Grid Point Models
Surface Data
Models: Types
• Spectral Models (AVN)– Data is not represented on grid– Data represented by wave functions– Resolution is a function of # waves used in model– Computational errors generally less– Not well-suited for mesoscale modeling
Models: Types
• Hydrostatic Models (ETA, AVN, NGM)– Cannot produce vertical accelerations– Vertical motions determined by the continuity
equation
• Non-Hydrostatic Models (Some MM5)– Can produce vertical accelerations– Calculate Vertical Motions explicitly– Used in mesoscale applications (conv)
Models: The Basics
• Domain: Area covered by the model– IDD grids
– Regional vs. Global– Nested models
Models: The Basics
• Resolution: Distance between grid points– High and low resolution models
• http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid211.gif• http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid215.gif
Models: Resolution
Model Resolution
Should have 5 to 7 grid points to resolve feature
Model Resolution
Should have 5 to 7 grid points to resolve feature
Models: The Basics
• What can’t models simulate?– Processes neglected in simplified equations– Processes unknown– Processes that are sub-grid scale
• How’s a model to cope?
Models: The Basics
• Parameterizations– Model’s attempt to ““simulate”” (incorporate)
important sub-grid scale processes – Examples:
• Convection• Microphysical processes of precipitation• Surface/Boundary layer fluxes
Model Parameterizations:CONVECTION
Model Parameterizations:CONVECTION
Model Parameterizations:CONVECTION
Model Parameterizations:CONVECTION
Why are model forecasts imperfect?
• Imperfect Initial Conditions– Too few observations– “Continuous atmosphere = Non-continuous sampling”
• some areas worse than others– Bad observations
• instrument error– Errors in the initialization procedure
• First guess & objective analysis
“GI = GO”
Imperfect Models: Accurate Ob = Good ob?
Good ObservationOr Bad Observation?
Why are model forecasts imperfect?
• Imperfect Models– Simplified equations
• many “unimportant” terms = 0– Neglected Processes
• that’s why we still have field projects!– Resolution
• can’t simulate small scale stuff• ‘good’ ob can be a bad ob
Trend of Numerical Models
• Resolution increasing!
• Run more frequently!
• More models!– Computer power increasing– Cost decreasing
Trend of Numerical Models
• Implications: Higher Resolution– Improved initialization– More small-scale effects will be predicted!– Will these small-scale phenomena be correct?
• If terrain-forced weather phenomena = YES!• Density obs VS. density grid points• Heightened sensitivity to initial conditions
Higher Resolution: Improves Initialization
Good ObservationOr Bad Observation?
Higher Resolutionwill help but not solve the problem!
Model Resolution
Should have 5 to 7 grid points to resolve feature
Higher Resolution: Improves Terrain-forced weather!
Model Terrain vs. Actual Terrain
Model Terrain
• ETA 80km– http://lnx21.wwb.noaa.gov/orog/80km_east_crop.gif
• ETA 32km– http://lnx21.wwb.noaa.gov/orog/32km_east.gif
• ETA 10km– http://lnx21.wwb.noaa.gov/orog/10km_east_crop.gif
• Actual terrain– http://fermi.jhuapl.edu/states/us/big_us_color.gif
Orographic: Differential Heating
Orographic: Differential Heating
Density of OBS vs. Grid points
What if grid density (aka. model resolution) exceeds observation density?
Sensitivity to Initial Conditions