Meteorological conditions contributing to the crash of a Boeing 737 at Denver International Airport Workshop on Aviation Turbulence 28 August 2013 NCAR, Boulder, CO Teddie Keller Bill Hall, Stan Trier, Bob Sharman, Larry Cornman, and Yubao Liu Research Applications Lab National Center for Atmospheric Research Boulder, CO

Denver Int. Airport accident 20 Dec 2008 - 1818 MST (21 Dec 2008 - 0118 UTC ) Boeing 737-500 departed left side runway 34R during take-off Strong westerly crosswinds at time of accident 34R 2

Denver Int. Airport accident 20 Dec 2008 - 1818 MST (21 Dec 2008 - 0118 UTC ) Boeing 737-500 departed left side runway 34R during take-off Strong westerly crosswinds at time of accident 34R 3

Denver International Airport LLWAS wind sensor network (pink circles): Runway 34 R (350.9o magnetic) 2 3 29

LLWAS winds 1:13 – 1:23 UTC (+/- 5 min of accident) x Cornman, Weekley

DIA – located east of Rocky Mtns: DIA lat 39.8617, lon -104.6732 **Mountain induced gravity waves common over Rockies

Mountain waves and windstorms From Lilly 1978 Windstorm 11Jan1972 Solid lines – potential temperature Large amplitude wave + marks where aircraft encountered turbulence Note turbulence near Stapleton (under lee waves in stable layer)

Visible satellite 2030 UTC 20 Dec 2008 Persistent clear notch near Denver Probably due to descending air in mountain wave Meteorological conditions: Strong westerly winds near mountain top Stable layer above mountain top Vertical speed shear

Turbulent pilot reports (PIREPs): Mountain waves with turbulence in blue 18 UTC 20 Dec - 06 UTC 21 Dec 2008 Significant turbulence reported Severe to extreme turbulence within 3 minutes of accident Moderate to severe mountain waves over Rockies Associated with altitude and airspeed changes Waves encountered from 8,000 to 40,000 ft

Example pilot reports: 20081220 1510 B737 37.75 -103.69 37000 /RM MOD MTN WAVE +/- 15 KT AND 200 FT UDDFS 20081220 1820 MD80 35.11 -105.04 26000 /TB SVR 20081220 2120 B737 35.11 -105.04 37000 /RM MOD MTN WAVE +- 15 KTS 20081220 2142 B737 39.41 -108.33 39000 /RM LGT-MOD MTN WAVE +/-500 FT 20081220 2240 CL60 39.40 -105.82 40000 /RM MOD MTN WAVE +/-200 FT 20081220 2255 BE35 39.57 -104.85 8000 /TB MOD /RM STRONG MT WAVE 20081220 2256 A319 36.77 -104.64 37000 /TB MDT MTN WAVE/RM + - 15 KTS 20081220 2256 LJ45 36.80 -104.71 47000 /TB MDT MTN WAVE/RM -+ 15 KT 20081221 0115 LJ25 40.36 -105.44 17000-19000 /TB SEV-EXTREME FL190-170 20081221 0118 LJ25 40.62 -105.12 19000 /TB SEV-EXTRM

Numerical simulation of meteorological conditions on the day of the accident Allows examination of three-dimensional structure of atmospheric flow Can help determine if gustiness at DIA possibly due to large amplitude mountain waves However - can not resolve exact strength and timing of gusts

Topography inner domain Clark-Hall nonlinear, nonhydrostatic model 5 nested domains Inner domain 250 m horizontal resolution Outer domain initialized with RUC data (20 Dec 2008 22:00 UTC) RUC 1 hr data used to drive boundary conditions Vertical grid stretching Domain 5 run 00:30 - 01:51 UTC with 1 min output X Topography inner domain

East-west wind U - vertical slice x X marks incident site Note regions of high velocity air moving downward Strong wave signature in potential temperature (black lines) X-Z slice at latitude 39.88 Inner domain 5 cont int 5.0 m/sec 1:00 – 1:39 UTC (+/- 20 minutes)

East-west wind U at 14.7 m AGL X marks incident site Red arrows -LLWAS winds Black arrows – model winds Runways– black lines Pink colors easterly flow Part of domain 5 cont int 2.5 m/sec 1:00 – 1:39 UTC (+/- 20 minutes)

High resolution Clark-Hall model results show: Lee waves in stable layer above DIA Amplification of lee waves associated with downward penetration of high velocity air Creates localized gustiness at airport May or may not be a common occurrence

High resolution Clark-Hall model results show: Lee waves in stable layer above DIA Amplification of lee waves associated with downward penetration of high velocity air Creates localized gustiness at airport May or may not be a common occurrence Are operational models high enough resolution to capture this event?

Resolution similar to HRRR X-Z slice at 1:30 UTC U wind speed (color) WRF model control simulation – 3 km horizontal resolution; single domain Resolution similar to HRRR X-Z slice at 1:30 UTC U wind speed (color) Wave over mountains Small wave downstream But high velocity air not penetrating to the ground near DIA X Near Time of Incident (0130 UTC)

WRF Simulations – control vs higher horizontal resolution X Near Time of Incident (0130 UTC) Near Time of Incident (0132 UTC) Control D = 3.3 km D = 367 m

Conclusions Both Clark-Hall and high resolution WRF simulations showed amplifying lee waves above DIA associated with higher velocity air penetrating close to the ground. For this event need higher resolution than current operational model (HRRR). Is it possible to forecast conditions favorable for this phenomena: From low resolution model output? From wind sensor data (LLWAS)? This research is in response to requirements and funding by the Federal Aviation Administration (FAA). The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA.

backup

WRF higher resolution run WRF version 3.4 Nested domains Inner domain 367 m horizontal resolution MYJ PBL WSM 6-class microphysics Noah land model RRTM (Rapid Radiative Transfer Model) longwave Goddard shortwave radiation schemes Horizontal mixing - Smagorinsky type diffusion

DIA numerical model configuration Clark-Hall 3-D nonlinear, nonhydrostatic terrain-following numerical model Initialized with 22:00 UTC RUC data (20 Dec 2008) 5 nested domains: domain DX, DY NX NY NZ domain started ---------- ---------- ---- ----- ------ --------------------- 1 18 km 130 194 81 22:00 UTC 12-20-08 2 6 km 194 290 140 22:00 UTC 3 3 km 194 290 134 23:00 UTC 4 1 km 290 434 128 00:00 UTC 12-21-08 5 .25 km 587 866 122 00:30 UTC Grid stretched in the vertical; domains 3-5 have 2:1 increase in z resolution compared with domains 1 and 2 Domain 5 run to 01:51 with 1 min output RUC 1 hour data used to drive boundary conditions