Sie sind auf Seite 1von 16

Journal of Wind Engineering

and Industrial Aerodynamics 87 (2000) 4560

Numerical simulation of wind ow over


hilly terrain
Hyun Goo Kima,b, V.C. Patela,*, Choung Mook Leeb
a
Iowa Institute of Hydraulic Research, The University of Iowa, 300 South Riverside Drive,
Iowa City, IA 52242-1585, USA
b
Advanced Fluids Engineering Research Center, Pohang University of Science & Technology,
Pohang, South Korea
Received 5 February 1999; accepted 16 March 2000

Abstract

Wind ow over hilly terrain is simulated by solutions of the Reynolds-averaged


Navier2Stokes equations without the hydrostatic approximation. The standard and RNG-
based k2e models are used together with wall functions to account for surface roughness. The
numerical model uses nite-volume discretization and boundary-tted coordinates to resolve
the terrain. Simulations were made for ow at four sites, namely, Cooper's Ridge, Kettles Hill,
Askervein Hill and Sirhowy Valley, for which eld data are available. Comparisons with wind
data show agreement with respect to the proles of local wind magnitude and direction. The
numerical model is therefore deemed suitable for reliable prediction of local-scale wind ow
over hilly terrain with regions of ow separation. # 2000 Elsevier Science Ltd. All rights
reserved.

1. Introduction

Prediction of local wind eld over complex terrain with hills and valleys provides
information that is critical to assess wind-related issues, such as wind loads on
structures, siting of airports, power plants, industrial projects, and wind mills, and
prediction of pollutant dispersion in the atmosphere. The global circulation or
mesoscale models of atmospheric ow are not suitable for such purposes for two
reasons. First, they are based on the hydrostatic approximation in which a balance is
assumed between the pressure eld and gravity in the vertical. The hydrostatic

*Corresponding author. Tel.: +319-335-5237; fax: +319-335-5238.


E-mail address: v-c-patel@uiowa.edu (V.C. Patel).

0167-6105/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 0 0 ) 0 0 0 1 4 - 3
46 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

assumption may be appropriate and convenient for length scales of the order of
hundreds of kilometers, but at scales required to address local wind eects it is not.
Pressure changes due to inertial eects in the vertical direction (the s-coordinate)
cannot be neglected at the local scales. Second, the mesoscale models are not able to
resolve variations in topography in the vertical direction that are important to
prediction of local wind patterns which typically involve ow separations and
recirculation eddies [1] on surfaces of varying roughness. For the purposes of
predicting local wind patterns, it is necessary to use microscale models. These are
usually based on numerical solution of the Reynolds-averaged Navier2Stokes
equations and a turbulence model in a boundary-tted coordinate system that
follows the local terrain. Non-hydrostatic models of this type have been reported by
Raithby et al. [2], Glekas and Bergeles [3], and Anagnostopoulos and Bergeles [4].
However, they have not been validated by detailed and extensive comparisons with
eld data.
This paper reports application of the model of Kim et al. [5] and extension of the
recent work of Kim and Patel [6] to a number of test sites for which wind data are
available. Kim et al. [5] presented the details of the model and some comparisons
with laboratory data in ows with separation, while Kim and Patel [6] tested a
number of turbulence models for their eectiveness in resolving such ows. Here we
present simulations of the ow over four representative sites for which eld data are
available. This study thus serves to validate the numerical model by full-scale
applications.

2. Outline of the numerical model

The ow solver used in the present investigation to predict an atmospheric ow is


based on the nite-volume discretization in non-orthogonal boundary-tted
coordinates and the SIMPLEC pressure-correction algorithm on a non-staggered
grid arrangement.
The atmospheric ow is assumed to be isothermal, incompressible and turbulent,
so that it is described by the Reynolds-averaged Navier-Stokes (RANS) equations,
comprising
Conservation of mass:
@Ui
0; 1
@xi
momentum:
 
@Ui Uj 1 @p @ @Ui
n ui uj 2eijk Oj Uk 2
@xj r @xi @xj @xj
where Ui are the mean velocities in the coordinate directions xi , p is pressure, r and n
are the air density and viscosity, respectively, and Oi is the angular-velocity vector of
the earth's rotation. Note that x represents the prevailing wind direction and y the
transverse direction while z indicates the vertical direction.
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 47

To close these equations, the Reynolds stress tensor ui uj in Eq. (2) is related to
the mean strain-rate tensor Sij by the Boussinesq approximation
 
