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

net/publication/321722213

Determination of cyclonic factor in different wind zones of India

Conference Paper · December 2017

CITATIONS READS
2 436

5 authors, including:

Gaurav Gugliani A. Sarkar

Indian Institute of Technology (Banaras Hindu University) Varanasi Indian Institute of Technology (Banaras Hindu University) Varanasi
13 PUBLICATIONS   84 CITATIONS    37 PUBLICATIONS   349 CITATIONS   

SEE PROFILE SEE PROFILE

Abhay Gupta Sasankasekhar Mandal

Skeleton consultants pvt ltd Banaras Hindu University
4 PUBLICATIONS   2 CITATIONS    33 PUBLICATIONS   102 CITATIONS   

SEE PROFILE SEE PROFILE

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

A Systematic Approach for seismic Evaluation and Retrofit of Reinforced Concrete Buildings in Earthquake Prone Area View project

Wind Energy View project

All content following this page was uploaded by Abhay Gupta on 12 October 2021.

The user has requested enhancement of the downloaded file.

 9th Asia-pacific Conference on Wind Engineering
Auckland, NewZealand
3-7 December 2017

Determination of cyclonic factor in different wind zones of India

G.K. Gugliani1, A. Sarkar2, A. Gupta3, S.Mandal4, Prince5

1,2,5 Department
of Mechanical Engineering, IIT B.H.U., Varanasi 221005, U.P., India
4Department
of Civil Engineering, IIT B.H.U., Varanasi 221005, U.P., India
3Director, Skeleton Consultants Private Limited

Mail id: [email protected], [email protected], [email protected], [email protected],

[email protected]

Abstract distributions, which have infinite upper tails, but rather by reverse
Weibull distributions which, like wind speed in nature, have finite
Cyclone has the highest turbulence intensity compared to other upper tails. This geophysical reasonable approach was also used in
storms in India. Therefore, structures are subjected to greater risk Kaspersky et al. (1997 (a, b)) and Holmes et al. (1997). Sarkar et al.
under cyclonic storms. To account for the enhanced risk, an (2013) compared the above-mentioned distribution and found that
enhancement factor also called cyclonic factor (k4), whose value as Fréchet distribution is more suitable distribution to fit extreme wind
mentioned in IS-15498:2004 is equal to 1 for dwellings, 1.15 for speed data.
industrial buildings, and 1.30 for structures of post-cyclone
importance shall be considered while determining the design wind In this paper, cyclonic factors for six zones of India have been
speed. Hence, the values of cyclonic factors depend on the types of redefined based on the modeling of extreme wind speed data for
structures, wherein, extreme wind climate modeling is not taken tropical cyclones. Since wind load is proportional to the square of
into consideration for the specification of these factors. In this the wind speed, a lesser underestimation in design wind speed
paper, cyclonic factors for six zones of India have been specified would cause a larger underestimation in wind load. Hence, this
based on the modeling of extreme wind speed data for tropical analysis would help to avoid the under estimation of wind loads on
cyclones. structures due to cyclones.

Keywords: cyclonic factors, risk factor, extreme value Mathematical model

distributions, return period, non-exceedance probability
Extreme Value distribution
Introduction The non-exceedence probability, F(x) of an extreme value
distribution with wind speed x is given by Eq.[1]
Cyclones are major storms in India. The statistics of damages in
 1

each year clearly indicate that the structures and their components,
 x  m  
F  x   exp   f1  f 2 .
 
