Aircraft-based Observations: Impact on weather forecast model performance Stephen S. Weygandt Eric James, Stan Benjamin, Bill Moninger, Brian Jamison, Geoff Manikin* NOAA Earth System Research Laboratory *NOAA National Centers for Environmental Prediction Friends and Partners of Aviation Weather NBAA Convention, Oct 10-11, 2017
ABO impact on weather models: Key points HRRR (and RAP) Future Milestones HRRR Milestones #1: Aircraft data most important observation type over North America for 3-12h (situational awareness) forecast accuracy (winds, temperature, Rel. Hum.) #2: Increased aircraft data has improved US (and global) forecast skill (2011-2015) #3: Geographical and temporal gaps in aircraft data provide opportunity for improved forecast accuracy through improved aircraft participation Ascent / descent 0-15 kft
RAP Rapid Refresh and HRRR NOAA hourly updated models Version 4 – GSD 13-km Rapid Refresh (RAP) RAP 3-km High Resolution Rapid Refresh (HRRR) AK HRRR Version 3 – NCEP implement – Aug 2016 Version 4 – GSD (including HRRR-AK) NCEP – Spring 2018 CONUS HRRR Hourly updating maximize use of ALL observations
Rapid Refresh hourly cycling improves guidance HRRR Milestones HRRR (and RAP) Future Milestones Get more accurate short-range weather forecasts for decision making Use latest observations EACH HOUR to obtain freshest, most accurate “snapshot” of atmospheric state WIND forecast errors 1h 3h 6h 12h 1-hr fcst 11 12 13 Time (UTC) Analysis Fields 3DVAR Obs Back- ground Wind forecast improvement from hourly updating for 1-h vs. 6-h forecast 1 Jan – 10 Oct 2017 vector wind RMS error for CONUS rawinsondes for different forecast lengths
Observations assimilated: RAP and HRRR HRRR Milestones HRRR (and RAP) Future Milestones Hourly Observation Type Variables Observed Observation Count Rawinsonde Temperature, Humidity, Wind, Pressure 120 / 12h Aircraft Wind, Temperature 2,000 -15,000 / hr Aircraft – WVSS, Tamdar (3 Aug ) Humidity 0 – 800 / hr Surface/METAR Temperature, Moisture, Wind, Pressure, Clouds, Visibility, Weather 2200 - 2500 Surface/Mesonet Temperature, Moisture, Wind ~5K-12K Buoys/ships Wind, Pressure 200 - 400 Profiler – 915 MHz Wind, Virtual Temperature 20-30 Radar – VAD Wind 125 Radar Radial Velocity 125 radars Radar reflectivity – CONUS 3-d refl Rain, Snow, Graupel 1,500,000 Lightning (proxy reflectivity) NLDN GOES AMVs 2000 - 4000 Polar Orbiter Satellite Radiances very large Geostationary Satellite large GOES cloud-top press/temp Cloud Top Height 100,000 GPS – Precipitable water 260 WindSat Scatterometer Winds 2,000 – 10,000
Variables measured by each observation type Types Temp-erature Wind Rel. Hum. Pres. / Height cloud/ saturation/ Hydrometeor Radiance Extent Raob Y --- 3D, 2x/day Surface 2D Aircraft Y (~15%) (icing pireps not used) 3D Radar Reflectivity (in precip) Radar – Vr, VAD Y” 3D, Column GPS-met Column GOES (cloud, AMV) Y’ Satellite Y* IN SITU OBS RADAR OBS SATELLITE OBS *Satellite radiance: a complex function of temperature and humidity profiles “Radar radial velocity: single wind component; AMV winds: height assignment issues
Regional Observation Impact studies with RAP - GSD Observation gaps are major source in limiting forecast accuracy, even over US RAP observation impact study 3 seasons, 8 (9) observation types Rawinsonde Aircraft obs Radar reflectivity VAD winds Surface obs GPS-Met AMV (winds) GOES clouds Aircraft observations most important observation type -- Ascent/descent and en route obs both important -- Water vapor observations (about 1/7 total) improve RH forecast accuracy Raob Coverage Radar Coverage (Satellite Radiance) Sample aircraft obs coverage