@Ui @Uj
ui uj 2nt Sij 23 kdij ; Sij 12 : 3
@xj @xi
The turbulent eddy viscosity nt Cm k2 =e in Eq. (3) is obtained from modeled
transport equations of the turbulence kinetic energy (k  ui ui =2) and its dissipation
rate e:
 
@Ui k @ nt @k
ui uj Sij e; 4
@xi @xi sk @xi
 
@Ui e @ nt @e e
C1 ui uj Sij C2 e: 5
@xi @xi se @xi k
These transport equations contain the ve constants listed in Table 1. The extra
term C1R in the source term of the e-equation of the RNG model accounts for the
eect of shear, which becomes more importantpin recirculating ows. Its modeling is

based on the innite-scale expansion in Z 2Sij Sij k=e, which is the ratio of the
turbulence time scale to the mean strain-rate scale.
In a previous study, Kim and Patel [6] investigated several two-equation
turbulence models and concluded that the RNG-based model oers the best
prospect for simulation of ow over complex terrain. The standard model is
employed here for comparison because of its widespread acceptance in diverse elds.
In both models, the local surface roughness is taken into account by means of wall
functions:
u z u2 u3
U ln ; k ; e ; 6
k z0 Cm
1=2 kz
where u is the friction velocity, z0 is the prescribed roughness height, and k is the
von Karman constant (=0.41).

3. Simulated wind and comparison with eld data

Although measured data from full-scale experiments are needed for validation of
numerical simulations, only a few such experiments have been carried out due to
practical diculties of full-scale measurements, selection of suitable sites, and the

Table 1
Constants in the turbulence modelsa

Models Cm C1 C2 sk se

Standard k2e model 0.090 1.44 1.92 1.0 1.3


RNG k2e model 0.085 1.42C1R 1.68 0.7179 0.7179
a
Note: C1R Z1 Z=4:38=1 0:015Z3 .
48 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

enormous expense. Among eld experiments conducted in recent years are those at
Brent Knoll Black Mountain [7], Kettles Hill [8], Blashaval Hill [9], Askervein Hill
[10], Nyland Hill [11], Cooper's Ridge [12] and Sirhowy Valley [13]. Unfortunately,
many of these contain signicant uncertainties and some do not provide the type of
information that is needed to carry out a meaningful simulation. In addition, many
of the eld experiments were conducted on what may be termed gentle-sloped hills
(with H=L51, where H is the height of the hill and L is one-quarter of its length)
with little or no signicant ow separation [14]. This type of terrain is ideally suited
for theoretical models, such as that of Jackson and Hunt [15], which solve linearized
perturbation equations.
For the present purpose, the experiments at Cooper's Ridge, Kettles Hill,
Askervein Hill and Sirhowy Valley were chosen. Their selection was based on the
following considerations: well-dened geometry (an accurate map for generating a
numerical grid), well-dened upstream proles to be used for the upstream boundary
condition, information about surface roughness, reliability of the measurements, and
a sucient number of measuring locations to describe the wind characteristics. The
principal features of these four sites are summarized in Table 2. The hill slope is
dened as the average slope of the top half of the hill upstream of the crest, i.e.,
s H=2L, the windward face of a hill being important for the speed-up features at
the hilltop. As a measure of the speed-up characteristics, we use the fractional speed-
up ratio, FSR (=U=U0 1), where U is the wind velocity at a given height above
ground and U0 is the approach velocity at that height.
In each simulation, eld measurements at a reference site were used to prescribe
the upstream conditions. Homogeneous Neumann condition was applied for all ow
variables on the outer boundaries (typically 10215 H from ground for the top
boundary, 5210 L for the upstream and side boundaries, and 15225 L for the
downstream boundary). For example, the calculations for the Askervein Hill were
carried out on a mesh of 1007030 (a surface grid of 75 m75 m), and converged
steady-state results were obtained after 1000 iterations which took about 1 h CPU
time on a CRAY-C90.

