This repository supplements the manuscript by
Sebastian Lehner 
This repository provides a toolkit to compute climate indicators for applications in climate impact research.
The implementation leverages xclim and self-implemented algorithms for index computation.
xclim is a library on top of xarray for the calculation of climate indicators.
Furthermore, this repository contains default configs for the following meteorological variables and their corresponding data source:
- Temperature: https://doi.org/10.60669/m6w8-s545
- Precipitation: https://doi.org/10.60669/m6w8-s545
- Evapotranspiration: https://doi.org/10.60669/f6ed-2p24
- Sunshine duration: https://doi.org/10.60669/m6w8-s545
- Snow and runoff: https://doi.org/10.60669/fsxx-6977
The routines are flexible and the source of the data can be configured to any data readable with xarray, by changing the configuration parameters in config.toml. Note though, that for different datasets, some adaptations to deal with differently structured data likely has to be done.
This project has been set up with uv. After cloning this repository, start with uv sync to install the dependencies into a local .venv environment (if local quota is an issue, you can also symlink .venv to a different location (e.g. your /perm, or /scratch); uv will still work correctly). Then, uv run src/main.py, or better, uv run pytask (see Automation below) can be used to execute the calculation of climate indicators. Note that all the configuration is done within config.toml.
In order to quickly verify if the calculation of indices was succesful, use uv run src/check_files.py. Thereby, based on the configuration within config.toml, the number of existing files in the output path is checked against the expected number depending on the years specified. As validity criteria, a filesize of 100 KiloBytes is used. Note that no checks on the data itself is being done.
This repository includes automation by using pytask, which is installed automatically by the given environment via uv. Simply run uv run pytask to execute all automated tasks. Some pytask configuration can be found and adapted within the pyproject.toml file. For parallel processing, either update the config file, or use uv run pytask -n NUM_WORKERS. Note that pytask by default captures all stdout and stderr, which means that it will not properly be logged within the logfiles. To prevent this and log all output correctly to the logfile, use the -s option: e.g. uv run pytask -s.
Climate indicators can be calculated by using the defined abbrevitation from the index column in doc/indices.csv, within the config.toml file in the [CLIMATE_INDICES] section to calculate only a subset of climate indicators. Otherwise, all climate indicators listed in doc/indices.csv are calculated.
The list of available climate indicators contains Climdex, BIOCLIM and various other indicators (full list in doc/indices.csv).
Aggregation and evaluation can be run by uv run src/climind_eval.py. This script also takes the configuration from the config.toml file for which indicators it should be executed. There are 3 steps that are executed: 1) calculation of spatial averages, 2) calculation of climatological means (for the two periods 1961 to 1990 and 1991 to 2020), and 3) a two-tailed Mann-Whitney U-test for statistical significant changes between the two climatological periods. The calculated data is saved into the output directory specified in config.toml.
Visualisation intakes the calculated climate indicators and/or the aggregation/evaluation data depending on the plots. There following visualisations are implemented and can be explored in this showcase document: 1) time series of spatial averaged anomalies for annual/seasonal aggregations, 2) spatial maps of the climatologies and their difference, 3) stampplots and anomaly stampplots of the climate indicators, 4) grouped significant changes plots, and 5) grouped warming stripes. The groupings for 4) and 5) are based on the base input parameter for each indicator (see the linked showcase above for more details).