See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/265216364

Computational Methods for Estimating Wind

Local Effects as Part of Wind Risk Assessment

Conference Paper · September 2005

CITATIONS READS

0 106

2 authors:

Xun Guo Lin Krishna Nadimpalli

Civil Aviation Safety Authority Geoscience Australia
40 PUBLICATIONS 144 CITATIONS 11 PUBLICATIONS 40 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Signal processing View project

Output structural controllability View project

All content following this page was uploaded by Xun Guo Lin on 02 September 2014.

The user has requested enhancement of the downloaded file.

 Computational Methods for Estimating Wind Local Effects as
Part of Wind Risk Assessment

*Xun Guo Lin1) and Krishna Nadimpalli2)

1)
CSIRO Mathematical and Information Sciences, Private Bag 10, Clayton South,
VIC 3169, Australia, E-mail: xunguo.lin@csiro.au
2)
Geoscience Australia, GPO Box 378, ACT 2601, Australia,
E-mail: krishna.nadimpalli@ga.gov.au

ABSTRACT
Wind risk assessment is to estimate the possible building damage due to severe winds.
Local terrain, topographic and shielding conditions can have major effects on local wind speeds.
Australian Wind Loading Standards have given mathematical formulae to estimate these local
wind effects, called multipliers. In order to apply these formulae on a large area such as a
metropolitan city, effective and efficient computational methods have to be developed. This
paper describes novel computational methods developed to estimate local wind multipliers. It
uses satellite remote sensing techniques, geographical information systems (GIS) and image
analysis softwares to effectively estimate surface roughness, terrain and shielding multipliers on
a very fine grid (25m) in eight cardinal wind directions. The digital elevation data provide
detailed local topographic characteristics. Efficient computational algorithms have been
developed to calculate the topographic wind multipliers in eight wind directions on each grid.
These multipliers will be used in the risk assessment of severe winds. A case study of
metropolitan Perth is also given.

BACKGROUND OF HAZARD LOSS ASSESSMENT

The basic quantitative method of hazard (natural or man-made) loss assessment includes
hazard modelling, element at risk, damage modelling, and cost modelling. See the diagram in
Fig. 1. The hazard modelling in our case is the modelling of extreme return wind speeds. The
element at risk is where and what the buildings are and possibly how many residents live within
these buildings. The damage modelling in our case is to develop the building vulnerability with
respect to wind speeds, that is, damage percentages at different speeds. The economic loss
modelling is to estimate direct (and maybe indirect) and tangible (and maybe intangible) losses,
for example, rebuild/repair cost per square-metre for different types of buildings. Finally, the
loss (cost) is aggregated from each building to a whole city or a region to give the total loss
(cost) with respect to a particular wind speed. This type of quantitative modelling is also called
consequence modelling.
In our case, because the damage to buildings is caused by extreme winds, it can be
modelled effectively by statistical modelling techniques. Hence the element of wind hazard
modelling can be further decomposed into several elements by using a statistical approach as
shown in Fig. 2. It includes wind data acquisition; local wind effect calculation; adjustment of
wind speed data to standard condition of 10m height in open terrain; extreme value distribution
fitting; return period speeds estimation, and finally return speeds interpolation to generate wind
hazard maps. The rest of this paper will describe the element of local wind effect calculation.
Developments of other elements can be found in Lin (2004), Lin and Courtney (2004) and Lin
(2005).

Hazard Element Damage Economic

Modelling at Risk Modelling Loss Modelling $
Fig. 1. Quantitative methodology of hazard loss assessment

Wind data
acquisition &
pre-processing

Return period
Wind data Distribution
wind speeds
adjustment fitting
Local wind effect estimation
calculation
•Terrain/height Return speeds
•Shielding interpolation
•Hill-shape
Hazard Maps

Fig. 2. Elements of wind hazard modelling

INTRODUCTION TO LOCAL WIND EFFECTS

The impact of severe wind varies considerably between structures at various locations,
because of the geographic terrain, the height of the structures concerned, the surrounding
structures and topographic features. These site wind exposure and speed modifications can be
numerically described by so-called wind multipliers. These multipliers give quantitative
measures of local wind conditions relative to the regional wind speed (defined as open terrain at
10m height) at each location. There are three wind multipliers, named terrain/height multiplier
(Mz), shielding multiplier (Ms) and topographic (also called hill-shape) multiplier (Mh). The
relationship between the regional wind speed (VR) and the local (site) wind speed (Vsite) and the
local wind multipliers is
 Vsite VR M z M s M h (1)

This is applied to each wind direction at each location.

