Changes (need to update README):

• new parameters for script including function arguments
• plot script included.

Calculate model trends on HPC

This was developed to calculate Mann-Kendall trend statistics on a per-pixel basis over any domain.

Your data requirements:

• Dataset must be in netcdf format
• The dataset must be stored in yearly files (e.g., 1995.nc, 1996.nc, etc.)

Step by step

1. Create your own data dictionary based on the template in data_config.py

Parameter	data type	Description
name	str	Any name of your choice, not relevant
version	tuple[int]	A tuple of the size three, not relevant
path	str	The path to the yearly data files
type_file	str	The data type format. Only netcdf is supported
year_start	int	The first year with available data
month_start	int	The month in the first year from which data is available
year_end	int	The last year with available data
month_end	int	The month in the last year until data is available
leapday	bool	True if leapday (29th February is present in the data)
resolution_time	str	time offset string of data time resolution
grid	str	name of the grid the data is based on, not relevant
variables	list[str]	All the variables available in the dataset
variable_names	dict[str, str]	dict relating common variable names to the ones in the dataset
variable_dimensions	dict[str, list]	The dimensions of the data. Lat and Lon must be last two dimensions
variable_units	dict[str, str]	units of the variables, not relevant
mask_values	None or int or float	Potential mask values in the data, e.g. -9999 that will be considered nan

2. Adjust the settings in calculate_trends.py

1. data: The name of the data dictionary you created before
2. variables: The variable names you want to calculate the trends for
3. outfile: The name of the output file
4. units: The units of the variable for the slope variable

3. Adjust the loaded environment in submit.sh

Please source your environment file in submit.sh in line 8. Importantly, the environment must contain the crusty PyPI package and its dependencies. It can be installed by pip install crusty.

To Do

• Probably it is more convinient to define all the settings in the submit.sh, and parse it as arguments to the pythons script.
• Functionally allow for different pymannkendall functions by parameter setting.