3.1. Cooper's Ridge

Cooper's Ridge, a north2south-oriented gentle-sloped ridge, is located in New


South Wales, Australia. Cooper's Ridge can be assumed as a quasi two-dimensional

Table 2
Summary of site characteristics

Terrain Height Base length Slope Wind direction Shape Roughness


H (m) 4L (km) s f z0 (cm)

Cooper's Ridge 115 1.6 0.143 2708 quasi-2D 3.0


Kettles Hill 105 2.4 0.1 2458, 2608 bell-shaped 3D 0.3  1.0
Askervein Hill 116 1.0 0.2 1808, 2108 elliptic 3D 3.0
Sirhowy Valley 200 1.8 0.4 908, 2708 cyclic 2D 1.0
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 49

topography considering the geometry and the incident wind direction (westerly
wind). Thus, only a two-dimensional computation was performed for this case.
Fig. 1 shows the horizontal distributions of FSR at 2 and 3 m above ground level.
The present numerical results are fairly close to the eld measurements especially on
the hilltop, while the linearized model of Jackson and Hunt shows signicant
dierences. As the upstream proles were not measured in the eld experiments [12],
the measurements at x 800 m were assumed for the upstream conditions in the
numerical simulation. The numerical solutions show signicant undulations in wind
speed on the leeside of the hill (x  1:2 km) but no data are available to verify this
behavior. It is important to note that the wavy distribution of FSR predicted by the
numerical model is related to the local elevation changes in the terrain, as shown by
the enlargement in Fig. 1. The theoretical predictions of Hunt et al. [16] deviate
considerably from the present results in the downstream segment. The linear theory
predicts a smooth distribution due to ltering of the terrain data.

3.2. Kettles hill

Kettles Hill in Alberta, Canada (40830'N 113850'W) is an isolated hill and has a
at and uniform upstream fetch of about 20 km. The eld experiments reported by
Mickle et al. [17] were conducted in neutral atmospheric conditions. The numerical
grid for the three-dimensional ow simulation was generated from the digital
topographical map (Fig. 2) created from a contour map [8]. The inow conditions
were prescribed from the measurements at the reference site located far upstream of
the hill.

Fig. 1. Fractional speed-up ratio (FSR) for wind over Cooper's Ridge at height of 3 m above ground.
Lower part of gure shows ridge topography, with a portion enlarged by factor of ve. Symbols: eld
measurements [12]; solid line: present numerical model (3 m); dashed line: linear theory [16].
50 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

Fig. 2. Three-dimensional topography of Kettles Hill.

Fig. 3 shows the normalized wind speed (U=U0 ) at two elevations (Dz) about the
ground level. In this case the numerical as well as the theoretical predictions by
the method of Jackson and Hunt (MS3DJH/3.1) [18] show good agreement with
the data. Recall that this is a very gentle-sloped hill (s 0:1). Vertical proles of
normalized wind speed at the hilltop are shown in Fig. 4, where the wind-tunnel
measurements [19] for f 2458 are compared with the present results. The
numerical results show that the speed-up increases as the wind direction changes
from 2608 to 2458. This is because the hill is steeper to the wind from 2458.

3.3. Askervein hill

The eld experiment at Askervein Hill in Scotland (57811'N 7822'N) was


conducted under an international collaboration project. This is regarded as the
most successful full-scale experiment [10]. Fig. 5 shows a contour map of the
Askervein Hill (the left-most hill which resembles an ellipsolid) and surrounding
area. The eld measurements [20] were made with several measuring devices along
the lines A2A, AA2AA and B2B at a height of 10 m above the hill surface. In the
numerical simulation, the computations of the three-dimensional ow over the
region were supplemented by two-dimensional computations along various lines to
better understand the three-dimensional eects. The three-dimensional ow
computations were performed with surface grids of 5050 m and 7575 m. The
two grids gave essentially the same results.
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 51

Fig. 3. Normalized wind speeds for ow over Kettles Hill along the east-west line for incident wind
direction of 2608 at two levels (Dz) from ground. Hollow circles: eld measurements (2558; [17]); solid
triangles: eld measurements (2638; [17]); solid lines: present numerical model (2608); dashed lines: linear
theory (2608; [18]).

Fig. 4. Vertical proles of normalized wind speed at the top of Kettles Hill. Symbols: wind tunnel data
(2458; [19]); solid line: present numerical model (2458); dashed line: present numerical model (2608).
52 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

Fig. 5. Topographical map of Askervein Hill and surrounding area (length units: m).

Fig. 6. Vertical proles of the atmospheric boundary layer at the reference site (Note: U10 U0 at
Dz 10 m.). Symbols: eld measurements by four instruments; solid lines: present numerical model (1-D);
dashed lines: Ayotte (1996).

