1. Page 1© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Variable resolution or lateral boundary conditions Terry Davies Dynamics Research Yongming Tang, Junichi Ishida

2. Page 2© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Convective scale NWP One-way nesting requires lateral boundary conditions from larger-scale model. Variable resolution LAM – still requires lbcs but high resolution region can be moved away from lateral boundaries.

3. Page 3© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik NWP Model Orography 12 km 4 km 1 km Height of model orography (m)

4. Page 4© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Variable resolution costs (relative to 4km) Area/  2 km1.85km1.5 km1 km 100% 8 10 19 64 80% 6.7 8.3 15 49 25% 5 - 10 25 10% - - - 10

5. Page 5© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik May 3 2002 Case ----- Nested model 1 km high resolution nested model and radar rainfall at 14 UTC 1 km4 km12 kmRadar

6. Page 6© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik May 3 2002 Case ----- variable resolution model Rainfall at 15 UTC. The three regions of the variable resolution domain are also shown Radar 1km

7. Page 7© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Variable Resolution Grid Structure Grid varies from coarse resolution at the outer boundaries smoothly to a uniform fine resolution in the interior of the domain Uniform High Res zone Var-Res 2 Var-Res 1 Uniform Coarse Res 1 Uniform Coarse Res 2 Typically, there are 3 regions, and inflation ratio R 1 = R 2 = 5~10% e.g. = 1 km, R 1 = R 2 = 10 % N_vr = 34 / 24 / 15 points  = 25 / 10 / 4 km R2R2 R1R1

8. Page 8© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Variable Resolution Grid Structure Grid varies from coarse resolution at the outer boundaries smoothly to a uniform fine resolution in the interior of the domain Uniform High Res zone Var-Res 2 Var-Res 1 Uniform Coarse Res 1 Uniform Coarse Res 2 Typically, there are 3 regions, and inflation ratio R 1 = R 2 = 5~10% e.g. = 1 km, R 1 = R 2 = 10 % N_vr = 34 / 24 / 15 points  = 25 / 10 / 4 km R2R2 R1R1

9. Page 9© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Variable resolution  Being built into latest version of UM (switch for variable resolution).  Uses look-up tables for semi-Lagrangian searches (departure points and trajectories).  Other optimisations to code and changes to lateral boundaries for variable grids.  Work towards 1.5km variable resolution model for UK LAM for 2009 (new supercomputer).

10. Page 10© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Forecast Domain and orography Regular UM (12km) Var-Res UM (1.5-4km) Global: 432x325L38 Regular UM (12km): 250x250L38 Var-Res UM (1.5-4km): 484x557L38 Regular UM (4km): 288x360L38 Regular UM (1.5km): 300x300L38 Domain of Var-Res UM Domain of 4km UM High Resolution Zone Domain of 1.5km UM

11. Page 11© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Typhoon track Domain of Var-Res UM (1.5-4km) Domain of 4km UM High Resolution Zone Domain of 1.5km UM 12UTC 05th 00UTC 05th 00UTC 06th The red lines and boxes show the observed track and position of the typhoon Nabi. The initial time is Regular 12km UM: 12UTC on 5 th Var-Res UM: 13UTC on 5 th Regular 4km UM: 13UTC on 5 th Regular 1.5km UM: 14UTC on 5 th The eye of the typhoon at initial time is located in the outside of the high resolution area and the typhoon moves into the area through the transition area (Var-Res UM) or lateral boundary (1.5km UM).

12. Page 12© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Global UM12km UMObservation 4km UMVar-Res UM Precipitation forecast The 12kmUM, 4kmUM and Var-Res UM well reproduce the rain caused by the orographic forcing over the Kyushu island. Hourly precipitation at 21 UTC on 5 th September

13. Page 13© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Precipitation forecast 4km UM1.5km UM Var-Res UMObservation Hourly precipitation at 21 UTC on 5 th September The rain band round the eye of the typhoon in the1.5km and Var-Res UM are better than that in 4km UM. The rain band in the Var-Res UM is slightly better than that in the 1.5km UM. The typhoon in the 1.5km UM moved from outside of the domain through the lateral boundary and the 1.5km UM seems not to have enough time to develop the rain band (spin up problem).

14. Page 14© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Lateral boundaries  To run variable resolution LAM will still need lbcs.  Current lbcs use standard merging technique and wrong lateral boundary condition for solving for the pressure increment

15. Page 15© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Lateral boundaries  Semi-Lagrangian advection applies lbcs naturally.  Apply appropriate lbcs to Helmholtz equation instead of zero pressure increment at lateral boundaries.  Reorganise lbc updating during time step.

16. Page 16© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik Lateral boundaries  For mpp, lateral boundary files do not need external halos – lbc files can be made 30% smaller.  It should also be possible to reduce the number of fields in the lbc file leading to further size reduction.  If interpolation to LAM is made on input, files could be even smaller.

17. Page 17© Crown copyright 2007SRNWP 8-11 October 2007, Dubrovnik The End