200 likes | 325 Views
Stan Benjamin, Bill Moninger, Tracy Lorraine Smith, Brian Jamison, Ed Szoke, Tom Schlatter NOAA Earth System Research Laboratory (ESRL), Global Systems Division (GSD) Boulder, CO. 2006(-07)TAMDAR aircraft impact experiments for RUC humidity, temperature and wind forecasts.
E N D
Stan Benjamin, Bill Moninger, Tracy Lorraine Smith, Brian Jamison, Ed Szoke, Tom SchlatterNOAA Earth System Research Laboratory (ESRL), Global Systems Division (GSD)Boulder, CO 2006(-07)TAMDAR aircraft impact experiments for RUC humidity, temperature and wind forecasts
Purpose for Rapid Update Cycle (RUC) model run operationally at NCEP • Provide high-frequency (hourly) mesoscale analyses, short-range model forecasts • Assimilate all available observations • Users: • aviation/transportation • severe weather forecasting • general public forecasting • Focus on 1-12 hour forecast range KEY QUESTION: Can TAMDAR add to RH/T/V forecast skill in RUC modelcompeting with other obs?
Observation needs for aviation NWP addressed bydeployment of automated weather data from regional carriers • Goal: Improved forecasts for aviation hazards – • Icing, low ceiling conditions, precipitation forecasts • Convective environment, especially for convective initiation (T, RH, V - all important) • Lower tropospheric wind profiles (e.g.,CTAS) • via • High density (regional airports) high-frequency (>6-8 soundings/day) • temperature and moisture profiles • winds from ascent/descent, enroute • => Examine tropospheric profile of T/RH/V forecast skill
Real-time TAMDAR impact experiment design • Parallel 20km RUC 1-h cycles • Uses latest code in RUC13 (but at 20km) • Dev cycle – all obs data but no TAMDAR • Dev2 cycle – dev + TAMDAR data • Lateral boundary conditions – same for Dev and Dev2 • Control design • Initialize Dev and Dev2 runs at exact same time – same observations used (except TAMDAR) • Reset dev and dev-2 background field at 1000z every • 48 h • Ensure against any computer processing differences between dev and dev2 cycles
Dev-Dev2 difference – 0h analysis Init 1800z 24 Aug 2005 – 500 mb 1000-1800z
Verification regions for GSD-RUC TAMDAR impact Large region (eastern half of US) -- 38 RAOB sites Small region (Great Lakes) includes 14 RAOBs
Key dates for RUC analysis/QC changes • 8 June 2005 - Introduction of improved aircraft reject list • 1 Sept 2005 – Updated reject list • 15 Sept 2005 – Incorporate RUC13 moisture analysis • 15 Nov 2005 – Add new QC step in RUC analysis using ob-background (O-B) difference – (important for isolated observations) • 1 Dec 2005 – Revised aircraft RH observation error to higher value (4% to 12%, now same as raob, was previously “overfitting” TAMDAR RH data) • 15 Dec 2005 – Flag TAMDAR winds on descent • 15-30 March 2006 – Incorrect use of TAMDAR-RH in dev2. Corrected on 31 March • June 2006 – Further improvement in O-B QC
Figure 3a. RMS RH at 500 hPa for 3h forecasts for the old verification system (centered at 15% RH). Figure 3b. RMS RH at 500 hPa for 3h forecasts for the new verification system (centered at 19% RH). New verification software – Bill Moninger - verify every 10mb, not just at mandatory levels - use all raob data, no QC screening Example - More improvement in 500mb RH forecasts from TAMDAR from including all obs.
Temp errors vs. raobs – 850 mb • 3h fcsts • Valid 00z • Gt.Lakes region • Running 30-day avg Dev –no TAMDAR Dev2- w/TAMDAR • Strongest effect at this level – 0.25 C improvement • Slightly improved in 2006
Temp errors vs. raobs – 1000-800 mb avg • 3h fcsts • Valid 00z • Gt.Lakes region • Running 7-day avg Dev –no TAMDAR Dev2- w/TAMDAR Diff – dev-dev2 • Most accurate results in 2006 since mid-Nov
Wind errors vs. raobs – 850-700-500 mb avg • 3h fcsts • Valid 00z • Gt.Lakes region • Running 30-day avg Dev –no TAMDAR Dev2- w/TAMDAR 3h persistence –dev2 • Consistent improvement from TAMDAR
Dev –no TAMDAR Dev2- w/TAMDAR 3h persistence –dev2 • 3h fcsts • Valid 00z • Gt.Lakes region • Running 30-day avg RH errors vs. raobs – 850-700-500 mb avg • Consistently positive since Jan 2006
RH errors vs. raobs – 850 mb • 3h fcsts • Valid 00z • Gt.Lakes region • Running 30-day avg Dev –no TAMDAR Dev2- w/TAMDAR • Consistent positive impact from TAMDAR in 2006, unlike in 2005
Dev –no TAMDAR Dev2- w/TAMDAR 3h persistence –dev2 • 3h fcsts • Valid 00z • Gt.Lakes region • Running 7-day avg RH errors vs. raobs – sfc-800 mb avg • Positive impact throughout 2006 • Average 1% in 1000-800mb layer
Dev –no TAMDAR Dev2- w/TAMDAR 3h persistence –dev2 • 3h fcsts • Valid 00z • Gt.Lakes region • Running 7-day avg RH errors vs. raobs – 900-500 mb avg • Increased impact in last 3 months
Temperature April-October 2006 Warm season + transition wTAM noTAM Dec06-Jan07 Cold season noTAM wTAM
Relative humidity April-October 2006 Warm season + transition wTAM noTAM Dec06-Jan07 Cold season wTAM noTAM
Wind April-October 2006 Warm season + transition wTAM noTAM Dec06-Jan07 Cold season wTAM noTAM
Forecast errors – RUCdev (no TAMDAR), RUCdev2 (w/ TAMDAR) Temp RH wTAM noTAM wTAM noTAM • TAMDAR impact study with RUC parallel cycles • 2005-2007 (ongoing) • Further improvement in • RH, temperature, wind Wind wTAM noTAM
GSD-RUC TAMDAR impact experiment results – updated 17Jan 2007 • Recent results – 2006 • RH impact improved in 2006 – Consistent 1-2%RH reduction of RMS error (20-30% reduction – peaks at 900-800mb, 600-400mb • Temperature impact • strongest from 950-800mb layer – inversion top! • 25% reduction of 3h forecast error (0.25K) • Better temp impact apparent w/ higher vert resolution (Nov06-current) • Wind impact – ~10% reduction of 3h fcst errors in 850-700-500mb layer. About same as in 2005. • Heading accuracy w/ Saab/Mesaba winds • Increase turboprop wind obs errors? • (now used in devRUC13). • Results (TAMDAR impact) have improved during continued TAMDAR evaluation