Climate-REF / Climate-REF/climate-ref
Tracking provenance of metrics
- Dominant language
- Python
- Stars
- 26
- Forks
- 14
- Avg merge
- 1d 10h
- Merged PRs (30d)
- 36
Description
## The problem
As a downstream user I want to be able to identify the datasets that were incorporated into a metric calculation.
We also need to know this information as part of the framework to correctly recalculate metrics.
An end user of this provenance information will also be IPCC authors.
Each metrics package may have slightly different approaches to this. ILAMB includes some information in the CMEC output, but not a ton. ESMValTool has a more advanced metrics
We need to have a position on what we capture.
The provenance capture could be done somewhat automatically for the input datasets/obs since that information is used to trigger a calculation. There might be additional information that we could capture inside the metric calculation.
We also need to think about how this info might be nested if we roll up metrics that depend on one another.
One option may be that the prefilled provenance information could be injected into the `Metric.run` function at runtime. Additional provenance options could be added to that object during execution.
## Definition of "done"
* [ ] Describe what we want to capture and how (RFC?)
* [ ] Implement an initial prototype
Contributor guide
Assessment
This issue has not been assessed yet.