Two-Stage Low-Level Wind Field Evolution and Fine-Scale Wind Shear Structures at Xining Caojiapu Airport Based on Multi-Source Observations

Abstract

To examine the fine-scale structure and evolution of the low-level wind field at a plateau valley airport under different weather conditions, this study analyzes a two-stage wind field event at Xining Caojiapu Airport on 8 April 2022. The analysis uses data from several scanning modes of a three-dimensional Doppler wind lidar (DWL), together with an automatic weather observation system (AWOS), sounding data, and the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5). The results show that: (1) From 13:25 to 13:45 BJT, downward momentum transport produced low-level wind shear. Under an upper-level jet, post-trough northwesterly flow, and stronger afternoon mixing in the boundary layer, west-northwesterly winds aloft descended and entered the runway area from west to east. Runway 11 responded about 1–2 min before Runway 29. The maximum wind vector difference between the runway ends was 9.42 m s⁻¹, and the maximum wind speed component difference along the glide path was 9.02 m s⁻¹. (2) From 20:15 to 20:35 BJT, the low-level wind field adjusted as a cold front moved into the airport. Strong easterly flow advanced westward from the eastern side as a shallow wedge, with local shear along its upper boundary. Runway 29 responded before Runway 11. The corresponding maximum differences at the runway ends and along the glide path were 6.44 and 4.01 m s⁻¹. (3) The two stages differed in airflow direction, the evolution of the shear interface, and the order of response at the runway ends. Combining the DWL scanning modes with AWOS observations gave a clearer view of the descending strong wind layer, the advance of low-level airflow, local shear, and wind changes over the runway and approach path. This case provides a reference for low-level wind monitoring and operational risk assessment at plateau valley airports.

1. Introduction

Wind shear is a marked change in wind speed or direction over a short time or distance. It is associated with a large horizontal or vertical gradient in the wind field [1,2]. Low-level wind shear (LLWS) generally refers to wind shear from near the surface to about 600 m (approximately 2000 ft) above ground level [3]. In this study, this height range is used only to describe the layer of main concern in aviation and is not treated as a threshold for identifying or grading the event. LLWS mainly affects aircraft during takeoff and landing. Rapid changes in wind speed and direction may alter airspeed and lift and, in severe cases, cause an aircraft to leave its intended flight path or go around [3,4]. Its small spatial scale, rapid change, and uneven distribution also make the full process difficult to observe [5,6]. More detailed observations are therefore needed to understand how these wind fields form and evolve.

LLWS can develop from processes acting at several scales. Changes in synoptic circulation and frontal activity provide the broader setting for low-level wind changes. Within the boundary layer, turbulent mixing and subsidence can strengthen momentum exchange between different heights and rapidly alter winds near the surface [7,8]. Local circulations related to surface heating, together with terrain blocking, flow splitting, and valley channeling, can make the local wind field more complex [9,10]. Around airports, LLWS is often related to convective outflows, downward transport of momentum from aloft, frontal passages, and terrain effects. The wind structures produced by these processes are not necessarily the same [11,12].

Strong winds and cold air occur frequently in winter and spring. Downward momentum transport and cold front intrusion are therefore two important causes of rapid changes in airport low-level winds [13]. Downward momentum transport is favored when strong winds or post-trough flow is present in the middle and upper atmosphere and turbulent exchange in the boundary layer is active. Horizontal momentum aloft can then be carried downward by turbulent mixing or subsidence, producing a rapid increase in low-level wind speed [14]. On clear afternoons, surface heating supports boundary layer growth and may further strengthen momentum exchange between upper and lower levels [15]. A cold front, by contrast, brings low-level cold air into the area and can quickly change wind speed and direction near the surface. Local shear may form where the advancing air meets the pre-existing flow [16].

Plateau airports are often surrounded by complex terrain, so their low-level winds may be affected by terrain blocking, flow splitting, and local circulations [17]. Xining Caojiapu International Airport lies in the Huangshui River valley in Qinghai Province, northwestern China. Valley channeling and local mountain–valley winds may add to the variability around the runway [11]. As a high-elevation valley airport in western China, Xining Airport is a useful site for examining low-level wind changes over complex terrain. Earlier studies at this airport have considered dry microbursts, convective LLWS, and frontal passages [16,18,19]. However, there has been little detailed observational work on a two-stage low-level wind field affected successively by downward momentum transport and a cold front on the same day.

