Deviations in ecoinvent2.2 and 3.0 datasets OR mistake in LCIA programming?

[dmutz]

Hello,

I noticed following problem while modelling an own energy mix with ecoinvent v3.0 data: there are some large deviations in LCIA results comparing to the same energy mix by using econinvent v2.2 data in LCA NXT. I used ReCiPe Hierarchist w/o LT for the comparison.

Following deviations in results seem pretty unrealistic only due to updates in the database (just an exerpt):
Windpower (onshore)

• freshwater ecotoxicity
       
  • v2.2 "electricity, at wind power plant 800kW [RER]": 1,1*10^(-4) kg 1,4-DCB/MWh
  • v3.0 "electricity production, wind, 1-3MW turbine, onshore [DE]": 0,6 kg/MWh
• climate change
       
  • v2.2 "electricity, at wind power plant 800kW [RER]": 0,17 kg CO2-eq/MWh
  • v3.0 "electricity production, wind, 1-3MW turbine, onshore [DE]": 22 kg CO2-eq/MWh

Lignite

• freshwater ecotoxicity
       
  • v2.2 "electricity, lignite, at power plant [DE]": 0,17 kg-DCB/MWh
  • v3.0 "electricity production, lignite [DE]": 40,9 kg DCB/MWh
• climate change (comparable results between new and old database!)
       
  • v2.2 "electricity, lignite, at power plant [DE]": 1,24 kg CO2/kWh
  • v3.0 "electricity production, lignite [DE]": 1,25 kg CO2/kWh

By analyzing the results, I noticed that ReCiPe (H) w/o LT, take also category factors into account for long term emissions. Is that an error in the LCIA programming within the software? By comapring LCIA data, ecoinvent v3.0 includes long-term emissions (and they have large impacts, which lead to the higher Impacts in most categories) and ecoinvent v2.2 does not include those long-term emissions. But both energy-mixes were calculated at the same time (in Umberto NXT LCA with ReCiPe (H) w/o LT)

If it is an error in the LCIA programming, it seems old ecoinvent v2.2 has no long-term emissions in their dataset (if they had, there would/should be still comparable results)?

Thank you in advance for answering!
Regards, D. Mutz

[pbeilschmidt]

Hello dmutz,

I have quickly checked on of the examples you mentioned

ecoinvent v3.0 (on website)
activity dataset "electricity production, lignite, DE"
ReCiPe Midpoint (H)/freshwater ecotoxicity
0.040201 kg 1,4-DCB-Eq per kWh

ecoinvent v2.2 (on website)
activity dataset "electricity, lignite, at power plant, DE [kWh]"
ReCiPe Midpoint (H)/freshwater ecotoxicity
0.038264 kg 1,4-DCB-Eq per kWh

ecoinvent v2.2 (from my local Umberto 5.6 installation)
activity dataset  "electricity, lignite, at power plant, DE [kWh]"
ReCiPe Midpoint (H)/freshwater ecotoxicity
0.0382631 kg 1,4-DCB-Eq per kWh
(see attached screenshot)

Without being a real expert on these datasets, I would still say that all values are plausible. the potential freshwater ecotoxicity value in ecoinvent v3 is slightly higher than in ecoinvent v2, but still in the same range (increase from 38.264 kg 1,4-DCB-Eq per MWh to 40.201 kg 1,4-DCB-Eq per kWh).
This change can be explained due to differences in the underlying activity datasets. The impact assessment system is exactly the same, as no updates were made to the LCIA methods by ecoinvent from v2 to v3.

There is a minimal deviation between the FETP value from ecoinvent v2.2 in Umberto (38.2631 kg 1,4-DCB-Eq per MWh) as compared to the value on the website (38.264 kg 1,4-DCB-Eq per MWh), but as far as I know the values on the website are shown rounded or truncated for legibility, but are exact in the downloaded dataset.

The value you are presenting (0,17 kg-DCB/MWh) is completely out of range, and I therefore assume that you are working with an outdated version of the ReCiPe method in your local installation.

Please verify that the root folder of your locally installed ecoinvent library is called "ecoinvent 2.2d".

The ecoinvent Centre made some changes to the impact assessment methods in 2010/2011, which led to the necessity of delivering updates to all our Umberto customers using this database. Please see this communication for more info (unfortunately the link to the PDF file is broken, but I can try to find a copy and send it to you).

We also informed all our registered users directly and sent out new installation CDs at that time.