  
especially those which are located in the east coast of India, are still (1)
vulnerable to cyclones. The probability distribution for cyclone  
wind speed comes under extreme value distribution. There are three
 
types of extreme value distribution depending upon the curvature The coefficient f1 and f2 depend on the curvature parameter τ and
parameter. are given as follow:
1. Gumbel distribution (Type I distribution)
2. Fréchet distribution (Type II distribution) f1  (1   ) ( 1a )
3. Reverse Weibull distribution (Type III distribution)
The Fréchet distribution and reverse Weibull distribution are types f 2  (1  2 )  f12 ( 1b )
of generalized value distribution when curvature parameter are
negative and positive respectively, whereas, in case of Gumbel
Where m is the mean value, σ is the standard deviation, τ is the
distribution the curvature parameter becomes 0. Simiu (1994)
curvature parameter and Γ(*) is the gamma function.
suggested that extreme wind speeds are not best fitted by Gumbel
The value of curvature parameter determines the tail behavior of the
distribution. For a negative curvature parameter, a finite lower tail 0.999
actual
theoretical
is obtained whereas for a positive value instead of a lower limit, an

non exceedence probability per year

0.995
upper limit is obtained (Kasperski, 2000). For τ=0, the extreme
0.990
distribution of Type I i.e. Gumbel distribution is obtained which has
both infinite lower and upper tails. A distribution similar to Gumbel
distribution was used in IS 875 Part 3 and it is popularly known as 0.900

Fischer-Tippet Type I distribution. The expression for Cumulative

Distribution Function (CDF) of FT-I distribution is given below: 0.500

0.100

   x  A  
FFT  I ( x)  exp   exp      (2)
0.001
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

   B   Reduced variate of maximum sustainable surface wind speed

m =39.24 m/s, -20.16 m/s, 
Figure 1. Non-exceedance probability per year versus reduced
Where A and B are coefficients mentioned in IS 875 Part 3 for six variate
different zones.
Figure 2 shows the plot of (-ln(p))τ versus reduced variate ((v-
Cyclonic Factor (k4) m)/σ). Figure 2 reveals that observed wind speed data is best fitted
Cyclonic factor, k4 is defined as the ratio of the extreme wind speed by Fréchet distribution with 98% coefficient of determination. The
obtained using Fréchet distribution, which has been found to be the slope and intercept of the theoretical line that best fitted the
best suitable model for probability distribution of cyclone wind observed data are -0.2196 and 1.5728 respectively.The high value
of R2 clearly indicates that the observed distribution is closely
speed data (Sarkar et al., 2014),  X revised  N ,P N
to the extreme following the theoretical trend.

wind speed using FT-I distribution, X N , PN .

2.0

 1
 Actual
f  ((1  PN ) N )  Theoritical
 f1  P  A ' B '  1 (3)
1.9
 
2
A ' B '  R =0.98
 X revised  N ,P   f2  Slope=-0.2196
k4  N
  f2     1.8
Intercept=1.5728
X N , PN A  B  ( ln( ln( P)))   1 
A  B  ln  ln 1  PN  
  N  1.7

(-ln(p))

1.6
where A’ and B’are the revised coefficients obtained using Fréchet
distribution. P is the annual non-exceedance probability, i.e., 1.5

1
P  1 ; N is the Return Period; PN is the exceedance 1.4

N 1.3
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
probability in N successive years.
(v-m)/

