You are on page 1of 10

The Eighth Asia-Pacific Conference on Wind Engineering,

December 10–14, 2013, Chennai, India

Numerical and Physical Simulation of a Thunderstorm Downburst

Michael Jesson1, Matthew Haines2, Nishant Singh3, Mark Sterling4 and Ian Taylor5
1
Research Fellow, School of Civil Engineering, University of Birmingham, Birmingham, UK,
m.a.jesson@bham.ac.uk
2
PhD Student, School of Civil Engineering, University of Birmingham, Birmingham, UK
3
Research Fellow, Department of Mechanical Engineering, University of Strathclyde, Glasgow, UK
4
Beale Professor of Civil Engineering, School of Civil Engineering, University of Birmingham,
Birmingham, UK
5
Lecturer, Department of Mechanical Engineering, University of Strathclyde, Glasgow, UK

ABSTRACT

This paper describes work undertaken as a joint project between the University of Strathclyde (UoS)
and the University of Birmingham (UoB) in the UK, aimed at simulating thunderstorm downbursts both
physically and numerically. The UoB downburst simulator uses a 1m diameter, impinging jet with computer-
controlled fans and flaps to physically simulate a downburst, with the data gathered used for calibration and
validation of the numerical models developed at the UoS. These models, which are in an early stage of
development, use novel inlet conditions to simulate the inflow, allowing more accurate and variable inflows.
Wavelet analysis is used to compare the full-scale and physical simulation data, with correlation analyses (phase-
plots) used to identify structures within the physical simulation flow field. Initial results show a good match
between the physical simulations and full-scale, field data, and between the two sets of simulations.

KEYWORDS: WIND ENGINEERING, CFD, DOWNBURSTS, IMPINGING JETS

Introduction

Thunderstorm downbursts are extreme wind events caused by the density change as
warm, rising air cools, resulting in the gravity-driven fall of the resulting cold air. Upon
contact with the ground, a high pressure stagnation region forms as the vertical motion is
stalled, resulting in a radial outflow (Fujita 1985). Further, a ring-shaped entrainment vortex
is created by the downdraft (Chay and Letchford 2002b), which then travels with the outflow.
At ground level, this vortex rotates in the direction of the outflow, resulting in a combined
wind speed which exceeds that of the outflow alone. The existence of a secondary vortex,
formed at the leading edge of the primary vortex, has been indicated by numerical simulation
(Kim and Hangan 2007; Mason et al. 2009) and leads to further localized flow acceleration.
Field data show that the velocity field associated with a downburst differs significantly
from that of synoptic wind events - velocity time-series are highly non-stationary,
characterized by rapidly increasing wind speeds which culminate in a short period (~5
minutes or less) of intense winds (Fujita 1990; Gast and Schroeder 2003; Choi 2004), while
the vertical profile shows a maximum near the ground (Fig. 1).
Fujita (1981), using his own scale developed in 1971, estimated that downburst events
were capable of causing “…SEVERE DAMAGE…Roofs and some walls torn off well-
constructed houses…steel-framed hanger-warehouse type structures torn” (Fujita 1971).
Indeed, structural failures in a number of countries have been attributed to downbursts (Chen
and Letchford 2004); downbursts are often the cause of design wind speeds but despite this,
synoptic wind profiles still tend to be used for wind loading calculations (Chay and Letchford
2002a). It is, therefore, unsurprising that downbursts have become the subject of research,

Proc. of the 8th Asia-Pacific Conference on Wind Engineering – Nagesh R. Iyer, Prem Krishna, S. Selvi Rajan and P. Harikrishna (eds)
Copyright c 2013 APCWE-VIII. All rights reserved. Published by Research Publishing, Singapore. ISBN: 978-981-07-8011-1
doi:10.3850/978-981-07-8012-8 P3 1139
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

with attempts made to record actual, full-scale downburst events, and the development of both
physical and numerical simulations. Full-scale data have been captured by airfield
anemometers as part of routine monitoring (e.g. at Andrews Air Force Base (AAFB) (Fujita
1990)), early projects such as NIMROD, MiST and JAWS (Fujita 1990) and, more recently,
the 2002 Thunderstorm Outflow Experiment which set out to make high resolution
measurements of a downburst, succeeding in recording a rear-flank downdraft on the 4th June
2002 (Gast and Schroeder 2003; Orwig and Schroeder 2007; Holmes et al. 2008).

