IMPACT OF CLIMATE CHANGE ON LOCAL WIND CONDITIONS

`Proceedings of HYDRO 2013 INTERNATIONAL,

4-6 Dec 2013, IIT Madras, INDIA

1 Research Scholar, Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai 400076, India, Email: [email protected] 2 Professor, Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai 400076, Email : [email protected] 3 Assistant Professor, Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai 400076, India, Email: subimal @ civil.iitb.ac.in

IMPACT OF CLIMATE CHANGE ON LOCAL WIND CONDITIONS

Sumeet Kulkarni1 , M C Deo2 and Subimal Ghosh3

Global warming could lead to a change in wind speeds both at global as well as regional levels. However at the regional scale a different trend over the future may happen than the one expected from the global mathematical models. An evaluation of changes in wind at specific sites is therefore necessary, especially when ocean structures including wind farms are to be designed and operated taking into consideration site-specific data. This paper presents authors’ experience at a few specific Indian offshore sites and also over the entire coastline. It is found that the most rigorous method of assessing the climate change involves downscaling the general circulation model (GCM) results at local level and deriving wind speeds with varying return periods from the projected wind conditions so obtained. A comparison of these values with those based on historical data provided the changed values of operational and design wind. There are so many uncertainties, however, that the modeler has to face in this process and these include alternative GCM’s, scenarios, distribution fittings and basis of comparison. The projected and the historical data sets can also be analyzed to know comparative wind behavior in terms of various statistics, probability distribution fits and trend lines. A look at many such works done over different locations of the world indicates highly varying effects including reduction in wind in future to a high increase of around 70 % in magnitudes in the coming decades. Authors downscaled the wind output yielded by the Canadian General Circulation model: CGCM4 that was run with the warming scenario: RCP-4.5 for a 30 yr time slice in future at two locations along the western Indian coastline. This involved use of the bilinear interpolation technique and quantile mapping-based correction to remove bias of the GCM data with respect to the observed data. Historical wind at these sites for the past 30 years was also collected from National Centre for Environment Protection / National Centre for Atmospheric Research (NCEP/NCAR) achieves. The reliability of projected and past data was ascertained and long term statistical analysis of these records was carried out for which the extreme value distributions were fitted with the help of the peak-over-threshold method. This analysis indicated that the daily wind speed with 10 and 100 years’ return would increase by 11 % to 15 % at the selected sites. In order to have an overall view of potential changes along the entire Indian shoreline and to confirm the mean trend instead of long term design values, changes in annual mean and maximum wind speeds were estimated along the entire eastern and western coastlines using




historical datasets available in the form of reanalysis data and a Canadian GCM for future predictions. A rising trend in the wind speeds of about 6 % to 8 % along Arabian Sea and 2 % to 4 % along Bay of Bengal was observed. Along some of the inland regions in India it was noticed that these speeds might decrease by 2 % to 3 %. Future design and operation of ocean structures should therefore consider the suggested changes in this paper. While these figures may suffer from statistical uncertainty and need confirmation from more detailed analysis the study points out to a relook on the safety margins kept in the design and operation of ocean structures in the light of global warming. Key words: Climate change, design wind speed, regridding, quantile mapping.




IMPACT OF CLIMATE CHANGE ON LOCAL WIND CONDITIONS

Sumeet Kulkarni1 , M C Deo2 and Subimal Ghosh3

Abstract: The increasing temperature of earth is likely to affect the wind speed and its pattern at global and regional levels. We have made an attempt in this paper to know how the design as well as operational wind conditions would change at specific sites off the Indian west coast as well as how mean and annual maximum wind would vary over the entire coastline. The past wind climate was simulated by the NCEP/NCAR wind speed information over 30 years from 1976 to 2005. The data for 30 years in future from 2006 to 2035 were generated from the Canadian General Circulation Model CGCM4 and the CMIP5 ensemble modeling experiment. Both sets of data were fitted to the Weibull distribution using the peak over threshold method to extract wind speeds with return periods of 10 and 100 years and this indicated an increase in operational and design wind of 11 % to 15 % at the selected sites. Changes in annual mean and maximum wind speeds were also observed along entire Indian shoreline. A rising trend in the wind speeds of about 6 % to 8 % along Arabian Sea and 2 % to 4 % along Bay of Bengal was noticed. These speeds might decrease by 2 % to 3 % along some of the inland regions in India. While these figures may suffer from statistical uncertainty the study points out to a relook on the safety margins kept in the design and operation of ocean structures in the light of global warming.

