The code can be tested using pytest:
pytest -s test_algorithm.py
This will run the functions contained in test_algorithm.py that begin with test_ (which is all of them).
There are two functions used for detecting the drought events. One uses a traditional looping structure, the other uses loop over a smaller number of pre-identified matching locations to speed up execution.
As tested it takes about 12 minutes to calculate the required metrics from a single CMIP5 dataset which is 128x64 pixels with 50735 data points at each pixel.
Run the program like so:
python seas_flash_drougth_count_CMS.py
There are three output files:
-
result_{model}.nc: the same dimensionality as the input data with moisture levels replaced by duration of a flash drought at the date the drought started (last wet day) -
result_{model}_seasonal.nc: a count of the number of flash droughts per season -
result_{model}_decadal.nc: a count of the number of flash droughts per decade