(b
) 2006)) and velocity time-series
Fig. 1 Example vertical velocity profiles (a, (Lin and Savory
(b, (Haines et al. 2013)) of synoptic and downburst winds.
While the full-scale data is invaluable in validating the simulation work, the near
impossibility of predicting precisely when and where downbursts will occur makes
measurements such as wind loading on structures impractical at full-scale. Laboratory-based,
physical simulations of downbursts have used two approaches – slot (e.g. (Lin and Savory
2006; Lin et al. 2007; Lin 2010)) and impinging jets (e.g. (Wood et al. 2001; Chay and
Letchford 2002a; Chay and Letchford 2002b; Mason 2003; McConville 2008; McConville et
al. 2009)). The experimental work described in this paper uses a modified version of the
impinging jet apparatus of McConville et al. (2009).
Computational Fluid Dynamics (CFD) is used in a variety of ways to numerical
simulate downbursts. Research on this topic is tending to fall broadly into two disciplines;
simulating the physical mechanisms of a downburst, using some form of meteorological
model (e.g. (Procter 1989; Orf et al. 1996; Lin et al. 2007), or simulating the physical
simulators being used for research (e.g. (Selvam and Holmes 1992; Zhou et al. 2012)).
Previous work by Letchford et al. (2002), Mason et al. (2005), Lin and Savory (2006) and
McConville et al. (2009) has shown that the physical simulations can match velocity time
histories reasonably well, and the data from them can be used to calibrate and validate the
numerical simulations.
The work described in this paper is part of an ongoing research project whose aim is to
investigate wind loading on structures subject to downburst winds, using a combination of
physical and numerical simulations. All of the simulations described above raise a
fundamental question which is discussed in Sterling et al. (2011) – i.e. “What is the
appropriate scaling to use in the physical simulations?” This problem is inherent to this type
of simulation, and it is envisaged that numerical studies may help to shed some light on the
subject. A brief description of the physical and numerical simulations undertaken for the
current work follows this introduction. Preliminary results are then given (more detailed
results and conclusions will appear in the full paper).

1140
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

Experimental Apparatus

The University of Birmingham (UoB) transient wind simulator is described by


McConville et al. (2009). Essentially, a 1m diameter impinging jet is created using computer
controlled fans and flaps, with the flaps controlling the onset of the downdraft. The flaps used
by McConville et al. have been replaced with more robust, plywood flaps, which are
counterweighted to prevent oscillation and hold them in the “open” position once released. A
jet velocity of Vj =13.7m/s is achieved, with a turbulence intensity of 13%. The jet impinges
onto a raised ground-plane, 2m below the jet outlet, with high frequency velocity and pressure
measurements made using 2000Hz Cobra probes and a bespoke, 500Hz, 64-channel pressure
measurement system respectively.

Numerical Simulation

The numerical simulation uses a numerical domain which is set up to closely match
the physical simulation. It uses an LES numerical method, similar to the work of Zhou et al.
(2012) with some important differences. The simulation is not of a jet which ramps up to a
steady state - in the same manner that the physical simulator turns the jet on and off again to
create a pulse so that only the vortex of the simulator is studied, the numerical model does the
same. In addition the inlet conditions can be changed from a non-steady jet, to a variety of
turbulent inlets. This includes a novel mixed-synthetic turbulent inlet developed by Singh
(2012). This works by selecting an appropriate correlation function for space and time and
also by selecting a true analytical representation of turbulence. For the synthetic LES
simulation, a mixed spectral method generates the turbulent velocity ¿eld with a designated
temporal correlation, a two point spatial correlation and a one point cross correlation. It is
envisaged that more accurately modelling the turbulence at the inlet, and also being able to
have some degree of control over it, will allow a range of conditions which aren’t possible in
the laboratory to be simulated.

