The ecoinvent data is regionalised in a number of ways, we have documented this here. You are absolutely correct, one of the largest factors influencing the updates would be the UK Grid. The other changes to the data would include:
- Updates to methodology. I don’t think this affects GWP between V14 and V19 but will be a factor in V20 when we move to EN15978 A2 / IPCC AR5 characterisation factors. For other indicators however there are often changes in methodology.
- Changes to the scope of the data, eg alignment to new standards
- New / improved / updated inventory data from supply chains. E.g. new supplier data becomes available for a product and the associated LCI models are updated to reflect this, then propagated throughout the supply chains of all the other processes in the data set
- Updates to align with new scientific knowledge. Since V14 was released I believe there’s been significant changes to the treatment of end of life of timber as research becomes available regarding degradation in landfill (regarding timber, see further comment below, more changes to come)
- Fixes to the LCI models in response to QA QC checks. These are less common in GWP but Cerclos’ QA QC checks invariably identify anomalies that require investigation. Whilst this is frustrating internally (huge effort on our part every QA QC round + delays between fixes) we have to accept that Life Cycle Assessment data sets are in a perpetual need of continuous improvement (and generally speaking those organisations tasked with this are very underfunded). We have come to accept that a reasonable expectation is that the data is always improving (rather than expecting it to be perfect). Since introducing our QA QC checks, Cerclos has identified over 50 issues in the Ecoinvent and BRE IMPACT data models that have been actioned for improvement.
- Alignment with new standards. A good example here is EN15804 A2 which moves from IPCC AR4 to IPCC AR5 characterisation factors for GWP. There is also a big shift in the way timber is treated in EN15804 A2 (essentially moving from the scientific knowledge of what happens to timber to a calculation method that always balances to zero biogenic carbon)
I would imagine that between each data update there are 100’s and possibly 1000’s of small changes to the data. Some changes are implemented by EcoInvent, some by national bodies (e.g. BRE, AusLCI), some by our data supplier, some initiated by us through QA QC activities. The underlying data is very well documented and referenced, however separately documenting each change would be a monumental task (particularly if one attempted to clarify which processes were affected by each change). Further to this, the data models themselves are extremely complex due to very deep modelling of underlying processes and it’s impossible to predict the effect a single change may have (so it’s not possible to reliably document the “important changes” only).
Lastly, a word on regionalisation. The method of applying the local grid throughout the supply chain, although used extensively (almost without exception) is quite flawed due to the globalisation of trade. An extreme example of where it is flaws is when 100% of a product’s supply is from overseas, but the LCI model presumes it is produced locally and applies the local grid factor to the electricity in production. This type of error is particularly obvious when a country has a low carbon grid (UK is a good example of this). We plan to address this with a “shallow regionalisation” model in an upcoming data update. This will account more accurately for global supply chains (and imported vs nationally produced flows of each material will be accounted for in the model). We think this will be an important update to improve data quality such that it better reflects reality.
I hope this answers your question.