Inquiry Regarding GCM Adjustment Thresholds in PyGEM

[Ruitangtang]

Hello PyGEM Community,

I encountered an error related to the GCM adjustment for future simulations based on CMIP6 data. This issue seems to stem from the code in gcmbiasadj.py (see lines 304-306).

I have a question regarding the threshold values set to 0.5 and 2. These thresholds appear to be causing errors for specific glaciers under certain GCMs and scenarios.

Could anyone provide insight into the rationale behind these threshold values?

@drounce , david, any suggestions or comments on how to resolve the error.

Thank you very much.

[drounce]

Yes, these thresholds were put in place to avoid drastically updating the precipitation data, which I defined as a factor of two. Where is this happening? When I put these assertions in place it ensured that we were only getting reasonable precipitation values and flagged glaciers where there were potential issues. Are you confident that the bias adjustment is doing what you're hoping it's meant to do? Have you checked that you want it to be adjusted by more than a factor of 2? If so, you could comment out and remove this assertion. Another alternative is to change the period that you're looking at. @btobers can help if you're using the pip install and thus don't have access to the code base to easily modify like you would in a development mode.

[btobers]

I think that values you are getting sound entirely reasonable - particularly since they're so close to the hard-coded threshold. As @drounce mentioned, the issue with this threshold is that a value of 0.49 such as you have is rejected while a value 0.50 is accepted.

And yes, the min will equal the max if you are processing a single glacier at once (or one glacier per core). If you are running multiple glaciers on a given core, then all the bias adjustment factors are calculated simultaneously and so there will be a range of values returned by this function.

[Ruitangtang]

That makes sense. I processed one glacier at a time per core, and I implemented parallel processing within each individual glacier's analysis. Thank you!

[drounce]

@btobers has set up a separate issue, so I believe this should be taken care of now.

[btobers]

This issue was addressed in PyGEMv1.0.4
https://github.com/PyGEM-Community/PyGEM/releases/tag/v1.0.4