Comparison of Physical Simulations to Full-Scale Data

In order to validate the physical simulations (“UoB”), comparison with the full-scale
data recorded during the Andrews Air Force Base (AAFB) downburst was made. The
velocities of both the UoB and AAFB data were non-dimensionalised using the respective
downdraft/jet velocity, Vj, while a dimensionless time, t*, was calculated as t* = tVj/D. For the
AAFB data, D = 1000m and Vj = 40ms-1 were used. The value for D is based on previous
estimates for the diameter of the AAFB downdraft (e.g. (Fujita 1985; Holmes and Oliver
2000)), while Vj is estimated from the Vj/D ~ 1.6 ratio seen in the physical simulations. It
should also be noted that the physical simulations do not replicate the translational motion of
a full-scale downburst. The non-dimensional time-series show close matching of the initial
peak (Fig. 2) – the second large peak in the AAFB data is not seen in the UoB data as it is
caused by the downburst passing over the airfield anemometer due to the translation
movement of the entire storm.
A Continuous Wavelet Transform (CWT) using the Morlet wavelet has been applied
to the non-dimensional forms of both the AAFB data and a selection of the UoB data, using
the algorithm of Büssow (2007). Before applying the CWT, the time-series were padded to a
length 2(n+1), where n is the smallest integer such that 2n exceeds the length of the time-series.
Padding reduces the effects of wrap-around on edge values, while padding to a power of 2
improves the efficiency of the fast Fourier transform used within the CWT. Generally,

1141
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

padding with zeroes is used; however in the case of the time-series processed here, the initial
and final values are non-zero and so zero padding introduces a step which manifests itself at
the edges of the true data range in the power spectrum. In their algorithm (not used here),
Torrence and Compo (1998) subtract the mean value from the time-series before applying the
CWT – for a stationary signal this will bring the initial and final values close to zero,
removing the step. For the non-stationary case seen here, half of the padding was added
before the true data, using the mean value of the first ten data points, with the remainder
added after the true data, using the mean value of the last ten data points. This was seen to
reduce step effects while leaving the remaining data unaffected. Finally, a lower bound of
[min. power + 0.01(max. power – min. power)]) was applied in order to improve clarity of the
power spectra, with values below this limit set to the boundary value.

Fig. 2 Full-Scale and Physical Simulation Velocity Time Series


Analysis of the CWT output (Fig. 3) shows a similar duration in dimensionless time
for the increase in power due to the initial peak of the downburst. The AAFB data shows a
wider peak due to the rear edge of the downburst passing over the anemometer. Further, the
same dimensionless frequency range of the high power region (approximately 0.1 – 2) is seen
for both the full-scale and physical simulation data. From this analysis, the physical
simulations appear to be capturing both the time-domain and spectral form of a full-scale
event.

Comparison of the Physical and Numerical Simulations

The numerical simulation work is in its early stages, but an initial examination of the
physical and numerical simulation results (Fig. 4) indicates good agreement between the two.
The initial velocity increases are negligibly different, and a plateau/dip is present in both at t =
0.3s. This dip is an indication of the presence of a secondary vortex. Although the numerical
simulation peaks slightly earlier than the physical, a maximum velocity of approximately
21m/s (1.6Vj) is seen in both.

1142
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

2
Dimensionless 10 -7.5
Frequency
-8

0 -8.5
10
-9

0 10 20 30 40
Dimensionless Time, t*
2
10 -6
Dimensionless

-6.5
Frequency

-7
0
10
-7.5

0 10 20 30 40
Dimensionless Time, t*
Fig. 3 Morlet CWT Power Spectra of the UoB (top) and AAFB (bottom) Data Sets

Fig. 4 Physical and numerical velocity time-series

Phase-Plot Analysis of the Physical Simulation Data


During a downburst the peak radial velocities are linked with the passing of the
primary, ring vortex. It would therefore be anticipated that a relationship between the vertical
velocity, w(t), and the radial velocity u(t) exists. As downburst winds are non-stationary,
standard calculations of correlation are not appropriate due to the use of the mean and

