Dealing with observation uncertainty

## The motivation

All observations have errors and uncertainties. Calculating model evaluation metrics against a single observation dataset means the result is a mix of model error and observation uncertainty.

We would like to do the evaluation in a way that accounts for the observation uncertainty so that the resulting metrics represent the model error only.

## The solution

This can be achieved using the Observation Range Adjusted statistics

J P Evans and H M Imran 2024 Environ. Res. Commun. 6 071001
DOI 10.1088/2515-7620/ad5ad8

It would be great if this method was available through the REF

## Alternatives

## Additional context