Key Words: Global warming, impact on wind, general circulation models INTRODUCTION

It is well known that the wind conditions at global as well as at regional scales are changing due to the effect of climate change brought about by global warming. It is believed that the actual rise in the global temperature is faster than IPCC 2007 projections. (See for example: http://nawindpower.com/e107_plugins/content/content.php?content.7130). The temperature gradient between the equator and the poles is responsible for driving the wind at global level. The change in such global wind pattern may alter local wind conditions due to the more influential local gradients. Further, location-specific factors such as topography, moisture, land use, water bodies, El Nino/La Nina phenomena can affect regional temperature and hence wind.



It was noticed by Ruggiero et al. (2011) that as the temperatures at poles are rising faster than at the equator, the temperature gradient is becoming smaller and this may reduce the wind speed, especially in upper latitudes. The storm tracks are found to be shifting northward in the Northern hemisphere and this may increase gusty or extreme winds at higher latitudes and reduce the same in the southern region. As poles are warming faster, storm tracks shift towards poles and mid-latitude storms reduce. Many studies have confirmed such change. For example, Graham and Diaz (2001) noted intensification of the storms in North Pacific during winter due to increased upper latitude winds, while McCabe et al. (2001) observed that while intensity may have increased the frequency has reduced due to the shifting of track, which was confirmed by Geng and Sugi (2003) for medium level storms. Impact of climate change on the regional design wind was assessed in the past by Debernard et al. (2002) over the northeast of Atlantic based on a control climate of 1980-2000 and projected climate of 2030-2050. Certain areas in Barents Sea, northern North Sea and some portions of Atlantic Ocean were found to undergo noticeable increase in wind speeds. Grabmann and Weiss (2008) analyzed extreme wind conditions in North Sea and noticed increased wind intensity and frequency. The long term 99 percentile wind speed in North Sea was predicted to increase by 7%. Merryfield et al. (2009) analyzed data of surface marine winds off Canada. The projected wind over two intervals, 2030-2049 and 2080-2099, were compared to the 1976-1995 baseline period. The projections were based on the warming scenario of A1B representing an intermediate projected rate of increase in greenhouse gas concentrations. An increase of 5 to 10 % in average summer wind speeds was noticed by the authors. For various locations in The Netherlands Raphel et al. (2012) found a change of 0.8 % to 2.3 % in 50-year return hourly wind. Authors also mention of an earlier study in which -1 % to 4 % change in the annual daily mean wind speed was reported. Cheng et al. (2012) studied potential changes in hourly and daily gust wind with the help of gust simulation models and statistical downscaling over Ontario, Canada. The results showed an increase up to 15 % and 25 % for annual mean wind exceeding 28 kmph and 70 kmph, respectively and further that these changes were greater than those due to GCM and scenario uncertainties. Over the North Sea Winter et al. (2013) used an ensemble of 12 CMIP5 general circulation model (GCM)s that were run for two climate scenarios of rcp4.5 and rcp8.5 to know changes in annual maximum and long term wind speeds. Authors however failed to notice any change in magnitudes but observed increased frequencies of occurrence. As regards the Indian sites, Deepthi and Deo (2010) considered 7 years’ daily measurements made at a location off the west coast of India as well as future projections of the daily wind at the same location corresponding to a general circulation model CGCM3 run for the worst global warming scenario: A2. This study however needs further validation by increasing the sample size and using more sophisticated statistical procedures. The present work is oriented in this direction. It considers a large sample of 30 years to represent the past wind and distribution fitting using the peak-over-threshold technique and also involves the use of recent CMIP5 scenarios and application of more sophisticated bias correction techniques.