1143
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

standard deviation in the calculation – in the case of a downburst both of these parameters are
dependent on the length of time-series used. Instead, the relationship between u(t) and w(t)
has been elucidated using phase-plots, in which points (u(t), w(t)) are plotted for all t. In the
case of two random signals the phase-plot will show a random pattern; in the case of
correlated signals the phase-plot will exhibit order. In the case of the work presented here, the
phase-plot analysis is limited by the Cobra probes, which are only capable of measuring the
speed of winds whose direction lies within a ±45° cone around the probe x-axis (TFI 2013),
which was aligned horizontally and radially. A qualitative analysis is still possible, however.
For clarity, only the phase plots for x/D = 1.5, z/D = 0.02 and 0.20 are shown (Fig. 5).
Due to the short duration of the downburst event, the majority of the points (those from before
the downburst and after the passing of the vortex) lie close to w = 0. There is some spread
along the u direction, due to the post-vortex outflow velocity. For z/D < 0.04 there is little
evidence of vertical velocities, with w ~ 0, as evidenced by the z/D = 0.02 graph in Fig. 5.
This may be caused by damping of vertical motion by the wall and, more importantly, the fact
that close to the wall the ring vortex velocity is primarily horizontal, which also means that
the ring vortex has its greatest effect on the radial velocity. For analogous reasons the
maximum value of w(t) increases with height, z – the vortex velocity moves from the
horizontal towards the vertical – while the maximum u(t) decreases with height.
It is clear from Fig. 5 that the peak u(t) is preceded by upward flow and followed by
downward flow. This is indicative of the passing of the ring vortex, which has upward flow at
its leading edge and downward flow at its rear edge.
From his examination of full-scale data from the JAWS experiments, Hjelmfelt (1988)
determined that the vortex was not fully formed at the time of the maximum velocity, and
similar results are seen in the current simulations, with the maximum w(t) occurring at x/D =
2.0 (Fig. 6). From this figure it is also clear that at x/D = 1.0 there is still downward velocity,
arising from the downdraft, possibly due to a widening of the downdraft with distance from
the jet outlet.
The initial rapid rise in u(t) is, for x/D > 1.0, interrupted by a brief dip (Fig. 7). The
approximate height over which this dip is seen is given in Table 1. This dip is hypothesised to
be caused by a secondary vortex which develops at the leading edge of the primary vortex –
the existence of such a vortex has previously been suggested by, for example, Kim and
Hangan (2007), Xu and Hangan (2008) and Mason (2009). For the z/D values where the dip is
present the velocity increases rapidly immediately following the dip, with all time-series
converging before the maximum velocity is reached (Fig. 7). By isolating a short sub-section
of the time series encompassing the reduction, phase-plots can be constructed to examine the
coherence between u and w (Fig. 8). As the phase-plots lack a time scale, the point
corresponding to the temporal midpoint of the sub-section is indicated. It is evident that at the
time of the reduction u shows a decrease while w is downwards, and the subsequent increase
in u is accompanied by an increase in w, though it is noted that magnitude of the downward w
greatly exceeds that of any upward w. This pattern is also seen in phase-plots for other
positions (not shown).
In the case of Mason’s numerical simulations over flat ground, the secondary vortex
did not form until x/D ~ 1.5, as in the current study. Mason also reported lifting of the
secondary vortex by the primary vortex from x/D ~ 2.0, which is consistent with the increase
in height of the reduction region as x increases.

1144
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

Fig. 5 u(t)-w(t) Phase Plot at x/D = 1.5, z/D = 0.02 and z/D = 0.2.

Fig. 6 u(t)-v(t) Phase Plot at z/D = 0.2

x/D Maximum z/D at


which u(t) dip is seen
1.0 Not seen
1.5 0.03
2.0 0.15
(though weak above 0.13)
2.5 0.05
(some weak reduction seen above 0.05)
Table 1: Regions in which the u(t) "dip" is seen

Conclusions

Thunderstorm downbursts have been simulated physically (using a 1m diameter


impinging jet with flap aperture control) and numerically using novel inlet conditions, though
the latter work is in the early stages of development. Comparison of the physical simulation
data with full-scale data, specifically the Texas Rear Flank Downdraft event, has been
performed using direct comparison of non-dimensional velocity time-series and wavelet

1145
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

analysis. These analyses show that the physical simulations are successfully capturing the
main features of a downburst.
20
x/D = 1.5 z/D = 0.01
x/D = 1.5 z/D = 0.02
15 x/D = 1.5 z/D = 0.03
x/D = 1.5 z/D = 0.04
u(t*) (ms )
-1

10

0
0 5 10 15 25 20
30 35 40 45
t*
Fig. 7 u(t) at x/D = 1.5, z/D = 0.01, 0.02 and 0.03. Vertical lines mark the dip sub-
section.

