the good, the bad, and the ugly: ron mctaggart-cowan, john gyakum, and lance bosart numerical...
Post on 19-Dec-2015
268 views
TRANSCRIPT
The Good, the Bad, and the Ugly:
Ron McTaggart-Cowan, John Gyakum, and Lance Bosart
Numerical Prediction of Hurricane Juan
Outline
• Case Background
• Operational
Forecasts and NWP
• Improved NWP
Guidance
• Summary/Discussion
Case Background• Depression crosses Atlantic 14-20
September with weak convection
• Isabel’s outflow suppresses devel-opment from 21-24 September
• During 22-24 September, genesis occurred along an extended front to the northeast of the Lesser Antilles
Case Background• Tropical transition
completed by 25th
• Accelerates and develops strong asymmetry by 0000 UTC 29th
• Landfall near Halifax at 0300 UTC 29th as a
Category 2 Hurricane
NHC Best Track
Case Background
0000 UTC 28 September
• Still near maximum intensity (obtained at 1800 UTC 27th)
• Minimum SLP 970 hPa with est. 90 kt winds
Case Background
0600 UTC 28 September
• Moving W-N-westward and weakening slightly
• Minimum SLP 972 hPa with est. 85 kt winds
Case Background
1200 UTC 28 September
• Moving W-N-westward and developing strongly asymmetric structures
• Minimum SLP 970 hPa with est. 90 kt winds
Case Background
1800 UTC 28 September
• Moving northward and accelerating rapidly
• Minimum SLP 972 hPa with est. 80 kt winds
Case Background
0000 UTC 29 September
• Accelerating rapidly towards Halifax (landfall 3h later)
• Minimum SLP 974 hPa with est. 70-85 kt winds
Forecasting and NWP• Operational forecasting for Juan was
provided primarily by NHC and CHC
• Track forecast errors were below the 5 yr mean for all lead times (NHC)
• Intensity forecast errors were larger than average, but still reasonable
Forecasting and NWP• Forecasters at
both the CHC and the NHC issued excellent forecasts
• Lead times of over 24h were given for winds and rain associated with Juan
Forecasting and NWP• Forecasters noted
a bifurcation in model guidance from the 0000 UTC 28th run
• Common features at mid-levels are used to produce a good forecast
Forecasting and NWP• Over the past 5 years, the major
American, British, and European models have all been upgraded to deal with tropical features …
So what caused this potentially dangerous split in the guidance?
Forecasting and NWPModel Centre Quality* Availablity
GFS NCEP Low/Med Yes
Eta coord NCEP Low Yes
GFDL NCEP High No
UK Met UKMO High Yes
NOGAPS US Navy Med/High No
GEM CMC Low/Med Yes
MC2 McGill Low Yes* Quality is defined subjectively based on 24h
surface forecasts from the 0000 UTC 28th run
Forecasting and NWPInitialization GFS Sat Image
GFS SLP and surface winds (kts) at 0000 UTC 28th
Composite vapour/IR at 0000 UTC 28th
satellite
Forecasting and NWP24h Forecast GFS Sat Image
satellite
Quickscat winds at 2300 UTC 28th
GFS SLP and surface winds (kts) at 0000 UTC 29th
Forecasting and NWPInitialization Eta Sat Image
Eta SLP and surface winds (kts) at 0000 UTC 28th
Composite vapour/IR at 0000 UTC 28th
satellite
Forecasting and NWP24h Forecast Eta Sat Image
satellite
Quickscat winds at 2300 UTC 28th
Eta SLP and surface winds (kts) at 0000 UTC 29th
Forecasting and NWPInitialization GFDL Sat Image
Not Available
Forecasting and NWP24h Forecast GFDL Sat Image
Not Available
Forecasting and NWPInitialization UK Sat Image
UKMet SLP and surface winds (kts) at 0000 UTC 28th
Composite vapour/IR at 0000 UTC 28th
satellite
Forecasting and NWP24h Forecast UK Sat Image
satellite
Quickscat winds at 2300 UTC 28th
UKMet SLP and surface winds (kts) at 0000 UTC 29th
Forecasting and NWPInitialization Navy Sat Image
Not Available
Forecasting and NWP24h Forecast Navy Sat Image
Not Available
Forecasting and NWPInitialization GEM Sat Image
satellite
GEM SLP and 1000-500 hPa thickness 0000 UTC 28th
Composite vapour/IR at 0000 UTC 28th
Forecasting and NWP24h Forecast GEM Sat Image
satellite
GEM SLP and 1000-500 hPa thickness 0000 UTC 29th
