Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Missing data in an observed flows will give a nan for the Nash-Sutcliffe #452

Open
Mayetea opened this issue Jan 11, 2023 · 2 comments
Open

Comments

@Mayetea
Copy link
Contributor

Mayetea commented Jan 11, 2023

I was testing the calibration of models with the HYSETS_2020_ERA5 dataset and I found out that when a data is missing in the observed streamflow, the calibration fails and the Nash-Sutcliffe is set to nan. It gives back the calibrated parameters but I'm not sure they can be considered valid since the "fitness" is null. I don't know if this behavior is just when a flow is missing or when any data is missing, since the temperature and precipitation are all filled properly.

Finally I don't know if its Ouranos or a backend issue, so I'll open the issue here, but we can move it if its a backend issue.

Here's some image of the issue :

Streamflow missing :

QObsMissing

Result :

image

@Mayetea
Copy link
Contributor Author

Mayetea commented Jan 11, 2023

If you want to reproduce the issue, the watershedId that I took was 23.

@huard
Copy link
Contributor

huard commented Jan 11, 2023

The NSE is computed by RavenC itself, not the Python wrapper, so I suspect there isn't much we can do here, except raising the issue with James.

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants