A tail strike event of an aircraft due to terrain-induced wind shear at the Hong Kong International Airport

METEOROLOGICAL APPLICATIONS Meteorol. Appl. 21: 504 511 (2014) Published online 14 March 2012 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/met.1303 A tail strike event of an aircraft due to terrain-induced wind shear at the Hong Kong International Airport P. W. Chan* Hong Kong Observatory, Kowloon, Hong Kong, China ABSTRACT: At about 0155 UTC, 22 February 2009, an aircraft departing from the south runway of the Hong Kong International Airport to the east experienced significant wind shear of headwind loss. This may contribute towards the tail strike of the aircraft. This paper documents the meteorological observations in this tail strike event. The case occurs in a background atmosphere with a stable boundary layer and fresh east to southeasterly winds near the surface. The surface anemometer readings do not indicate the occurrence of significant headwind drop over the south runway. The glide-path scan data of a Doppler Light Detection and Ranging (LIDAR) system over the runway corridor concerned also do not indicate significant changes of the headwind due to blind zone of the LIDAR and geometrical constraint. However, the wind data measured on board the aircraft show that the tail strike may be due to a wind change from headwind of 19 knots to a tailwind of 5 knots over the western and middle parts of the south runway when the aircraft was on rotation. The headwind drop appears to be due to a jet of more easterly component near the surface occurring over the western part of the south runway. This drop could be captured if the LIDAR s headwind profiles over different runway corridors could be combined, and the present case suggests that, for wind shear alerting purposes, it may be necessary to try out combinations of headwind profiles over different parts of the same runway. The possibility of forecasting the significant wind shear in this case is also studied using a numerical weather prediction (NWP) model. The model results show that it may not be possible to forecast the event by using the direct model output alone, but would need to consider both the simulated wind direction as well as the forecast gust near the surface. Here the gust is forecast based on a physical approach as applied to the NWP model output. KEY WORDS tail strike; windshear; LIDAR Received 24 October 2011; Revised 22 December 2011; Accepted 31 January 2012 1. Introduction The Hong Kong International Airport (HKIA) is situated in an area of complex terrain. To its south is the mountainous Lantau Island with peaks rising to about 1000 m above mean sea level and valleys as low as 400 m in between. The complex terrain may lead to airflow disturbances in the vicinity of the airport when the air climbs over the mountains. Terraindisrupted airflow is common when the boundary layer is stable in the background atmosphere, which often occurs in the spring time. The airflow disturbances over HKIA could be hazardous to the landing/departing aircraft by causing low-level wind shear, which refers to wind shear occurring within a height of 1600 feet (about 500 m) over the airport or within a distance of 3 nautical miles (about 5.6 km) from the runway end. In aviation meteorology, the wind shear is considered to be significant if it leads to headwind/tailwind change of 15 knots (7.7 m s 1 ) or more. Significant wind shear may cause the aircraft to deviate from its intended flight path by changing the lift, which could be hazardous to the operation of the aircraft when it is close to the ground, e.g. during rotation for departing aircraft. At HKIA, terrain-induced wind shear is the most common type of low-level wind shear experienced by the aircraft. Based * Correspondence to: P. W. Chan, Hong Kong Observatory, 134A Nathan Road, Kowloon, Hong Kong, China. E-mail: pwchan@hko.gov.hk on the pilot reports, about 70% of the wind shear is due to terrain disruption of the airflow. The other causes of wind shear include sea breeze, microburst, gust front, low-level jets as well as building effects on the airflow. Terrain-induced wind shear could occur in the spring time, when the background easterly winds of continental origin flow over Lantau Island under a stable atmosphere, and it could also appear in the summer time, with the cross-mountain airflow brought about by intense