A significant wind shear event leading to aircraft diversion at the Hong Kong international airport

Authors


Abstract

A significant wind shear event occurred in the early morning of 27 December 2009 at the Hong Kong International Airport (HKIA), leading to the diversion of three aircraft to the airport of Shenzhen (about 39 km to the north-northwest of HKIA). This paper documents the meteorological observations and predictions for the event. The significant wind shear arose from disruption of the prevailing easterly to southeasterly flow by the complex terrain to the south of the airport. It was well captured by the Doppler Light Detection and Ranging (LIDAR) systems at HKIA and, as such, the aircraft were given sufficient warning about the occurrence of significant wind shear. The headwind changes along the glide paths as measured by the LIDAR were consistent with the aircraft measurements. The observations of the remote-sensing instruments during the event, including the LIDAR, are described in the paper. Moreover, this paper discusses the possibility of including wind shear forecast information for flight planning purpose for this particular case. The forecast wind field near HKIA based on a high-resolution numerical weather prediction (NWP) model is described. For the present wind shear event, it appears that the occurrence of terrain-disrupted airflow near HKIA may be forecast about 6 h ahead. The timing of significant wind shear is, however, not reproduced well in the NWP forecast. Copyright © 2011 Royal Meteorological Society

1. Introduction

Terrain-induced airflow disturbances are a major cause of significant wind shear at the Hong Kong International Airport (HKIA). Most of them occur in the spring when east to southeasterly winds climb over the complex terrain over Lantau Island to the south of the airport in the stable boundary layer of the atmosphere. In the morning of 27 December 2009, an early-season easterly airstream affected the southeastern coast of China (synoptic pattern in Figure 1). Between 0000 and 1100 LST on 27 December 2009 (Hong Kong time = UTC + 8 h), there were 25 pilot reports of encountering significant wind shear. The aircraft providing wind shear reports included landing and departing flights, and involved both the north and the south runways of HKIA. Due to significant low-level wind shear, three aircraft landing at HKIA from three separate airlines had to divert to the nearby airport at Shenzhen. According to what could be recalled from the airlines, this is the first time that aircraft at HKIA needed to divert to another airport due to significant wind shear in non tropical cyclone and non-thunderstorm situations.

Figure 1.

Surface isobaric chart at 0000 UTC (0800 LST), 27 November 2009

This paper documents the meteorological observations during this low-level wind shear event. In particular, the measurements from ground-based, remote-sensing equipment will be discussed to look for signature of terrain-disrupted airflow disturbances. It turns out that the Doppler Light Detection and Ranging (LIDAR) systems (locations in Figure 2) at HKIA successfully captured the headwind changes along the glide paths during the event and they are generally consistent with the aircraft observations. Technical specifications of the LIDAR can be found in Shun and Chan (2008).

Figure 2.

Surface observations around HKIA at 0532 LST, 27 December 2009 (2132 UTC, 26 December 2009). The green wind barbs give the surface wind speeds and directions. Temperature, dew point and sea surface temperature are shown in blue, purple and brown respectively (in degrees Celsius). Pressure is shown in black, e.g. 142 means 1014.2 hPa. The LIDAR systems over HKIA are indicated by blue dots. Each runway has a length of about 3.8 km (as a scale for the map)

Moreover, from the feedback of an airline, there is a suggestion to provide wind shear forecast information for flight planning purpose for such a significant wind shear event. The possibility of forecasting the occurrence of terrain-disrupted airflow at HKIA as well as the associated headwind changes along the flight paths are discussed in this paper using the results from a high-resolution numerical weather prediction (NWP) model.

2. Meteorological observations

A moderate to fresh easterly airstream affected the south China coastal area in the early morning of 27 December 2009. Synoptically, this airstream was associated with a ridge of high pressure along the southeastern coast of China (Figure 1). Meanwhile, a cold front over the inland areas of Guangdong was moving southward towards the coast. Winds veered with altitude in Hong Kong so that southeasterly winds prevailed at the mountain tops of Lantau Island—a hilly island to the south of HKIA. The surface observations at 0532 LST, 27 December 2009, when significant wind shear was experienced by a landing aircraft, are shown in Figure 2.

