Help to build a basic HindcastEnsemble with correct metric calculation.

[gmfarache]

Hello,

I want to get a MSE metric from a forecast dataset in comparison with a observation dataset. For this I'm using the HindcastEnsemble class.

Both datasets are simple. The simulation dataset has one init coordinate value (which is the simulation date) and 5 lead coordinate values (the forecast time steps). So I have the model forecast for days 2022-12-02, 2022-12-03, 2022-12-04, 2022-12-05 and 2022-12-06. The observation dataset has 5 time coordinate values (which are the dates that I want to calculate the MSE). The problem is that the metric is being calculated for each position of the variable list, not for all of them as a group.

The result that I was expecting is array([47.05431175]). What am I doing wrong?

Obs: if I use data.transpose() instead of data as parameter to HindcastEnsemble, the result is the same

[aaronspring]

Hi @gmfarache, thanks for reaching out and using climpred.

Sounds like you expect to get an average mse over all leads. However climpred loops over all leads and therefore you get the 5 values by design.

I might be able to help more if you outline your use case a bit more, I.e. what's your real data use case with dimensions etc

Not sure whether just taking the mean over lead is helpful advice.

[gmfarache]

You're right. I was understanding wrong the climpred idea, now it's working as it should.
Thank you so much