High-Resolution Flow Simulations Around a Steep Mountainous Island in Korea Using a CFD Model with One-way Nested Grid System

Abstract

High-resolution flows around a steep mountainous island (Ulleungdo) in Korea were simulated by a computational fluid dynamics (CFD) model. To cover entire Ulleungdo and to resolve the topography around the Ulleungdo automatic synoptic observing system (ASOS) with high resolution, one-way nested grid system with large (60 m), and small (20 m) grid sizes was applied in the CFD model simulations. We conducted the numerical simulations for 16 inflow directions, and, for each inflow direction, we considered six different wind velocities(5, 10, 15, 20, 25, and 30 m s^-1) at the reference height (1,000 m). The effects of topography on surface wind observations were well reflected in the observed wind roses for the period of January 01, 2012 ~ December 31, 2016 at the Ulleungdo ASOS and marine buoy. Wind roses at the Ulleungdo ASOS was reproduced based on the CFD simulations. The changes in surface winds at the Ulleungdo ASOS caused by surrounding topography were relatively well simulated by the CFD model. The simulated wind-rose indicated that south-southwesterly and northeasterly were the dominant wind directions, which were also observed at the Ulleungdo ASOS. We investigated the flow characteristics around the Ulleungdo ASOS for northwesterly, south-southwesterly, and northeasterly winds in detail.

1. Introduction

People living in islands are sensitive to the weather due to geographical particularity (Gong et al., 2013; Seetanah and Fauzel, 2019). Detailed and accurate weather information may help prevent the islanders’ breakaway in the long term, by relieving their sense of alienation and improving their quality of life. Practically, reliable weather information for islands helps visitors arriving through vessels to establish travel itineraries to the islands, thereby invigorating travel to the islands and, subsequently, their economy (Scott and Lemieux, 2010).

Ulleungdo, located in the East Sea, is the ninth largest island in Korea. More than 300,000 people visit Ulleungdo every year by a vessel. However, visitors and residents experience inconvenience due to approximately 20% cancelation of vessels caused by severe weather and a high-sea watch. Recently, the Korean government decided to construct an airport in Ulleungdo to resolve the inconvenience and to enhance accessibility. Because Ullengdo is steep mountainous terrain and strong winds are frequent around Ullengdo, we need to investigate the characteristics of the wind environment around Ullengdo spatially in detail for the safe of airplane take-off and landing. The Korea Meteorological Administration (KMA) currently operates the local data assimilation and prediction system (LDAPS) based on the unified model (UM) of the United Kingdom’s national weather service (Met office). The LDAPS coverage includes Ullengdo, but it has relatively coarse a horizontal resolution (1.5 km) for reflecting the effects of Ulleungdo’s complicated terrain on the detailed wind environment around Ullengdo.Computational fluid dynamics(CFD)models would be one of the appropriate tools for simulating airflows over high and narrow topography such as Ulleungdo, because it is able to employ very high spatial resolutions within the order of 1 m (Iisuka and Kondo, 2004; Parker and Kinnersley, 2004; Balogh et al., 2012; Casterllani et al., 2015; Nedjari et al., 2017; Uchida, 2018).Most previous CFD-modeling studies have been conducted to simulate airflows over small parts of building-congested urban areas because of the limitations of computing power (Ashie and Kono, 2011; Gousseau et al., 2011; Ng et al., 2011).

A nested grid system is commonly employed in most of the atmospheric weather prediction models (Sullivan et al., 1996; Nozu et al., 2008; Park et al., 2014) but rare in the CFD models. The Envi-met model based on the Reynolds Averaged Navier-Stokes’ equations (RANS) employs a nested grid system to minimize the impact of lateral boundaries(Kong et al., 2016; Acero and Herranz-Pascual, 2015). Piomelli et al. (2003) investigated the near-wall turbulent flows using a hybrid RANS-LES (large-eddy simulation) model with a nested grid system. Vonlanthen et al. (2016) proposed a LES-LES nested grid system and conducted numerical simulations in a narrow area with a horizontal resolution of about 10 m. The use of a nested grid system enables the CFD models to cover wider areas, satisfying the CFD-model guidelines for the numerical domains (Franke et al., 2007).

