Development of a scripting system in python to run numerical weather forecast model F/H

As a public expert in weather and climate, Météo-France is here to help keep you safe every day and assist you in making the best decisions in a changing climate. With dangerous weather events becoming even more intense and frequent due to climate change, our mission to keep you safe is crucial. We mobilize our expertise and scientific and technological excellence to enable you to anticipate and adapt to challenging weather and climate events.

Find us online: https://meteofrance.com/carte-didentite-de-meteo-france

The position will be at CRNM in Toulouse, in the GMAP group. (GMAP is responsible for operational models at Météo-France), with collaborations on this theme with the ECMWF and with others ACCORD country.

Why join us?

Embark on a stimulating adventure that serves everyone, alongside men and women who are committed every day to tackling the challenges posed to our society by weather and climate.

And enjoy the following benefits:

• flexible working hours
• RTT (reduced working time)
• teleworking
• staff restaurant or meal vouchers
• 75% contribution to public transport costs
• contribution to mutual insurance
• sports and cultural associations depending on the site concerned (climbing, gym, pottery, theater, etc.)
• Other benefits await you, come and discover them

The main objective of the position is to contribute to the evolution of the current Deode-Workflow scripting system, which is used to run the On-demand Extremes DT, and to take part in the operation of the DT.

The DT is meant to provide relevant meteorological information to impact sectors such as water, food, energy resource and health management, and to demonstrate the benefit of the DT framework for extremes risk management, adaptation and mitigation decision making by key stakeholders at regional or local scale.

The model that is launched is the ACCORD numerical prediction system that is used in many European countries under 3 different versions (AROME, HARMONIE-AROME, ALARO). Those models share a lot in common but also have some differences.

The Deode-Workflow scripting system (also named Tactus) has been developed during Phases 1 and 2 of the project. It is mainly written in python with configuration files in TOML and YAML formats, and it is able to launch weather forecasts for the 3 models AROME, HARMONIE-AROME and ALARO with their specificities.

In Phase 3 it remains the backbone of the operational workflows, and it has to evolve: refining scripting conventions and configuration formats, integrating the evolutions of the NWP models and of the impact-sector applications, and improving its documentation and testing.

The system is run on the ECMWF HPC and on the EuroHPC platforms LUMI and LEONARDO, which both have a GPU partition. One goal of the project is to be able to run on mixed CPU/GPU architectures, though today, with our model, GPUs can only be used for some specific routines.

The work is organised in the following tasks (durations are indicative):

1. Evolution of the Deode-Workflow scripting system (about 6 months). One first task will be to familiarize oneself with the Deode-Workflow scripting system, then to contribute to its evolution, in particular for the parts designed to launch the AROME model. This includes the system infrastructure: upgrading the NWP component to the new cycles of the ACCORD model, implementing new functionalities that are required, and improving the documentation of the system. This will involve collaborating with the other contributors of this task in the ACCORD and DE_330 teams.
2. HPC platform porting (about 2 months). The Deode-Workflow scripting system will be installed on the Météo-France HPC machines (Belenos and Taranis, and the new CPU/GPU machines that will replace them from 2027), with the maintenance of the corresponding compilation and platform configuration files, so that the DT configurations can be run and tested at Météo-France. This also includes contributing to the porting and testing of new NWP releases on the EuroHPC platforms LUMI and LEONARDO.
3. Operations and real-time observation data (about 5 months). The successful applicant will be a member of the operation team, both for the real-time operation of the DT on the ECMWF ATOS HPC and for the climate simulations on LUMI: launching and monitoring runs, daily maintenance of the monitoring tools, handling issues. This also includes the acquisition of real-time observation data: monitoring of the observation data workflows (ingestion, processing, quality control) and inclusion of new types of observations.
4. Dissemination, data governance and user uptake (about 3 months). This task consists in ensuring that the real-time outputs comply with the standards (WMO GRIB2, metadata, naming conventions), and in helping the national meteorological services and other users to retr