Tropical cyclone simulations and predictions with GFDL s prototype global cloud resolving model S.-J. Lin and GFDL model development teams NOAA/Geophysical Fluid Dynamics Laboratory Workshop on High-Resolution Global Modeling Fort Collins, June 15, 2010
Objectives: To develop a global cloud-resolving weather-climate model (GFDL HiRam) with applicable resolution from 1 to 100 km (uniform and/or variable via nesting/ stretching) Evaluate the prediction skill of the model for mediumrange (5-10 days), seasonal, to climate time-scale.
The GFDL High-Resolution Atmosphere Model (HiRam) is derived from the GFDL AM2.1 used in the IPCC AR4 except the following major modifications for highresolution: Non-hydrostatic FV dynamics re-formulated on the Cubed-sphere 6-category bulk cloud microphysics (similar to Lin et al. 1983). Formation of sub-grid clouds is accounted for via the PDF approach No deep convective parameterization. Instead, an essentially non-precipitating shallow convection scheme is used (based on Bretherton et al. 2004) Surface fluxes modified for high-wind situation over ocean (Moon et al. 2007); wave model coupling coming soon! doubly periodic f-plan configuration
Noteworthy attributes of the non-hydrostatic finite-volume dynamical core Formulated on the Cubed-Sphere with a truly equal distance Gnomonic projection Transport of vorticity, potential temperature, air mass, and tracer mass are consistently performed with the same conservative & monotonic scheme (Lin & Rood 1996; Putman & Lin 2007) Vertically Lagrangian discretization: free of vertical CFL condition (Lin 2004) Vertically propagating sound-waves are treated (too) accurately by solving the interaction of two Riemann invariants; 3D divergence damping is needed if the grid aspect ratio is far from one. A stretched grid capability with the Schmidt transformation (nearly working) 2-way regional-global nesting (ongoing work)
Some idealized tests of the dy-core 2D thermal bubble ( =5 m) 3D Density current over the equator on a very small planet ( =100 m) Held-Suarez benchmarks: C2000 (~5 km); non-hydrostatic, true-size earth C100 (~100 km); hydrostatic, true-size earth C100 (~ 4 km) 25X small-planet; non-hydrostatic
Adiabatic tests of the global non-hydrostatic model Density current (cold bubble) on a very small planet Radius: 1000 X smaller than the Earth resolution: ~100 m, model top at 6.4 km Simulation time = 10 min = 7 earth days Vertical cross section: Θ 6.4 km No rotation PS 1000X faster rotation PS
Held-Suarez forcing True-size earth or a 25 X small planet Rotation rate, thermal & mechanical damping re-scaled by the radius Integration length: 500 days 4.5 km, true-size earth with real terrain
Scaling of the global cloud-resolving prototype model C2000L32 (6 x 2000 x 2000 x 32) IBM BG/P at DOE ANL
500 days integrations with Held-Suarez forcing 25X small-planet Hydrostatic; C100 (~100 km) non-hydrostatic; C100 (~4 km)
Small-planet: C100 (~4km) Zonal mean vertical velocity No divergence damping divergence damping
No 3D divergence damping 3D divergence damping
Simulations of tropical-cyclone climatology (Zhao, Held, Lin, and Vecchi 2009, J. Climate) 4-member AMIP-type runs (25-yr each) with observed SST C180 (~50 km) resolution No deep convective parameterization UW shallow convection Simplified cloud microphysics (similar to GFDL AM2/ AM3)
Observed cyclone tracks: 1981-2005 Simulated tracks: 1981-2005 (C180 model) One realization
Seasonal cycle of hurricanes (1981-2005) (red: 4-member ensemble)
Seasonal (Jul-Oct) hurricane hindcasts: Resolution: C360 ( ~25 km); 32 Levels with top at 1 mb ICs: initialized by nudging the dynamical fields toward NCEP T382 analysis; each member is one-day apart External forcing: O3, green-house gases, etc, are climatology Land model: no special initialization (ICs from climatology) SST: constructed by adding a persistent anomaly (from forecast time) to the monthly climatology value. No information of any kind is used beyond the forecast time (June 1)
2005 hurricane season (Jul-Aug-Sep-Oct) Atlantic basin # of hurricanes: Obs: 12 Model: 9.8
2005 hurricane season (Jul-Aug-Sep-Oct) Western Pacific # of hurricanes: Obs: 13 Model: 11.3
7-Day hindcasts during 2008 & 2009 hurricane seasons HFIP funded, 3 post docs hired in 2010 7-days forecasts initialized by nudging and vortex breeding (whenever storm data is available); to be replaced by Ensemble Kalman Filter (EnKF) Resolution: C360L32 (to be upgraded to C720 next year)
Average track errors: Atlantic basin 2009 hurricane season 2008 hurricane season (Tracker developed by Machok for HFIP)
Average intensity errors: Atlantic basin 2009 hurricane season 2008 hurricane season
Ongoing & future works: 1. For 7-day forecasts (HFIP demo project), upgrade to global 13 km resolution. 2. Cloud-resolving stretched grid model (3-5 km over CONUS or MDR & GOM) i. Regional climate (National Climate Service) ii. Seasonal hurricane predictions 3. Couple to ocean & wave models 4. Climate runs with global cloudresolving resolution (4.5 km) to start soon