Fig. 6 compares upstream proles measured in the eld experiment at the


reference site located 2.7 km upstream from the hill with one-dimensional numerical
simulations (solid lines) for ow over a at terrain and proles obtained by
application of optimal control theory (dashed lines) by Ayotte [21]. The eld data at
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 53

the reference site were analyzed as fully developed boundary layer data by Mickle
et al. [22]. It is found that proles obtained from a one-dimensional, fully developed
ow simulation are much closer to the eld data than proles obtained by control
theory.
Distributions of FSR at a height of 10 m above the ground level along lines A2A,
AA2AA and B2B (see Fig. 5) are presented in Fig. 7. Comparisons are
made among the wind-tunnel experiments [22], full-scale measurements [23],
two-dimensional computations, three-dimensional computations, and theoretical
predictions [24]. There is substantial agreement among all the results on the
windward slope of the hill, but signicant dierences are found on the leeside
along line A2A where the full-scale measurements show a rapid decrease
of FSR down to 0.7. Only the three-dimensional simulations predict this
behavior although there is some deterioration of agreement with data at
the peak. It is conjectured that this trend arises from the three-dimensional
ow separation [23] on the lee-slope due to the inuence of neighboring hills.
For reference, it should be noted that ow separation on the leeside of Askervein
Hill was not predicted by a three-dimensional computation when the neighboring
hills were excluded. In that case the FSR distributions were quite similar to those
predicted by the two-dimensional computation. The wind-tunnel experiment, which
employed a smooth-surface model, failed to depict the decrease of FSR on the
lee-slope [23].
The horizontal velocity vectors at a height of 1 m above the ground level in the
vicinity of Askervein Hill are plotted in Fig. 8. The three-dimensionality of the ow
on the leeside is clearly evident. The ow over the downward face of the hill makes
an abrupt turn due to the blockage eect of the downstream hill. The marked
decrease of wind speed along line A2A (see Figs. 5 and 7) is associated with this
local feature of the ow. The turbulence kinetic energy distribution 10 m above the
ground along line A2A is shown in Fig. 9. Signicant discrepancies are found near
the hilltop and on the leeside slope. The higher values in the eld measurements may
be due to an intermittent or unsteady ow separation, which is not modeled by the
Reynolds-averaged equations. However, somewhat better agreement is seen with the
wind-tunnel data. Fig. 10 shows the distributions of the deviation angle of the local
ow along line A2A. This angle is dened as the angle dierence in clockwise
direction between the approaching wind and the local wind. The agreement between
the present predictions and the data is somewhat better than that shown by Raithby
et al. [2] on the leeside. This indicates a more accurate representation of the three-
dimensionality of the ow in the present model.
For the incident wind direction of 1808, the 3D prediction of FSR along line
A2A, shown in Fig. 11, is quite similar to the prediction of the nite-dierence
model of Beljaars, Walmsley, and Taylor [24] based on the linear theory of Jackson
and Hunt [15]. The proles of mean horizontal velocity and turbulence kinetic
energy at the hilltop, which are presented in Fig. 12, show that the present numerical
model underestimates the wind speed at ground level by about 30%. Zeman and
Jensen [25] reported the same trend from their wind-tunnel experiments and argued
that the roughness around the hilltop is less than that over the upstream fetch. It is
54 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

Fig. 7. Fractional speed-up ratios for wind ow over Askervein Hill 10 m above ground level (f 2108).
Hollow rectangles: wind-tunnel data [23]; solid rectangles: eld measurements [20]; solid lines: present
numerical model (3-D); dotted lines: present numerical model (2-D); dashed lines: theoretical model [24];
(a) Along the line A2A. (b) Along the line AA2AA. (c) Along the line B2B.
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 55

Fig. 8. Streamlines over Askervein Hill 1 m above ground level from present numerical model with
incident wind direction of 2108 (Note: This gure is rotated relative to Fig. 5.).

Fig. 9. Turbulence kinetic energy 10 m above ground level along line A-A (f 2108). Hollow rectangles:
wind-tunnel data [23]; solid rectangles: eld measurements [20]; solid line: present numerical model (3-D).
56 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

Fig. 10. Deviation in ow angle 10 m above ground level along line A2A (f 2108). Symbols: eld
measurements [20]; solid line: present numerical model (3-D); dashed lines: numerical model [2].