summer monsoon or tropical cyclones. Though it is not common, there have been reports of tail strikes of the departing aircraft at HKIA due to terrain-induced wind shear. According to the records of an airline in Hong Kong, there were two such reports in the recent years. One such case occurring in March 2010 has been documented in Chan (2011). In the present paper, another case of tail strike of an aircraft due to terrain-induced wind shear occurring in February 2009 is described. The peculiarities of this event as compared to the event in March 2010 include: 1. while the cause of the tail strike in the present case again appears to be related to headwind loss occurring near the ground, this headwind loss was not recorded by the surface anemometers, contrary to the 2010 case, and, 2. contrary to the 2010 case, while the headwind profile measured by the Doppler Light Detection And Ranging (LIDAR) system over the runway corridor concerned does not give the headwind loss leading to the tail strike, it may be possible to combine the headwind profiles obtained over 2012 Royal Meteorological Society

Tail strike event of aircraft 505 different runway corridors in order to capture the headwind loss. This may suggest another approach in using the LIDAR data for alerting wind shear occurring at very low level close to the ground, which is not attempted with the present geometrical setup of the LIDAR Wind shear Alerting System (LIWAS) (Shun and Chan, 2008). Moreover, as in the 2010 case, the possibility of using a high resolution numerical weather prediction (NWP) model for forecasting the significant wind shear in the present case is considered in order to find out the possibility of providing an earlier alert to the aviation weather forecasters about the chance of occurrence of significant low-level wind shear. The approach of using an NWP model includes numerical simulation down to a horizontal resolution of 50 m so as to resolve the fine details of the terrain of Lantau Island, and the consideration of gust forecast using a physical method as described in Brasseur (2001). 2. Synoptic situation The tail strike event occurred at 0155 UTC (0955 HKT, which is 8 h ahead of UTC), 22 February 2009. The aircraft concerned departed from the south runway of HKIA to the east. The surface weather chart at 0000 UTC, 22 February 2009 is shown in Figure 1. There was a ridge of high pressure over the part of the southeastern coast of China closest to Hong Kong, bringing easterly winds to Hong Kong and the adjacent areas. The easterly winds were not particularly strong. In fact, over the south China coast there were only three isobars (with a pressure increment of 2 hpa between two isobars). This is a typical synoptic pattern of easterly winds with continental origin over Hong Kong in the spring time. Higher up in the boundary layer of the atmosphere, for instance at 925 and 850 hpa levels (not shown), south to southwesterly winds prevailed along the coastal area, and they had an origin from the more humid and warmer airmass over the South China Sea. The undercutting of the warmer southerlies by the cooler surface easterlies resulted in a stable boundary layer as well as the occurrence of low clouds and light rain patches. Close to the time of the event (0200 UTC, 22 February 2009), the weather observer at HKIA reported 6 oktas of clouds, with a cloud base height of 1000 feet (about 300 m). 3. Local meteorological observations The surface wind data around the time of the event (0155 UTC, 22 February 2009) are shown in Figure 2. Southeasterly winds basically prevailed over HKIA and the sea to the west. On the other hand, the winds had a more easterly component at the northeastern corner of the airport island as well as the sea to the east. Over the south runway, all four anemometers recorded south-southeasterly winds of 15 25 knots (7.7 12.9 m s 1 ). There was generally a decrease of wind speed from the western to the eastern end of the south runway, but the corresponding headwind change did not appear to be significant (i.e. far less than 15 knots, or 7.7 m s 1 ). The background wind profile in the lower atmosphere may be indicated by the measurements of the radar wind profiler at Cheung Chau (location in Figure 2). This wind profiler operates at a central frequency of 1299 MHz. It measures the three Figure 1. Surface isobaric chart at 0000 UTC, 22 February 2009.

