null Learning how to configure Delft-FEWS

Learning how to configure Delft-FEWS

I still remember the first time I looked into the configuration of a Delft-FEWS application. It was fascinating seeing all these different XML files which piece and puzzle together a forecasting system. Configuring Delft-FEWS can be challenging in the beginning, but it is definitely a skill that everyone can learn. It normally starts with looking at the FEWS wiki and following a basic configuration course. After the COVID pandemic we quickly realized that we could no longer offer Delft-FEWS courses the traditional way. Travel restrictions made it impossible to organize instructor-led courses. Existing plans to improve course content and moving towards an online learning experience where accelerated. This has led to the development of the Delft-FEWS eLearning platform.

On this platform you will find a:

  • Basic Configuration Course: with many basic modules from configuring modules, workflows to displays and the forecast tree

  • Advanced (New) Features Course: with different advanced modules on existing and new Delft-FEWS configuration, including spatial data displays, dashboards and timeseries statistics.

Check the video below for a brief introduction on the Delft-FEWS eLearning platform.

The platform contains many modules on how to configure Delft-FEWS, below are some expert tips to improve the quality of your configuration!

  • Don’t invent everything yourself, but copy from other applications. Build or manage your own Delft-FEWS system, but use other systems as inspiration.

  • Become acquainted with the Delft-FEWS Wiki and know where to find example configuration files.

  • Don’t try to configure everything at once, but break the configuration down in parts that interact with each other: Modules vs Workflows: DisplayGroups vs Topology, General Adapter vs Models.

  • Use process flow diagrams to visualize how different workflows and processes within workflows are connected.

  • Make use of templates within your configuration for repeatable tasks.

  • Split configuration of workflows into different categories, historical vs forecasts, pre-processing vs model runs.

  • Make use of ensemble IDs to differentiate between different NWP products (e.g. DWD, ECMWF, GFS) both for deterministic as well as ensemble products.