Conventional point observations have limited spatial coverage and cannot continuously show the structure of the wind field around an airport [20]. A three-dimensional Doppler wind lidar was installed at Xining Airport in 2017. With several scanning modes, it provides wind information over the airport and along the glide path and is well suited to examining fine low-level structures [14]. Here, DWL, surface observations, sounding data, and reanalysis data are combined to study the two-stage evolution of the low-level wind field at Xining Airport on 8 April 2022. The first stage was an LLWS event caused by downward momentum transport. The second was mainly a low-level wind adjustment and local shear associated with cold front intrusion. The analysis focuses on the weather background, wind structure, and evolution of the two stages, with the aim of improving our understanding of low-level wind changes at high-elevation valley airports in western China.

2. Data and Methods

2.1. Study Area and Observational Data

Xining Caojiapu International Airport (ZLXN; hereafter Xining Airport) is located approximately 28 km southeast of Xining, Qinghai Province, at an elevation of 2184 m within the Huangshui River valley. The runway is 3800 m long and 45 m wide, with orientations of 110°/290°, corresponding to Runway 11 and 29, respectively. The DWL is located near the runway, and AWOS stations are installed at both runway ends (Figure 1c).

The study examines the two-stage evolution of the low-level wind field at Xining Airport on 8 April 2022 using Doppler wind lidar (DWL), Automatic Weather Observation System (AWOS), sounding, and ERA5 reanalysis data. The DWL specifications, scanning modes, and processing methods are described in Section 2.2. The other datasets are as follows. (1) AWOS observations were available every 30 s for wind and every 60 s for temperature, pressure, and relative humidity. Wind observations from the two runway ends were matched at common times. (2) Sounding data came from Xining station (WMO station 52866), about 32 km northwest of the airport. The 08:00 and 20:00 BJT soundings were used to examine the vertical atmospheric structure. Based on the 08:00 profile, the surface parcel temperature was adjusted to the daily maximum temperature to estimate surface-based convective available potential energy at 13:00 BJT. (3) Hourly ERA5 data were provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) [21] at a horizontal resolution of 0.25° × 0.25°. The 13:00 and 21:00 BJT fields on 8 April 2022 were used to examine wind, temperature, geopotential height, vertical velocity, and sea-level pressure at 200, 500, and 700 hPa and at the surface. Unless otherwise stated, all times are Beijing Time (BJT; UTC + 8), and DWL heights are above ground level (AGL).

2.2. DWL Scanning Modes, Quality Control, and Wind Field Retrieval

This study used the FC-III all-fiber coherent Doppler wind lidar developed by the Southwest Institute of Technical Physics. The lidar operates at a wavelength of 1.55 μm and measures radial velocity along the laser beam from the Doppler shift of aerosol backscatter. Its vertical detection range is 30–3000 m, the horizontal range is 30–10,000 m, and the measurable wind speed range is 0–75 m s⁻¹. The main specifications are listed in

Table 1. Main specification of the FC-III wind lidar.

Parameter	Specification
Power consumption	≤150 W
Wavelength	1.55 μm
Detection range	30–3000 m (Vertical) 30–10,000 m (Horizontal)
Wind speed range	0–75 m·s−1
Range resolution	15–120 m
Scanning modes	PPI, RHI, DBS, GP
Scanning range	0–360°/−90–+90°
Data update rate	1 s–10 min

.

The observation sequence included plan position indicator (PPI), range height indicator (RHI), Doppler beam swinging (DBS), and glide path (GP) scans. A full cycle took about 12 min. PPI scans described the horizontal distribution of radial velocity around the airport, while RHI scans showed the vertical structure along selected azimuths. DBS provided profiles of horizontal wind and vertical velocity, and GP scans were used to examine wind changes along the approach path. For this event, the vertical spacing was 30 m in DBS mode. The radial spacing was 100 m for PPI and RHI, and the sampling interval along the GP was also 100 m.

Quality control was applied to the raw radial velocity before plotting and wind field analysis. Previous studies have shown that signal-to-noise ratio or carrier-to-noise ratio is useful for judging the reliability of DWL radial velocity. A single threshold, however, may leave isolated noise and abnormal values, so additional statistical and neighborhood checks are often needed [22,23,24].

After tests with different thresholds and consideration of the FC-III signal characteristics, data with a signal-to-noise ratio (SNR) below 6 dB were removed. A 3 × 3 moving window was then applied to the time–range field to screen the remaining isolated noise. If the central point was valid but fewer than half of the surrounding points were valid, it was treated as unreliable and removed. Isolated abnormal values within the wind field were identified using a neighborhood consistency index C and a standardized deviation Z:

$$C={\displaystyle \frac{1}{1+\sigma }}$$

(1)

$$Z={\displaystyle \frac{\left|x_{\mathrm{c}}-\mu \right|}{\sigma }}$$

(2)

where μ and σ are the mean and standard deviation of the valid points in the moving window, excluding the central point, and x_c is the radial velocity at the center. When more than half of the points in the window were valid and both C > 0.8 and Z > 2 were met, the central point was treated as an outlier and removed. Isolated gaps after quality control were filled with the mean of neighboring valid points only when more than half of the window remained valid. Continuous unreliable areas and gaps with too few valid neighbors were left unfilled.