Ulleungdo, which corresponds to the peak of a seamount, has steep slopes, and its mountainous topography is very complicated. Therefore, airflows over Ulleungdo would be severely distorted by such complex topography. Besides, selecting a part of Ulleungdo as a numerical domain may result in unrealistic simulation results.In this study, to cover the entire Ulleundgo as well as adjacent seas, we adopted a one-way nested grid system to a CFD model. We conducted the simulations of the flows around Ulleungdo for 16 inflow directions and different inflow speeds using the CFD model with the one-way nested grid system. We evaluated how well the CFD model reproduced the wind speeds and directionsin Ulleungdo by comparing the simulations with the measurements at the Ulleungdo automated synoptic observation station. Also, we analyzed the simulation results in detail for the selected inflow directions(northwesterly, south-southwesterly, and northeasterly cases).

2. Methodology

1) Ulleungdo ASOS and Ullengdo marine buoy

Ulleungdo, located at 37°29′N and 130°54′E (about 217 km the northeast of Pohang, Korea), is a seamount rising from the seabed and the entire island corresponds to the peak of the mountain. The mountainous topography has complicated shape toward the coastal direction, with Seonginbong (984 m) located at the center of Ulleungdo. In this study, we analyzed the wind speeds and directions observed at the Ulleungdo ASOS (ASOS 115) and the Ulleungdo marine buoy (21229) for January 01, 2012 ~ December 31, 2016. Fig. 1 shows the satellite photographs around Ulleungdo. The Ulleungdo marine buoy is located about 19 km southeast of the Ulleungdo ASOS (Fig. 1(a)). The Ulleungdo ASOS lies in the southeast part of Ulleungdo, and its altitude is 222.4 m above sea level. There are Manghyangbong (317 m above sea level) and Hangnambong (281 m above sea level) near the southeast coastline of Ulleungdo. In the northwest region of the two peaks, a low valley-like topography is formed from the southwest to the northeast directions. The Ulleungdo ASOS is located inside this shallow valley (Fig. 1(b)). In this study, we analyzed hourly averaged wind speeds and directions measured at the Ulleungdo ASOS and marine buoy. For the analysis, we excluded wind speeds less than 1 m s^-1.

Fig. 1. Satellite photographs around (a) Ulleungdo, and (b) the Ulleungdo ASOS (ASOS 115). The red inverted triangle in (a) and (b) indicate the Ulleungdo ASOS and the yellow inverted triangle in (a) indicates the Ulleungdo marine buoy (21229).

2) Numerical model

The CFD model used in this study is the same as that in Kim et al. (2014). The model is based on the Reynolds-averaged Navier-Stokes(RANS) equations, assuming a three-dimensional, incompressible, and non-rotating airflow system. The model employs a staggered grid system and the finite volume method. For numerically integrating the governing equations, the semi-implicit method for pressure-linked equation algorithm(SIMPLE) is used. For turbulence parameterization, the model includes a renormalization group (RNG) k-ε turbulence closure scheme.

3) Simulation setup

In this study, we used two one-way nested domains with respective horizontal(vertical) grid intervals of 60 m (15 m) and 20 m (5 m)(Fig. 1).The grid dimensions (east-west × south-north × vertical directions) of the outer and inner domains were 400 × 350 × 250 and 100 × 100 × 300, respectively. The outer domain (domain 1) covered the entire Ulleungdo and its adjacentseas, with the Ulleungdo ASOS located at the center ofthe inner domain (domain 2).The topography in the model was constructed using the contours in a vector format of the geographic information system established by the National Geographic Information Institute of Korea in 2016 (Fig. 2). We linearly interpolated the contours to make a Lego-type terrain at each grid cell.

Fig. 2. Three-dimensional topographic configurations forthe areas around(a) Ulleungdo, and (b) Ulleungdo ASOS.

