You are on page 1of 8

Journal of Wind Engineering

and Industrial Aerodynamics 79 (1999) 201—208

Short Communication
Further note on directional assessment of extreme winds
for design
Nicholas J. Cook *, Craig A. Miller
Wind Engineering Consultant, 10 Arretine Close, St. Albans AL3 4JL, UK
 Risk Management Solutions Inc., Peninsular House, 30-36 Monument Street, London, UK

Received 3 January 1998; accepted 11 March 1998

Abstract

We review a method for the directional analysis of extreme winds, proposed in this journal in
1983, and extend it to include correction of anemograph exposure by direction. We correct an
error in a critical parameter of the previous work and discuss the possible consequences of this
error for design in the UK. We present new evidence for a geographic variation of the
correlation of extreme wind speeds between adjacent directional sectors but can offer only
a tentative explanation for this observation.  1999 Elsevier Science Ltd. All rights reserved.

1. Introduction
In 1983 the first analysis methodology for directional and seasonal analysis of
extreme wind speeds was reported in this Journal [1] by the first author of this note
(NJC). This work used a new analysis procedure [2] that improved analysis accuracy
by utilising the sectorial extremes in individual storms and fitting them to
a Fisher—Tippett type 1 (FT1) model for dynamic pressure, rather than wind speed. It
led to a re-analysis of the UK extreme wind climate [3], in terms of the “base map” of
50 yr return wind speed and associated direction and seasonal climate factors which
are currently used in UK design Standards [4,5].
In 1996—1997, the UK Meteorological Office funded an investigation to determine
if current engineering models for assessing the exposure of buildings are a suitable
basis for assessing the exposure characteristics of UK anemographs, and to revise the
“base map” and associated climate factors. A preliminary account of this work [6],
executed by the second author (CAM) and supervised by NJC and R. H. Barnard,

* Corresponding author. E-mail: nicholas—cook@compuserve.com.


 I.e. looking over his shoulder and making helpful comments.

0167-6105/99/$ — see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 9 8 ) 0 0 1 0 9 - 3
202 N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208

was given at 2EACWE, in Genoa, June 1997, and has been reported in this Journal
[7]. Further, more detailed, papers are in preparation.
The purpose of this note is to update and correct the published record of the
directional methodology introduced by the original 1983 Short Communication [1].

2. Methodology

In 1983, the use of dynamic pressure in place of wind speed in the storm analysis
methodology was a novel and unproven concept, so the analysis was performed for
both variables at 50 sites distributed across the UK. The directional analysis was
performed on the raw anemograph data without any corrections for exposure,
because practical correction procedures, based on the Deaves and Harris ABL model
[8—10] for assessing site exposure in heterogeneous terrain, were still under develop-
ment [11]. It was assumed that the directional climate factors would be constant
across the UK so that variations in exposure by direction should be cancelled out by
averaging the directional characteristics for the 50 sites. However, the values contribu-
ting to the “base map” of 50 yr return wind speed, irrespective of direction, are affected
by non-uniform site exposure by direction. In addition to calibration errors, there is
the likelihood that differences in exposure by direction may lead to selection of the
“wrong” sectorial extreme in some storms.
The 1983 note [1] introduced a method for assessing the difference between risk in
a directional sector and the overall risk from all sectors that accounted for the
unknown degree of correlation between adjacent sectors in each storm. This was
expressed in terms of the factor, k , defined as:
O
dynamic pressure with 0.02 annual risk overall, equally distributed by sector
k"
O dynamic pressure with 0.02 annual risk in 30° wide sector
(1)

Determining k requires transforming the original wind data by dividing it by the


O
directional climate factors to give an equal risk in each sector, before selecting each
storm extreme irrespective of direction. This procedure is also sensitive to the addi-
tional direction variation caused by non-uniform site exposure.
In the new work we seek to eliminate errors associated with exposure by correcting
the raw data for exposure before reanalysing the storm extremes by 30° wide
directional sectors and irrespective of direction, using the FT1 model for dynamic
pressure only. However, we report the results in terms of wind speed, since wind speed
remains the basis for design procedures. We represent the observed wind speed, » , in
F
a given sector by the expression

