

 Differences in agricultural processes, something expected since most models have different location of soybean/palm production;

 Substitution procedure in GHGenius in opposition to the allocation methods considered in the other LCA models, which contributes to either considerably decrease or increase final emissions depending on the feedstock or industrial pathway;

 High use of renewable energy sources in industrial processes considered in the VSB model;  High variation of energy intensity between models for the industrial pathway;

 Differences in modals and distances of feedstock transportation that are specific for each country.

The location of feedstock production and industrial process are specific for each model and, therefore, variable results are expected in terms of GHG emissions per MJ of biofuel. In general, there are differences in the inputs data and, also, in methodological choices. Some of these differences are justified by the particularities of each model, while others can be harmonized. It is also worth noting that this study has proceeded to consider both HVO and HEFA routes as producing a hydrocarbon mix with similar applications, since the production of either renewable diesel or renewable jet fuel through them are practically identical. It was assumed that the impact of an extra consumption of hydrogen for an isomerization step would be minimal and that the energy-based allocation employed in the HVO-based models BioGrace and New EC would minimize the influence on determining the carbon intensity of biojet fuel when a fractionation of the hydrocarbon mix is carried out.

To harmonize the models, default data and parameters (such as agricultural and industrial inputs, emission factors and allocation procedure) were retrieved from the VSB database and entered on three other models. With this approach, it was possible to identify the main differences and to reach similar impacts from different LCA models considering the same production system, as shown in Figure ES1. Only the soybean FAME pathway was harmonized in this study. It is important to highlight that the New EC was not included in the harmonization procedure: despite the data for several scenarios being available online (and an external user would be able to “rebuild” the calculation structure, if needed), the spreadsheet with the calculation tool is locked for edition by users. This led to the removal of New EC from this specific section of the study since the purpose of a harmonization exercise is not only identifying the differences between assumptions and input data from each model, but also understanding the underlying features of the calculation mechanism itself. 

In this sense, there is room for discussion and standardization of models in order to decrease the variance of input data and approaches (e.g. necessity of a collection and pre-processing phases for UCO) and thus “pre-harmonize” all models. An effort to build a harmonized data set of input data for the technological pathways and to update the databases of the main models would benefit the community and deliver better GHG emission results for the life cycle assessment of biofuels production.