Repository Authors: Jhayron S. Perez-Carrasquilla and Maria J. Molina (University of Maryland, College Park)
Weather regimes are large-scale, persistent, and recurrent atmospheric patterns useful for prediction at subseasonal timescales due to their impact on surface weather anomalies (e.g., Molina et al., 2023; Perez-Carrasquilla and Molina, 2025).
These weather regimes are often defined using 500-hPa geopotential height anomalies over a specified domain (e.g., North America) and k-means clustering.
Here, we demonstrate how seemingly minor differences in preprocessing choices can lead to a different number of preferred subseasonal weather regimes over North America with different spatial characteristics.
Weather regimes are computed using ERA5 data that is regridded/resampled following 0_ResampleRegridERA5.py.
To highlight preprocessing sensitivity, we will calculate the North American weather regimes in three different ways contained in the following Jupyter Notebooks:
ExpC_WRCompClean.ipynbExp1_StdDevOfEachPixel.ipynbExp2_19792023.ipynb
The Python environment used to run associated software is contained in wrpreprocessing_env.yml.
We would like to acknowledge high-performance computing support from the Derecho system (doi:10.5065/qx9a-pg09) provided by the NSF National Center for Atmospheric Research (NCAR), sponsored by the National Science Foundation.
- Computational and Information Systems Laboratory. 2023. Derecho: HPE Cray EX System (University Community Computing). Boulder, CO: NSF National Center for Atmospheric Research. doi:10.5065/qx9a-pg09.
- Paper/preprint citation forthcoming.