Formulas to estimate these wind multipliers for a given location are given in AS/NZS
1170.2 (2002) (referred as the Standard in this paper). However, efficient and effective
computational methods to estimate these wind multipliers across a larger area with high
resolution need to be developed. Satellite remote sensing techniques, geographic information
systems (GIS), image analysis software and a digital elevation dataset have been used to apply
these formulas to calculate the three wind multipliers. The proposed techniques will be
demonstrated in a case study in metropolitan Perth as part of wind risk assessment.

AUSTRALIAN WIND STANDARD

Australian Wind Standard provides formulas to calculate the three local wind multipliers in
document AS/NZS 1170.2 (2002). The rest of this section will give a brief description on how
to estimate each of these multipliers.

Terrain/height wind multipliers

The terrain/height multiplier (Mz) is dependent on the terrain roughness as well as the
height of a building/structure concerned. There are four terrain categories defined in the
Standard:
Category 1 í exposed open terrain with few or no obstructions including smooth water or
flat snow surfaces;
Category 2 í rough water surfaces, uncut grassland, isolated trees from 1.5m to 10m;
Category 3 í numerous closely spaced obstructions 3m to 5m high such as suburban
housing;
Category 4 í numerous large, high (10m to 30m) and closely spaced obstructions such as
large city centres and well-developed industrial complexes.
Each category has a defined terrain roughness length. There are some classes of terrain,
such as forests, sitting in between two categories and their roughness lengths are also given in
AS/NZS 1170.2 Supp 1 (2002).
Table 1 gives Mz values in each terrain category for various building/structure heights up to
500 m for regions A1 to A7 and B specified in the Standard, which include most Australian
capital cities except Darwin.
When considering a wind approaching direction across ground with changes in terrain
categories, it is necessary to take an averaging process with a formula given in the Standard.
The distance over which to average is also given which is dependent on the building/structure
height concerned.

Shielding wind multipliers

The shielding wind multiplier (Ms) has a value between 0.7 and 1 according to the
Standard. A value of 0.85 is typically used for design of Australian suburban housing. If the
ground gradient upwind is greater than 0.2 (20% or 11.31o), then the effect of shielding is not
applicable or there is no shielding effect, so it has the value of 1 for this wind direction at the
location. When the ground slope is less than 0.2, Ms can be estimated by the Standard using
information of buildings such as heights and breadths within a 45o upwind sector of radius 20
times the building height, referred to as the shielding zone in this paper.

Table 1. Mz for different building heights, from Table 4.1 (A) in AS/NZS 1170.2 (2002)

Height (z) Terrain/height multiplier (Mz)

m Category 1 Category 2 Category 3 Category 4
”3 0.99 0.91 0.83 0.75
… … … … …
10 1.12 1.00 0.83 0.75
… … … … …
50 1.25 1.18 1.07 0.90
… … … … …
500 1.38 1.37 1.35 1.31

Topographic wind multipliers

Standing on an uphill slope will normally feel windy comparing to standing in a flat area.
This is called the topographic (or hill-shape) wind effect. In wind engineering, it is named the
topographic (or hill-shape) multiplier Mh. It can be estimated using the following formula from
the Standard:

­
°
° 1.0 for H /( 2 Lu )  0.05
°° § H ·§ x ·
Mh ®1  ¨¨ ¸¸¨1  ¸ for 0.05 d H /( 2 Lu )  0.45 or within the local topo zone when H /( 2 Lu ) ! 0.45
3. 5 z  L  ¨ L ¸
° © 1 ¹© 2 ¹
° § x ·
° 1  0.71¨1 - ¸ for H /( 2 Lu ) ! 0.45 and within the separation zone
°¯ ¨ L ¸
© 2 ¹ (2)

where
H vertical distance from base to crest
x  horizontal distance upwind or downwind from structure to crest
z  height of the structure above the local ground level
Lu  horizontal distance from crest to 1/2 H below crest
L1  the greater of 0.36 Lu and 0.4 H
L2  local topographic zone, to be 4*L1 for upwind, and for downwind is also 4*L1 for
the cases of ridges or 10*L1 for escarpments.

A “separation zone” starts at the crest with the size of H/4. It is only defined for the case of
slopes greater than 0.45.
 Eq.(2) specifies that slopes below 0.05 are assigned a Mh of 1 and slopes equal and greater
than 0.45 are assigned a maximum Mh of 1.71 within the local topographic zone. For the slopes
between 0.05 (2.9o) and 0.45 (24.2o), Table 2 gives Mh values for some different slopes.
It can be seen that to calculate these multipliers for a large region there will be a large
amount of tedious computations involved requiring details of terrain and building information
which may not be available on a large scale. Hence, effective and efficient computing
techniques will be necessary which will be described in the next section.

Table 2. Hill-shape multiplier at crest (|x| = 0) when z = 0, from Table 4.4 of AS/NZS 1170.2 (2002)

Hill Slope Hill Slope (in degrees) Mh

< 0.05 < 2.9 1.0
0.05 2.9 1.08
0.10 5.7 1.16
0.20 11.3 1.32
0.30 16.7 1.48
t 0.45 t 24.2 1.71

COMPUTATIONAL METHODS DEVELOPED

To calculate Mz and Ms, Remote Sensing techniques were used effectively to extract
necessary terrain details automatically. Next, GIS techniques were used to efficiently
manipulate these information to estimate Mz and Ms. Digital elevation data provided grid
heights which were used to calculate the ground slopes on grid in 8 cardinal wind directions.
Scientific computing software allowing matrix operations was used to calculate Mh efficiently.
An overview of the methodology to develop the three multipliers for 8 cardinal wind directions
is shown in Fig. 3.

Terrain classification mapping

The terrain classification mapping is to identify terrain categories automatically. This was
done using satellite remote sensing techniques for the region. A list of terrain features within
the region such as grassland, forests and suburbs was decided according to four terrain
categories. Next, digital samples of these terrain features in terms of their spectral reflectance
were identified and then they were used to classify the region using maximum likelihood
algorithms. Then, an image software (for example, ERDAS Imagine) was used to match up
these features automatically over the whole region. Note that this approach has some
limitations, for example, concrete building roofs and airport runways have similar reflectance so
they were classified as the same class. Therefore, critical features such as runways and city
buildings were mapped separately and overlayed.
 Fig. 3. Diagram of methodology used in estimation of local wind modification multipliers

Averaging method
When there are different terrain classifications upwind of a structure of primary interest, the
method of averaging described in the Standard has been applied. To achieve this in a GIS
environment, a moving window/kernel was established that contains an array of weighting
factors (Lillesand and Kiefer, 1994). The kernel was moved throughout the original image, and
the value at the centre of the kernel in output image was obtained by averaging all the pixels
considered for averaging. Eight kernels were created for 8 cardinal wind directions to average
the terrain multiplier values.

Computation of terrain/height wind multipliers

The terrain map interpreted using satellite remote sensing was reclassified into the
terrain/height multiplier. Then the wind directional kernels were applied to estimate the terrain
multiplier for all 8 cardinal wind directions. If we know the exact location of each building and
its height, then the averaging and hence the Mz can be calculated. However, for a large area
such as a metropolitan city, this condition might not be satisfied. Hence, nominated building
heights were specified with examples given in the Case Study later.

Computation of shielding wind multipliers

The shielding multiplier (Ms) of a structure depends on the number of buildings upwind in
a shielding zone with at least the same height as the structure concerned. Again, since detailed
information of each building, such as its height, width and orientation, may not be available for
a large area, some assumptions have been made. Particularly, nominal values of Mz for different
housing classifications have been assumed which were within the range given by the Standard.
These values have been verified using limited available building details (from building surveys)
or by simulation. Then, the initial Mz were calculated based on the terrain classification map
using the nominated multiplier value for each housing type such as suburb and city. Finally, an
averaging process was undertaken for each wind direction. The shape of the averaging area was
the same as the shielding zone and its radius compatible to the housing type (that is, 20 times the
nominated height of the housing type).

Computation of topographic wind multipliers

The topographic wind multiplier Mh was calculated by applying Eq.(2) using grid elevation
data. However, the calculation could not be based on adjacent grid elevation, rather it has to be
based on adjacent hills and valleys. A non-recursive algorithm has been developed to pair each
pair of valley and hill in order to calculate the ground slope. Extra care has to be taken when
two or more hills or valleys are close together. The developed algorithm was applied in each
direction at a time, with each direction represents a row or a column or a diagonal of the
elevation matrix. An effective and efficient method has been developed to deal with the 8
cardinal wind directions by transposing, flipping and rotating the elevation matrix.