The vertical profiles of wind and turbulence incoming toward Ulleungdo would be very complicated, and their types would be diverse, influenced by atmospheric stability and local circulations around Ulleungdo. However, in this study, we assumed neutral atmospheric stability for simplicity. The vertical profiles of wind (U, V, W), turbulence kinetic energy (TKE, k), and the dissipation rate of TKE (ε) are given as follows:

\(U(z)=\left(\frac{u_{*}}{k}\right) \ln \left(\frac{z}{z_{0}}\right) \cos \theta\)       (1)

\(V(z)=\left(\frac{u_{*}}{k}\right) \ln \left(\frac{z}{z_{0}}\right) \sin \theta\)       (2)

\(W(z)=0\)       (3)

\(k(z)=\frac{u_{*}^{2}}{c_{\mu}^{\frac{1}{2}}}\left(1-\frac{z}{\delta}\right)^{2}\)       (4)

\(\varepsilon(z)=\frac{c_{\mu}^{\frac{3}{4}} k^{\frac{3}{2}}}{\kappa z}\)       (5)

Here, u_*, z₀ (= 0.05 m), κ (= 0.4), and θ denote the friction velocity,roughness length, von Karman constant, and wind direction,respectively. U_refrepresentsthe wind velocity at the reference height, z_ref(= 1,000 m). The vertical profiles of wind, TKE, and the dissipation rate of TKE above z_ref are the same as those at z_ref . We conducted the numerical simulations for 16 inflow directions (22.5° ≤ θ ≤ 360° with an increment of 22.5°) and, for each inflow direction, we considered six different wind speeds (U_ref = 5, 10, 15, 20, 25 and 30 m s^-1) at the reference height in the domain 1(hereafter,referred to as UD1). For providing the initial and boundary conditions of the domain 2(hereafter, referred to as UD2), we linearly interpolated wind speeds, TKEs, and the dissipation rates of TKEs at the boundary between UD1 and UD2 in the horizontal and vertical directions. The CFD model was numerically integrated up to 7,200 s with 1 s interval in the UD1 and UD2.

3. Results and discussion

1) Analysis of surface winds at the Ulleungdo ASOS and marine buoy

The surface winds measured at the Ulleungdo ASOS are likely to be distorted by its complicated surrounding topography. To investigate how much the airflow at the Ulleungdo ASOS was affected by its surrounding topography, we initially compared the averaged wind speeds and directions measured at the Ulleungdo ASOS with those at the Ulleungdo marine buoy. Then, we conducted numerical simulations for 16 inflow directions and compared the surface wind changes between the Ulleungdo ASOS and marine buoy in simulations with those in observations.

We analyzed the wind roses at the UlleungdoASOS and marine buoy for January 01, 2012 ~ December 31, 2016 (Fig. 3).In the case ofthe Ulleungdo marine buoy, wind direction frequencies were relatively evenly distributed, with maximum (south-southwesterly) and minimum (east-southeasterly) frequencies of 8.54% and 3.56%, respectively (Fig. 3(a)). Most recorded wind speeds were in the range from 3.3 to 5.4 m s^-1 for all wind directions (27.02%). We analyzed the diurnal [daytime (06 to 17 h) and night-time (18 to 05 h)] and seasonal [spring (March to May), summer (June to August), autumn (Septemberto November), and winter (December to February)] variations of the wind roses. The results showed that the diurnally and seasonally analyzed wind roses were similar to that in Fig. 3(a) (not shown). In the case of the Ulleungdo ASOS, northeast (17.79%) and southwest (15.99%) were the dominant wind directions (Fig. 3(b)). The prevailing wind directions are associated with the valley-like topography lying from the southwest to the northeast. At the Ulleungdo ASOS, the most frequent wind speed range was the section of 1.5 to 3.3 m s^-1 for all wind directions (32.78%). Northeasterly and south-southwesterly (southwesterly and northeasterly) winds were dominant in the daytime (night-time). Seasonally, southwesterly and northeasterly winds were dominant in the spring, northeasterly and south-southwesterly winds were dominant in the summer, northeasterly and southwesterly winds were dominant in the autumn, and northeasterly and southwesterly winds were dominant in the winter. The wind speeds at the UlleungdoASOS were lower than at the Ulleungdo marine buoy, which indicates that the wind speeds of the airflows moving toward Ulleungdo were affected by ground friction and the complicated surrounding topography.

Fig. 3. Wind-roses measured at (a) the Ulleungdo marine buoy and (b) the Ulleungdo ASOS for 2012. 01. 01 ~ 2016. 12. 31.