With the prevalence of cooler easterly winds near the surface and warmer south to southeasterly winds aloft, the boundary layer was stable in the morning of 27 December 2009. In fact, from the radiosonde measurements at King's Park (location in Figure 2) at 0000 UTC on that day (0800 a.m. LST), there was a temperature inversion of about 3° between 700 and 900 m above mean sea level (Figure 3). The occurrence of cross-mountain airflow and a stable boundary layer is a typical meteorological setup for terrain-induced wind shear at HKIA in autumn and spring time.

Figure 3.

Vertical profiles of temperature (blue) and dew point (purple) measured by the upper-air ascent at King's Park at 0000 UTC, 27 December 2009

The terrain-disrupted airflow disturbances were captured well by the LIDARs at HKIA. Figure 4 shows a radial velocity plot from the Plan Position Indicator (PPI) scan of the south runway LIDAR with an elevation angle of 3° from the horizon. Against the prevailing east to southeasterly airflow, there was an extensive area of winds with opposite direction (coloured green in Figure 4) to the west and south of the point 3 nautical miles (5.6 km) west-southwest of the western end of the south runway. Furthermore, there was also a streak of higher wind speed between 1 and 2 nautical miles (1.9–3.7 km) to the west of the south runway. As the aircraft landed at HKIA from the west, they would encounter significant wind shear when the winds changed from westerly (tailwind) to easterly (headwind). The reverse flow appears to be associated with the mountains on Lantau Island.

Figure 4.

Radial velocity at 3.2° PPI scan as measured by the south runway LIDAR at 2133 UTC, 26 December 2009 (0533 LST, 27 December 2009). The marks on the extended runway centrelines correspond to distances of 1, 2 and 3 nautical miles (1.9, 3.7 and 5.6 km) away from the runway ends

3. Comparison with aircraft data

The north runway of HKIA was closed in the early morning of 27 December due to maintenance, and an aircraft had to land on the south runway from the west. As indicated in HKO, IFALPA and GAPAN (2010), wind shear is expected to be more frequent over the south runway due to its closeness to the hills of Lantau Island. Three aircraft had to divert to Shenzhen airport because they could not land at HKIA. The flight recorder data from one of the aircraft is shown in Figure 5. This aircraft had made two attempts to land at the south runway of HKIA but failed to do so. From the flight recorder data, during the two landing attempts, there were sudden increases of headwind of 20–25 knots (10–13 m s−1) when the aircraft was about to touch down on the runway. Such headwind changes are consistent with the pilot reports of headwind gains of about 25 knots (13 m s−1) at short final.

Figure 5.

The measured headwind and radio altitude of an aircraft conducting two missed approaches at the south runway of HKIA from the west in the morning of 27 December 2009: (a) 2132 UTC, 26 December 2009 and (b) 2147 UTC, 26 December 2009. The headwind data of the aircraft are compared with those measured by the south runway LIDAR. Five knots is about 2.6 m s−1 and 500 feet is about 152 m.

Thumbnail image of

is the radio altitude of the aircraft,

Thumbnail image of

is the headwind measured by the aircraft and

Thumbnail image of

is the headwind measured by the south runway LIDAR

Flight recorder data are only available from a couple of local airlines in Hong Kong. As such, it is not possible to analyse the data of all the flights that reported encountering significant wind shear in the event. However, based on the data from the available flights (not shown), the headwind changes appeared to be similar to those given in Figure 5 and, as such, the wind shear feature as shown in Figure 5 is believed to be persistent during the event.

The glide-path scans are implemented for the LIDAR systems at HKIA for the measurement of headwinds along the flight paths in clear-air situations and the detection of low-level wind shear. Details of the LIDAR Wind Shear Alerting System (LIWAS) are described in Shun and Chan (2008). It can be seen from Figure 5 that the headwind profiles from the south runway LIDAR are generally consistent with those from the aircraft. In general, both datasets indicate that the headwind does not change significantly beyond 1 nautical mile (1.9 km) away from the western end of the south runway. In between 1 nautical mile (1.9 km) and the runway threshold, the headwind increases abruptly, starting from a few knots up to 15–20 knots (8–10 m s−1). Based on the headwind profiles, LIWAS gave wind shear alerts of headwind gains of 20–25 knots (10–13 m s−1) to the aircraft. In fact, all the aircraft encountering significant wind shear in the morning of 27 December 2009 had been provided with wind shear warnings/alerts, which were basically all based on the LIDAR observations. In the present case of significant wind shear, the LIDARs at HKIA had provided sufficient alerting of the wind shear to the aircraft.