506 P. W. Chan Figure 2. Observations from the surface weather stations at 0955 HKT, 22 February 2009. The green wind barbs give the wind observations. The value at the upper right corner of a station is the pressure in 0.1 hpa (after deducing 1000 hpa), that at the upper left corner is the temperature, and those at the lower left corner are the dew point and the sea surface temperature. Height contours are in 100 m. A full wind barb equates to 10 knots (5.1 m s 1 ) and a half wind barb equates to 5 knots (2.6 m s 1 ). The blue dots are the locations of the two LIDARs. components of the wind up to a height of 6000 m above ground by using three radar beams, namely, one to the vertical, and two oblique beams with an angle of about 15 from the vertical. Wind data are available every 60 m in the low mode (up to 1500 m above ground), and every 200 m in the high mode (up to 6000 m above ground). The wind profiles presented in this paper are obtained by combining the data from the two modes. The data are basically 10 min averages. The wind profile at the time of the event (0155 UTC, 22 February 2009) from the Cheung Chau wind profiler is given in Figure 3. In the lowest 600 m or so above mean sea level, wind speed increased slightly from 7 to 11 m s 1.The winds then weakened a bit aloft up to 1000 m, and increased again in the upper part of the boundary layer. For wind direction, it was veering steadily with height, from southeasterly near the surface becoming southwesterly at the top of the boundary layer. The cross-mountain airflow of south to southeasterly winds, together with a wind speed in the order of 10 m s 1 within the first few 100 m above ground, could be favourable to the occurrence of terrain-disrupted airflow downstream of Lantau Island, namely, in the region of HKIA. The background temperature profile is obtained by radiosonde ascent at King s Park, the radiosonde station in Hong Kong located at about 25 km to the east of HKIA. The radiosonde data are shown in Figure 4, including the temperature and the dew point profiles. It can be seen that the atmosphere was close to saturation above 200 m from the mean sea level, which is explained in the previous section as a result of the cutting of the warmer and more humid airflow with maritime origin by the cooler continental airflow near the surface. Moreover, this undercutting results in a temperature inversion of about 2 C between 200 and 300 m above mean sea level. Figure 3. Wind profile given by the wind profiler at Cheung Chau at 0155 UTC, 22 February 2009: wind speed (a) and wind direction (b). The wind data refer to 10 min averages. The height of this temperature inversion is well below the altitude of the peaks of Lantau Island. Combining with the results from the Cheung Chau wind profiler, it is noted that the

Tail strike event of aircraft 507 Figure 4. Temperature (blue) and dew point (pink) profile from the radiosonde ascent at King s Park at 0000 UTC, 22 February 2009. cross-mountain airflow in a stable atmosphere with the temperature inversion below the height of the mountains could be favourable for the occurrence of terrain-induced airflow disturbances over the airport. 4. LIDAR observations The Hong Kong Observatory (HKO) operates two LIDARs at HKIA, with one serving each of the two runways of the airport. The LIDAR uses an infrared laser beam with a wavelength of 2 μm for measuring the line-of-sight (or radial) velocity of the wind. There is a blind zone of about 400 m from the LIDAR, in which no data could be collected due to constraints of the optics of the system. The main application of the LIDAR is the alerting of low-level wind shear experienced by the aircraft, using an algorithm called LIWAS developed by HKO (details could be found in Shun and Chan, 2008). The locations of the LIDARs are shown in Figure 2. LIWAS alerts wind shear based on a special scanning pattern of the laser beam, called the glide-path scan. In this kind of scanning, the elevation and the azimuthal motors of the LIDAR s scanner rotate in unison to slide the laser beam along the glide path of the aircraft, which is taken to be a line with an elevation angle of 3 originating from the runway end for arriving aircraft and a line with an elevation angle of 6 originating from the middle of the runway for departing aircraft. The radial velocities so collected along a glide path are used to construct the headwind profile to be encountered by the aircraft, as long as the radial direction and the runway orientation are sufficiently close together. A threshold of 30 is adopted for the angle between the measurement radial and the runway direction, i.e. when the angle is less than 30, the radial velocity is used directly to represent the headwind in the direction of the glide path. The headwind profile