SITE SPECIFIC STUDIES The Locations We have assessed in the first part of this study the impact of climate change on operational and design wind speeds corresponding to different return periods at two selected offshore locations in India where commercial interests exist and where past wind observations made by National Institute of Ocean Technology (NIOT), India, with the help of wave buoys were available. The two locations in the present study are along the west coast of India and code named DS1 and DS2 which are off Goa and Lakshadweep, respectively. (Fig. 1).The site: DS1 has the latitude and longitudes of 15.4470 N and 69.2360 E, respectively. For the other location: DS2 the latitude, longitude values are 10.6510 N and 72.5250 E, respectively. Both the sites are in deep water.

Fig.1: Locations of NIOT wave rider buoys considered for present study

The Data The wind measurements made over the monsoon season by National Institute of Ocean Technology at Chennai, India over a period of 8 years from 1998 to 2006 at both the locations were used to validate the reanalysis data at the local scale. From the 3- hourly observations the values of daily mean wind speed (referred to as wind speed only hereinafter) were derived. The buoy data were available for 8 years, however for applying the technique of quantile mapping type bias correction at least 30 years data were necessary. The NCEP/NCAR dataset, being more accurate and reliable than a general GCM output, was used as additional equivalent observations in the present study. The information over the grid size of 2.5 x 2.5 degrees-squared at a single grid each surrounding DS1 and DS2 and corresponding the duration of 1976-2005 was extracted



To obtain the projected climate over the years: 2006-2035 the Canadian General Circulation Model: (CGCM)-4, run for the RCP-4.5 scenario was used. The grid size was 2.81 x 2.81 dgerees-squared. The information of meridional and zonal wind over a 3x3 grid area centered over DS1 and DS2 was extracted. The use of nine grids facilitated interpolation at a single grid of NCEP/NCAR described later. The outcome of the Canadian GCM called 20C3M experiment was extracted to simulate the past wind conditions over the years: 1976-2005 at both the sites for applying the quantile mapping type bias correction technique for future data. The information over the same nine grids (of size: 2.81 x 2.81) as above was extracted. The Methodology The procedure to arrive at the operational and design wind speeds at the selected locations using the above mentioned datasets consisted of following steps: Interpolation of the NCEP/NCAR data at DS1 and DS2 was done using the bi-linear interpolation method for those eight years for which the buoy records were available. Comparison of the resulting interpolated wind speed with the buoy measurements was done to ensure compatibility. The CGCM-RCP-4.5 data (from 2.81 x 2.81) were re-gridded to the NCEP/NCAR grid size (2.5 x 2.5) using the bi-linear interpolation method. The quantile mapping was performed for CGCM-RCP-4.5 and 20C3M data as per the NCEP/NCAR data. At the end of the quantile mapping RCP-4.5 and 20C3M datasets got corrected for the bias at the resolution of 2.5 x 2.5. A bi-linear interpolation of the above unbiased data was carried out at the buoy location levels. Having generated bias-free and interpolated GCM data as above for past as well as future, the long term statistical analysis was done to obtain the wind speed with varying return periods as mentioned below. The Weibull distribution was fitted to the partial series data generated by the Peak over Threshold method. A particular threshold value was selected and all the data points greater than it were shortlisted for further processing. The selection was by trials made such that the distribution fit satisfied the Kolmogorov-Smirnov (KS) goodness of fit test at the significance level of 0.05 to 0.10 and also had the largest number of data values above the threshold. The Weibull distribution used was as given below:

P (Ws) = 1-푒[ ( )] (1)

Where, P (Ws) = probability distribution function of Ws; Ws = wind speed; µ = location parameter (selected threshold), = scale parameter; = shape parameter. For a given return period, Tr, P(Ws) can be obtained by:



P (Ws) = 1 –

∗ (2)

where γ = average number of selected wind speed values above the threshold per year The P (Ws) value obtained from the left side of eq. (2) was substituted in eq.(1) to obtain the unknown wind speed Ws corresponding to return period: Tr.

The Results In order to use the long range (and past) reanalysis data as reference data in place of the short duration (but accurate) buoy data it was necessary to confirm the accuracy of the former data at the scale of buoy location. After performing the bilinear interpolation on NCEP/NCAR data for monsoon period the resulting time series was statistically compared with the buoy data for the same period. It was observed that the two datasets resembled well with each other. Fig 2(a) and 2(b) that show the time history comparisons at DS1 and DS2 as well as Tables 1(a) and 1(b) that provide corresponding quantitative error measures of correlation coefficient, root mean square error and mean absolute error, and mean and standard deviations, respectively, confirm the same.