5
x/D = 1.5 z/D = 0.01
Temporal midpoint
x/D = 1.5 z/D = 0.02
Temporal midpoint
w(t*) (ms )
-1

x/D = 1.5 z/D = 0.03


0 Temporal midpoint
x/D = 1.5 z/D = 0.04
Temporal midpoint

-5
0 5 10 15 20
-1
u(t*) (ms )
Fig. 8 Phase-plots for the dip sub-section indicated in Fig. 7. An additional separation of 0.5ms-1
has been added in the w direction for clarity.

A phase-plot analysis of the physical simulation data shows evidence of both a


primary, ring vortex and a preceding, weaker, secondary vortex. The existence of this
secondary vortex has previously been indicated by the results of other numerical simulation
studies.
Early results from the numerical simulations show consistency between the physical
and numerical simulations. These numerical simulations are intended to allow the
examination of cases which cannot be simulated physically, and their ability to reproduce the
physical simulations is encouraging.
References

1146
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

BÜSSOW, R. 2007. An algorithm for the continuous Morlet wavelet transform. Mechanical
Systems and Signal Processing, 21 (8), 2970-2979,
http://dx.doi.org/10.1016/j.ymssp.2007.06.001.
CHAY, M. T. & LETCHFORD, C. W. 2002a. Pressure distributions on a cube in a simulated
thunderstorm downburst - Part A: stationary downburst observations. Journal of Wind
Engineering and Industrial Aerodynamics, 90 (7), 711-732.
CHAY, M. T. & LETCHFORD, C. W. 2002b. Pressure distributions on a cube in a simulated
thunderstorm downburst - Part B: moving downburst observations. Journal of Wind
Engineering and Industrial Aerodynamics, 90 (7), 733-753.
CHEN, L. & LETCHFORD, C. W. 2004. Parametric study on the along-wind response of the
CAARC building to downbursts in the time domain. Journal of Wind Engineering and
Industrial Aerodynamics, 92 (9), 703-724.
CHOI, E. C. C. 2004. Field measurement and experimental study of wind speed profile during
thunderstorms. Journal of Wind Engineering and Industrial Aerodynamics, 92, 275-
290, 10.1016/j.jweia.2003.12.001.
FUJITA, T. T. 1971. Proposed characterization of tornadoes and hurricaines by area and
intensity. University of Chicago.
FUJITA, T. T. 1981. Tornadoes and downbursts in the context of generalized planetary
scales. Journal of the Atmospheric Sciences, 38 (8), 1511-1534.
FUJITA, T. T. 1985. Downburst: Microburst and Macroburst, Chicago, Illinois, University of
Chicago Press.
FUJITA, T. T. 1990. Downbursts: Meteorlogical features and wind field characteristics.
Journal of Wind Engineering and Industrial Aerodynamics, 36, 75-86.
GAST, K. D. & SCHROEDER, J. L. 2003. Supercell rear-flank downdraft as sampled in the
2002 thunderstorm outflow experiment. 11th International Conference on Wind
Engineering. Lubbock, Texax: ICWEIA.
HAINES, M., SINGH, N., STERLING, M. & TAYLOR, I. 2013. The simulation of non-
synoptic effects with applications to wind loading using CFD. European-African
Conference on Wind Engineering. Cambridge, UK.
HJELMFELT, M. R. 1988. Structure and life cycle of microburst outflows observed in
Colorado. Journal of Applied Meteorology, 27 (August), 900-927.
HOLMES, J. D., HANGAN, H. M., SCHROEDER, J. L., LETCHFORD, C. W. & ORWIG,
K. D. 2008. A forensic study of the Lubbock-Reese downdraft of 2002. Wind and
Structures, 11 (2), 137-152.
HOLMES, J. D. & OLIVER, S. E. 2000. An empirical model of a downburst. Engineering
Structures, 22, 1167-1172.
KIM, J. & HANGAN, H. 2007. Numerical simulations of impinging jets with application to
downbursts. Journal of Wind Engineering and Industrial Aerodynamics, 95, 279-298,
doi:10.1016/j.jweia.2006.07.002.
LETCHFORD, C. W., MANS, C. & CHAY, M. T. 2002. Thunderstorms - their importance in
wind engineering, a case for the next generation wind tunnel. Journal of Wind
Engineering and Industrial Aerodynamics, 90, 1415-1433.
LIN, W. E. 2010. Validation of a novel downdraft outflow simulator: a slot jet wind tunnel.
The University of Western Ontario.
LIN, W. E., ORF, L. G., SAVORY, E. & NOVACCO, C. 2007. Proposed large-scale
modelling of the transient features of a downburst outflow. Wind and Structures, 10
(4), 315-346.
LIN, W. E. & SAVORY, E. 2006. Large-scale quasi steady modelling of a downburst outflow
using a slot jet. Wind and Structures, 9 (6), 419-440.

