This repository provides a Python implementation of the Gaussian Mixture Model (GMM) algorithm for detecting extreme events in time series data.
Paçal, A., Hassler, B., Weigel, K., Kurnaz, M. L., Wehner, M. F., & Eyring, V. (2023). Detecting Extreme Temperature Events Using Gaussian Mixture Models. Journal of Geophysical Research: Atmospheres, 128, e2023JD038906. https://doi.org/10.1029/2023JD038906
Corresponding Author: Aytaç Paçal, [email protected]
To use this implementation of the GMM algorithm, you will need to have Python 3 installed on your system. You can use this repository with ESMValTool environment. Follow the installation steps on the ESMValTool installation page. Alternatively, the environment_v2.5.0.yml file for the ESMValTool version(v2.5.0) used in this study is provided in this repository. You can create a new environment using this file with the following command:
mamba env create -f environment_v2.5.0.yml
The CMIP6 datasets, which are a collection of climate model simulations produced by research institutions worldwide, are available on the ESGF servers.
Regional grid files are generated using the recipe with ESMValTool, which is a community-driven Python-based evaluation tool for Earth system models. ESMValTool provides a set of predefined recipes for the evaluation of various aspects of Earth system models. Shapefiles for IPCC (Intergovernmental Panel on Climate Change) regions are freely available at the ATLAS GitHub repository. Separate shapefiles for each region should be placed into the auxiliary_data directory of the ESMValTool.
The GMM algorithm is implemented in the diagnostic script for ESMValTool v2.5.0. You can use the ESMValTool tutorial to learn how to run a recipe. The recipe can be run using the following command:
esmvaltool run recipe_gmm_ssp.yml
This script takes the CMIP6 data, applies preprocessing functions, and creates regional output files. Then, the diagnostic script takes the ESMValTool output as input and produces the GMM results for each grid cell and each model.
Return periods for regions under different Global Warming Levels (GWL) are calculated using the return_period.py script. This script takes the GMM results from the ESMValTool recipe and produces the return periods for each region and each model under GWL scenarios. INPUT_PATH is the path to the ESMValTool output directory, and OUTPUT_PATH is the path to save the results.
return_period.py [-h] [INPUT_PATH] [OUTPUT_PATH]
Figures in the paper can be plotted using the result_plotter.py script. This script takes the return periods as input and produces the figures in the paper. [INPUT_PATH] is the path to the directory containing the regional return period files. Files should be named as {region_abbreviation}.csv. The script will produce the figures and save them in a new plot directory.
result_plotter.py [-h] [INPUT_PATH]
This code is released under Apache 2.0. See LICENSE for more information.