Wind Resource Characterization in the Arabian Peninsula

Wind resource characterization in the Arabian Peninsula

Item Type Article

Authors Yip, Chak Man Andrew; Gunturu, Udaya; Stenchikov, Georgiy L.

Citation Wind resource characterization in the Arabian Peninsula 2016, 164:826 Applied Energy

Eprint version Post-print

DOI 10.1016/j.apenergy.2015.11.074

Publisher Elsevier BV

Journal Applied Energy

Rights NOTICE: this is the author’s version of a work that was accepted for publication in Applied Energy. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Applied Energy, 28 December 2015. DOI: 10.1016/j.apenergy.2015.11.074

Download date 02/10/2021 22:34:00

Link to Item http://hdl.handle.net/10754/596964 Wind Resource Characterization in the Arabian Peninsula

Chak Man Andrew Yipa,∗, Udaya Bhaskar Gunturua, Georgiy L. Stenchikova

aKing Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia

Abstract

Wind energy is expected to contribute to alleviating the rise in energy demand in the Middle East that is driven by population growth and industrial development. However, variability and intermittency in the wind resource present significant challenges to grid integration of wind energy systems. These issues are rarely addressed in the literature of wind resource assessment in the Middle East due to sparse meteorological observations with varying record lengths. In this study, the wind field with consistent space-time resolution for over three decades at three hub heights (50 m, 80 m, 140 m) over the whole Arabian Peninsula is constructed using the Modern Era Retrospective-Analysis for Research and Applications (MERRA) dataset. The wind resource is assessed at a higher spatial resolution with metrics of temporal variations in the wind than in prior studies. Previously unrecognized locations of interest with high wind abundance and low variability and intermittency have been identified in this study and confirmed by recent on-site observations. In particular, the western mountains of Saudi Arabia experience more abundant wind resource than most Red Sea coastal areas. The wind resource is more variable in coastal areas along the Arabian Gulf than their Red Sea counterparts at a similar latitude. Persistent wind is found along the coast of the Arabian Gulf. Keywords: Wind Energy, Variability, Intermittency, Middle East, Resource Assessment, Reanalysis

1 1. Introduction 22 the latest expansion of the renewable energy market 23 [5]. Among net oil importers such as Jordan, energy

2 The potential adverse impacts of climate change and 24 insecurity and dependence on expensive oil imports 3 energy insecurity have encouraged countries worldwide 25 have led to an expansion of the renewable energy 4 towards adopting renewable energy as an integral 26 program. Renewable energy has grown from 0.4 TW h 5 part of their future energy mix. Near-surface wind 27 in 2008 to 1.2 TW h in 2011 among net oil importers 6 energy has the potential to power the world; it allows 28 [4]. In the net oil exporting countries, renewable 7 extracting energy at a rate of at least 400 TW [1]. It 29 energy has grown from 0.8 TW h in 2008 to 1.6 TW h 8 is suggested that following a moderate wind energy 30 in 2011 [4]. This growth has been the result of rising 9 deployment plan by 2050 would delay the crossing 31 opportunity cost of oil and gas accompanied by an 10 of the 2 °C threshold for 1 to 6 years [2]. Wind 32 increase in urbanization and a rapid rise in domestic 11 energy provides a viable alternative energy source 33 demand for energy. These expansions are evident 12 to energy intensive countries such as China, where 34 in the countries’ recent large-scale procurement of 13 it is estimated to be suﬃcient to replace 23% of the 35 renewable energy systems to fulﬁll national renewable 14 electricity generated from coal [3]. In the Middle East 36 targets [5]. The surging interest in renewable energy 15 and North Africa (MENA), population growth has led 37 calls for a better understanding of the spatial and 16 to increases in demand for fuel and electricity for air- 38 temporal characteristics of the resource. This paper 17 conditioning and desalination. Regional annual Total 39 focuses on the wind energy resource, the most variable 18 Primary Energy Supply (TPES) increased by 14.9% to 40 and intermittent source of renewable energy in the 19 800 millions Mtoe (million tonnes of oil equivalent) in 41 Arabian Peninsula. 20 2010 compared to the TPES of 2007 [4]. These steady 21 increases in domestic consumption of energy drive 42 There are two key challenges in assessing the wind 43 resource in the Arabian Peninsula. Most of the

