Flood Forecasting and Early Warning: An Example from the UK Environment Agency - Reducing Disaster: Early Warning Systems for Climate Change - page 190

Delft FEWS
Database Layer

General Adapter

XML
Model Forcing
Input State Files
Runtime Parameters

XML
Model Forcing
Output State Files
Run Diagnostics

Execute
Execute

Execute

Model Adapter

Model Adapter

Native Model Format
Model Forcing
Input State Files
Runtime Parameters

Native Model Format
Model Forcing
Output State Files
Runtime Diagnostics

Model
Executable
- or -
Java Class

Fig. 10.8 Concept behind linking Delft-FEWS with external models and the fl ow of data through
XML and native model formats using solid lines. Dashed lines indicate executable commands
(Werner et al. 2013 )

This is undertaken fi rst on a 'standalone' system. This is then uploaded to the 'oper-
ational' system. The NFFS operational system is set up as a client-server applica-
tion. Forecasters use a client application from their offi ce PC to visualise data,
schedule forecast runs and generate “what if” scenarios. All the 'effort' is computed
on the server side. To build resilience, the server architecture has been designed with
parallel duty servers located at the Environment Agency's two data centres.
Real-time hydrological and meteorological measurements, weather forecasts
and climate predictions are fed into the server. Computational processes occur, and
the results are returned and visualised to the client at their PC. The philosophy
behind this system is to provide a shell through which an operational forecasting
application can be developed specifi c to the requirements of an operational forecast-
ing centre (Werner et al. 2013 ).