Aircraft observations -- most important data source for weather prediction skill Impact on WIND in 1000-100 hPa layer Forecast degradation for withholding each obs type Withhold: A – ALL Rawinsonde B – ALL Aircraft C – Aircraft above 350 hPa D – Aircraft below 350 hPa E – Aircraft temp/humidity F – ALL Profiler G – ALL Radar Reflectivity H – ALL VAD winds I – ALL surface obs J – ALL GPS-Met PW K – ALL AMVs winds L – GOES (winds/clouds) Raob AC <350 mb AC >350 mb AC Temp+RH Rawinsonde Aircraft Aircraft Reflectivity GOES Surface VAD GPS AMV Radar Resolving the coastline over time and thunderstorm scale GOES 3/6/9/12 hr impact for each obs type Aircraft obs most important for wind accuracy at all forecast lengths Significant impact also from rawinsonde, surface observations, GOES observations (likely from clearing of spurious convection)
Most Observation impact: Aircraft, raobs, GOES Wind impact (m/s) Sat. radiance Aircraft GOES Rawinsonde GPS Surface VADs AMVs Radar reflect. Sat. radiance Aircraft GOES Rawinsonde GPS Surface VADs AMVs Radar reflect. Most Observation impact: Aircraft, raobs, GOES HRRR (and RAP) Future Milestones HRRR Milestones 25% Error reduction (6-h) 25% Error reduction (6-h) Sat. radiance Aircraft GOES Rawinsonde GPS Surface VADs AMVs Radar reflect. Radar Reflectivity Radiance data Rawinsonde GPS-Met PW Surface obs VAD winds RARS data Aircraft Profiler GOES AMVs Relative humidity impact (%) Wind impact (m/s) (1000-100 mb) Spring retrospective (1000-400 mb) 25% Error reduction (6-h) Temperature impact (k) (1000-100 mb)
Aircraft observation also important for global model forecast skill (satellite data dominates) Raob/balloon Aircraft Cloud-drift vecs Surface obs Aircraft Surface obs Global model obs impact: John Eyre UK Met Office - Impact on 24h global forecast error (FSOI) Sat obs Contributions to the total observation impact on a moist 24-hour forecast-error energy-norm, surface-150 hPa.
Global aircraft observation importance will increase as global models start hourly cycling Global aircraft observation density
RUC skill improved from more aircraft obs HRRR (and RAP) Future Milestones HRRR Milestones RUC model frozen, skill improvement solely from more aircraft obs Monthly aircraft obs counts RUC RAP 6-h upper-level Wind RMS error Worldwide count # of reports 2011 2015 US count 2011 2015 2008 2010 2012 2014 2016 Date http://www.wmo.int/pages/prog/www/GOS/ABO/data/statistics/aircraft_obs_cmc_mthly_ave_daily_reports_by_program.jpg
FAA aircraft obs study (GSD, AvMet) GOALS: Quantify gaps in airborne observations (spatial, temporal, parameter, etc,) Identify most cost effective ways to obtain airborne data Study sponsored by FAA Tammy Farrar – project sponsor AMDAR obs density CONUS (obs/km**2) per 24h period 2017 MWR article by James and Benjamin
AMDAR obs densitry -- time of day -- CONUS Morning 12z – 18z Evening 00z – 06z Afternoon 18z – 00z Overnight 06z – 12z
AMDAR obs by elevation – full day – CONUS All levels 15-30 kft Ascent / descent 30-45 kft 0-15 kft
Regional Observation Impact studies with RAP - GSD Significant gap to achieve profiles every 300 km/3h -- achieving frequent aircraft profiles out of regional airports estimated to significantly improve forecast accuracy Airborne observation study sponsored by FAA (GSD, AvMet) 2017 article by James and Benjamin, MWR Ascent / descent Data for an entire day What are the coverage expansion priorities For improved model skill? 0-15 kft
Improved mesoscale fields will lead better small-scale forecasts Previous results: Increased RAP skill leads to improved HRRR skill for thunderstorms and other weather hazards Improvements will be extended by ongoing storm-scale ensemble data assimilation for HRRR and future storm-scale ensemble forecast system Improved short-term skill from HRRR ensemble HRRR ensemble skill HRRR current skill Wiring chart for HRRR Ensemble
ABO impact on weather models: Key points HRRR (and RAP) Future Milestones HRRR Milestones #1: Aircraft data most important observation type over North America for 3-12h (situational awareness) forecast accuracy (winds, temperature, Rel. Hum.) #2: Increased aircraft data has improved US (and global) forecast skill (2011-2015) #3: Geographical and temporal gaps in aircraft data provide opportunity for improved forecast accuracy through improved aircraft participation Ascent / descent 0-15 kft