Single directional approach In a single directional approach, all the hills and valleys were
identified and paired with each other in the order of valley-hill, valley-hill, and so on. Next, the
ground slope was calculated for each pair of valley-hill and then this hill and the next valley.
After that, the length of the local topographic zone (L2) was determined and the grid values of
Mh were calculated around this hill. The algorithm then processed the next pair of valley-hill.
When multiple valleys and hills were close by, their local topographic zones might be
overlapping. So the final Mh was the maximum of all multipliers from each overlapping zone.

Rotating data method A rotating data method was developed to use the single directional
approach to achieve multi-directional computation by transposing, flipping or rotating the
dataset. In particular, the method of 45o matrix rotation has been developed to calculate the Mh
for 4 diagonal winds, that is, northeast to/from southwest and northwest to/from southeast. The
method of 45o rotating will be described next.
Assume we have a 4 by 4 matrix representing an elevation grid with the wind direction
indicated by arrows from southwest to northeast.

1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 16
 The original matrix can be augmented diagonally to make it having the same number of
rows elements on each diagonal ‘column’, see below. Note that, the value of L will be taken as
the minimum of the grid value and H as the maximum. Hence this augmentation will not
introduce extra valley-hill pairs.

1 2 3 4 L L L
H 5 6 7 8 L L
H H 9 10 11 12 L
H H H 13 14 15 16

Now, this augmented matrix is 4 by 7 and can be easily rearranged to the following normal
form with its calculation direction given by arrows.

1 2 3 4 L L L
H 5 6 7 8 L L
H H 9 10 11 12 L
H H H 13 14 15 16

In general, a n by m matrix will be augmented to a n by m+(ní1) matrix. Note that, the

grid size for the augmented matrix has been changed to S¥2 from the original S.
The other diagonal direction winds can be dealt with by similar methods.

CASE STUDY

Background
Extreme wind is one of the major natural hazards experienced in Perth. These extreme
winds are generally produced by cold fronts and not by cyclones (see, for example, Lin and
Courtney, 2004) or thunderstorms related downbursts. The wind study area covers Perth City
and its surroundings including Rottnest Island. The geographic extent of the area in the
Universal Transverse Mercator (UTM) coordinate system lies between 354500 to 416475
longitude and 6437995 to 6498970 latitude, covering an area of about 3767 square km. See the
red box at the top left in Fig. 4.

Perth terrain classification map

In order to estimate the terrain/height multiplier and the shielding multiplier, a terrain
classification map described earlier needed to be developed. Terrain classes from AS/NZS
1170.2 Supp 1 (2002) were used to classify the Perth metropolitan area. The following classes
were relevant to our wind hazard assessment.
x City buildings
x Forests
x Centres of small towns and industrial area
x Suburban and wooded country
x Long grass with few trees
x Air fields and uncut grass
x Water
x Sandy beaches
x Cut grass

Fig. 4. The setting of the wind study area in metropolitan Perth

 Fig. 5. LANDSAT satellite database of Bands 432 of metropolitan Perth

To map the terrain of the study area, Landsat Thematic Mapper data was used as input. It
has 25 meters spatial resolution and has 6 frequency bands. (Further information about the data
specifications can be obtained from the Australian Centre for Remote Sensing (ACRES) website
http://www.ga.gov.au/acres/prod_ser/landdata.htm). The satellite database of bands 432 (RGB)
is depicted in Fig. 5. The estimated terrain classification map using image software is shown
in Fig. 6.

Perth terrain/height multipliers

The terrain/height multiplier (Mz) can be estimated by applying Table 1 for a specified
structure using upwind terrain classifications over a design standard specified distance related to
the height of the structure. However, since we do not have data giving the heights of individual
buildings, apart from a small set of surveyed ones, nominated heights have been assumed for
buildings and vegetations in each terrain category.
Table 3 lists the nominated heights with respect to various terrain classifications and the
calculated terrain/height multipliers according to Table 1. Note that some interpolation was
necessary in order to estimate the values of Mz for some terrain classifications that were between
two terrain categories, such as forest.
The averaging process described earlier was to deal with different terrain classifications
upwind of a structure of primary interest. The distance for averaging depends on the height of
the structure. However, since the nominated heights for various terrain classifications in the
Perth wind study area are all less than 50 metres, an averaging distance of 1 km was adopted,
following Table 4.2(A) of AS/NZS 1170.2 (2002). The terrain/height multipliers were
estimated after the averaging process. Fig. 7 shows the Mz for the wind direction from West
to East.

