How to carry out a Monte Carlo analysis correctly using the ECOINVENT 3.9.1 database

Question

Hello everyone, I wanted to ask how to carry out a Monte Carlo analysis correctly using the ECOINVENT 3.9.1 database. Trying to do an analysis, very different results come out compared to those of the calculation. I have selected the uncertainty curves for all the flows used in my analysis (LCA).

Can you help me, thanks.

Answer 1

Good post and question, there were indeed two issues:

1, there are mistakes in the original ecoinvent files regarding the uncertainty, which led to negative results in the simulation and thus often to totally off results; we fixed this recently with a new release of ecoinvent 3.9.1, ecoinvent plans to fix this in the next release on their side if I understand correctly. You need to download the new version of 3.9.1 from nexus to fix this.

2, in openLCA 2.0.0 there is a memory leak, which leads to ever increasing memory use in the simulation. We fixed this in the new 2.0.1 version (available on openLCA.org).

Thank you!

Andreas

Comments

Thanks a lot for the tips Andreas

---

did this impact just monte carlo simulation, or all results?

---

the memory leak was not really visible in a normal calculation, the wrong uncertainty values are of course also there in normal calculation but maybe do not look at them much without a Monte Carlo simulation.

---

Dear Andreas and Cipriano,
I believe I downloaded the latest Ecoinvent 3.9.1 on Nexus (ecoinvent 3.9.1 Cutoff UNIT Regionalized), however I still get results that are way off when I run a Monte Carlo simulation. Did it work for you Cipriano? Andreas, was it fixed for all 3.9.1 databases?
Thanks a lot!
Sophie

Answer 2

Hello Cipriano,

I think initially it was you who reported the bug with uncertainties in ecoinvent. In your current example, I can see from you picture that your providers are mostly system processes (LCI). You need to use the unit process (UPR) database from ecoinvent in order to have the uncertainties of all the product flows in the supply chain, which is then used for the monte carlo analysis.

Best wishes,

Conrad

Comments

Thanks a lot for the tips Conrad!!