» "S k » , (2)
F F F 
where, S is the directional climate factor for 50 yr return wind speed in sector, k is the
F F
gain factor for site exposure in that direction, and » is the base, 50 yr return wind

speed, irrespective of direction.
N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208 203

We calculated the site gain factor for exposure, k , in two stages to account for:
F
(1) Changes in ground roughness in the 30° wide sector for 200 km upwind of the
site using the BRE program STRONGBLOW [12], supported by a database of UK
ground roughness, and
(2) Local topographic effects using the program MsMicro, which is a three-
dimensional implementation of Jackson and Hunt’s method [13] developed for the
Atmospheric Environment Service of Canada by Taylor et al. [14], in conjunction
with Ordinance Survey’s ¸andform PANORAMA digital terrain database.
We report only those aspects relevant to the 1983 note [1] here. For an assessment
of the effectiveness of the exposure corrections, see Ref. [6] or [7]. The seasonal
characteristics are unaffected by directional variations in exposure and do not need to
be reconsidered.

3. Climatic directional factor for sectorial wind speeds, Sh

Although the analysis was performed in terms of dynamic pressure, we choose to


present the direction factor on sectorial wind speed, S , which is equivalent to c in
F F
Ref. [1].
Analysis was performed in two phases: on 43 of the original “A” sites, which had
been selected as being the best exposed, and on a further 17 poorer “B” sites. In both
phases, CAM assessed the exposure factors of each site without prior knowledge of the
corresponding site observations. Table 1 gives the values of the climatic direction
factor averaged for the sites, S , and the standard deviation from the average, p .
F 1

Table 1
Value and standard deviation of directional climate factor S
F
Direction 1983 analysis, A sites New analysis, A sites New analysis, A and
B sites

°True S p S p S p
F 1 F 1 F 1
0 0.722 0.105 0.747 0.090 0.753 0.091
30 0.673 0.099 0.685 0.084 0.681 0.084
60 0.681 0.095 0.687 0.094 0.690 0.097
90 0.692 0.098 0.700 0.090 0.695 0.092
120 0.683 0.104 0.693 0.100 0.696 0.105
150 0.742 0.098 0.755 0.098 0.760 0.097
180 0.794 0.104 0.814 0.092 0.809 0.101
210 0.870 0.087 0.880 0.077 0.874 0.082
240 0.951 0.049 0.948 0.055 0.942 0.057
270 0.929 0.051 0.934 0.051 0.938 0.056
300 0.851 0.096 0.858 0.096 0.860 0.097
330 0.763 0.103 0.775 0.085 0.781 0.084
All 0.093 0.086 0.088
204 N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208

Fig. 1. Comparison of original and new climatic direction factors.

As may be seen in Fig. 1, the climatic directions factors, S , obtained after exposure
F
correction from the Class A stations are virtually identical to the original values.
Including the additional Class B stations makes no significant difference, suggesting
that the 1983 procedure of averaging the 50 sites to cancel out exposure variations,
assuming no geographic variation, was sound. Exposure correction removes this need
to average and enables the assumption of no geographic variation to be tested. We did
this by sub-dividing the exposure-corrected data into geographical sets: firstly with
three north—south sets — Scotland, northern and southern England, then with two
east—west sets; and found no significant differences between the sets.
Comparing the standard deviations around the mean values in Fig. 2 shows that
the variance associated with the Class A stations is slightly lower than for the original
set, but that inclusion of the poorer Class B stations reduces this improvement. This
indicates that there are still components contributing to the site exposure of the
poorer Class B sites that have not been accounted for. In our judgement, the principal
missing component is likely to be direct sheltering from the wakes of upwind
obstructions, as addressed by Taylor and Salmon [15].

