This repository contains examples for using FDB and Polytope to access ICON forecast data and process it using meteodata-lab.
The directory notebooks contains the Jupyter notebooks and the directory nwp_polytope_demo contains examples of processing data as a Python service.
Forecasts available in FDB
[!NOTE] The realtime FDB normally contains just the latest day of forecasts. This means the FDB requests should usually use date = today and time as either 0000, 0300, 0600, 0900 etc. (given the forecasts run every 3 hour) for ICON-CH1-ENS and 0000, 0600, 1200 etc. for ICON-CH2-ENS. The data is usually available in the FDB a couple of hours after the forecast start time. If the requests to FDB return no data, see these instructions for how to query yourself which data is currently archived in the FDB.
FDB allows to retrieve any multi-dimensional dataset of the recent real-time ICON NWP forecasts. FDB is designed to access full horizontal fields (feature extraction is not supported) and it is only accessible from within CSCS.
FDB data access requires the FDB libraries. In order to facilitate the use of the notebooks, we provide a jupyter kernel configuration for VSCode that will load the environment with the required libraries:
bash host/install_kernel.sh
(you might need to reload the window Ctrl-Shit P -> ‘Developer: Reload Window’ in order to let VSCode recognize the newly installed kernel)
Ctrl-Shift P -> ‘Notebook: Select Notebook Kernel’ -> ‘Select Another Kernel’ -> ‘Jupyter Kernel’ -> ‘Polytope demo’
If you would like to develop python examples outside of the juypeter notebooks, see TODO how to use the FDB uenv environment .
The following notebooks demonstrate how to access ICON model data (ICON-CH1-ENS & ICON-CH2-ENS) through FDB:
Polytope allows to efficiently extract specific features from the same real-time ICON NWP forecast, such as grid point data, time-series, vertical profiles or polygons. Polytope is an HTTP service and therefore access is not restricted to CSCS (it supports access from LabVM and ACPM).
You can follow the same installation of jupyter kernel as with FDB or deploy your own python environment and
poetry install
The following notebooks demonstrate various use cases to access model data (ICON-CH1-ENS & ICON-CH2-ENS):