Forum for ecoinvent Version 3

USEtox characterisation factors - deviations between v2.2 and v3.1

Written on: 16.02.2015#1

Author:
dmutz, Daniel Mutz

Hello,

for human toxicity potential and ecotoxicity potential I want to use USEtox model.
In some projects I am currently using ecoinvent v3.1 database. But in other projects ecoinvent v2.2.

Now, I noticed strong deviations in CF by using both databases (separately). For instance, the characterisation factors for same materials with (more or less) the same emission type are completely different, e.g. water/river or water/surface, respectively.

Is this described fact just an error within the database or does ecoinvent v2.2 really account USEtox indicators with other CF than in v3.1?

An example (for ecotoxicity CF for USEtox w/o LT):

1,4-Butanediol:[list]no CF (v2.2) - emission into water, river[/list][list]7,31 CTUe (v3.1) - emission into water, surface water[/list]
Cadmium, ion:[list]no CF (v2.2) - emission into water, river[/list][list]9 712 CTUe (v3.1) - emission into water, surface water[/list]
Mercury:[list]22 090 (v2.2) - emission into water, river[/list][list]22 090 CTUe (v3.1) - emission into water, surface water[/list]
In this example: Mercury does have the same CF, but Cadmium has not. Why?

First, I had the suspicion that it is an issue of the software Umberto, I am currently using. But the support of Umberto referenced me to you (ecoinvent), since they noticed the same problem already.


Hopefully, you can give me some advise!

Kind regards!
Daniel

Written on: 17.02.2015#2

Author:
guillaumebourgault, ecoinvent

Dear Daniel,
There has been a revision of the implementation of the methods between the two versions, and the difference you notice were introduce to correct mistakes.
We will publish enhanced documentation with the next release (3.2) that will make it easier for the users to understand why the elementary exchanges are paired with their CF.
In the meanwhile, I can only tell you that the implementation in 3.1 is more reliable than the previous ones, but this is not a guaranty that there is no mistakes in the 3.1 implementation. The methods amount to a total of 170 000 CFs, so we need users with an eye for the details, like you, to correct them.
Could you send me the list of CF you have noticed have change, I will look at it and make sure that if you spotted a mistake, I will correct it in the next implementation.
Kind regards,
Guillaume Bourgault
Project Manager, ecoinvent
bourgault@ecoinvent.org