4. Ratio of overall risk to sectorial risk values

4.1. Overall values

A novel feature of the 1983 note [1] was the method for determining the factor
k of Eq. (1). Here we choose to present the equivalent factor on wind speed, k ,
O 4
N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208 205

Fig. 2. Comparison of standard deviations for original and new climatic direction factors.

defined as
windspeed with 0.02 annual risk overall, equally distributed by sector
k "
4 windspeed with 0.02 annual risk in 30° wide sector
(3)
This was determined in exactly the same manner as before, but from the extreme wind
speeds predicted by the analysis of dynamic pressures. In this context, k +k ,
4 O
differing only that the average of the square root is not quite the same as the square
root of the average, leading to the expectation of a value of k "1.12 from the 1983
4
note [1].
CAM was surprised to discover that the analysis of exposure corrected values gave
a value of k "1.185, much higher than anticipated. Up to this point CAM had
4
naturally assumed that NJC was infallible, but decided anyway to repeat the original
analysis on the raw data to confirm the 1983 value, and obtained a revised value of
k +raw,"1.202. This confirmed that the original value of k "1.24 given in the 1983
4 O
note is wrong. The closeness to the revised value for k +raw, to original value of k ,
4 O
the analysis differing only in the number of stations used, suggests that this error was
due to confusion between values obtained from the two original analyses: one for wind
speed and the other for dynamic pressure.
In the 1983 procedure for determining the numerator in Eq. (3) for k , the values of
4
wind speed are divided by the corresponding climatic direction factors to give an
equal risk of exceedance in each 30° wide sector. If the extreme wind speeds in
adjacent sectors in each storm were independent, the overall risk would be equivalent
to twelve trials of the sectorial risk. Eq. (3) would become

 
5!ln (!ln (1!0.02)#ln (12)) 
k " "1.13 (4)
4 5!ln (!ln (1!0.02))
206 N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208

from the FT1 model for dynamic pressure with P "5 [5,11]. The 1983 value of 1.12,
being lower, would indicate that the observed correlation between adjacent sectors
reduces the risk slightly, which is counter-intuitive. The new value of 1.185, being
higher, correctly indicates that correlation increases the risk.

4.2. Geographic variation

As already noted, application of exposure corrections enables the geographic


variation to be tested. Values of k for the individual stations have been plotted in
4
Fig. 3 as contours on a map of the UK, in which Class A stations have been marked

Fig. 3. Values of k for class A and B stations.


4
N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208 207

by circle symbols and Class B stations by triangle symbols, respectively. Whilst there
appears to be a high degree of random variation, a clear trend to higher values
towards the west is evident. The highest values appear to be located mostly adjacent
to straits — between the Orkneys, between SW Scotland and NE Ireland and the
Bristol Channel. We interpret higher values to mean stronger correlation between
adjacent sectors. Passage of the depressions, generally from west to east, appears to
weaken this correlation. A possible mechanism for the highest correlation in straits is
preferential “steering” of the depressions by the landmasses.

4.3. Impact on design

In correcting the published record, we must examine the effect the error may
have had on engineering design from 1983. The Designer’s guide to wind loading of
building structures, Part 1 [11] quotes k "1.24 in the text, but uses the value 1.12
4
in the table of S . Unfortunately, the erratum published in Part 2 of the Guide
F
changes the text value and not the table. How directional risk should be incorporated
in design is still a matter of debate, but it is generally accepted that it should relate to
the overall risk. The drafting committee for the current UK Standard [5] took the
pragmatic view that, use of directional factors should not alter the highest design
loads obtained irrespective of direction, so they scaled the original direction factors
to be unity in the direction of the strongest winds, h"240° True. These design values
are compared in Table 2, which shows that the error has not compromised design
safety in the UK. It is clear that the design risk given by the UK Standard lies
somewhere between an annual risk of 0.02 in sector and an annual risk of 0.02 overall,
but instead of lying about halfway between these two risks, it lies closer to the risk in
sector.