As discussed in Shun and Chan (2008), LIWAS generates wind shear alerts by examining locations of significant headwind changes based on the headwind profiles measured by the LIDAR systems. For terrain-induced wind shear, headwind loss and gain may occur over the same runway corridor, and there could be multiple occurrences of wind shear features. LIWAS gives higher priority for headwind losses compared to headwind gains, and the wind shear features are also prioritized based on a wind shear severity parameter. The LIWAS alerts include the location of the wind shear feature, the magnitude of headwind change and the sign (plus for headwind gain and minus for headwind loss) for the feature of the highest priority.

4. Forecasting of wind shear by NWP model

Though the aircraft had been given adequate warning, there is a suggestion of the possibility of providing forecasts of such a severe wind shear event at an earlier time, e.g. at the pre-flight planning stage. The LIDAR and other meteorological observations at HKIA could mostly be useful in the detection and very short-term nowcasting of the occurrence of low-level wind shear, e.g. in the next several minutes or so. The longer term forecasting of wind shear occurrence would only be possible with the use of NWP model at high spatial resolution, sufficient to resolve the complex terrain in the vicinity of HKIA. This kind of NWP forecast of low-level wind shear and turbulence has been described in, for instance, Szeto and Chan (2006) and Chan (2009).

For the wind shear event under study in this paper, a special NWP model run has been performed with a horizontal resolution of 200 m, which is sufficient to resolve the mountains of Lantau Island. The model in use is Regional Atmospheric Modelling System (RAMS) version 4.4 (the same as that in Szeto and Chan, 2006) and it is nested within the 20 km resolution Operational Regional Spectral Model (ORSM) of the Hong Kong Observatory (HKO). RAMS was run with nested grids having spatial resolutions of 4 km, 800 m and 200 m with two-way nesting. The model simulation domains are shown in Figure 6. The Mellor–Yamada turbulence parameterization scheme was used in the first grid, and the Deardorff scheme in the other two grids. As mentioned in Chan (2009), the Deardorff scheme appeared to give the best results in simulating the terrain-disrupted airflow at HKIA among the turbulence parameterizations available in RAMS. The model run started at 1200 UTC, 26 December 2009 and was carried out for 12 h. It turns out that, starting at about 1800 UTC, 26 December, a vortex with reverse flow appears over the sea west of HKIA. A sample ‘LIDAR plot’ of the forecast wind data is given in Figure 7. The corresponding headwind profile along the arrival glide path to the west of the south runway of HKIA is shown in Figure 8. The forecast temperature profile within the boundary layer (not shown) is consistent with the actual radiosonde measurement at 0000 UTC of 27 December. However, the model does not forecast the higher wind speed streak occurring between 1 and 2 nautical miles (1.9 and 3.7 km) to the west away from the western end of the south runway.

Figure 6.

The model simulation domains: grid 1, grid 2 and grid 3. Height contours are in every 200 m

Figure 7.

The simulated horizontal wind field at a height of 30 m above sea level as resolved along the measurement radials of the south runway LIDAR, i.e. the ‘LIDAR plot’ of the simulated wind field at 2040 UTC, 26 December 2009, when the mountain wake to the west of HKIA develops into the maximum spatial extent in the model simulation

Figure 8.

The simulated headwind profile at the arrival glide path to the west of the south runway (i.e. over 07RA runway corridor) at the time of Figure 7 is shown as the black curve. ‘Range’ refers to the distance from the western end of the south runway. The wind barbs at the bottom of the plot are the simulated horizontal winds along the glide path. The horizontal, broken line in black is used to mark the range of headwind change of 15 knots. Five knots is about 2.6 m s−1

From the model simulation (Figure 7), the location of the simulated mountain wake (between 22.27 and 22.28°N, 113.84 and 113.89°E) is similar to that of the actual wake observed in Figure 4. On the other hand, the tiny area of reverse flow (coloured green in Figure 4) between 0 and 1 nautical mile (0 and 1.9 km) away from the western end of the south runway in the actual observation (Figure 4) is not reproduced in the numerical simulation. The southeasterly jet over the airport in the model simulation is also weaker than actual.