obtained from the glide-path scan is the basis for wind shear detection and alerting. An algorithm has been developed by HKO to automatically look for abrupt changes of the headwind in this profile. Apart from the glide-path scan, the LIDARs also perform the more conventional scanning patterns such as Plan-position Indicator (PPI) scans to give an overview of the wind distribution over the airport area to support the work of the aviation weather forecasters. Since the tail strike event in the present study occurs over the south runway, the data from the south runway LIDAR are considered here. The 6 PPI scan of this LIDAR at the time of the event is given in Figure 5(a). The 6 PPI data are used because it is close to the 6 glide path of the departing aircraft. It could be seen that a southeasterly airstream with a speed of 14 m s 1 prevailed over the airport. There are some small-scale features of reverse flow, e.g. appearing as green dots over the eastern part of the south runway, each having a spatial scale of a couple of 100 m, against the background away-from-the-lidar flow (in brown and yellow). However, such features are small in both magnitude and scale compared to the change in headwind apparent in the aircraft data (see Figure 6, to be discussed below). The headwind profile obtained by the south runway LIDAR for the runway corridor concerned, 07RD corridor (departing from the south runway to the east), is shown in Figure 6(a). It is compared with the headwind measured onboard the aircraft that experienced the tail strike. The corresponding heights of the laser beam and the aircraft are shown in Figure 6(b). It could be seen that the two headwind profiles are generally consistent with each other for the distance of 1 nautical mile from the runway end (i.e. 1 nautical mile to the west of the eastern end of the south runway) and further to the east. Over

508 P. W. Chan Figure 6. Headwind profile (a) from the aircraft that experienced tail strike, the south runway LIDAR, and the model simulation; the altitudes of the aircraft, LIDAR scan and model data extraction are given in (b). Imperial units are used because they are more commonly adopted in aviation meteorology. They correspond to metric units as follows: 1 knot = 0.5144 m s 1, 1 nautical mile = 1.8519 km, and 1 feet = 0.3048 m. The data are taken at 0155 UTC, 22 February 2009. Figure 5. Radial velocity image from the south runway LIDAR as measured in 6 conical scan (a) and 3.2 conical scan (b) at 0155 UTC, 22 February 2009. that region, the headwind had minor fluctuations only, and there did not appear to have significant wind shear. It turns out that the significant wind shear that may lead to the tail strike appears at the distance of about 1.5 nautical miles to 1 nautical mile, namely, with a headwind drop of 19 knots (9.8 m s 1 )to 5 knots ( 2.6 m s 1 ) over a distance of 0.5 nautical mile (0.9 km). Due to the blind zone of the LIDAR and the geometrical constraint (i.e. the angle between the runway orientation and direction of the laser beam is more than 30 ), no headwind data are available from the 07RD glide-path scan itself. In order to supplement the headwind profile over 07RD, the headwind profile obtained over the western part of the south runway, 07RA runway corridor, is considered (07RA means arriving at the south runway from the west). The corresponding 3.2 PPI scan imagery of the south runway LIDAR (mainly used for monitoring the airflow to be encountered by the arriving aircraft) is shown in Figure 5(b). It could be seen that the low-level southeasterly jet over the western part of airport as shown in the 3.2 PPI scan had more easterly component than that shown in the 6 PPI scan. As a result, the corresponding 07RA glide-path scan data (Figure 6(a)) had rather large headwind value at a distance of 1.5 nautical miles, namely, about 18 knots (9.3 m s 1 ). The altitude of the laser beam in the 07RA glide-path scan is shown in Figure 6(b). If the glide-path scan data of 07RA and 07RD are combined, basically the shape of the headwind profile is very similar to that recorded onboard the aircraft (Figure 6(a)). Unfortunately, due to the blind zone of the LIDAR and geometrical constraint, the headwind data at the distance between 1.2 and 1 nautical miles from the eastern end of the south runway are not available. Nonetheless, the available headwind data obtained by glide-path scans show that there would be a significant drop of headwind over the western part of the south runway. The combination of headwind data from different glide-path scans is not made in the existing LIWAS algorithm. Based on the present case, it may be necessary to do such headwind profile combination in order to capture the significant wind shear that occurs at very low level near the ground.