We numerically reproduced the surface wind changes in the Ulleungdo ASOS measurement that might be affected by the complicated surrounding topography. Fig. 4 shows the changes in wind speeds and directions at the UlleungdoASOS in the UD1 and UD2 simulations with U_ref = 5, 10, 15, 20, 25, and 30 m s^-1 . The wind speeds at the Ulleungdo ASOS almost linearly increased with the increase of the inflow wind speeds in both the UD1 and UD2 simulations (Fig. 4(a) and Fig. 4(c)). The wind speeds at the ASOS were high (low) at the south-southwesterly and north-northeasterly (east-southeasterly and west-northwesterly) inflows of which directions were parallel (perpendicular) to the shallow-valley-like topography around the Ullengdo ASOS. The changes in wind directions at the Ulleungdo ASOS in both the UD1 and UD2 simulations were similar to each other, except for the west-northwesterly and northwesterly inflows in the case with U_ref = 5 m s^-1 (Fig. 4(b) and Fig. 4(d)).

Fig. 4. Wind speeds and directions simulated at the Ulleungdo ASOS for the 16 inflow directions in the UD1[(a) and (b)] and UD2 simulations [(c) and (d)].

Fig. 5(a)shows the measured and simulated average percentiles of the wind speeds at the UlleungdoASOS (10 m above the ground surface) to those at the Ulleungdo marine buoy (6 m above the sea level) for 16 inflow directions. The measurements showed that the wind speeds decreased at the Ulleungdo ASOS in all inflow directions (67.16% on average), compared to the Ulleungdo marine buoy.This decrease is because the airflow became affected by surface friction of the complicated topography, after getting onshore. The wind-speed reduction at the Ulleungdo ASOS was relatively small for the northeasterly and southwesterly of which directions were parallel to the orientation ofthe shallow valley (northeasterly: 84.5%, southwesterly: 83.32%). On the other hand, the reduction was relatively significant for the southeasterly and northwesterly of which directions were perpendicularto the orientation of the shallowvalley (southeasterly: 53.03%, northwesterly: 42.76%). The UD1 and UD2 simulations reproduced the measured wind-speed reduction between the Ulleungdo ASOS and marine buoy for the alongand cross-valley inflow directions. The reduction was relatively small in the southwesterly (UD1: 104.14%, UD2: 93.15%) and northeasterly (UD1: 76.35%, UD2: 52.75%) inflows and large in the southeasterly (UD1: 43.90%, UD2: 46.37%) and northwesterly (UD1: 25.41%, UD2: 22.87%) directions. Besides, in both UD1 and UD2 simulations, the west-northwesterly and east-southeasterly winds became also significantly weakened [(west-northwesterly: UD1 – 14.61%, UD2 – 13.23%),(east-southeasterly: UD1 – 18.91%, UD2 – 21.71%)]. Note that the wind speeds measured at the Ulleungdo ASOS were 46.74 and 60.41% of those measured at the Ulleungdo marine buoy in the west-northwesterly and east-southeasterly directions, respectively. The wind speeds at the UlleungdoASOS for upslope wind directions(southwesterly in the UD1 and south-southwesterly in the UD1 and UD2) are excessively estimated because the CFD model employing the Cartesian coordinate system simulated stronger wind speeds along the steep upslope terrain. We analyzed the change of averaged wind direction caused by the topography around the Ulleungdo ASOS, by comparing the wind directions observed and simulated at both the Ulleungdo ASOS and marine buoy (Fig. 5(b)). For this, we assumed that the wind directions at the Ulleungdomarine buoy represented the background wind directions for the analysis period. Northeasterly (southwesterly) winds were measured at the Ulleungdo ASOS when the wind directions were between 0° and 112.5° (180° and 270°) at the Ulleungdo marine buoy. The consistent wind directions at the UlleungdoASOS somewhat dependent on the inflow directions (at the Ulleungdo marine buoy) of 0° ~ 112.5° and 180° ~ 270° were associated with the orientation ofthe shallow valley (from the southwest to the northeast). The UD1 and UD2 simulations qualitatively well reproduced the characteristics of the measured wind speeds and directions at the ASOS, despite the overestimation of the wind speed reductions for the east-southeasterly and northwesterly directions.