1(a)

1 (b)

Fig. 2(a) and (b). Comparison of NCEP/NCAR data with wave buoy observations.

0

5

10

15

20

0 100 200 300 400 500 600 700 800

Win

d Sp

eed(

m/s

)

Data Points

Comparision of NCEP Data vs Buoy Data at DS1

NCEP Reanalysis data

NIOT Buoy data

01/06/1998 30/09/2005

05

101520

0 200 400 600 800

Win

d Sp

eed(

m/s)

Comparision of NCEP Data vs Buoy Data at DS2

NCEP reanalysis data

NIOT buoy data

30/09/200501/06/1998



Table 1(a) and 1(b). The error statistics to compare NCEP/NCAR wind data with wave buoy observations

1(a)

Error statistics DS1 DS2

Correlation coefficient 0.62 0.66 Root mean square error(m/s) 3.77 1.86 Mean Absolute error (m/s) 2.90 1.37

1(b)

DS1 DS2 Statistical

coefficients Buoy

data(m/s) Reanalysis data(m/s)

Buoy data(m/s)

Reanalysis data(m/s)

Mean (m/s) 7.54 9.95 6.66 6.95 Standard deviation (m/s)

3.00 3.50 2.07 2.07

The long term wind speed values used for operational and design purposes obtained by following the methodology discussed above and applied for both past as well as future data are shown in Table 2. The Table shows that for these return periods the daily mean wind speed would increase in future for both the sites and that such an increase would be 11 % to 14 % at DS1 and 12% to 14 % over the various return periods.

Table 2 : Changes in the long term wind speed by the interpolation methodology

Return period (Years)

Wind speed based on past climate

(m/s)

Wind speed based on projected climate (m/s)

% increase

DS1 DS2 DS1 DS2 DS1 DS2

10 18.71 12.91 20.25 14.36 14.41 11.20

100 20.14 14.14 22.60 16.15 12.21 14.17



STUDY OVER THE ENTIRE COASTLINE

The Data and Procedure The site specific changes in operational and design wind speeds have been discussed in previous section. In the subsequent part of the study the change in mean and annual maximum wind speed over entire coastline of the country is predicted. For this purpose more recent wind data: ERA 40 reanalysis, was considered as equivalent to an observed dataset. The duration of 1971 to 2000 was considered since data beyond 2002 are not available. As in the preceding study, the outcome of running the Canadian GCM with RCP 4.5 scenario was used to create the future 30 years’ wind while the CGCM 20C3M dataset was used to simulate the past wind. The regrinding and bias correction was performed on the GCM datasets and thereafter the temporal nodal mean for all the grids was considered. The maximum and mean annual wind speeds were extracted and averaged out along the temporal dimension which produced 2-D spatial wind speed plots.

The Results

The results of the above exercise are summarized in Fig. 3, 4 and 5. Fig. 3a and b show the 20C3M mean and annual maximum wind over the past 30 years, respectively. Fig. 4a and b indicate the CGCM mean and maximum annual wind over the future 30 years while Fig. 5a and b depict the percentage change in mean and annual maximum wind due to the impact of climate change. These figures indicate a significant difference between land based and offshore wind scenarios in future. Areas away from the land and toward the coast would see considerable increase in the speeds. This trend can be mainly because of smaller frictional resistance faced by the gust wind at the sea surface as compared to the land. Also in certain regions of India it is seen that the wind speed would decrease by 2 to 3 %. But as we approach the coast this figure may go up to 6 to 8 % along Arabian Sea and 2 to 4 % along Bay of Bengal. Interestingly it is seen that as we go offshore in Bay of Bengal a high rise of 6 % to 7 % can be experienced. Thus an overall rising trend in offshore Arabian Sea and Bay of Bengal can be expected. The mean wind speed at most of offshore locations ranges from 4 to 8 m/s and an increase of about 6 to 8% in this value will result in more wind coming under the extractable spectrum of wind potential (5 to 20 m/s). This may be beneficial in future offshore wind power extraction.



