Core Concepts — Overview

If you learn nothing else before opening a config file, learn this section. Delft-FEWS has a reputation for being hard to configure. Most of that difficulty isn’t the XML — it’s that the XML only makes sense once you hold a particular mental model in your head. This section installs that model.

The one-sentence version

Delft-FEWS is a machine for moving time series from the outside world, through a standardized internal representation, into forecasts and displays — and every config file you’ll ever write is configuring one hop of that journey.

Read that again. Almost every question — “where does this value come from?”, “why isn’t my data showing up?”, “which file do I edit?” — is really a question about which hop you’re looking at.

The four things FEWS needs to know

Before FEWS can do anything useful, its configuration has to answer four questions. Every configuration, however large, is mostly these four answers repeated at scale:

1. What are the things I’m tracking? — the locations (a gauge, a reservoir, a catchment) and parameters (discharge, water level, precipitation). See the data-flow model.

2. Where does data come from, and what is it called out there? — imports and ID mapping, which translate an external system’s names into FEWS’s names.

3. What should I compute from it? — transformations, models, and the workflows that run them in order.

4. How should people see it, and where does it go next? — displays, thresholds, and exports.

How this section is organized

Page	What it gives you
How data flows through FEWS	The central diagram: external source → import → internal time series → process → display/export. The backbone of everything.
The configuration directory	What the folders and files on disk are, and how FEWS finds them.
The forecasting lifecycle	What actually happens when a forecast runs, and where each config file plugs in.
Glossary	Plain-language definitions of the terms the official docs assume you know.

Tip

Read them in order the first time. After that, the Glossary and data-flow diagram are the two you’ll come back to.

A note on vocabulary

FEWS uses some words in specific ways that trip people up. Three worth knowing before you continue — full definitions are in the Glossary:

• A time series in FEWS is never just “some numbers.” It’s numbers plus a full identity: which location, which parameter, which module produced it, which time step, and whether it’s observed or forecast. That identity is called a time series header, and it’s how FEWS keeps millions of series straight.
• A module is a configured unit of work — an import, a transformation, a model run. Modules are the verbs.
• A workflow is an ordered list of modules. It’s the recipe that says do this, then this, then this.

Once these click, open the data-flow model →