Fig. 5. (a) Observed and simulated percentiles of the wind speeds at the Ulleungdo ASOS (ASOS 115) and those at the Ulleungdo marine buoy for the 16 inflow wind directions in UD1 and UD2 simulations and (b) the relationship of the observed and simulated wind directions at the Ulleungdo ASOS and those at the Ulleungdo marine buoy. In the simulations, the wind speeds and directions at the inflow boundaries are assumed to be the same as those at the Ulleungdo marine buoy.

We made wind-roses at the UlleungdoASOS based on the simulated wind speeds and directions for 16 inflow directions. The limitation of the computational resources precluded the CFD simulations at every one hour for five years. Instead, we used the frequency of each wind direction for the six wind-speed sections measured at the Ulleungdo marine buoy (Fig. 3(a)) as that of the inflow directions. The combination of the wind speeds and directions at theASOS corresponding to 16 inflow directions and the frequencies of 16 wind directions for the six wind-speed sections produced wind-roses simulated at the ASOS by the CFD model without the whole five-year simulations (Fig. 6). The simulated wind-rosesindicated thatsouth-southwesterly (UD1: 27.33%, UD2: 39.99%) and northeasterly (UD1: 21.95%,UD2: 21.93%)were dominant attheUlleungdo ASOS, and the most frequently simulated wind-speed section was 1.5 ~ 3.3ms^-1 for all wind directions(UD1: 31.24%,UD2: 38.24%).Compared to themeasurements (Fig. 3(b)), the UD1 and UD2 simulated the excessive south-southwesterly and northeasterly and the deficient southwesterly,north-northeasterly, andwest-southwesterly at the Ulleungdo ASOS.

Fig. 6. Wind-roses at the Ulleungdo ASOS reconstructed on the basis of wind-direction frequencies at the Ulleungdo marine buoy and the 96 results in (a) UD1 and (b) UD2 simulations.

When reproducing the wind-roses using numerical simulations, the wind directions were biased in windroses as the wind directions were divided by 16 directions. The following reasons may cause a discrepancy between measurement and simulation. In this study, we assumed that the wind directions at the Ulleungdo marine buoy were the inflow directions. However, there might be slight changes in wind direction while the airflows reached Ulleungdo because of the changes in synoptic weather patterns, local circulations around Ulleungdo, and the effects of turbulence and surface friction.Besides, the relatively coarse topography using UD1 may confer less realistic boundary conditions compared to UD2. Finally, the limitation of the performance of the CFD model may cause such discrepancy. It is well known that the RANS equationtype CFD models overestimate the length of the recirculation zone behind a cubical obstacle (Lateb et al., 2014). When the Ulleungdo ASOS lies in the overestimated recirculation zone, the CFD model will simulate the wrong wind direction. Nevertheless, the CFD model reproduced the dominant wind directions at the UlleungdoASOS reasonably well, reflecting the effects ofthe topography around the UlleungdoASOS.

2) Analysis of the flows around the Ulleungdo ASOS

In this section, we analyzed the detailed airflows around the Ulleungdo ASOS for the (1) northwesterly case, in which a large discrepancy was apparent between the observed and simulated wind directions; (2) south-southwesterly case, in which the wind speed increased in the simulations; and (3) the northeasterly case, which occurred most frequently