After quality control, velocity–azimuth processing (VAP) was used to estimate the background wind field [25], followed by a least-squares retrieval [26]. The VAP background field constrained wind direction, and the retrieved wind speed from the previous scan was used to maintain temporal continuity.

2.3. Wind Shear Indices and Stage Definition

To account for differences in both wind speed and direction, the horizontal wind vector difference between Runway 11 and 29 was used to describe the wind difference across the runway. It was calculated as follows:

$$I=\sqrt{V_1^2+V_2^2-2V_1{V}_2\cos \mathsf{\Delta }\theta }$$

(3)

where V₁ and V₂ are the horizontal wind speeds at the two runway ends, Δθ is the difference in wind direction, and I is the horizontal wind vector difference between the runway ends. Following the operational use of low-level wind shear alert systems (LLWAS), 7.7 m s⁻¹ (15 kt) was adopted as a reference threshold [27,28]. When I reached or exceeded this threshold, the difference between the two runway ends was deemed to satisfy the LLWAS reference criterion and was used as surface evidence of runway-scale LLWS.

This value describes the difference between fixed surface observation points. It is not used as the only criterion for identifying local wind shear above the airport. The vertical structure, horizontal distribution, and wind changes along the glide path observed by the DWL were also considered.

For each GP scan, the horizontal wind was projected onto the aircraft approach direction to obtain the wind component along the glide path, V_H. Positive values indicate headwind and negative values indicate tailwind. Within the 0–5 km approach path, the maximum and minimum values were extracted, and their difference, ΔV_GP, was calculated as follows:

ΔV_GP(t) = V_H,max(t) − V_H,min(t)

(4)

where t is the observation time, and V_H,max(t) and V_H,min(t) are the maximum and minimum wind components along the glide path. Based on previous lidar studies of glide path wind shear, ΔV_GP ≥ 7.7 m s⁻¹ (15 kt) was taken to indicate a wind change at the reference level [29]. The locations of the maximum and minimum values were also recorded to identify the part of the approach path where the largest change occurred.

The main periods of the two stages were defined from both DWL and AWOS observations. The start of a stage was identified when the DWL showed a spatially continuous descent of strong winds or an advancing low-level airflow, followed by corresponding changes in wind speed, wind direction, or wind vector difference at the runway ends. The end was placed when the wind vector difference decreased clearly and the two runway ends became more similar, or when a new low-level wind state had largely formed.

3. Synoptic Background and Observations

3.1. Synoptic Situation

ERA5 data were used to examine the large-scale circulation at representative times during the two stages.

At 13:00, two jet streams were present at 200 hPa (Figure 2a). The northern jet between 45° N and 60° N and the southern jet between 20° N and 30° N merged near 140° E. The jet core extended from the southeastern coast of China toward Japan, with a maximum wind speed above 80 m s⁻¹. Xining Airport lay in the area of strong winds north of the southern jet, where wind speed exceeded 50 m s⁻¹. At 500 hPa (Figure 2b), a trough extended from Gansu to Sichuan. The thermal trough lagged behind the height trough, and cold advection continued behind the trough. Xining Airport was under northwesterly flow. At 700 hPa (Figure 2c), the airport was near a weak ridge behind the trough. At the surface (Figure 2d), pressure centers above 1025 hPa were located to the northwest and southeast of the Tibetan Plateau, and pressure near the airport was about 1020 hPa. This setting was associated with subsidence and clear, dry weather, which favored the downward transfer of momentum and a rapid increase in surface wind speed.

At 21:00, the intensity and position of the 200 hPa jet had changed little (Figure 3a), while the 500 hPa trough had moved farther east (Figure 3b). The thermal trough still lagged behind the height trough, and cold advection continued behind the trough.

At 700 hPa (Figure 3c), wind direction changed across the area near the airport. The surface field (Figure 3d) showed a relatively strong pressure gradient near Xining Airport together with marked changes in wind direction. Cold advection in the middle and upper levels and the near-surface pressure gradient provided a favorable background for easterly cold air to move toward the airport.

