Challenge 33 - Adjusting Climate Projections

[jwagemann]

Challenge 33 - Adjusting Climate Projections

Stream 3 - Applied data science for weather, climate and atmosphere

Goal

Develop a Python package that supports the estimation of biases and uncertainties of future climate projections.

Mentors and skills

• Mentors: Edward Comyn-Platt, Chiara Cagnazzo, James Varndell
• Skills required:
  • Coding skills in Python
  • A good level of statistical mathematics

Note: Challenge is funded by Copernicus. Only nationals from European Union (EU) Member States and countries associated with EU’s Space Programme (currently Iceland and Norway) are eligible to participate (see Terms and Conditions).

---

Challenge description

The Climate Model Intercomparison Project 6 (CMIP6) provides a suite of models which are the state-of-the-art representation of the Earth system and simulate how the climate conditions are going to evolve over the next decades. For several applications, it is essential that the CMIP6 model outputs biases and uncertainties are correctly accounted for.

This project will create a python package which contains the tools required for estimating the biases and uncertainties, that use past observations to adjust future projections. The python-package should be stand-alone and operate on the data at a numeric level such that it can be used on data from any source (for both the reference and projection components). Such a package could be imported into any python script/session and would allow use of the specific bias correction methods required for their application without the need for explicit bias-correction processing. This package would be used for a range of ECMWF/C3S applications and could become a standard toolset for performing such analyses.

The current ISIMIP software does this, however, this is implemented as a tool for batch processing data in a specific format. This software should be turned into a python package that can be hosted publicly on the ECMWF Github and installed via PyPi.

References

• ISIMIP3b, used to bias correct and downscale CMIP6 output – description and code
• ISIMIP2b, used to bias correct and downscale CMIP5 output – description and code

[vidurmithal]

Hello!

The current implementation of these calculations in the ISIMIP software is done using CDO. If I understand correctly, then in the developed Python package, the data processing and calculations should be encoded explicitly (using core Python packages like xarray and numpy) rather than calling the CDO operators. Is this understanding correct?

Thank you!

[EddyCMWF]

Hello!

The current implementation of these calculations in the ISIMIP software is done using CDO. If I understand correctly, then in the developed Python package, the data processing and calculations should be encoded explicitly (using core Python packages like xarray and numpy) rather than calling the CDO operators. Is this understanding correct?

Thank you!

Hi @vidurmithal ,

Yes that is the idea, we would like the tools to be more stand-alone and flexible for use on any input data.

The ISIMIP3b does not use cdo for the conversion, it uses iris which is already a head-start in doing this. I have just seen that the code is also included in the ISIMIP3 link, so I have updated the project description.

ISIMIP2 code, which does use cdo, is a little older than ISIMIP3, so we would like to start with the ISIMIP3 code and create a generalised python package.

All the best,
Eddy