NUMERICAL SIMULATION ON THE WIND FIELD STRUCTURE OF A MOUNTAINOUS AREA BESIDE SOUTH CHINA SEA DURING THE LANDFALL OF TYPHOON MOLAVE

The coastal region in southern China contains abundant wind energy potential^[1], however, frequented typhoon visits pose a serious threat to the development and utilization of wind energy in these areas^[2-4]. Since the terrain along the south China coast is relatively complex with non-uniform wind speed distribution, the analysis of three-dimensional wind structure over complex coastal terrain during a typhoon landfall is crucial in maximizing wind power generation efficiency and taking reasonable measures against the threat of strong winds.

Over complex terrain, wind fields are distributed with a high degree of non-uniformity within the surface layer, therefore, observational data at a single point is of very limited applicability in representing the exact conditions for a large area. Consequently, the application of numerical simulation is valuable in obtaining high-resolution wind field data over complex terrain. Traditionally, wind field numerical simulation research has relied on mesoscale simulation^[5-7], however, the finest spatial resolution of a mesoscale model is roughly on the order of 100 m due to the limitation of its numerical algorithm. In addition, mesoscale model simulations require a pre-processing stage to smoothen terrain in order to achieve calculation stability^{[8, 9]}. In cases with very complex and steep terrain, mesoscale models often fail to obtain a convergent numerical solution.

As an alternative choice for wind field simulation the Computational Fluid Dynamics (CFD) approach has attracted more and more attention in recent years Relative to mesoscale modeling, CFD modeling has a higher spatial resolution (minimum horizontal grid spacing can be reduced to 10 m), so the model can describe more realistic terrain in a finer scale. Some experimental studies have demonstrated that CFD is capable to construct realistic fine-scale surface layer wind structure over extremely complex terrain^[10-12].

This study is based on the basis of previous studies and uses CFD techniques to analyze the 3D wind structure over Paiya Mountains and its surroundings in Dapeng Peninsula near Shenzhen, a coastal city beside the South China Sea, during the landfall of Typhoon Molave, in order to enhance the understanding of 3D fine-scale wind structure over complex coastal terrain.

Typhoon Molave is the strongest typhoon ever to directly hit Shenzhen in the most recent decade. It tracked from southeast and made landfall at around00:50 LST (same below) on July 19, 2009 on the Dapeng Peninsula near Shenzhen. Even though the eye of Molave was not clearly visible from satellite images, the strongest winds over Dapeng Peninsula still reached Beaufort force 12 during landfall. With central pressure of only 970 h Pa, Molave brought strong winds and heavy rains to Shenzhen. Based on observed data, the timing of the simulation was set to00:00 and 02:00 of July 19, the two on-the-hour markers before and after Molave's landfall with substantial changes in wind direction. Performing simulations at these two moments can provide an understanding of the 3D wind structure over complex terrain under strong wind conditions.

The simulation domain includes the Paiya Mountains and its surrounding areas on Dapeng Peninsula near Shenzhen. The geographic scope and topography are as shown in Fig. 1a. The area is 8 km long on the east-west direction and 7 km wide on the south-north direction, and contains two automatic weather stations (AWSs), labeled as BG and DZ, and one wind measurement tower (location marked by asterisk and labeled as"Tower"in Fig. 1a). The observed data from the BG/DZ AWSs and the wind tower are used to verify the simulation results.

Figure 1.  Simulation domain, topography and mesh.

This study used the commercial CFD software FLUENTas simulationtool. Theauthors demonstrated in a series of previous studies that FLUENT is capable of describing fine-scale wind structure over complex terrain^[10-12]. For details on equations and physical models setting used in FLUENT simulations, refer to Li et al.^[11].

A computer-aided design (CAD) model based on the terrain elevation data was built by using GAMBIT, the preprocessor for FLUENT. The model was then discretized with high quality structured grids, which will help to ensure stable numerical calculation. The structured grids within the simulation domain are shown in Fig. 1b.

The FLUENT solution domain is 3000 m high There are 80 girds along the east-west (x) direction and 70 along the south-north (y) direction, with a constant grid spacing of 100 m. There are 70 grids in the vertical direction with an increasing grid spacing ratio of 1:1.08, which allows near-surface layer grids to have a vertical spacing in the order of magnitude of1 m.

For boundary conditions, this study used wind data from a radar wind profiler (RWP) around 60 km west to Dapeng Peninsula. Wind profiles in the vertical direction can be reconstructed using wind speed measured by the RWP coupled with wind speed and direction observed at the measurement tower in Paiya Mountains. This wind profile was then input into FLUENT through the User-Defined Function (UDF) module to initiate FLUENT calculations Although this method does not guarantee the full accuracy on boundary conditions, it is more accurate than just using a simple exponential wind profile fitting, since it at least correctly describes the vertical wind speed distribution of a typhoon. Wind velocity profiles served as boundary conditions in this study are shown in Fig. 2.