3.2. AWOS Observations

AWOS observations were used to describe near-surface wind changes at the two runway ends. Figure 4 covers 13:25–13:45 and 20:15–20:35. In Figure 4a,c, the red and green wind barbs represent winds at Runways 29 and 11, respectively. The black line is the wind vector difference between the runway ends, and the blue dashed line marks the 7.7 m s⁻¹ reference threshold. Figure 4b,d show temperature, pressure, and relative humidity at Runway 29.

During the first stage, the strong winds linked to downward momentum transport reached Runway 11 at about 13:29 and then affected Runway 29. The wind vector difference between the runway ends increased three times, near 13:31, 13:36, and 13:40 (Figure 4a). The values at 13:31 and 13:40 exceeded 7.7 m s⁻¹, while the value near 13:36 was close to the threshold.

At 13:31, northwesterly winds above 6 m s⁻¹ were already present at Runway 11, whereas winds at Runway 29 remained below 1 m s⁻¹. The difference between the two runway ends was therefore clear. Wind speed at Runway 29 then increased, but the response lagged Runway 11 by about 1–2 min and developed more slowly, producing another large difference near 13:36. Around 13:40, wind speed at Runway 29 rose from below 1 m s⁻¹ to above 8 m s⁻¹ within about 5 min, while Runway 11 was near 6 m s⁻¹. The wind vector difference reached its maximum of 9.42 m s⁻¹. After 13:42, wind speed and direction at the two runway ends became more similar, and the difference decreased.

Between 13:30 and 13:41, temperature at Runway 29 rose by about 1.4 °C, while relative humidity fell by about 2% and pressure decreased by about 0.5 hPa (Figure 4b). These changes accompanied the adjustment of the near-surface wind field during the afternoon momentum transport stage.

The second stage occurred from 20:15 to 20:35. As downward momentum transport weakened, the earlier northwesterly flow gradually faded and easterly flow began to affect the airport low levels. Before 20:27, both runway ends were under easterly flow. Wind speed at Runway 11 remained near 2–3 m s⁻¹, while Runway 29 increased from about 4 m s⁻¹ to above 8 m s⁻¹. The wind vector difference reached 6.44 m s⁻¹ but remained below the 7.7 m s⁻¹ reference value. After 20:27, the easterly flow moved farther west, and wind speed at Runway 11 increased to about 6 m s⁻¹ within roughly 2 min. The difference between the runway ends then decreased quickly and became more stable. At Runway 29, temperature rose by about 0.5 °C from 20:18 to 20:27 and then fell by about 0.4 °C by 20:35 (Figure 4d). Overall, the surface thermodynamic changes during this period were relatively small.

3.3. Sounding Analysis

Sounding data were used to examine the vertical atmospheric structure related to the wind changes. Figure 5 shows temperature–log-pressure diagrams from Xining station on 8 April 2022. The red, green, and black lines represent environmental temperature, dew point, and parcel ascent, respectively. Wind barbs on the right show wind direction and speed at different pressure levels.

At 08:00 (Figure 5a), the dew point depression was small from the surface to 718 hPa, indicating relatively moist low-level air. Above 718 hPa it increased, giving a moist lower layer and a drier layer aloft. The low-level lapse rate was smaller than the dry adiabatic lapse rate, indicating stable stratification. At 700 hPa, a southeasterly wind of about 4 m s⁻¹ veered to a westerly wind of about 14 m s⁻¹ at 500 hPa, showing clear directional shear with height. Winds at 400 hPa were southwesterly at about 18 m s⁻¹, so relatively strong winds were already present in the middle and upper levels. Convective available potential energy (CAPE) was 0 J kg⁻¹, suggesting little potential for surface-based parcel ascent during the morning.

Using the 08:00 sounding and adjusting the surface parcel temperature to the daily maximum, the surface-based CAPE at 13:00 was estimated as 423.8 J kg⁻¹. This suggests that instability increased during the afternoon. Surface heating also favored boundary layer growth and turbulent mixing, which provided a more favorable thermal setting for momentum aloft to reach the lower levels. By 20:00, the atmospheric structure had evolved substantially (Figure 5b). The low-level dew point depression was larger, the lapse rate was close to the dry adiabatic rate, and the wind at 400 hPa had decreased to about 10 m s⁻¹. Winds in the middle and upper levels were therefore weaker than in the morning.

