Home
Welcome to the PyCPT wiki!
PyCPT is a Python library that provides interface and extra functionalities to IRI's Climate Predictability Tool (CPT), with a special focus on mass-production of seasonal and sub-seasonal forecast skill assessment maps and probabilistic flexible forecasts.
The user will need to install Anaconda (Python3), the Climate Predictability Tool and the Python extension of CPT (PyCPT). Detailed instructions are available in this document.
For Windows, we presently recommend to install a Virtual Machine with all needed packages. Instructions and necessary resources can be found here. Read the PyCPTUbuntuREADME.txt file located in that folder for details.
[ONLY needed for the s2s version] Create the file .IRIDLAUTH in the main PyCPT folder. Its content must be the Data Library S2S key obtained via help@iri.columbia.edu
PyCPT is broadly divided in four steps.
Most of the configuration of your particular case happens in the namelist section, found at the beginning of the PyCPT Jupyter notebook. This section enables the user to select predictor and predictand datasets, calibration methods and spatial and temporal domains.
The download and CPT execution section deals with preparing all the needed input datasets and running CPT to produce skill assessment and forecast files, conducting the calibration process selected by the user.
The skill assessment section visualizes predictive skill metrics as maps or text.
And the forecast section deals with the ensemble generation and production of forecast maps and other figures, with a especial focus on flexible formats (the use of the entire forecast probability density function).
Several examples are provided via the Jupyter notebooks available in the Code section.