Please make sure you are running the latest version of the ecoinvent v2.2 library on your computer. If you are running the latest version already, and still get much different values, please contact our help desk again (support@umberto.de), so that we can try to figure out possible other reasons.

[dmutz]

Hello Mr. Beilschmidt,

thank you a lot for your quick reply. Now, I think I get the problem!
I checked the figures, you've provided for lignite. It is not a failure in the installation of the database in my case. It is a problem of the chosen database (and possibly an error within the ecoinvent v3.0 database?):

In the attached screenshots (from ecoinvent website) you can see a difference in versions 2.2 and 3.0:

Use of "electricity, lignite, at power plant [DE]" (ecoinvent v2.2) leads to:

ReCiPe Midpoint (H)/freshwater ecotoxicity
0.038264 kg 1,4-DCB-Eq per kWh
and
ReCiPe Midpoint (H) w/o LT/freshwater ecotoxicity w/o LT
0.00018148 kg 1,4-DCB-Eq per kWh

Use of "electricity production, lignite, DE" (ecoinvent v3) leads to:
ReCiPe Midpoint (H)/freshwater ecotoxicity
0.040201 kg 1,4-DCB-Eq per kWh
and
ReCiPe Midpoint (H) w/o LT/freshwater ecotoxicity w/o LT
0.040201 kg 1,4-DCB-Eq per kWh

Since I used the ReCiPe Midpoint method without longterm emissions it has changed significantly in my case. Or there is a problem with the "ReCiPe Midpoint (H) w/o LT" method...?! Comparing all indicators, there is no change between with LT and w/o LT, which leads to the question, why there is a distinction between longterm and no-longterm emission?

[pbeilschmidt]

For my checks I used the "with LT" version of the LCIA indicator only, and didn't look into the "without LT".

Now that I understand better that your questions relates to why there IS a difference between the two time perspectives ("with LT"/"without LT") in ecoinvent v2.2, and there IS apparently NO difference between the two time perspectives in ecoinvent v3 for the dataset used as example, I will have another look into this issue.

I think I can report back on this tomorrow. (Maybe another forum visitor is faster in giving an explanation...)

[pbeilschmidt]

Didn't want to go home without quickly checking this. Since we have both the ecoinvent v2.2 and the ecoinvent v3 datasets available in Umberto NXT LCA I did a simple comparison by dragging these two activities into a model. I then chose the two different LCIA factors we were discussing: 'FETPinf with LT' and 'FETPinf w/o LT' from ReCiPe (H).

Have a look at the two results in the screenshot: The values with the little blue marks are for the ecoinvent v3 activity (shown on the right side above), the other pair with the red marks are for the ecoinvent v2.2 activity.

It confirms what you have already shown in your two screenshots. It also shows that even though you said that the value for 'FETPinf with LT' and 'FETPinf w/o LT' is identical in ecoinvent v3, it is not identical (40.20084014 kg 1,4-DCB-Eq per MWh for 'FETPinf w/o LT' and 40.20087480 kg 1,4-DCB-Eq per MWh for 'FETPinf with LT'), but they are very close to each other. On the ecoinvent website this value is shown rounded to less significant digits, but the dataset as imported into Umberto NXT LCA has all the detail.

This will be a starting point to find the answer to your questions tomorrow
* what are the elementary flows that are not considered in the 'FETPinf w/o LT'
and
* why is there a large difference in the results between 'FETPinf with LT' and 'FETPinf w/o LT' in ecoinvent v2.2, and why has this apparent difference almost completely been leveled in ecoinvent v3.

Very interesting indeed. Having checked that Umberto NXT LCA has the same values as in the original data on the ecoinvent website, this is actually not an Umberto support case any more, and would probably best be answered by somebody from ecoinvent support.

But you got me hooked there! I find this exciting and will try to find out myself.

[dmutz]

Thank you for your reply!
I started to ask ecoinvent support for an answer, as soon as I get an information about the topic, I will let you know.

[dmutz]

Hello,

still I have no answer to the question above. I am still waiting...

But another question emerged to the issue of LCIA in Umberto:

Is there a possibility to edit the existing LCIA valuation systems or to create own valutation systems similar to the valuation editor in Umberto 5.6?

[pbeilschmidt]

Hello,

in the present version of Umberto NXT there no possibility to create own LCIA methods (as it could be done in the Valuation System Editor in Umberto 5.6). However, this feature is high up on our development priority list and you can expect to have such a feature in a future release. (I can not promise though, if when this will be). We have had several requests from the LCA community, and of course we see the need for such a feature.

Regards,