4. Development and Evolution of the Low-Level Wind Field

4.1. Time–Height Evolution of Wind Profiles

DBS observations were used to examine the time–height evolution of horizontal wind and vertical velocity during the day (Figure 6). Relatively continuous observations were obtained from 0 to 3000 m AGL, with usable wind information also available above 2 km. Because the main shear structures in both stages and their influence on the airport wind field were concentrated below 2 km, the following discussion focuses mainly on this layer. Before the main afternoon event, two periods of vertical wind shear appeared between 01:00 and 13:00. The first was mainly below 500 m. From 01:00, a weak west-northwesterly flow of about 2 m s⁻¹ developed near the surface beneath easterly winds above 8 m s⁻¹. Between 01:00 and 04:00, this lower layer deepened to about 500 m and strengthened to 6–8 m s⁻¹. The shear layer rose to around 500 m and remained there until about 10:00. From 10:00 to 12:00, the flow below 500 m changed to easterly, and the shear weakened. A second period developed from 10:00 to 13:00 as stronger west-northwesterly winds aloft extended downward over the easterly flow below. The shear layer descended from about 2 km to below 500 m. At 13:00, the airport was already under post-trough northwesterly flow, but the stronger momentum aloft had not fully reached the surface, where wind speed remained near 2 m s⁻¹. From 13:00 to 13:40, the strong wind layer continued to descend, and wind speed near the surface rose rapidly to about 16 m s⁻¹. Strong west-northwesterly flow then persisted until 18:00, with speeds of 10–20 m s⁻¹ between 13:40 and 16:00. Vertical motion was active during this period and reached about ±3 m s⁻¹. Short-lived vertical motions also appeared below 500 m from 12:00 to 13:00. These motions were probably related to afternoon surface heating. The resulting reduction in low-level stability may have strengthened turbulent mixing under the post-trough subsiding flow and favored momentum exchange between the upper and lower levels. In the evening, surface heating and turbulent exchange weakened. From 18:00 to 20:00, winds over the airport decreased quickly, the low-level strong wind region weakened and shifted upward, and vertical motion also became weaker. Around 20:30, DBS showed easterly winds of about 14 m s⁻¹ over the airport. These winds formed clear vertical shear with weaker flow above, and the shear layer was located at about 1.5–2.0 km.

4.2. Horizontal Evolution of the Wind Field

Figure 7 shows PPI radial velocity and retrieved horizontal wind fields at selected times during the two stages. Shading represents radial velocity, black vectors show the retrieved horizontal wind, and the black boxes mark the glide path areas near the two runway ends. Each box has a side length of 1 nautical mile (approximately 1.852 km). The airflow direction and horizontal wind speeds discussed below are taken mainly from the retrieved wind field; the radial velocity shading is used to help show the spatial structure.

During the first stage, the wind near the airport was generally weak at 13:15. The retrieved vectors showed northwesterly flow of about 4 m s⁻¹. Only scattered stronger radial velocity areas appeared beyond 2 km to the northwest and southeast of the lidar, suggesting that the strong winds linked to downward momentum transport had not yet clearly reached the runway area (Figure 7a). By 13:31, west-northwesterly flow dominated near the airport and the stronger wind area on the western side had expanded eastward into the runway area. Wind speed near Runway 11 exceeded 10 m s⁻¹, while Runway 29 remained near 2 m s⁻¹. Runway 11 was therefore affected earlier, and the wind difference across the runway was clear (Figure 7b). At 13:39, the radial velocity pattern on the two sides of the lidar became more symmetric, and the retrieved wind field showed west-northwesterly winds above 10 m s⁻¹ over most of the runway area. The difference between the runway ends had decreased (Figure 7c).

During the second stage, weak northeasterly flow of 2–4 m s⁻¹ was present near the airport at 20:13, while stronger easterly flow appeared east of the lidar. The radial velocity pattern and retrieved vectors both showed that this flow was moving westward and that a clear transition in wind speed and direction had formed against the earlier weak flow (Figure 7d). At 20:21, the easterly flow expanded into the runway area, and winds were stronger on the eastern side of the airport. East-southeasterly winds near Runway 29 reached about 12 m s⁻¹, whereas winds near Runway 11 remained near 6 m s⁻¹. This shows that Runway 29 was affected earlier and that local wind speed and direction differed along the runway (Figure 7e). By 20:29, the stronger easterly flow had moved farther west and largely covered the runway area. Winds above 10 m s⁻¹ were present near both runway ends, and the difference across the runway had decreased (Figure 7f).

In summary, stronger west-northwesterly flow crossed the airport from west to east during the first stage, and Runway 11 responded before Runway 29. In the second stage, easterly flow moved from east to west and Runway 29 responded first. The retrieved vectors show that the order of response at the runway ends agreed with the direction of airflow in each stage.

