Welcome to MurCSS User Guide!

Installation

Download and install MurCSS via pypi:

pip install murcss

Or you can clone this repository:

git clone https://github.com/illing2005/murcss.git

Requirements

MurCSS needs some additional python packages. If you install MurCSS using pypi, they will be downloaded and installed automatically.

• Matplotlib >= 1.1.0
• Basemap
• NumPy >= 1.5.0
• SciPy >= 0.8.0
• CDOpy (is now included in MurCSS sources)
• scientificpython

Additionally MurCSS requires a working binary of the Climate Data Operators ( CDO >= 1.5.4 ).

Usage

After installation you can use the tool typing murcss in your terminal.

Help:

murcss --help

Now you can see all parameters accepted by MurCSS.

Outside of MiKlip:

If you are using MurCSS outside of MiKlip, you have to set the file_system = ‘custom’ and specify your data structure in murcss_config.py.:

#If you are working on your own filesystem specify the folder structure:
DRS_STRUCTURE = {
      "root_dir":"/scratch/ROOTDIR_OF_YOUR_DATA/",
      "parts_dir":"project/product/institute/model/experiment/time_frequency/realm/variable/ensemble/file_name".split('/'),
      "defaults" : {}
     }

Generate Sample output:

To generate the sample files in /sample_output/ download the files in /sample_data/ . Adjust the DRS_STRUCTURE in murcss_config.py to your system. For the comparison with real observations you need to download e.g. a HadCRUT dataset from the Met Office Hadley Centre. Webpage and user informations: http://www.metoffice.gov.uk/hadobs/hadcrut4/data/current/download.html Download and unzip e.g. the median of the HadCRUT4 dataset:

cd /observations/path/
wget http://www.metoffice.gov.uk/hadobs/hadcrut4/data/current/gridded_fields/HadCRUT.4.2.0.0.median_netcdf.zip .
unzip HadCRUT.4.2.0.0.median_netcdf.zip

Now you can generate the sample output using:

murcss variable=tas project1=miklip product1=initialized institute1=mpi-m model1=mpi-esm-lr experiment1=initialized project2=miklip product2=uninitialized institute2=mpi-m model2=mpi-esm-lr experiment2=uninitialized decadals=1960,1965,1970,1975,1980,1985,1990,1995,2000 leadtimes=1,2-5 result_grid=r72x36 observation=/PATH/TO/OBSERVATION_FILE metrics=all

Inside a Python Script:

MurCSS constits of two types of metrics, accuracy and ensemble_spread. The first is located in metrics.msss and the second in metrics.crpss Here is a simple script to calculate the accuracy part:

from metrics.msss import Msss
msss = Msss(**parameter_dict)   #Create an instance of Msss Class
msss.prepareInput()             #Search and prepare input Files (ie. remapping)
msss.analyze()                  #Calculates the actual skill scores
msss.deleteCache()              #Delete all temporary files

First we create an instance of Msss. Msss takes a bunch of keywords arguments, which can be separated in two parts:

1. arguments to find the input data
2. specific options for calculation

For a complete list and description of the keyword arguments see metrics.msss.Msss. Then we search and prepare the input data for the final calculation. The prepared files will be stored in msss.input1Remapped, msss.input2Remapped and msss.observationRemapped for later use. The main calculation is done in msss.analyze() (see metrics.msss.Msss.analyze()). After the calculation is finished we delete the working direcory.

And here for the ensemble spread part:

from metrics.crpss import Crpss
crpss = Crpss(**parameter_dict) #Create an instance of Crpss Class
crpss.prepareInput()            #Search and prepare input Files (ie. remapping)
crpss.analyze()                 #Calculates the actual skill scores
crpss.deleteCache()             #Delete all temporary files

crpss.prepareInput() stores the input files in crpss.inputRemapped and crpss.observationRemapped. As you can see the scripts looks almost the same. The major difference is that crpss expects only data information about one model (see metrics.crpss.Crpss).

Unittests

Download the files in /integration/tests/ and navigate to the directory. You should also download the sample_data and adapt the DRS_Structure to run the tests for the file input component. Now run the tests using:

python -m unittest discover . '*_test.py'

Currently some of the tests are designed for the MiKlip file system. Therfore they are commented out.

Support, Issues, Bugs

Please open an issue on GitHub or write an email to sebastian.illing@met.fu-berlin.de