It can be seen from the above results that the NWP model appears to show some success in forecasting the occurrence of terrain-disrupted airflow in the vicinity of HKIA about 6 h ahead. There is potential to use the model forecasting results to indicate the chance of the occurrence of significant wind shear. However, by comparing the headwind profiles of Figures 5 and 8, the location and magnitude of significant headwind change in the actual observations are not reproduced well in the numerical simulation. Of course, the time difference between Figures 5 and 8 may be a factor. Considering the transient and sporadic nature of terrain-disrupted airflow, it may not be possible for the numerical simulation to reproduce the terrain-induced wind shear feature at exactly the same time as its actual occurrence.

In order to find out if further increase of model resolution is able to capture the terrain-induced wind shear, an additional domain (grid 4) with a horizontal resolution of 50 m is considered. The technical details of the model run are the same as those in Chan (2009). The Deardorff parameterization scheme in RAMS is used in grid 4. The model-simulated field of horizontal wind speed is shown in Figure 9(a). It could be seen that with higher spatial resolution the model successfully captures the higher wind speed areas over the threshold at the western part of the south runway of HKIA as well as about 1 nautical mile (1.9 km) to the west of the threshold. The corresponding headwind profile is shown in Figure 9(b), together with aircraft data in Figure 5(b). The profiles at different times are compared because it may not be possible for model simulation to reproduce the terrain-induced wind shear feature at exactly the same time as its actual occurrence. Though the times of the model simulation and the landing aircraft are different, their headwind profiles look very similar. It appears that, with sufficiently high spatial resolution, numerical model simulation has the potential to capture wind shear in highly complex terrain. Though the actual timing of wind shear occurrence may not be successfully predicted, the model simulation results could give an indication of the potential of the occurrence of terrain-induced wind shear.

Figure 9.

The simulation results for grid 4. Part (a) shows the simulation domain and the wind speed distribution at a height of 10 m above sea level. Streamlines of the airflow and the simulated winds at the locations of the ground-based automatic weather stations are shown. Part (b) is the situation in Figure 5(b) together with the model-simulated headwind profile at the time 2006 UTC, 26 December 2009 (

Thumbnail image of

)

5. Conclusions

The significant wind shear event in the morning of 27 December 2009 is discussed in the present paper. Based on what could be recalled from the airlines, it is the first time that aircraft had to divert to another airport due to low-level wind shear in nontropical cyclone and non-thunderstorm situation. The meteorological setup of east to southeasterly airflow in a stable boundary layer appears to be conducive to the occurrence of low-level wind shear due to terrain-disrupted airflow. The LIDAR systems worked well in capturing the wind shear and provided sufficient alerts to the aircraft.

The possibility of providing a longer-term forecast for such a severe wind shear event has been examined by a special NWP model run at sufficiently fine grids for resolving the complex terrain in the vicinity of HKIA. It turns out that the model may give an indication of terrain-disrupted airflow at the airport up to 6 h ahead. With sufficiently high spatial resolution (e.g. horizontal resolution of 50 m), the occurrence of terrain-induced wind shear in areas of complex terrain could be captured. It is envisaged that, when provided with such model simulation data, the aviation weather forecaster may mention the chance of low-level wind shear for HKIA in the pre-flight planning information, but it could be difficult to highlight the actual timing of the event. It would be useful to run a suitable high resolution model on a regular basis so that the statistical skill of forecasting such an event could be determined quantitatively. The trial of such an ‘aviation model’ would be a major development area of HKO in the years to come.

Acknowledgements

The author gratefully acknowledges the support of Cathay Pacific Airways Ltd which provided the QAR data used in this study and the assistance of pilots for filing wind shear reports to HKO for the purpose of enhancing flight safety.

Ancillary