4.3. Evolution of the Vertical Wind Field

RHI radial velocity and retrieved wind fields along the runway direction were used to examine the vertical structure during the two stages. Figure 8a–c show the first stage and Figure 8d–f the second. Shading represents radial velocity, and black vectors show the retrieved wind field. Valid radial velocity data were also obtained in parts of the 2–4 km layer. Because the shear structures most relevant to the runway and approach area were mainly located below 2 km in both stages, the following analysis focuses on the lower-level wind field.

During the first stage, weak westerly winds of 2–4 m s⁻¹ were present below 1 km at 13:10, while wind speed above 1 km exceeded 12 m s⁻¹. A strong vertical gradient was therefore present near 1 km (Figure 8a). At 13:28, the stronger winds in the middle and upper levels extended downward. Wind speed on the western side of the lidar increased first to above 8 m s⁻¹, while weak winds of 2–4 m s⁻¹ remained below 1 km on the eastern side. The interface between the strong and weak wind regions rose eastward from near the surface on the western side to about 1 km. This pattern shows that the descending strong winds affected the western side first and then spread downward and eastward (Figure 8b). By 13:36, wind speed over most of the eastern side had also increased to above 8 m s⁻¹, and the stronger flow had reached more of the runway area. Local winds on the western side were weaker than at 13:28, showing some short-term variation within the low-level wind field (Figure 8c).

During the second stage, wind direction changed near 1.5 km at 20:09. The retrieved vectors showed westerly flow above and easterly flow below, both generally at 2–4 m s⁻¹. Easterly winds above 6 m s⁻¹ were already present at the far eastern end of the scan, showing that the cold front related flow was approaching the airport (Figure 8d). At 20:18, stronger easterly flow advanced westward through the low levels east of the lidar. Wind speed below 1 km increased to above 8 m s⁻¹, and the flow formed a clear wedge. Its upper boundary rose to about 1.3 km. Near this boundary, the retrieved vectors turned upward and a marked wind speed gradient and local shear were present (Figure 8e). By 20:27, the easterly flow had expanded farther west, with local speeds above 12 m s⁻¹. The interface between the stronger low-level flow and the weaker flow above extended from about 0.5 km to nearly 2 km, showing a clear adjustment of the low-level wind field during the cold front intrusion (Figure 8f).

The RHI observations show a clear contrast between the two stages. In the first, strong winds from the middle and upper levels extended downward and the shear interface descended toward the surface. In the second, easterly flow advanced westward through the low levels, forming an inclined shear interface between the wedge and the weaker flow above and ahead. The direction of strong wind entry and the change in the shear interface were therefore different.

4.4. Wind Characteristics Along the Glide Path

The wind component along the glide path retrieved from the DWL GP scans shows how wind changed along the aircraft approach path. Figure 9 presents the Runway 11 glide path during the two stages. Positive values indicate headwind and negative values indicate tailwind. The difference between the maximum and minimum wind components within 0–5 km was calculated and compared with the 7.7 m s⁻¹ reference value.

During the first stage, tailwind dominated the glide path. At 13:12, the tailwind component was generally 1–5 m s⁻¹ and changed little along the path. The maximum difference was 3.41 m s⁻¹, showing that the strong winds linked to downward momentum transport had not yet clearly entered the glide path region. At 13:20, the tailwind strengthened. The more noticeable changes occurred about 1500–2000 m and 3300–3700 m from the runway end, and the maximum difference rose to 4.58 m s⁻¹ but remained below the reference value. By 13:28, tailwind had increased further as downward momentum transport strengthened. Beyond about 2000 m from the runway end, the tailwind component was generally above 8 m s⁻¹ and extended toward the runway. The maximum difference reached 9.02 m s⁻¹, indicating a marked wind change along the glide path.

During the second stage, the glide path changed from tailwind to headwind as downward momentum transport weakened and easterly flow reached the airport. At 20:18, headwind was about 2–4 m s⁻¹ and the maximum difference was 3.04 m s⁻¹. At 20:26, headwind strengthened to 5–6 m s⁻¹ within about 2 km of the runway end and then decreased with distance. The maximum difference rose to 4.01 m s⁻¹. At 20:34, headwind over most of the glide path reached 6–8 m s⁻¹, weakened locally near 3300–3700 m, and then increased again. The maximum difference was 3.44 m s⁻¹. None of the three scans reached the 7.7 m s⁻¹ reference value. Although the easterly flow increased the overall headwind, the change along the approach path remained relatively small.