Quickscat winds at 2300 UTC 28th
Forecasting and NWPInitialization MC2 Sat Image
MC2 SLP and 1000-500 hPa thickness 0000 UTC 28th
Composite vapour/IR at 0000 UTC 28th
satellite
Forecasting and NWP24h Forecast MC2 Sat Image
satellite
MC2 SLP and 1000-500 hPa thickness 0000 UTC 29th
Quickscat winds at 2300 UTC 28th
Forecasting and NWP
So what happened? …
• Forecast models initialized without a strong enough vortex failed to develop or maintain a closed circulation and favoured the western trough as the weather-maker
Forecasting and NWP• Weak initial
vortex without troposphere-deep PV column
• Weak poten-tial stability ahead of trough
Forecasting and NWP• PV tower tilts
downshear as development occurs over the Bay of Fundy
• Surface circu-lation tracks westward
Forecasting and NWP• Bogusing methodologies are very different
at the various operational centres
Centre Model Vortex Initialization
NCEP GFDL Grown balanced vortex
UKMO UK Met Synthetic observations
Navy NOGAPS Developed vortex insertion
NCEP GFS, Eta Vortex relocation
CMC GEM None
McGill MC2 None
complexity
most
least
Forecasting and NWP• Who produced “Good” guidance for
Hurricane Juan?
Centre Model Vortex Initialization
NCEP GFDL Grown balanced vortex
UKMO UK Met Synthetic observations
Navy NOGAPS Developed vortex insertion
NCEP GFS, Eta Vortex relocation
CMC GEM None
McGill MC2 None
complexity
most
least
Forecasting and NWPGFDL vortex initialization
• Vortex in GFS initial fields is removed using filters and optimal interpolation
• New vortex is grown in model
• Full treatment of SST evolution and interaction
• Strong, well-balanced vortex in initial fields
Forecasting and NWPUKMO vortex initialization
• Manual intervention when satellite or in-situ data is significantly different from first guess field
• Synthetic wind observa-tions are added
• Analysis cycle balances mass field to produce a realistic vortex
Forecasting and NWPUS Navy vortex initialization
• Vortex generation based on balanced hurricane vortex of Kurihara
• Fully-developed vortex is inserted in HPC-analyzed position
Forecasting and NWPNCEP vortex initialization
• Vortex removal is made using multiple filter passes
• Hurricane compon-ent moved to HPC-analyzed position
• Eta adds bogus winds (for now)
Juan?
Improved Guidance• Dense observations to the N/W/S of
the storm over the continent mean that upstream analysis should be good
• Quality of forecast depends on the strength/coherence of the initial vortex
Can we improve a bad forecast with a better initial vortex alone?
Improved GuidanceMcGill vortex initialization
• Insert a fully-developed Kurihara-type vortex at 0000 UTC 28th
• Rerun 48h forecast without any other modifications to the initial or boundary conditions (supplied by GEM)
• Both 12km and 3km domains are rerun
Improved Guidance
• Modified initial conditions at 0000 UTC 28th September
satellite
satellite
Improved Guidance
• Modified forecast SLP and winds fields valid 0000 UTC 29th September
satellite
satellite
Improved Guidance• Initial PV tower
is well-defined throughout the troposphere
• Convective instability over the continent is not changed
Improved Guidance• Vortex tube
remains upright and wraps potential instability into the circulation
• Much improv-ed track and structure
Improved Guidance• How much could
have been done in real time at high resolution?
Summary• The operational forecasts for Hurricane
Juan were of very high quality
• Operational models without a realistically- bogused vortex gave poor sfc guidance
• Synthesis of a strong (Category 2) vortex in the initial conditions of one of the “bust” models results in a much-improved forecast
Discussion• The NWP for Hurricane Juan reduces to
an initial value problem involving the tropical vortex itself:
A realistically-strong, balanced synthetic vortex was ALL that
was required for accurate numerical prediction of Juan in
real time