1147
Proc. of the 8th Asia-Pacific Conference on Wind Engineering (APCWE-VIII)

MASON, M. S. 2003. Pulsed jet simulation of thunderstorm downbursts. MSc Thesis, Texas
Tech University.
MASON, M. S., LETCHFORD, C. W. & JAMES, D. L. 2005. Pulsed wall jet simulation of a
stationary thunderstorm downburst, Part A: Physical structure and flow field
charaterisation. Journal of Wind Engineering and Industrial Aerodynamics, 93, 557-
580.
MASON, M. S., WOOD, G. S. & FLETCHER, D. F. 2009. Numerical simulation of
downburst winds. Journal of Wind Engineering and Industrial Aerodynamics, 97,
523-539.
MCCONVILLE, A. C. 2008. The Physical Simulation of Thunderstorm Downbursts.
University of Birmingham.
MCCONVILLE, A. C., STERLING, M. & BAKER, C. J. 2009. The physical simulation of
thunderstorm downbursts using an impinging jet. Wind and Structures, 12 (2), 133-
149.
ORF, L. G., ANDERSON, J. R. & STRAKA, J. M. 1996. A three-dimensional numerical
analysis of colliding microburst outflow dynamics. Journal of the Atmospheric
Sciences, 53 (17), 2490-2511.
ORWIG, K. D. & SCHROEDER, J. L. 2007. Near-surface wind characteristics of extreme
thunderstorm outflows. Journal of Wind Engineering and Industrial Aerodynamics, 95
(7), 565-584, http://dx.doi.org/10.1016/j.jweia.2006.12.002.
PROCTER, F. H. 1989. Numerical simulations of an isolated microburst. Part II: sensitivity
experiments. Journal of the Atmospheric Sciences, 46 (14), 2143-2165.
SELVAM, R. P. & HOLMES, J. D. 1992. Numerical simulation of thunderstorm downdrafts.
Journal of Wind Engineering and Industrial Aerodynamics, 44 (4), 2817-2825.
SINGH, N. K. 2012. Large Eddy Simulation of acoustic propagation in turbulent flow
through ducts and mufflers. PhD, University of Hull, UK.
STERLING, M., BAKER, C. J., HAINES, M. & QUINN, A. D. 2011. Scaling a thunderstorm
downburst simulator. In: 13th International Conference on Wind Engineering, 2011
Amsterdam, The Netherlands. ICWE, RAI.
TFI. 2013. Cobra Probe [Online]. Victoria, Australia: Turbulent Flow Instrumentation Pty
Ltd. Available: http://www.turbulentflow.com.au/Downloads/Flyer_CobraProbe.pdf
[Accessed 29th August 2013].
TORRENCE, C. & COMPO, G. P. 1998. A Practical Guide to Wavelet Analysis. Bulletin of
the American Meteorological Society, 79 (1), 61-78, 10.1175/1520-
0477(1998)079<0061:apgtwa>2.0.co;2.
WOOD, G. A., KWOK, K. C. S., MOTTERAM, N. A. & FLETCHER, D. A. 2001. Physical
and numerical modelling of thunderstorm downbursts. Journal of Wind Engineering
and Industrial Aerodynamics, 89 (6), 535-335.
XU, Z. & HANGAN, H. 2008. Scale, boundary and inlet condition effects on impinging jets.
Journal of Wind Engineering and Industrial Aerodynamics, 96, 2383-2402,
doi:10.1016/j.jweia.2008.04.002.
ZHOU, X., WANG, F. & LIU, C. 2012. Wind pressure on a solar updraft tower in a simulated
stationary thunderstorm downburst. Wind and Structures, 15 (4), 331-343.

1148