Based on the combined AWOS, DBS, PPI, RHI, and GP observations, 13:25–13:45 and 20:15–20:35 were selected as the main analysis periods. Before the first stage, the 13:15 PPI and 13:20 GP scans showed that the descending strong winds had not yet clearly affected the runway or glide path. Around 13:28, RHI and GP observations showed stronger flow extending into the low levels and approach path, followed by an increase in wind speed at Runway 11 around 13:29. Therefore, 13:25 was selected as the start of the first main analysis period to include the transition as the strong winds entered the runway area. After 13:39, strong winds had largely covered the runway, and after 13:42, wind speed and direction at the two runway ends became more similar and the vector difference decreased. Accordingly, 13:45 was selected as the end of the first main analysis period. Strong west-northwesterly winds continued after this time, so 13:45 marks the weakening of the difference across the runway rather than the end of downward momentum transport.

Before the second stage, the 20:09 RHI and 20:13 PPI scans already showed easterly flow moving westward from the eastern side of the airport. Around 20:18, RHI observations showed cold air entering the airport low levels as a wedge, and corresponding changes also appeared in the AWOS and GP observations. Therefore, 20:15 was selected as the start of the second main analysis period. After 20:27, the easterly flow reached Runway 11, and the difference between the runway ends decreased quickly. By 20:29–20:34, the easterly flow had largely covered the runway and glide path, and a new low-level wind state had formed. Accordingly, 20:35 was selected as the end of the second main analysis period. These intervals represent the times when the two stages most clearly affected the runway and approach path, not the full duration of the weather processes.

5. Discussion

5.1. Comparison with Previous Studies of Hazardous Airport Wind Fields

Huang et al. (2024) [11] examined LLWS at Xining Airport under several weather backgrounds and showed that thunderstorm outflows, strong winds, cold air, and complex terrain can all produce rapid wind changes near the runway. The increase in near-surface wind speed, changes in wind direction, and local wind shear observed in the present study are broadly consistent with earlier findings. However, unlike previous cases, downward momentum transport and cold-front intrusion affected the airport sequentially on the same day, resulting in two distinct stages with different wind structures and intensities. Thus, the present event shares several near-surface features with previously reported hazardous wind events at Xining Airport, while the sequential occurrence of the two processes on the same day has been less commonly documented.

Earlier case studies at complex terrain airports in northwestern China have mainly focused on convective outflows. Gu et al. (2025) [19] examined LLWS at Xining Airport caused by downdrafts and a gust front. Han et al. (2020) [30] described the fine structure of a gust front at Lanzhou Airport, and Feng et al. (2023) [31] observed near-surface divergent outflow from a dry microburst at Yinchuan Airport. The first stage in the present study differed from these convective events. No typical downdraft core or near-surface divergent outflow was observed, and the main feature was the descent of strong winds from the middle and upper levels. The second stage involved rapid wind changes as easterly cold air moved into the airport. Different weather processes can therefore produce similar changes near the surface even when their sources and internal structures are different. Multiple DWL scanning modes help distinguish these wind field structures.

In the first stage, both the wind vector difference between the runway ends and the wind component difference along the glide path reached the 15 kt reference value. Neither measure reached this value in the second stage, although the DWL still showed clear local shear. The fixed-point difference at the runway ends, the wind change along the glide path, and local shear above the airport describe different spatial scales and do not have to occur at the same time. Because this study considers only one event, the climatological frequency and wider applicability of this two-stage pattern need to be examined using more cases.

To assess the sensitivity of the results to small changes in the reference thresholds, we conducted sensitivity analyses for both the AWOS-based wind vector difference between the two runway ends and the GP-based wind speed component difference along the glide path. The 7.7 m s⁻¹ reference value was varied by −20%, −10%, +10%, and +20%. The corresponding test values were 6.16, 6.93, 8.47, and 9.24 m s⁻¹. For the AWOS observations, the maximum value during the first stage was 9.42 m s⁻¹ and exceeded all tested reference values. The maximum value during the second stage was 6.44 m s⁻¹ and exceeded only the reference value reduced by 20%. For the GP observations, the maximum value at 13:28 during the first stage was 9.02 m s⁻¹. It exceeded all tested reference values except the value increased by 20% (Figure S1). The values at 13:12 and 13:20, together with all three values during the second stage, remained below the minimum test value of 6.16 m s⁻¹. These results show that the overall difference between the two stages remained clear. However, the classification of observations close to the reference level changed when the threshold was adjusted. Therefore, the reference thresholds should be considered together with the observed wind structure and should not be used as the only basis for event identification.