(1) Northwesterly (315°) case

In the northwesterly case, the simulated wind directions (southeasterly) at the ASOS were quite different from the measured wind directions (north-northwesterly) (Fig. 5(b)). The streamlines over Ulleungdo were extremely complicated in patterns (Fig. 7), indicating that the complicated topography of Ulleungdo induced quite differentflow patternsfrom the background flow. There was a convergence zone behind the south and east coastlines (denoted by the white dashed line in Fig. 7).Wind vectorsin the southeast part of Ulleungdo showed that there was a secondary circulation, and the dominant wind directions simulated in the UD1 and UD2were southeasterly (opposite to the inflowdirection, northwesterly) (Fig. 8(a) and Fig. 8(c)).As a result, the southeasterly flows were simulated at the Ulleungdo ASOS for the northwesterly inflows in the UD1 and UD2. Wind speeds were overall weaker in UD2 than UD1. Similarly, the wind speed at the UlleungdoASOS of UD2 was weaker than UD1 and, however, wind direction at the Ulleungdo ASOS was similar (UD1: 2.16 m s^-1 , 114.5°, UD2: 1.18 m s^-1 , 120.2°). The wind speeds at the Ulleungdo ASOS markedly decreased, compared with the inflow wind speeds(UD1: 25.41%, UD2: 22.87%). In the observation, the wind speed at the Ulleungdo ASOS was about 42.76% of that at the Ulleungdo marine buoy (Fig. 5(a)). The wind speeds were accelerated along the mountain ridgesin both the UD1 and UD2 (Fig. 8(b) and Fig. 8(d)). In the UD2, the simulated surface wind was stronger at the side of Manghyangbong (inside the white dashed circle in Fig. 8(d)) than the UD1. This is because the detailed topography including the mountain ridges stretching from the Manghyangbong and the small valleys between the ridges were resolved in the UD2 and the channeling effects along the valleys like the Venturi effects in urban areas occurred therein.

Fig. 7. The streamlines at z = 10 m in the UD1 simulations in the northwesterly case.

Fig. 8. The wind vectors (left panel) and distribution of the wind speeds (right panel) at the surface (z = 10 m) inside the domain 2 in the [(a) and (b)] UD1, [(c) and (d)] UD2 simulations in the northwesterly case.

(2) South-southwesterly (202.5°) case

In the south-southwesterly case, the wind speed at the Ulleungdo ASOS was larger than the inflow wind speed (Fig. 5(a)). As in the northwesterly case, the topography of Ulleungdo induced complicated airflow pattern there in and secondary circulations were generated behind the downwind coastlinesin the north and east of Ulleungdo (Fig. 9).Wind speeds around the Ulleungdo ASOS in the UD1 were overall stronger than the UD2 (Fig. 10(a) and Fig. 10(c)). In the south-southwesterly case, the wind direction in the UD2 (203.4°) at the Ulleungdo ASOS was similar to that in the UD1 (197.9°), with smaller wind speed (7.73 m s^-1) than the UD1 (8.01 m s^-1). Note that, in both the UD1 and UD2, the wind speeds at the Ulleungdo ASOS were slightly largerthan the inflow wind speeds (UD1: 110.5%, UD2: 105.9%). Because there was no terrain higherthan the observation height in the upwind region,relatively strong airflowsreached the Ulleungdo ASOS on the upwindmountain slope and the simulated wind speedswere higherthan the inflowwind velocities. In the observation, the wind speed wasreduced to about 81.95% of the inflow wind speed (Fig. 5(a)). The surface wind was relatively strong along the mountain ridgesstretching from Seonginbong and in the nearthe southeast coast (Fig. 10(b) and Fig. 10(d)). However, in most other regions, the wind speeds markedly decreased compared with the inflow wind speeds.Wind speeds over the sea near the coast in the UD2 were similar to those in the UD1, with weaker inland wind speeds.

Fig. 9. The same as in Fig. 7 except for the south-southwesterly case.