Results and Discussions

Cyclone wind data from Regional Meteorological Centre, Chennai Figure 2. The plot of observed data fitted with Fréchet distribution.
with 56 ensemble size have been fitted by extreme value Table 1 shows the different values of initial and revised coefficients
distributions. Figure 1 shows the plot of non-exceedence for six zones of India. The initial coefficients have been determined
probability versus reduced variate of maximum sustainable surface from wind speed data using FT-I distribution, whereas, revised
wind speed. From Figure 1, it has been found that they are best coefficients, A’ and B’ can be determined using Fréchet
fitted by Fréchet distribution with m=39.24 m/s, σ=-20.16 m/s and distribution.
τ=-0.10 (Sarkar et al., 2014).
Zone (m/s) A (km/h) B (km/h) A’ (km/h) B’ (km/h) Zone (m/s) R2 RMSE K-S
33 83.2 9.2 88.51 -11.88
33 0.9996 0.0060 0.0323
39 84.0 14.0 92.07 -18.07
39 0.9995 0.0075 0.0325
44 88.0 18.0 98.38 -23.24
47 88.0 20.5 99.83 -26.47 44 0.9994 0.0085 0.0324
50 88.8 22.8 101.95 -29.44 47 0.9993 0.0090 0.0325
55 90.8 27.3 106.547 -35.25 50 0.9992 0.0095 0.0324
Table 1. The initial and revised coefficients for six different zones of 55 0.9990 0.0104 0.0325
India Table 2. The result of goodness of fit test showing Fréchet and FT-I
distribution.
The probabilities have been determined both from FT-I and Fréchet
distributions using their respective coefficients as mentioned in Figure 4 shows the plot of –ln(-ln(p)) versus wind speed for six
Table 1 for six basic wind speed zones of India. To check the different zones of India with six sets of different coefficients (A and
similarity between FT-I and Fréchet distribution, the Probability- B). This plot is used to check whether the simulated wind speed
Probability (P-P) plot as well as goodness of fit test, viz., R2, RMSE data are exactly following the trend of theoretical FT-I distribution
and K-S test have been performed (Chang.T.P., 2011). For different or not. Initially, 106 random wind speed data have been generated
wind speeds the Cumulative Distribution Function (CDF) of by inverse cumulative distribution function method with six known
Fréchet and FT-I distributions can be generated. These CDFs are values of A and B and from the simulated data these coefficients
then used as to fit into the accuracy judgment criteria as mentioned are estimated. Figure 4 reveals that the simulated wind speed data
in Figure 3. obtained from FT-I distribution is exactly following the trend of the
theoretical FT-I distribution. Now from these wind speed data
actual probabilities have been determined using order statistics.
1.0
33(m/s) After that theoretical probabilities have been estimated by Fréchet
39(m/s)
44 (m/s) distribution with τ=-0.10. The least square method is employed to
0.8 47(m/s)
50(m/s) estimate the parameters of Fréchet distribution. The parameters
55(m/s)
obtained from Fréchet distribution are used to estimate the revised
Fre'chet distribution

0.6

coefficients, A’ and B’.

0.4

16
0.2
14

12
0.0
0.0 0.2 0.4 0.6 0.8 1.0
10
Fisher Tippet Type(I) distribution
A=83.2;B=9.2
8
A=84.0;B=14.0
-ln(-ln(P))

A=88.0;B=18.0
Figure 3. Probability-Probability plot of Fréchet and FT-I 6
A=88.0;B=20.5
distribution. 4 A=88.8;B=22.8
A=90.8;B=27.3
2
A=83.19;B=9.2
Figure 3 shows the P-P plot of Fréchet distribution versus FT-I 0 A=83.9,4;B=13.99
A=88.0;B=18.0
distribution for six different zones. This plot is used to determine -2 A=88.0;B=20.5
A=88.76;B=22.79
-4
the concurrency between Fréchet and FT-I distributions in basic A=90.80;B=27.3

