- 17th June 2020
- Posted by: Emilia Gates
- Category: Projects
Client: Vietnam Meteorological and Hydrological Administration
Partner: Kisters AG and Harmony Soft Environmental
This project formed part of a large World Bank funded project for the Vietnam Meteorological and Hydrological Administration to strengthen weather forecasting and early warning systems. We delivered a national scale forecasting system integrating models of both the river and the sea.
As we see in so many organisations, over time VNHMA had built up a significant number of modelling tools and techniques for coastal and flood forecasting across Vietnam. Data acquisition and preparation for models was time consuming and potentially prone to errors. Computational resources to execute model simulations were mostly on local computers and sharing of simulation results was difficult. This means that the system relies on particular individuals to execute a forecast, rather than a robust and reliable systematic process.
The solution was to deliver a client-server based forecasting system on dedicated hardware. This means that daily tasks such as importing and processing data, executing models, and sharing results are all done automatically. Results are held in a central database, so all forecasters can share the latest data and model simulations. A user interface configured to the needs of the forecaster groups, gives a consistent experience irrespective of the underlying model.
The system integrates with the Cray supercomputer, which means that ROMS coastal simulations which previously took 2 hours to complete are now executed in 12 minutes, and the SWAN wave model from 25 mins to 2 mins. The client-server architecture offers multiple computation nodes which means that tasks such as ensemble runs can be executed in parallel.
The solution is based on the Delft-FEWS platform which is used by national forecasting agencies around the world, giving VNHMA the confidence that they are using a system which is tried and tested.
Working with local partners HarmonySoft, we delivered a package of training to build capacity within VNHMA. This included not only the operation of the system but also system configuration to allow them to make changes to their system in the future without vendor lock-in.
Efficiencies in the forecasting process mean that forecasters can now focus on analysing, interpreting and improving models rather than administrative tasks. The nature of the system makes it more robust to failure from both a technological and organisational perspective.
The flexible architecture means that the number of computational nodes available on the system can be scaled to meet future demand.