Pathfinder

A Bayesian-inferred simple climate model.

How-to

Download a release. Read The Fine Manual.

Pathfinder has been developed in Python 3.7 and run preferentially through IPython. Currently, packages required to run it are numpy (v1.19.2), scipy (v1.5.2) and xarray (v0.16.0), and it has hard-coded dependencies on pymc3 (v3.8) and theano (v1.0.4) that are in fact used only for calibration. Newer versions of Python or these packages are likely to work, although they were not tested.

Known issues

• The model requires a high number of substeps to remain stable under high CO2 (because of the ocean carbon cycle). This can be set using the nt argument when calling run_xarray.

• The temperature-driven mode (Tdriven) is extremely sensitive to its forcings: it can be very difficult to make it transition smoothly from historical to projections. This is unavoidable because mathematically it requires the second derivative of T and the first derivative of ERFx as input.

• Unclear whether the my_AR1 class from cls_calib is actually needed.

Changelog

v1.0.1

Exact same physical equations and numerical values as v1.0.

• Added: best-guess parameters and outputs (in internal_data/pyMC_calib/) for single-configuration runs.
• Improved: README and MANUAL files.

v1.0

Exact model described by Bossy et al. (subm).

v1

First release!

References

v1.0 (full) | Bossy, T., T. Gasser & P. Ciais. "Pathfinder v1.0: a Bayesian-inferred simple carbon-climate model to explore climate change scenarios." Geoscientific Model Development (submitted). doi:10.5194/egusphere-2022-802.