Regional GEM 15 km

1. From Mesoscale to Microscale 12 UTC 00 UTC 12 UTC Regional GEM 15 km OPERATIONAL 48-h RUN (00 or 12 UTC) 13 UTC Global Variable resolution 576 x 641 Timestep = 7.5 min 58 levels (for NWP) 1D turbulence No TEB IC + LBC 36-h run GEM-LAM 2.5 km LAM 201 x 201 Timestep = 60 s 53 levels (two levels of packing near surface) 1D turbulence TEB IC + LBC T-3 T+12 GEM-LAM 1 km 15-h run IC + LBC LAM 201 x 201 Timestep = 30 s 53 levels (two levels of packing near surface) 1D turbulence TEB • Coupling variables • Surface layer coupling • Lateral boundary conditions • 250-m vs 1-km results T+5 T-1 MC2-LAM 250 m 6-h run IC + LBC LAM 201 x 401 (long axis oriented along the low-level wind direction) Timestep = 10 s 53 levels (two levels of packing near surface) 3D turbulence TEB Microscale Urban flow Models (urbanSTREAM) EVENT

2. Coupling Between Mesoscale and Microscale T-3 T+12 GEM-LAM 1 km 15-h run IC + LBC T+5 T-1 MC2-LAM 250 m 6-h run IC + LBC Inflow boundary conditions • Horizontal wind • Vertical motion • Turbulent Kinetic Energy • Boundary-layer height Microscale Urban flow Models (urbanSTREAM)

3. The Surface in GEM The Concept Atmospheric model First atmospheric level, about 50 m above the surface Surface layer FLUX AGGREGATION zatm SURFACE = TOP of CANOPY Flat surfaces Vegetated canopy Urban canopy

4. GEM’s Surface: How it connects with the Real World Atmospheric model Logarithmic wind profiles over each type of surfaces First atmospheric level, about 50 m above the surface zatm+(zblg-zveg) zatm zatm+zblg Flat surfaces Vegetated canopy Urban canopy

5. GEM’s Vertical Wind Profiles vs Observations at Station ANL (IOP9) RADAR GEM SODAR

6. GEM’s Vertical Wind Profiles vs Observations at Station ANL (IOP9) RADAR GEM SODAR CANYON NATURAL

7. IOP9 (Night): Daytime turbulence More turbulent One hour later, the turbulence is more homogeneous Less turbulent 26 July 2003 20 UTC 26 July 2003 21 UTC

8. Automation and Testing We propose to run the prototype in a fully automatic manner at a regular frequency (once a week or once a month?). • Specify location of event (lat, lon) • Generation of computational grids (250-m grid is oriented along the mean daytime low-level winds) • Production of surface fields using an interpolation from pre-processed large grids (for Montreal, Toronto, Ottawa, and Vancouver) • Integration of 2.5-km, 1-km, and 250-m runs • Production of outputs for microscale models (with adaptation to local surface characteristics) FULLY AUTOMATIC PROCESS

9. Computational Cost On CMC’s current operational machine… Grid size ni x nj x nk x nsteps Wall clock time 2.5 km 200 x 200 x 53 x 2160 (36h) 35 min with 200 cpus 1 km 200 x 200 x 53 x 1800 (15h) 30 min with 200 cpus 250 m 200 x 400 x 53 x 2160 (6h) 65 min with 400 cpus Total of 110 min On new computer (early 2007), 2.5 times faster, i.e., 45 min.

10. CRTI-1 Further Work • Urban cover classification • finalize vector data (MTL and VAN) • examine hybrid approach? (satellite + vector) • other cities (TOR, Ottawa,?) • Pre-processing of large grids for selected cities (MTL, VAN, OTT, TOR) • Urban anthropogenic fluxes • generation for MTL (+ validation with Quebec Region data) • modify GEM inputs to include anthropogenic fluxes • sensitivity study on OKC: impact on mixing in boundary layer • Prototype • complete OKC and apply to MTL (with MUSE) • output adaptation for microscale modeling (coupling issue) • current prototype with MC2-250m (waiting for updated GEM vertical discretization) • start running prototype (e.g. once a week?) • 3D turbulence • finalize code validation with LES-type runs • impact study with 250-m runs • MUSE • continue analyses of MUSE-1 and MUSE-2 data • TEB • cascading of TEB prognostic variables (for initialization purpose) • snow treatment • Publications…