EGU21-9672_presentation.ppt
- 1. DYAMOND-II simulations with IFS-FESOM2 Thomas Rackow1, Nils Wedi2, Kristian Mogensen2, Peter Dueben2, Helge F. Goessling1, Jan Hegewald1, Christian Kühnlein2, Lorenzo Zampieri1, and Thomas Jung1 1) Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI) 2) European Centre for Medium-Range Weather Forecasts (ECMWF) SST snapshot tropical cyclones
- 2. 2 • A digital twin is a digital replica of a living or non- living physical entity. • Expert and non-expert users get access to high-quality information, services, models, scenarios, forecasts and visualisations (e.g. in climate modelling and weather forecasting, hurricane evolution). • Digital twins rely on the integration of continuous observation, modelling and high performance simulation, resulting in highly accurate predictions of future developments. Motivation: Digital Twin? From EU Destination Earth website: https://ec.europa.eu/digital-single-market/en/destination-earth-destine
- 3. 3 Motivation: Digital Twin? • A digital twin is a digital replica of a living or non- living physical entity. • Expert and non-expert users get access to high-quality information, services, models, scenarios, forecasts and visualisations (e.g. in climate modelling and weather forecasting, hurricane evolution). • Digital twins rely on the integration of continuous observation, modelling and high performance simulation, resulting in highly accurate predictions of future developments. Science website: https://www.sciencemag.org/news/2020/10/europe-building-digital-twin-earth-revolutionize-climate-forecasts 1km IFS (Integrated Forecasting System by ECMWF)
- 4. 4 Motivation: DYAMOND++? ERA5-forced until 20 Jan 2020 CORE2, “ORCA025“; soon “Rossby4.2“ FESOM2 grid
- 5. 5 Motivation: ocean coupling SST forecasts (Neoguri) using the uncoupled model (left) and the coupled model (right). SST observations from ships and buoys (circles); ‘best track’ cyclone data: black line K.Mogensen (ECMWF) 1. heat loss to atmosphere 2. vertical mixing in the ocean 3. upwelling by Ekman pumping
- 6. 6 Motivation: sea-ice coupling Sea-ice concentration in the Baltic Sea evolved considerably between 1 April (left) and 11 April 2018 (middle), according to ECMWF’s OCEAN5 analysis. A coupled 10-day forecast for 11 April (right) captured this. K.Mogensen (ECMWF)
- 7. 7 … in configurations with the OASIS coupler (2 executables) Our usual coupling schematic coupling step (e.g. every hour) ECHAM tasks or OpenIFS, e.g. 1152 FESOM tasks, e.g. 1920 - FESOM2 uses fluxes from the atmosphere that were averaged over the previous coupling time step -> introduces some lag; - It is desirable that both models take the same amount of time 00:00 -> 01:00am 01:00 -> 02:00am 02:00 -> 03:00am
- 8. 8 … in IFS-FESOM2 (single executable) A different coupling schematic coupling step (every 6hrs, or 1h recently) Cores are shared, e.g. 1152 • FESOM2 uses fluxes from IFS for the same period; FESOM surface fields are forwarded in time for 1h and used in IFS’ next time step • Important for fast processes like tropical cyclones etc.? • Technically difficult: Hybrid MPI/OpenMP parallelization vs MPI-only IFS calls FESOM2 as a subroutine (after every timestep) 00:00 -> 01:00am 00:00 -> 01:00am 01:00 -> 02:00am (including ocean feedback)
- 9. 9 Split main into 3 parts (parts callable from within IFS) Creating a FESOM2 “library” • The stand-alone FESOM2 is not significantly affected by this technical change because the new main just calls these 3 parts one after the other • If FESOM2 is compiled as library, __ifsinterface is True, and the “main” block to the left is neglected
- 10. 10 One example how the interface routines look like Creating a FESOM2 “library” ocean
- 11. 11 With ‘prepIFS’ and ‘ecflow’: Controlling the experiment (with GUI)
- 12. 12 SST over 1 year with IFS-FESOM2 FESOM2 on ORCA025 grid; coupled with the full IFS (here about 18km; now about 4.5km)
- 13. 13 • New IFS-FESOM2 DYAMOND simulations are currently in the machine; most severe technical difficulties are overcome – The DYAMOND models will produce a wealth of information, overlapping with (part of) MOSAiC and the EUREC4A tropical field campaign, which is 20 Jan 2020 until 1st Mar 2020 (40 days) – Potential candidate model for the 4-5 Digital Twins in DestinE? – Rossby4.2 ocean grid is set up, with resolution following the Rossby radius of deformation (capped at 2km) -> impact of eddies/sea ice leads on 1km atmosphere? Outlook