Figure 2.  Boundary conditions for inflow velocity based on wind profiler measured data.

Table 1 shows the comparisons at both BG and DZ AWSs for both 00:00 and 02:00 on July 19, in which the measured data at AWSs were 2-minute average data.Table 1 shows that simulated wind speeds are stronger than observed results, but simulated wind directions are generally consistent with observed ones. Wind speed simulations for the BG station are somewhat less accurate, with a relative error of 29.8%and 42.9%for the 00:00 and 02:00results, respectively. Results for the DZ station are better, with a relative error of 15.9%and 27.6%, respectively.

Figure 3 shows the comparison between simulated and measured wind speeds at the wind measurement tower located in Paiya Mountains. The simulated wind profile displayed logarithmic characteristics at the surface layer, and provided wind speeds generally similar to measured data at different elevations for both simulation times.

Figure 3.  Observed and simulated wind speeds (solid line:simulated; box:observed value).

There are three possible reasons leading to the differences between simulated results and measured data: (1) due to limitations with the observational environment, both BG and DZ AWSs are installed at the top of buildings and the building itself might affect the observed wind speed data, wherein this effect is beyond the capability of this simulation; (2) boundary conditions are computed from RWP measured data, but uncertainties arise due to the distance between the RWP instrument and the simulation region; and (3) uncertainties are also introduced by FLUENT physical model settings, such as selection of turbulence models, as different settings might produce different results from measured data. Nevertheless, comparative analysis at the AWSs and the wind measurement tower showed that simulated results were generally consistent with measured data before and after the change in wind direction during a typhoon landfall. Hence, simulated results are sufficient to provide basic and realistic three-dimensional data for performing a wind structure analysis over mountainous area under strong wind conditions.

The wind field around the Paiya Mountains at different elevations from the ground at 00:00 on July19 is shown in Fig. 4. Paiya Mountains and their surroundings were entirely dominated by northerly winds, such that wind structures were generally uniform across different elevations under 150 m relative to the ground. In addition, Fig. 4 shows that northerly winds shifted significantly to become westerly winds after passing through the Paiya Mountains, especially in the southeastern part of the simulation region. This shows that wind fields within the surface layer under 150 m from the ground were all considerably affected by the terrain.

Figure 4.  Simulation results for wind field at various elevations. Gradient:terrain elevation, unit:m.

Comparing Fig. 4a to Fig. 4d, there are still significant differences in the wind field at different elevations:at 10 m above the ground, despite the overall strong wind, the wind field seems to be rather turbulent. The impact of the terrain is very prominent and the changes in air flow direction were more dramatic, especially around the peak of Paiya Mountains. However, the variation of the wind field becomes smoother and gentler in the higher elevations above the ground.

Wind speeds at 10 m, 50 m, 100 m and 150 m above the ground are shown in Fig. 5, in which all the drawings were superimposed with terrain data.Fig. 5shows that maximum wind speed all reaches about 38m/s across different elevations within the surface layer and areas with high wind speed values are concentrated along the edge of the east-west ridge of the Paiya Mountains. Wind speed distribution have a high degree of agreement with the topography, and this shows that terrain is a dominant factor determining surface layer wind speed distribution under neutral high wind speed conditions.

Figure 5.  Simulation results for wind structure at various elevations. Gradient:wind speed, unit:m/s; contour lines:elevation.

Analysis on the wind fields at different elevations also suggests a number of differences in detailed features: (1) although the overall characteristics for wind fields at different elevations are relatively similar, average wind speed is lower at the near-surface layer than at higher altitude, reflected by the darker color regions in Fig. 5, which gets progressively larger with higher altitudes above the ground; (2) at 10 m above the ground, high wind speed values are concentrated in a relatively narrow vicinity of the windward side of the ridge, but high-value areas cover the entire ridge at higher altitudes above the ground; (3) wind speeds are distributed more uniformly at higher altitudes, reflected by the decreasing difference between maximum and minimum wind speed values against increasing elevation; and (4) complex terrain can intensify maximum wind speed as in Fig. 2a, wind profile data indicate maximum inflow wind speed was30 m/s under the altitude of 1000 m, but maximum near-surface wind speed at the ridge in Paiya Mountains exceeded 36 m/s. Analysis on the simulation results for 02:00 on July 19 indicate similar characteristics and confirmed the results from the 00:00 simulation regarding the changes in wind structure and wind speed at various elevations above the ground (not shown).

To analyze the wind structure in the vertical direction, especially for wind field blowing through a ridge, Fig. 6 shows the wind structure in horizontal cross-sections within the solution domain for both00:00 and 02:00 July 19 simulation times. Fig. 6a shows the simulated wind speed at 00:00 on July 19in which the horizontal cross-section is generally accordant to the dominant wind direction over the solution domain. The cross-section passed through three control points, namely A (x=5000, y=7000, z=0), B (x=7000, y=0, z=0), C (x=7000, y=0, z=3000). Fig. 6c shows the simulated result at 02:00 on July 19, and since a southeastern wind is the prevailing wind at this time, the cross-section passed through a different set of control points, namely A (x=5860, y=3810, x=0), B (x=5860, y=3810, z=3000), C (x=3000, y=7000, z=0).