wind speed ranges. A 450 slope between Fréchet and FT-I 0 50 100 150 200 250 300 350 400 450 500 550 600
Wind speed (km/h)
distributions reveals their concurrency behavior. Table 3 shows the
accuracy judgement criteria to show how suitably the Fréchet
distribution describes the FT-I distribution in lower wind speed Figure 4. The simulated and observed data following Fisher-
ranges. The R2 is above 99% and The RMSE and K-S tests Tippett Type-I
supplement the conclusion made by the R2 test.
By varying the return period from 5 to 10000 years, probability of
non-exceedance in one year can be estimated (IS-875 Part 3). Using
these proababilities of non-exceedance, extreme wind speeds can be
obtained by two distributions, firstly, by Fréchet distribution with
the revised coefficients A’ and B’ and secondly, by FT-I Conclusions
distribution with the initial coefficients A and B. The ratios of these
two extreme wind speeds for different return periods are used to 1. The Fréchet distribution is a suitable distribution to model
estimate the cyclonic factors k4. both basic as well as extreme wind speed.
2. The cyclonic factors (k4) vary from 1 to 1.31 for six different
Table 3 shows the cyclonic factors (k4) for different return periods. zones based on return periods. Based on the design criteria, the
Table 3 reveals that cyclonic factor enhanced the design wind speed design wind speed can be enhanced by a factor of k4.
from 1 to 1.31 for different return periods.
3. The cyclonic factor varies from 1.5 to 2.82 for the region of
Cyclonic Coefficient (k4) Bay of Bengal and had a huge impact on the structures situated
in this region, as the variation of percentage error in wind load
Zone (m/s)
years 33 39 44 47 50 55 varies from 74.69 to 501.31% for 5 to 10000 years return
5 0.988 0.984 0.982 0.979 0.978 0.976 periods.
25 1.004 1.005 1.006 1.006 1.006 1.007
References
50 1.016 1.022 1.024 1.026 1.027 1.029
100 1.032 1.041 1.046 1.049 1.052 1.055 Kasperski, M. (2000), Specification and Codification of Design
250 1.058 1.073 1.081 1.086 1.089 1.095 wind loads, Habilitation Thesis, Ruhr-University Bochum,
500 1.080 1.101 1.111 1.117 1.121 1.129 Germany.
1000 1.106 1.131 1.144 1.151 1.157 1.295 Kasperski, M., Koss, H. and Sahlmen, J. (1997a), “BEATRICE
10000 1.218 1.254 1.274 1.286 1.295 1.308 Joint project: Wind Action on Low-Rise Buildings, Basic
Table 3. Cyclonic factors for different return periods Information and First Results”, Journal of Wind Engineering and
Table 4 shows the cyclonic factor (k4) for the Bay of Bengal region Industrial Aerodynamics, 64,101-125.
for different return periods. The revised coefficients obtained after Kasperski, M. (1997b), “Climate Change and Design Wind load
fitting cyclone wind speeds into the Fréchet distribution, A’ and B’ Concepts”, Proc. 2nd European and African Conference on Wind
are 39.24 and -20.16 (m/s) respectively. The effects of cyclonic Engineering, Genova, Italy, June 22-26,147-154.
Simiu, E. (1994), Estimation of Extreme Wind Speeds, State of the
factors on wind speeds and wind loads are also determined based Art in Wind Engineering, Wiely Eastern Limited, New Delhi.
on percentage error. The initial value of k4 was taken from IS Sarkar. A. Kumar N. & Mitra D. (2014). Extreme wind climate
15498:2004 for industrial building (k4=1.15). modeling of some locations in India for the specification of the
design wind speed of structures. KSCE Journal of civil engineering.
18(5):1496-1504.
Return Cyclonic Error (%) in Error (%) in Sarkar A., Mahato J., & Mitra D. (2013), Specification of design
Period (yrs) factor (k4) wind speed wind load wind speed for cyclone Events in India. VIII APCWE, Dec. 10-14,
5 1.52 32.17 74.69 Chennai, India.
25 1.80 56.52 144.99 IS: 875 (Part 3)-2015, Design Loads (Other than Earthquake) for
50 1.91 66.08 175.84 Buildings and Structures-Code of Practice, BIS New Delhi..
100 2.02 75.65 208.53
250 2.17 88.69 256.06 IS: 875 (Part 3 - A.1)-2016, Design Loads (Other than Earthquake)
for Buildings and Structures-Code of Practice, BIS New Delhi.
500 2.28 98.26 293.07
1000 2.40 108.69 335.53 IS:15498-2004 (Reaffirmed-2010), Guidelines for Improving the
10000 2.82 145.22 501.31 Cyclonic Resistance of Low Rise Houses and Other
Buildings/Structures, BIS New Delhi.
Table 4. Cyclonic Factor for different return period and percentage
Chang, T.P. (2011). Performance comparison of six numerical
error in wind speed and wind load for Bay of Bengal. methods in estimating Weibull parmeters for wind enrgy
applications. Applied Energy, 88:272-282.

View publication stats