Tail strike event of aircraft 509 Figure 7. The simulated wind magnitude at the surface at 0200 UTC, 22 February 2009 for the model domain with a spatial resolution of 200 m; the colour scale of the wind magnitude is given at the bottom of the figure. The simulated winds at the weather stations are given in wind barbs. The runways are indicated by the black straight lines. 5. Numerical simulation The possibility of forecasting the present case of significant wind shear is studied by using Regional Atmospheric Modelling System (RAMS) version 4.4. It is nested with the Operational Regional Spectral Model (ORSM) of HKO with a horizontal resolution of 20 km. Four nested runs of RAMS are performed, with a horizontal resolution of 4 km, 800, 200 and 50 m. High-resolution terrain data of Hong Kong are used in the model, with a horizontal resolution of 100 m. Details of RAMS could be found in Cotton et al. (2003). In the simulation, The Deardorff turbulence parameterization scheme (Deardorff, 1980) is employed. The simulated wind field at a height of 10 m above ground for the 200 m resolution domain is shown in Figure 7. It can be seen that the wind directions at the western and the middle parts of the south runway are very similar, southeasterly winds. However, the wind speed at the western part is slightly higher than that over the middle part. This slightly higher wind speed leads to a higher value of headwind. The model also simulates that the wind at the eastern part of the south runway has more easterly component. As a result, the headwind value over there is also higher. This pattern is basically maintained in the 50 m resolution run (Figure 8). As a result, the simulated headwind profile over the south runway takes on the shape as given in Figure 6(a), higher values of headwind at a distance of 1.5 nautical miles and 0.5 nautical miles from the eastern end of the south runway, and lower value at a distance of about 1 nautical mile. The corresponding altitudes of the model grid points taken to construct the headwind profile are shown in Figure 6(b), basically following the height of the aircraft concerned. It can be seen from Figures 7 and 8 that the model-simulated wind field near the surface has two major discrepancies as compared with actual observations (e.g. Figures 2 and 5). First, the region of east-southeasterly winds is more extensive over the airport island in the simulation results, as compared with the actual surface observations from the anemometers. The former cover the north runway and the eastern part of the south runway, while in reality the east-southeasterly winds only appear at the northeastern corner of HKIA. Second, there is an extensive area of cyclonic flow to the west of HKIA, but it is not present in the actual observations. The model may have over-reacted to the low level temperature inversion (Figure 4) in the simulation of the near-surface wind field. Despite the above discrepancies, the model basically reproduces the board features of the terrain-disrupted southeasterly airflow over the airport island. However, by using the direct model output only, the simulated headwind profile does not show the significant headwind change as experienced by the aircraft, Figure 6(a). Similar to Chan (2011), the authors would like to see the possibility of using the forecast gust as an indication of stronger headwind change. The physical modelling approach for gust as described in Brasseur (2001) is adopted. It is applied to the modelling results for the horizontal resolution of 200 m. The forecast gust near the surface is given in Figure 9. It could be seen that, probably due to airflow disruption by Nei Lak Shan, the gust could be higher over the western part of the south runway, reaching about 27 knots (14 m s 1 ). By using the simulated wind direction over that region, the simulated headwind could reach 21 knots (11 m s 1 ). Together with the simulated headwind minimum of about 1 knots (0.5 m s 1 )at a distance of about 1 nautical mile from the runway end, the headwind change (21 knots to 1 knots) becomes comparable with that from the actual aircraft data. Therefore, by considering the spatial distribution as well as the value of the simulated

