Difference in results

Question

Hello

I have a problem due to a big difference in results obtained by the same model.

In the file enclosed you can see the results one obtained by the model using Ecoinvent 3.12 (the higher results) and the lower one obtained using Ecoinvent 3.12 library.

I used the same method (IPCC 2021 ISO 14067) and the same allocation.

I create a new database and then I add the Ecoinvent 3.12 library created by a fresh Ecoinvent 3.12 one.

In the same database I import my model with its parameters and method and finally I perform calculation of product system.

Why there is the difference? Is something wrong in my flow?

Thank you

https://ask.openlca.org/?qa=blob&qa_blobid=9954992828345608931

Answer 1

Do you have no validation errors (right-click database "validate")? Are all your imported foreground processes correctly linked to the same provider from ecoinvent (please click "Check linking properties" and make sure that there is no processes without default provider set)? Is the required amount and the calculated inventory (the total elementary flow results) the same?

Comments

Database is validated.
Processes are linked but when I check the reply states there are some of them unlinked. I checked twice and there are the correct providers.
Yes required amount and inventory are correct

---

What means "but when I check the reply states there are some of them unlinked"?

When you click check linking properties, there should be no processes without a default provider set. If there are, you should set them. Otherwise they will be set automatically to the first provider in the provider list of the exchange, which can of course lead to different results as when you set the default provider by yourself.

---

All processes are linked and I checked the correct provider.
But the results are diferent

---

Difficult to say where it is coming from without seeing the model. You can check the contribution tree and look at "C4: Manutenzione infrastruttura" to see why there is a 20% difference. We must know if it is due to a difference in inventory or due to a difference in characterization (factors for the environmental impacts).

---

Checked again and infact there was a different link in provider, solved, but now the difference is around 14%.
It's the only value out of control (C4: Manutenzione infrastruttura) all the other are ok now.

---

If this is the only one, and it is only for one sub-processes, but the other ones are fine, it is very likely that the provider are still not the same.

To be on the safe side, you can create a new product system and select "only link default provider". And make sure that all provider are the same and none is missing by using the "check provider linkings" feature.

---

I did as you suggested, but nothing the difference is still evident.
Product system created from the original model gives the result of 6.97385E7 kgCO2e
Product system created from the parametric model + Ecoinvent 3.12 library gives 5.85276E7 kgCO2e.
Both product systems are created by the process C4: Manutenzione infrastruttura

---

Just take a random ecoinvent process and do the same test. I'm sorry but I can't help for your own processes, where I don't know the details and where you first wrote  that you checked and the provider are correct and then later wrote that there was a different link in the provider.

You can also see in the contribution tree where the difference is coming from. You can see there if the difference is from a flow, from an amount value, from a characterization factor, from a different provider etc.

If nothing works and if the issue is not only with your own processes (that I don't know), you can send a mail with more details on the process. But because you said that "It's the only value out of control (C4: Manutenzione infrastruttura) all the other are ok now.", it is most likely an issue that is related to your own process that you created.

---

Thanks a lot for your help.
First wrote  that I checked and the provider are correct and then later wrote that there was a different link in the provider because at a first sight everything seem ok but then I drilled down the search and I discovered that some links were missing and it's a real problem when it happens.
I think it's a bug in the system because if I create a model using Ecoinvent dataset I must be sure that these links will be mainteined when I import it in a database which cointains the same Ecoinvent library.
I do not have changed nothing and nothing must be lost or modified. I do not have to check it again.
I'm trying to build a tool to be used for the systematic approach under ISO 14067 so I need a database mixed with Ecoinvent library (I do not want changes made by any users) and a parametric model which allows studies of different products in the same family.
The tool must be error proof and I can not imagine to check everytime links because it's wasting time.
I particular C4: Manutenzione infrastruttura is the original dataset of Ecoinvent called railway construction and I'm using it as manintenance, the only things I did is change it's name and erased copper inside because I do need it. Thank you very much for your help and patience.

---

I don't know all the details. But you only need to check the linkings once. And as I said, there is also the check linking properties tool in openLCA. Also what I mean with checking an ecoinvent process is that if you get the same results in the ecoinvent library and the ecoinvent database (which you should), the reason for your differences are most likely in your own imported and exported foreground processes. And a copied and modified ecoinvent process is also an own foreground process. Only what is immutable in the ecoinvent library counts as background.

I will test the transfer for bugs, since I don't use the ecoinvent libraries often. But I understand your use case that you want to create an immutable, error proof tool. By the way, you can also transform any database that you want, also your own models or full databases, into a library by yourself. This functionality is in openLCA. And with one of the next versions of openLCA, there will be also a migration tool included, which makes migrating own processes from one database (even if different ecoinvent version) to another much easier.

---

Thank you very much.
Your help is precious to improve skill to use openlca.

---

I worked and drilled the question and I find that in Ecoinvent 3.12 library downloaded from Nexsus and library created from Ecoinvent 3.12  database without modifications there are two strange things.
In processes in both libraries
market for used railway track | used railway track | Cutoff, U CH and
treatment of used railway track | used railway track | Cutoff, U
have in output two times the flow used railway track.
But in the same processes inside Ecoinvent database the flow "used railway track" appears in output only one time.
I corrected in the libraries and the result now differs only for 0,21% - in real world nothing.
I think it's the problem in creating Json file during library creation.