Fig. 10. The same as in Fig. 8 except for the south-southwesterly case

(3) Northeasterly(45°) case

Northeasterly is the most frequently observed wind direction at the Ulleungdo ASOS for the analysis period (Fig. 3(b)).Asin the northwesterly and south-southwesterly cases, the airflow pattern was complicated, and the wind speeds were markedly reduced. In the downwind region of Ulleungdo (the region along the southwest coastline), secondary circulations were generated (Fig. 11). Compared with UD1 simulations, UD2 simulated relatively weak wind speeds over Ulleungdo. (Fig. 12(a) and Fig. 12(c)). In both the UD1 and UD2 simulations, the airflows ascending the mountain upslope in the northeast of Ulleungdo ASOS were accelerated after reaching the mountain ridge and the speed-up waslargerin the UD1 than the UD2 (the ellipses in black dashed line in Fig. 12(a) and Fig. 12(c)).The wind speed at the Ulleungdo ASOS in the UD2 was smaller than the UD1. On the other hand, the wind directions at the UlleungdoASOS in the UD1 and UD2 were similar (UD1: 5.54 m s^-1, 38.89°, UD2: 3.99 m s^-1 , 40.27°).Because there was no terrain higherthan the observation height in the upwind region, in both UD1 and UD2 the wind directions at the Ulleungdo ASOS was similar to the inflow direction (northeasterly). The Ulleungdo ASOS was located at the downwind region of the mountain ridge in this case. As the airflow moved over the mountain ridge,the surfacewindwasweakened,with the simulated wind speeds at the UlleungdoASOS about 76.35% and 52.75% of the inflow wind speed in the UD1 and UD2(Fig. 12(b) and 12(d)).In the observation, the wind speed wasreduced to about 84.50% ofthe inflow wind speed (Fig. 5(a)). The streamlines and distribution of the vertical wind componentsin the crosssection taken along the (\(\overline{AB}\)) and (\(\overline{CD}\)) in Fig. 12(a) and Fig. 12(c) were analyzed (Fig. 13). Recirculation zones were generated at the downwind regions of Hangnambong and the Ulleungdo ASOS in both the UD1 and UD2 simulations.

Fig. 11. The same as in Fig. 7 except for the northeasterly case.

Fig. 12. The same as in Fig. 8 except for the northeasterly case.​​​​​​​

Fig. 13. The streamlines and contours of vertical wind components taken along (a) (\(\overline{AB}\)) in Fig. 12(a) and (b) (\(\overline{CD}\)) in Fig. 12(c).​​​​​​​

4. Summary and conclusions

In this study, we investigated the topographic effects of Ulleungdo on wind speeds and directions observed at the Ulleungdo ASOS using a CFD model with oneway nested grid system. Numerical simulations were conducted for 16 inflow directions and six wind speeds (U_ref = 5, 10, 15, 20, 25 and 30 m s^-1) at a reference height (z = 1,000 m).

To determine the effects of topography on the observed surface winds, we compared the observed wind speeds and directions at the Ulleungdo ASOS with those at the Ulleungdo marine buoy, which were used as the background wind speeds and directions. The changes in wind speed and direction at the Ulleungdo ASOS caused by the topography were relatively well simulated by the CFD model with one-way nested grid system. The wind rose for the period of January 01, 2012 ~ December 31, 2016 at the Ulleungdo ASOS was reconstructed using the frequencies for each wind direction observed at the Ulleungdo marine buoy and the relationships between the wind speeds and directions at the inflow boundaries and the Ulleungdo ASOS in UD2 simulations, with a high spatial resolution

The CFD model simulated the excessive south-southwesterly and northeasterly inflows and the deficient southwesterly, north-northeasterly, and west-southwesterly inflows. This discrepancy between the observation and simulation may have been caused by various factors. (1) The assumption that the wind directions at the Ulleungdo marine buoy were the same as the inflow directions (before airflows reached the boundaries of the numerical domain, wind directions may have changed due to the changes in synoptic weather patterns, effect of local circulations around Ulleungdo, effects of turbulence and surface friction, and otherfactors).(2)The difference in spatialresolution (UD1 resolved a relatively coarse topography, which may confer less realistic boundary conditions than UD2). (3) The poor performance of the RANS-type CFD model in simulating recirculation zones (RANStypeCFDmodels overestimate the length ofrecirculation zone behind a cubical obstacle). Nevertheless, the CFD model with nesting grid system was considered to reconstruct the wind rose at the Ulleungdo ASOS reasonably well, reflecting the effects of local topography at the site.

The characteristics of flows around the Ulleungdo ASOS were investigated forthree representative inflow directions (northwesterly, south-southwesterly, and northeasterly) in the UD1 and UD2 simulations. The simulated wind rose indicated thatsouth-southwesterly and northeasterly was the dominant wind directions, which was also observed at the Ulleungdo ASOS. We confirmed that the CFD model with one-way nested grid system reflected the local topographic effect included in the Ulleungdo ASOS observation data. Also, we investigated the flow characteristics around the Ulleungdo ASOS for the northwesterly, south-southwesterly, and northeasterly inflows in detail.