Author Topic: Can I use ecoinvent v2.2 and ecoinvent v3 datasets in one Life Cycle model?  (Read 8681 times)

Offline pbeilschmidt

  • Administrator
  • Hero Member
  • *****
  • Posts: 586
    • View Profile
Question: "Can I use ecoinvent v2.2 and ecoinvent v3 datasets in one Life Cycle model?"
Answer: "Well, you could, but you will have to face a number of practical problems". 

We have heard this question from Umberto users quite a few times recently, so I will try to explain a little more what the practical problems are. Also I would like to give some recommendations on how to handle this.

As pointed out in this post, the underlying data format for the ecoinvent v3 database was changed and expanded significantly (from EcoSpold v1 to EcoSpold v2).

Should you wish to use datasets from ecoinvent v2.2 in a model that also has ecoinvent v3 datasets, you must be aware, that your "old" datasets may lack some information, which has been expanded in ecoinvent v3. You should also be aware that the changes made also affected naming of materials, basic units, and of course the actual values in the datasets, partly due to the application of different system models for the handling of allocation.

I'll focus on these four topics as the ones that have the biggest impact on the question, whether you can use ecoinvent v2.2 and ecoinvent v3 datasets in one Life Cycle model.

1) naming of materials / newly introduced exchanges
Many material names (both intermediate and elementary) were renamed during the switch from ecoinvent v2.2 to v3:
  • intermediate exchanges: As mentioned in this post, the identity of the process name and the reference flow (product) of the process has been dissolved. This does, however, not really constitute a problem for the usage of v2.2 and v3 data in one model as intermediate exchanges by definition are internal flows in the Life Cycle Model, and have no characterization factors, hence will not be part of a LCIA method provided in ecoinvent. An Excel file with a so-called 'correspondence table' can be downloaded from the ecoinvent website to see changes made to flow names.
  • elementary exchanges There were few changes to names of elementary exchanges during the changeover from ecoinvent v2.2 to ecoinvent v3. Elementary flow names stay the same. However, the compartments and subcompartments, which are considered an identifying part of the elementary exchange name did change. For example: "[resource/in air]" was renamed to "[natural resource/in air]". As a consequence the elementary flow identified previously as "Xenon, in air [resource/in air]" can now be found under "Xenon, in air [natural resource/in air]". This might possibly cause double entries in the list of elementary flows that have to be merged or considered separately in the impact assessment, most likely with identical characterization factors.
    Another example are the subcompartments of resource in water. While in ecoinvent v2.2 there were eight (!) subcompartments, these were reduced to five. As a consequence - when using datasets from the different versions - the level of detail in the inventory is not consistent any more in regard to the subcompartments.
  • new elementary exchanges Some new elementary flow have been added. An example is "Carbon dioxide, to soil or biomass stock" in the [soil/unspecified] compartment. A list of newly added elementary flows will be provided when available. Note that when using ecoinvent v3 activities and ecoinvent v2.2. activities in the same model, the new elementary flows will not be accounted for in the old dataset. Hence the contribution of such a substance or elementary exchange to an impact category is only based on the contribution from the new activity datasets. This must be considered when interpreting results. Possibly the impact assessment method must also be expanded to cover the new elementary exchanges.

2) basic units
As pointed out in this post, ecoinvent has decided to discontinue some basic units. Therefore, if you "mix" datasets from ecoinvent v2.2 and ecoinvent v3 you will have to tackle the question of basic units that are used in one dataset, but not the other.

3) actual values/coefficients
Updates to activity datasets were made from ecoinvent v2.2 to ecoinvent v3. Since the unit datasets are linked and aggregated into a (system terminated) dataset, any change in a unit dataset also results in different values for the result dataset. Hence it is unlikely that a result dataset from ecoinvent v2.2 and the corresponding ecoinvent v3 result dataset have the same values.
This is something that occured already in previous version steps (e.g from v2.0 to v2.01 to v2.1 to v2.2) as errors were removed, or datasets were updated. It is therefore a general question, whether one wishes to constantly update a life cycle model as new data becomes available, or, if you rather stay with one database release version, to use the data consistently in regard to the data source within one LCA study.

4) From multi-output processes (combined production) to single-output datasets
In ecoinvent v2.2 the way single-output processes (i.e. basically all ecoinvent datasets with only few exceptions) were derived from multi-output processes was left in the hands of the dataset editor (creator). In ecoinvent v3 dataset editors are now encouraged to submit combined production datasets as multi-output processes with mathematical relations and leave the partitioning or development of the single-output processes to an internal linking algorithm and the application of a system model (see below). The linking algorithm takes care of supplementing the appropriate input datasets, the choice of geography, and the handling of by-products depending on the system model. So the sub-division of combined production (e.g. in a refinery) is most probably handled differently in ecoinvent v2.2 and v3, yielding other values in the single-output datasets (diesel, petroleum, naphtha, etc.)

5) system models
The choice of system model in ecoinvent v3 determines the way how result datasets ('system terminated' or 'accumulated system' datasets) are created and how allocation is being handled in this process. In ecoinvent v2.2 the creation of (geography specific) processes was left to the dataset editor. In case an intermediate input was not readily available, the dataset creator would most likely use a stand-in dataset as input, that he/she considered most appropriate. The handling of by-products and allocation on the process level was also left to the dataset creators responsibility.
In ecoinvent v3  there is now a consistent handling of the creation of how result ('system terminated' or 'accumulated system') datasets. The linking algorithm takes into account the question whether market Several different system models are available and in fact the output of each system model application in the calculation algorithm is a different result process (in regard to the coefficients).

The good news is: ecoinvent transferred/transfromed most, if not all of the datsets from ecoinvent v2.2 to ecoinvent v3. These datset can be identified through a remark in the "General Comment" field that reads "[This is a dataset transferred from ecoSpold v1 / ecoinvent database version 2. It may not in all aspects fulfill the requirements of the ecoinvent data quality guideline for version 3.]".
In other words: the whole content of ecoinvent v2.2 is part of ecoinvent v3. There is - from our point of view - no need to try to mix "old" datasets with "new" ones.

At present we therefore do not plan to "downgrade" ecoinvent v3 activity datasets to the old EcoSpold v1 format, in order to make them available in the old Umberto 5.6 for Life Cycle Assessment product version together with 'ecoinvent v2.2 for Umberto 5.6'. We also don't intend to transform ecoinvent v2.2 data to the EcoSpold v2 data format, and then provide these datasets as an additional separate master database, since most of this work has already been daone, and the data is available in ecoinvent v3 already (although updated).
We strongly encourage our users to switch to the new Umberto NXT LCA version to be able to benefit from the new, updated data and avoid facing the issues describe above.
« Last Edit: October 30, 2014, 11:34 by pbeilschmidt »
Peter Müller-Beilschmidt
myUmberto Moderator
ifu Hamburg GmbH