Perth shielding multipliers

Using the practical method described earlier to estimate Ms, an averaging distance of 300 m
was adopted for suburban and industrial area, while 1 km was used for city buildings. These
distances were the radiuses of the shielding zone (the area of averaging) which were 20 times
the nominated building heights given in Table 3. Fig. 8 shows the estimated shielding
multipliers for winds from West to East in the study area.

Fig. 6. Terrain map of Perth wind study area

 Table 3. The nominated height in each terrain category

Terrain category Terrain classification Nominated height in metres Mz

4 City 50 0.9
Forest 20 0.794
Town centre/Industrial 20 0.92
3 Suburb buildings 3-10 0.83
Trees <5 0.89
2 Water/Airport <3 0.91
Cut grass <3 0.92

Fig. 7. Terrain/height multipliers Mz for the wind direction from West to East
 Fig. 8. Shielding multipliers Ms for the study area

Perth topographic features

The topographic features of metropolitan Perth have been captured by a DEM (digital
elevation Model) generated from spot height data at a spacing of 1 minute UTM. The DEM of
metropolitan Perth in a 25m by 25m grid is shown in Fig. 9. It can be seen that the majority
of metropolitan Perth is gently undulating, except the Darling Range Scarp which is more than
200m above mean sea level. Note that, the Gooseberry Hill weather station was built on the
Darling Range Scarp with the purpose of catching the easterly winds.
 Fig. 9. The elevation (in m) of metropolitan Perth in a 25m by 25m grid

Perth topographic multipliers

In the application of Eq.(2), the value of z (height of the structure above the local ground
level) was set to zero. There were three reasons to do this. First, the estimate of a topographic
multiplier is required for every point within the study area of metropolitan Perth, whether it
currently has a building on it or not. So a building height may not currently exist. Secondly,
even in the case of an existing building, in general we do not know the height of the building
without a physical building survey having been done. Thirdly, the Mh value has been found to
be insensitive to the value of building height.
The 8-directional topographic wind multipliers for metropolitan Perth on a 25 by 25 metres
grid have been estimated. Among them, the Mh for West to East is shown in Fig. 10.
 Fig. 10. The Mh of metropolitan Perth from West to East

The combined wind multipliers for Perth

The combined local wind effect (named M3) was obtained by multiplying three individual
multipliers Mz, Ms and Mh together on grid in each direction. Fig. 11 displays the M3 of the
study area with the wind direction from West to East. M3 can be used to convert site wind
speeds recorded to regional speeds and vice versa.

CONCLUSIONS
Computational techniques have been developed to estimate local wind effects effectively
and efficiently. These local wind effects include the terrain/height multipliers, shielding
multipliers and topographic multipliers at each wind direction on a fine grid. The proposed
methods use satellite remote sensing techniques, geographic information systems, digital image
analysis software, digital elevation dataset and scientific computational software. A case study
of metropolitan Perth was given to demonstrate the developed techniques.
 Fig. 11. Multiplier M3 for wind from West to East for the study area

REFERENCES
AS/NZS 1170.2 (2002), Structural Design Actions, Part 2: Wind Actions, Australian/New
Zealand Standard.
AS/NZS 1170.2 Supp 1 (2002), Structural Design Actions ʊ Wind Actions ʊ Commentary,
Australian/New Zealand Standard.
Lillesand, T.M. and Kiefer, R.W. (1994), Remote Sensing and Image Interpretation, John Wiley
& Sons Inc., 3rd Ed., 524-634.
Lin, X. (2004), Probabilistic framework of cyclone risk assessment, 5th Asia-Pacific Industrial
Engineering and Management Systems Conference, Gold Coast, December 12-15.
Lin, X. and Courtney, J. (2004), Statistical spatial analysis of severe wind gusts in Perth,
International Conference on Storms: Storm Science to Disaster Mitigation, Brisbane, July
5-9, 118-119.
Lin, X. (2005), How reliable are estimates of likely extreme wind speeds for metropolitan Perth?
4th International Symposium on Business and Industrial Statistics, Cairns, April 13-16.

View publication stats