Fig. 11. Normalized wind speed over Askervein Hill along line A-A 10 m above ground level (f 1808).
symbols: eld measurements [20]; solid line: present numerical model (3-D); dashed line: theoretical model
[24].
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 57

Fig. 12. Vertical proles of (a) normalized wind speed (U=U10 ), and (b) normalized turbulence kinetic
2
energy (k=U10 ) at hilltop of Askervein Hill (f 1808). Symbols: eld measurements [22]; solid lines:
present numerical model (3-D); dashed line: theoretical model [24].

expected that a calculation with a non-uniform distribution of surface roughness


would improve the prediction accuracy.

3.4. Sirhowy valley

The Sirhowy Valley in South Wales is an extensive topography composed of


nearly two-dimensional periodic ridges and valleys. Fig. 13 shows the valley shape at
the eld site. For the numerical simulation, periodic conditions were imposed on the
upstream and downstream faces of the computational domain.
The numerical simulations were carried out both for westerly and easterly winds.
The predicted separation and reattachment points are summarized in Table 3. The
eld data [26] are compared with the numerical results in Fig. 13 with respect to
velocity measurements at a height of 8 m above the valley surface for easterly wind,
and the locations of the separation and reattachment points. This gure and Table 3
show that the numerical model with the RNG turbulence model oers satisfactory
prediction of the size of the recirculation region.

4. Conclusions

Numerical simulations of atmospheric wind ows over hilly terrain were carried
out and compared with data from eld experiments at four sites. The numerical
predictions were found to be in reasonable agreement with the measurements with
58 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

Fig. 13. Variation of wind velocity 8 m above the surface of Sirhowy Valley for easterly wind, and
separation regions for westerly and easterly winds (Note: US U at x 0, Dz 8 m.). Symbols: eld
measurements [26]; lines: present numerical model.

Table 3
Separation and reattachment points in Sirhowy Valley

Wind directions Classication Field data k2e model RNG model

Westerly wind Separation point (m) 245 293 270


Reattachment point (m) 1023 596 1055
Length (m) 778 303 785
Easterly wind Separation point (m) 845 870 800
Reattachment point (m) 1254 1205 1215
Length (m) 409 335 415

respect to the wind magnitude and ow separation when present. In the most
extensive simulation, that of the wind ow over Askervein Hill, inclusion of the
neighboring hills in the computational domain was found to be important in the
prediction of the three dimensionality of the ow on the leeside. Simulation of
separated ow in the Sirhowy Valley showed that the RNG-based k2e turbulence
model gave better results than the standard model.
The present numerical model can be readily extended to simulate eects of
stratication of the atmosphere, and to add pollutant-transport equations to
simulate realistic and practical situations. Validation of such a comprehensive model
would clearly require additional eld data.

Acknowledgements

This research was supported under a collaborative agreement between the Iowa
Institute of Hydraulic Research (IIHR) of The University of Iowa and the Advanced
H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560 59

Fluids Engineering Research Center (AFERC) of Pohang University of Science and


Technology. The authors acknowledge Drs. Walmsley and Mickle for providing the
eld data and Miss Y.J. Noh who participated in the early phase of this work.