5.2. Terrain Effects and Data Limitations

Xining Airport lies in the Huangshui River valley, and the surrounding terrain may affect the movement of airflow near the surface. During the first stage, strong west-northwesterly flow crossed the airport from west to east. Runway 11 responded first, followed by Runway 29 about 1–2 min later. During the second stage, easterly flow moved from east to west, and Runway 29 responded before Runway 11. These directions suggest that the valley may guide the low-level airflow and contribute to the difference in response time. The present observations, however, cannot separate the influence of the weather system from valley channeling, constriction effects, and local mountain–valley winds. Their individual contributions cannot yet be quantified.

DWL measurements depend on aerosol backscatter. When aerosol concentrations are low, the return signal becomes weak and the effective detection range decreases. Precipitation and poor visibility can attenuate the laser signal and increase the occurrence of low-SNR data and data gaps, especially at longer ranges and higher altitudes. Under these conditions, some parts of the wind field may not be fully captured. Scanning geometry may also limit the spatial coverage of the observations. SNR filtering, neighborhood checks, and outlier removal were used to reduce the influence of unreliable data. However, these quality control procedures cannot recover observations lost because of weak or attenuated signals. Therefore, only valid areas after quality control were used in this study. The DWL represents winds within range gates and scan volumes, whereas the AWOS measures near-surface winds at fixed runway end locations. Their observation heights and spatial scales are therefore different. In this study, DWL was used to examine wind structures above the airport and along the glide path, while AWOS described the surface response at the runway ends.

ERA5 was used mainly to describe the synoptic circulation. Its spatial resolution is too coarse to fully represent the mountains and valleys around the airport or small-scale processes such as valley channeling, constriction effects, and the advance of cold air. ERA5 was therefore used to identify the upper-level jet, trough, cold advection, and surface pressure pattern.

5.3. Potential Operational Application

Following the approach illustrated for the present test case, multiple DWL scanning modes, AWOS measurements at the runway ends, ERA5, and sounding data can be combined in a simple framework from observation to feature identification and potential operational support (Figure 10). DBS, PPI, RHI, and GP scans reveal the descent of strong wind layers. These methods also identify the advance of low-level airflow, local shear, and wind changes along the approach path. AWOS provides the near-surface response at the two runway ends. ERA5 and the soundings provide the synoptic background and the vertical thermodynamic and dynamic conditions. Used together, these datasets provide a fuller description of the airport low-level wind field.

In an operational setting, the DWL can identify a descending strong wind layer or airflow moving toward the runway. Subsequently, AWOS can be used to compare responses at both runway ends, while GP scans show wind changes along the approach path. ERA5 and sounding data help place the event in its weather background. If the wind vector difference between the runway ends or the GP wind component difference approaches or reaches its respective reference value, this information may help forecasters, air traffic controllers, and airport operations staff determine when closer monitoring is needed.

The proposed framework is intended as a decision-support guide rather than an automatic warning system. It is worth noting that Figure 10 shows only a possible application framework based on this single case, but it is not a validated warning algorithm. In fact, pilot reports, flight data, and go-around records were not available for this event. Therefore, the reference thresholds, lead times, and operating rules require further testing with more events and local flight data before the framework can be used in operational applications.

6. Conclusions

This study combined DWL, AWOS, sounding, and ERA5 data to examine the two-stage evolution of the low-level wind field at Xining Airport on 8 April 2022. The main conclusions are as follows.

(1) The first stage was mainly caused by downward transport of strong post-trough momentum. DBS showed the strong wind layer descending from the middle and upper levels. PPI and RHI showed that the stronger flow entered the airport low levels from the west and moved eastward across the runway. Runway 11 responded first, followed by Runway 29 about 1–2 min later. The maximum wind vector difference between the runway ends was 9.42 m s⁻¹ and exceeded 7.7 m s⁻¹ at 13:31 and 13:40. The maximum wind component difference along the glide path was 9.02 m s⁻¹, showing clear LLWS in both the airport low levels and the approach path.

(2) The second stage was associated mainly with cold front intrusion. PPI and RHI showed easterly cold air moving westward into the airport as a low-level wedge. This caused rapid changes in low-level wind speed and direction and produced local shear. Runway 29 responded first, followed by Runway 11. The maximum wind vector difference between the runway ends was 6.44 m s⁻¹, and the maximum difference along the glide path was 4.01 m s⁻¹. Both remained below 7.7 m s⁻¹. This stage was therefore mainly a low-level wind adjustment with local shear rather than runway-scale LLWS at the reference level.

(3) The combined DBS, PPI, RHI, GP, and AWOS observations showed the descent of the strong wind layer, horizontal airflow movement, local shear, wind changes along the glide path, and surface responses at the runway ends. Together, they provide a more comprehensive view of the low-level wind structure and its evolution, thereby supporting wind monitoring.