Figure 6.  Wind speed distribution and wind structure at cross-sections.

Figure 6a and 6c show the detailed wind speed distributions around complex mountainous surface under strong wind conditions. The figures clearly show that there was a region with significantly increased wind speed at the windward side and the area close to the peak of the mountain. This effect was particularly evident within the surface layer. At both00:00 and 02:00, simulated results demonstrate local wind speed intensification in the surface layer, and this once again confirms that protruding mountainous bodies do increase local surface layer wind speed under strong wind conditions.

Figure 6b and 6d show the wind structure around the mountain peak. The figures show that when surface-layer air flows across the ridge, its speed reaches maximum value near the peak, and then its energy dissipates rapidly, which results in a low wind speed area behind the ridge. In terms of flow field structure, a less apparent vortex forms behind the ridge. In the area behind the ridge, vertical wind shear is very strong, wind speed at the surface layer is relatively low, and wind direction is opposite to that at higher altitudes due to the effects of the vortex. Since the variation of the wind field is quite drastic in the leeward side of the ridge, any artificial structures located in this area will suffer from strong wind shear.

Three locations surrounding the peak of Paiya Mountains are selected to analyze the relative size of turbulence, and their positions are depicted by black lines and shown in Fig. 7. The maximum values of turbulence intensity at each respective position are denoted as I_a, I_b and I_c. For the simulation time at00:00 on July 19, I_a, I_b and I_c values were 145%, 257%and 246%, respectively, and maximum turbulence all occurred in the surface layer. Turbulence with intensity higher than 100%is very strong and uncommon in field observations, but it is not unusual for that to be observed in some wind tunnel experiments^[13].

Figure 7.  The three locations analyzed for turbulence intensity. (a) windward side; (b) mountain top; (c) leeward side.

Despite the lack of observed turbulence intensity data to quantitatively verify simulated results, a comparison between the relative values at these three locations showed the impact of a mountain on the spatial distribution of turbulence intensity. Turbulence intensity appears quite strong within the surface layer around the mountain ridge, but turbulence is even stronger at the peak and the leeward side than at the windward side of the mountain. Analyzed in conjunction with Fig. 6, stronger turbulence intensity at the peak is due to airflow acceleration when climbing over the ridge, resulting in stronger vertical wind shear and hence forming stronger mechanical turbulence. On the other hand, turbulence is stronger behind the ridge due to the existence of vortices. Vortices create strong vertical wind shear in the region with lower average wind speed, resulting in higher turbulence intensity at the leeward side of the mountain than at the windward side, in which the vertical wind shear is weaker. In the analysis of the simulation result for 02:00 on July 19, since southerly winds had become dominant, position c became the windward side of the mountain. I_c was the smallest among the three locations in this case, and I_b and I_a were significantly higher than I_c. The maximum values of turbulence intensity for all three locations exceeded 100%, which confirmed the analysis of the00:00 simulation results.

Based on the analysis of this study, a simple conceptual model of the wind structure over complex mountainous topography under strong wind condition is shown in Fig. 8.

Figure 8.  Conceptual model for wind structure over mountainous area under strong wind conditions.

Figure 8 shows that under strong wind conditions there is a special strong wind zone in the windward side and over the top of a mountain, wherein wind speed may even exceed wind speeds at the higher altitudes. Above the mountain peak, there is a crown-shaped region wherein wind speed is significantly intensified due to the terrain and is higher than that away from the mountain region at the same elevation. Vortices are easily formed at the leeward side of a mountain, and since average wind speed is lower in this region, vertical wind shear and turbulent intensity become stronger.

Based on CFD technique, the fine-scale three-dimensional wind structure over the Paiya Mountains on the Dapeng Peninsula near Shenzhen, a city beside the South China Sea, during the landfall of Typhoon Molave is simulated and analyzed. The analysis supports the following conclusions:

(1) FLUENT can reasonably describe the three-dimensional wind structure over complex terrain under strong wind conditions. It can also describe the effect of terrain on wind field and turbulence field.

(2) Wind speed distributions, under strong wind conditions, exhibit generally similar characteristics at different elevations above a mountain within the surface layer. However, non-uniformity for wind speed becomes more prominent in the area with relatively low altitudes.

(3) The kinetic effect induced by the presence of a tall mountain can intensify local wind speed and turbulence intensity at the windward side of the mountain and areas over the mountain top.

(4) There is a special zone at the leeward side of the mountain, wherein wind speed is relatively low with relatively strong wind shear and turbulence.