References

[1] B.W. Atkinson, Introduction to the uid mechanics of meso-scale ow elds, in Diusion and
Transport of Pollutants in Atmospheric Mesoscale Flow Fields, Kluwer Academic Publishers,
Dordrecht, 1995, pp. 1220.
[2] G.D. Raithby, G.D. Stubley, P.A. Taylor, The Askervein Hill Project: A nite control volume
prediction of three-dimensional ows over the hill, Boundary-Layer Meteorol. 39 (1987) 2472267.
[3] J.P. Glekas, G.C. Bergeles, Dispersion under neutral atmospheric conditions, Int. J. Num. Methods
Fluids 19 (1994) 2372257.
[4] J.S. Anagnostopoulos, G. Bergeles, A numerical model for wind eld and pollutant concentration
calculations over complex terrain. Application to Athens, Greece, J. Wind Eng. Ind. Aerodyn. 73
(1998) 2852306.
[5] H.G. Kim, C.M. Lee, H.C. Lim, N.H. Kyong, An experimental and numerical study on the ow over
2-D hills, J. Wind Eng. Ind. Aerodyn. 66 (1997) 17233.
[6] H.G. Kim, V.C. Patel, Test of turbulence models for wind ow on terrain with separation and
recirculation, Boundary-Layer Meteorol. 94 (2000) 5221.
[7] E.F. Bradley, An experimental study of the proles of wind speed, shearing stress and turbulence at
the crest of a large hill, Quart. J. Roy. Meteorol. Soc. 106 (1980) 1012124.
[8] J.R. Salmon, H.W. Teunissen, R.E. Mickle, P.A. Taylor, The Kettles Hill Project: Field observations,
wind-tunnel simulations and numerical model predictions for ow over a low hill, Boundary-Layer
Meteorol. 43 (1988) 3092343.
[9] P.J. Mason, J.C. King, Measurements and predictions of ow and turbulence over an isolated hill of
moderate slope, Quart. J. Roy. Meteorol. Soc. 111 (1985) 6172640.
[10] J.L. Walmsley, P.A. Taylor, Boundary-layer ow over topography: impacts of the Askervein study,
Boundary-Layer Meteorol. 78 (1996) 2912320.
[11] P.J. Mason, Flow over the summit of an isolated hill, Boundary-Layer Meteorol. 37 (1986) 3852405.
[12] P.A. Coppin, E.F. Bradley, J.J. Finnigan, Measurements of ow over an elongated ridge and its
thermal stability dependence: the mean eld, Boundary-Layer Meteorol. 69 (1994) 1732199.
[13] P.J. Mason, J.C. King, Atmospheric ow over a succession of nearly two-dimensional ridges and
valleys, Quart. J. Roy. Meteorol. Soc. 110 (1984) 8212845.
[14] P.A. Taylor, P.J. Mason, E.F. Bradley, Boundary-layer ow over low hills, Boundary-Layer
Meteorol. 39 (1987) 1072132.
[15] P.S. Jackson, J.C.R. Hunt, Turbulent wind ow over a low hill, Quart. J. Roy. Meteorol. Soc. 101
(1975) 9292955.
[16] J.C.R. Hunt, S. Leibovich, K.J. Richards, Turbulent shear ows over low hills, Quart. J. Roy.
Meteorol. Soc. 114 (1988) 143521470.
[17] R.E. Mickle, J.R. Salmon, P.A. Taylor, Kettles Hill'84: Velocity prole measurements over a low hill.
Research Report AQRB-84-012-L, Atmospheric Environment Service, Toronto, Canada, 1984.
[18] P.A. Taylor, J.L. Walmsley, J.R. Salmon, A simple model of neutrally stratied boundary-layer ow
over real terrain incorporating wavenumber-dependent scaling, Boundary-Layer Meteorol. 26 (1983)
1692189.
[19] H.W. Teunissen, Wind-tunnel and full-scale comparisons for mean ow over an isolated low hill,
J. Wind Eng. Ind. Aerodyn. 15 (1983) 2712286.
[20] J.R. Salmon, A.J. Bowen, A.M. Ho, R. Johnson, R.E. Mickle, P.A. Taylor, The Askervein Hill
Experiment: Mean wind variations at xed heights above the ground, Boundary-Layer Meteorol. 43
(1988) 2472271.
60 H.G. Kim et al. / J. Wind Eng. Ind. Aerodyn. 87 (2000) 4560

[21] K.W. Ayotte, Optimization of upstream proles in modeled ow over complex terrain, Boundary-
Layer Meteorol. 83 (1997) 2852309.
[22] R.E. Mickle, N.J. Cook, A.M. Ho, N.O. Jensen, J.R. Salmon, P.A. Taylor, G. Tetzla,
H.W. Teunissen, The Askervein Hill Project: Vertical proles of wind and turbulence, Boundary-
Layer Meteorol. 43 (1988) 1432169.
[23] H.W. Teunissen, M.E. Shokr, A.J. Bowen, C.J. Wood, D.W.R. Green, Askervein Hill Project: Wind-
tunnel simulations at three length scales, Boundary-Layer Meteorol. 40 (1987) 1229.
[24] A.C.M. Beljaars, J.L. Walmsley, P.A. Taylor, A mixed spectral nite-dierence model for neutrally
stratied boundary-layer ow over roughness changes and topography, Boundary-Layer Meteorol.
38 (1987) 2732303.
[25] O. Zeman, N.O. Jensen, Modication of turbulence characteristics in ow over hills, Quart. J. Roy.
Meteorol. Soc. 113 (1987) 55280.
[26] P.J. Mason, Diurnal variations in ow over a succession of ridges and valleys, Quart. J. Roy.
Meteorol. Soc. 113 (1987) 111721140.

Das könnte Ihnen auch gefallen