510 P. W. Chan Figure 8. The simulated wind magnitude at the surface at 0200 UTC, 22 February 2009 for the innermost domain (spatial resolution of 50 m); the colour scale of the wind magnitude is given at the bottom of the figure. The simulated winds at the weather stations are given in wind barbs. The runways are indicated by the black straight lines. gust, it may be possible to give the aviation weather forecaster an earlier alert about the chance of occurrence of significant headwind drop over the western and the middle parts of the south runway. 6. Conclusion A tail strike event of an aircraft due to terrain-induced wind shear at HKIA on 22 February 2009 is presented in this paper. From the aircraft data, the tail strike could occur as a result of significant headwind drop from 19 knots to 5 knots over the western and the middle parts of the south runway during rotation. This significant wind shear is not captured by surface anemometer data as well as the LIDAR s headwind profile over 07RD runway corridor alone. However, by combining the LIDAR s headwind profiles over 07RA and 07RD, it is possible to recover this headwind loss. The higher headwind value at the western part of the south runway may be due to a more easterly component of the southeasterly jet in the lower part of the boundary layer, after the jet passes through the gaps of Lantau Island. Based on the present case, there appears to be a need to combine the headwind profiles from the LIDAR as obtained by glide-path scans over different runway corridors in order to capture the significant wind shear that occurs close to the ground. Combining the different LIDAR scans clearly shows that there is a drop of headwind component in excess of 15 knots (the threshold for a warning) although from these data alone it is not possible to say exactly where this drop occurs. However, due to the blind zone of the LIDAR as well as geometrical constraints, there are still many missing headwind data in the combined headwind profile. It may be necessary to fill in this data void by using, for instance, an additional LIDAR or surface anemometer readings. The possibility of forecasting this event of terrain-induced wind shear is studied by using RAMS with a horizontal resolution down to 50 m. The direct model output of the simulated headwind profile does not capture the headwind loss concerned. However, by considering the simulated surface wind field as well as the forecast gust, it may be possible to give the aviation weather forecasters an earlier indication about the chance of occurrence of significant wind shear near the ground. The present results as well as Chan (2011) serve to document the meteorological observations in the only two cases of tail strikes due to terrain-disrupted airflow over HKIA in the recent years, according to the record of an airline. The common

Tail strike event of aircraft 511 Figure 9. The simulated gust at the surface at 0200 UTC, 22 February 2009 for the model domain with a horizontal resolution of 200 m; the colour scale of the gust is given at the bottom of the figure. feature of both cases is the occurrence of significant headwind drop near the ground during the rotation of the aircraft. The major difference between the two events is the height of the occurrence of the significant wind shear: the case of Chan (2011) occurred very close to the ground so that the wind shear could be picked up from the ground-based anemometer measurements, whereas in the present case the wind shear may have occurred at higher location so that it could not be captured by the anemometer data but by the LIDAR data which scanned slightly higher above the runway. More cases of tail strike would be investigated in the future to see if there are any common features about the winds shared by these events. Tail strike is a hazardous condition as a danger to the aircraft and its passengers. Fortunately, in the two events documented so far, the aircraft involved could continue its flight (departing from HKIA). There does not appear to be an emerging trend in more frequent occurrence of terrain-induced wind shear leading to tail strike. References Brasseur O. 2001. Development and application of a physical approach to estimating wind gusts. Mon. Weather Rev. 129: 5 25. Chan PW. 2011. An event of tail strike of an aircraft due to terrain-induced wind shear at the Hong Kong International Airport. Meteorol. Appl. DOI: 10.1002/met.264. Cotton WR, Pielke RA Sr, Walko RL, Liston GE, Tremback C, Jiang H, McAnelly RL, Harrington JY, Nicholls ME, Carrio GG, McFadden JP. 2003. RAMS 2001: current status and future directions. Meteorol. Atmos. Phys. 82: 5 29. Deardorff JW. 1980. Stratocumulus-capped mixed layers derived from a three-dimensional model. Boundary Layer Meteorol. 18: 495 527. Shun CM, Chan PW. 2008. Applications of an infrared Doppler Lidar in detection of wind shear. J. Atmos. Oceanic Technol. 25: 637 655.