(a) (b)

Fig. 3(a) and (b): 20C3M mean and maximum annual wind speeds over past 30 years (1976-2005)

(a) (b)

Fig. 4 (a) and (b): CGCM mean and maximum annual wind speeds over future 30 years (2006-2035)

(a) (b)

Fig. 5 (a) and (b): % change in mean and maximum annual wind speeds due to climate change impact.



More details of the present work are under consideration for possible publication elsewhere.

CONCLUSIONS

The assessment of the effect of climate change based on the analysis of the wind speeds projected using state of the art GCM data and statistical fitting procedures at two offshore locations along the west coast of India indicated that the values of operational and design wind corresponding to 10 and 100 years’ return period at these locations might increase by around 11 % to 15% in future.

The analysis for the entire coastal region of the country showed a significant difference between land based and offshore wind scenarios in future. Areas from the land toward the coast would see considerable increase in the speeds. In certain land regions of India the wind speed would decrease by 2 to 3 %, however as we approach the coast this figure may go up to 6 to 8 % along Arabian Sea and 2 to 4 % along Bay of Bengal. Thus an overall rising trend in offshore Arabian Sea and Bay of Bengal can be expected. The mean wind speed at most of offshore locations ranges from 4 to 8 m/s and an increase of about 6 to 8% in this value will result in more wind coming under the extractable spectrum of wind potential (5 to 20 m/s). This may be beneficial in future offshore wind power extraction.

The procedures and tools adopted to assess the effect of climate change significantly affect the magnitude of the final outcome and hence continuous re-working of the effect is necessary as and when more accurate and latest data become available.

REFERENCES

Debernard, J., Sætra, O. and Røed, L. P. (2002), “Future wind, wave and storm surge climate in the northern North Atlantic”, Climate Research 23, 39-49

IPCC (2007b), “Climate Change 2007 - The Physical Science Basis: Contribution of Working Group I to the Fourth Assessment Report of the IPCC”, Cambridge University

IPCC (2000), “IPCC special report: Emission scenarios; Summary of policymakers”, A special report of IPCC working group III

Ruggiero P, Brown C A, Komar P D, Allan J C, Reusser D A, Lee H, (2011): “Impacts of climate change on Oregon’s coasts and estuaries” Merryfield, W.J., Pal, B., Foreman, M.G.G., (2009). “Projected future changes in surface marine winds off the west coast of Canada”. Journal of Geophysical Research 114, C06008, doi:10.1029/2008JC005123.



Graham N E, Diaz H F (2001): “Evidence for intensification of North Pacific winter cyclones since 1948, Bulletin of American Meteorological Society, 82, 1869-1893. McCabe G J, M P Clark, M C Serreze (2001): “Trends in Northern hemisphere surface cyclone frequency and intensity, Journal of climate, 14 (2763-2768). Geng Q and M Sugi (2003): “Possible change of extra-tropical cyclone activity due to enhanced greenhouse gases and surface aerosols: Study of high resolution AGCM, Journal of Climate, 16, 2262-2274. Grabmann, I. and Weisse, R. (2008). “Climate change impact on extreme wave conditions in the North Sea: an ensemble study”, Ocean Dynamics, 58(2008), 199-212. Raphel D J M, Steenbergen, Tessa Koster, Chirs P W Geurts (2012): “The effect of climate change and natural variability on wind loading values for buildings”, Building and Environment, Elsevier, 55(2012), 178-188. Cheng, Chad Shouquan, Guilong Li, Qian Li, Heather Auld, Chao Fu, 2012: Possible Impacts of Climate Change on Wind Gusts under Downscaled Future Climate Conditions over Ontario, Canada. J. Climate, 25, 3390–3408. Winter, R.C. de, A. Sterl and B.G. Ruessink, Wind extremes in the North Sea basin under climate change: an ensemble study of 12 CMIP5 GCMs. J. Geophys. Res., Atmos, 2012. Deepthi R and M C Deo. (2010). “Effect of climate change on deisgn winds at the Indian offshore locations”. Ocean Engineering, 37(2010). 1061-1069.

IMPACT OF CLIMATE CHANGE ON LOCAL WIND CONDITIONS

Documents