Table 2
Climatic direction factors used in design

Direction BS6399 1983 New


h S k S k S
 4 F 4 F
0° 0.78 0.809 0.885
30° 0.73 0.754 0.812
60° 0.73 0.763 0.814
90° 0.74 0.775 0.830
120° 0.73 0.765 0.821
150° 0.80 0.831 0.895
180° 0.85 0.889 0.965
210° 0.93 0.974 1.043
240° 1.00 1.065 1.123
270° 0.99 1.040 1.107
300° 0.91 0.953 1.017
330° 0.82 0.855 0.918
208 N.J. Cook, C.A. Miller/J. Wind Eng. Ind. Aerodyn. 79 (1999) 201–208

Acknowledgements

This work was financially supported by the UK Meteorological Office and was
executed under the auspices of the University of Hertfordshire — Building Research
Establishment Graduate School of Construction.

References
[1] N.J. Cook, Note on directional and seasonal analysis of extreme winds for design, J. Wind Eng. Ind.
Aerodyn. 12 (1983) 365—372.
[2] N.J. Cook, Towards better estimation of extreme winds, J. Wind Eng. Ind. Aerodyn. 9 (1982) 295—323.
[3] N.J. Cook, M.J. Prior, Extreme wind climate of the United Kingdom, J. Wind Eng. Ind. Aerodyn. 26
(1987) 371—389.
[4] British Standard: Lattice Towers and Masts: Part 1, Code of Practice for Loading, BS8100:1:1986.
British Standards Institution, London, 1986.
[5] British Standard: Loading for Buildings: Part 2, Code of Practice for Wind Loads, BS6399:2:1997.
British Standards Institution, London, 1997.
[6] C.A. Miller, N.J. Cook, R.H. Barnard, Calibration of the exposure of UK anemographs, in: G. Solari
(Ed.), Proc. 2nd European and African Conf. on Wind Eng., Genoa, 22—26 June 1997, Servizi Grafici
Editoriali, Padua, 1997, pp. 171—178.
[7] C.A. Miller, N.J. Cook, R.H. Barnard, Calibration of the exposure of UK anemographs, J. Wind Eng.
Ind. Aerodyn. 74—76 (1998) 153—161.
[8] D.M. Deaves, R.I. Harris, A mathematical model of the structure of strong winds, CIRIA Report 76,
Construction Industry Research and Information Association, London, 1978.
[9] R.I. Harris, D.M. Deaves, The structure of strong winds, Wind Engineering in the Eighties, Proc.
CIRIA Conf. 12/13 November 1980, Paper 4, Construction Industry Research and Information
Association, London, 1981.
[10] D.M. Deaves, Computations of wind flow over changes in surface roughness, J. Wind Eng. Ind.
Aerodyn. 7 (1981) 65—94.
[11] N.J. Cook, The Designer’s Guide to Wind Loading of Building Structures, Part 1, Butterworths,
London, 1985.
[12] N.J. Cook, B.W. Smith, M.V. Huband, BRE Program STRONGBLOW: User’s Manual, BRE
Microcomputer Package, Building Research Establishment, Garston, 1985.
[13] P.S. Jackson, J.C.R. Hunt, Turbulent wind flow over a low hill, Quant. J. Roy. Meteorol. Soc. 101
(1975) 929—955.
[14] P.A. Taylor, J.L. Walmsley, J.R. Salmon, A simple model of neutrally stratified boundary-layer flow
over real terrain incorporating wave-number dependent scaling, Boundary-Layer Meteorol. 26 (1983)
169—189.
[15] P.A. Taylor, J.R. Salmon, A model for the correction of surface data for sheltering by upwind
obstacles, J. Appl. Meteorol. 32 (1993) 1683—1694.

You might also like