44 ∗Corresponding author observations available are sparse in space and in- Email address: [email protected] (Chak Man 45 consistent in time: spatially scattered observations Andrew Yip) 46 with varying record lengths come from meteorological November 15, 2015 47 stations that are located mainly in clustered coastal 101 Moreover, previous resource characterizations have 48 and inland settlements. Hourly wind speeds were 102 focused on average wind abundance and annual energy 49 collected by 293 weather stations in the Peninsula 103 production estimates. Wind power variability and 50 during our period of study from 1979 to 2013. Among 104 intermittency present significant challenges to grid 51 the stations, 42 collected data for at least half of the 105 integration of wind energy systems, as identified by 52 time. Only 17 stations have observations available 106 wind integration studies in the United States [17]. 53 for more than 80% of the record length [6]. Despite 107 Variability and intermittency have been considered, 54 these challenges, Ansari et al. [7] constructed the 108 most commonly using tower measurements where data 55 Saudi Arabian Wind Energy Atlas in 1986 using hourly 109 are limited in the spatial and temporal dimensions 56 observations from 20 airport weather stations from 110 [18–20]. Rehman and Halawani [8] provided a wind 57 1970 to 1982. They described diurnal and seasonal 111 persistence measure via auto-correlation and auto- 58 variations of wind speed at measurement height at 112 regression for ten weather stations. Rehman and 59 these locations and mapped prevailing wind directions. 113 Ahmad [21] presented a wind availability analysis 60 Rehman and Halawani [8] described diurnal, monthly, 114 for 5 coastal locations in Saudi Arabia in terms of 61 and inter-annual wind speed variations at 10 weather 115 frequency of wind speed within a specified interval. 62 stations. Most of the studies focused on prominent 116 Wind speed time series from meteorological stations 63 sites of assessment that are mainly coastal. A similar 117 were fitted by Weibull distributions to investigate the 64 tendency is observed for the MENA region [9–11], 118 monthly variation of wind speed and their changes 65 with the exception of Ohunakin et al. [12] where the 119 with hub height in Saudi Arabia [22, 23] and Bahrain 66 focus was inland. These analyses concentrated on 120 [24]. Ouarda et al. [25] fitted multiple distributions 67 wind speed time series from meteorological stations 121 and assessed their goodness-of-fit with wind speed 68 with different record periods. The wind resource is 122 measurements in the United Arab Emirates (UAE). 69 frequently characterized by average wind speed or 123 However, the variability and intermittency of the wind 70 wind power density (WPD) that is at measurement 124 resource have not been studied in the entire Peninsular 71 height or is adjusted to hub height. Recent works have 125 region. 72 attempted to study the spatial variation of the wind 126 The primary goal of this study is to overcome the 73 resource. Jervase and Al-Lawati [13] performed an 127 limitation of sparse station observations with varying 74 areal analysis of wind resource abundance in Oman 128 record lengths by constructing the wind field using a 75 using the NASA Surface Meteorology and Solar Energy 129 gridded reanalysis dataset with a multi-decadal record 76 (SSE) Release 6.0 dataset with a spatial resolution of 1° 130 period to arrive at a characterization of wind variability 77 × 1°. Al-Yahyai and Charabi assessed wind resource 131 and intermittency. We characterize the wind resource 78 in Oman using a nested ensemble numerical weather 132 using metrics proposed in Gunturu and Schlosser [26] 79 prediction (NWP) approach, where two global models 133 (United States), Cosseron et al. [27] (Europe), Fant 80 were used as boundary conditions to drive two local 134 and Gunturu [28] (South Africa), and Hallgren et al. 81 area models. The wind abundance has been assessed 135 [29] (Australia). 82 at the scale of a country [14] and a city [15]. Moreover, 83 Charabi et al. [16] demonstrated that NWP models 136 This work aims to answer the following questions: 84 at 7 km are effective in resolving finer structures such 85 as the sea breezes in this region. However, without 137 • What methodology can be used to assess the wind 86 well-formulated boundary conditions based on a long- 138 energy resource in a region where observational 87 term and spatially and temporally consistent dataset, 139 data are sparse and non-concurrent? (section2) 88 an NWP model would not capture the impact of

89 large-scale circulations such as the El Niño. Since 140 • Where are the areas with wind power potential 90 these circulations are of low frequency, they have 141 that were not previously located due to lack of 91 higher spectral power and, therefore, have a significant 142 observations? (section 3.1) 92 impact on the wind resource. Existing meteorological 93 observations contain missing data due to handling 143 • How do wind variability and intermittency differ 94 errors or malfunctioning of the instrument. These 144 in spatial distributions from conventional metrics 95 observations are spatially sparse, for instance, located 145 of resource abundance? (sections 3.2 and 3.3) 96 mainly at airports and urban areas. These factors 97 led to some potentially resource-rich regions being 146 In the following sections, we explore regional wind 98 overlooked, which prevented prior studies in developing 147 resource and compare our results with results from 99 a comprehensive characterization of the wind resource 148 prior studies. We first describe wind resource abun- 100 for the entire Arabian Peninsula. 149 dance as characterized by the median WPD. Our wind 2 150 field reconstruction shows a qualitative agreement with 195 ψ = 0. Neutral stratification assumes negligible 151 previous studies. We then describe the regional wind 196 effects of external heating on the vertical distribution 152 resource using metrics of variability and intermittency. 197 of temperature. Although neutral stratification is 198 widely assumed in previous studies of wind resource 199 assessment to adjust wind speed to hub height from 153 2. Material and Methods 200 measurement height, the neutral boundary layer may 201 not occur in the presence of surface heating [33]. 202 Intense surface heating causes warm air near the 154 2.1. Data 203 surface to rise and create turbulence as in the case 155 The spatial domain of interest spans the Arabian 204 of an unstable boundary layer. It implies a smaller 156 Peninsula bounded between 10°N and 35°N in latitude 205 change in wind speed with height than in a neutral 157 and 35°E and 60°E in longitude. This spatial extent 206 case. WPD is the wind power available per unit swept 158 allows investigation of the Red Sea, the Gulf of Aden, 207 area of a turbine defined as 159 the Arabian Gulf, and part of the Arabian Sea along 160 with inland areas. 1 WPD = ρu3, (2) 2 161 The WPD field is reconstructed using the Modern 162 Era Retrospective-Analysis for Research and Appli- 208 where ρ is the air density and u is the wind speed 163 cations (MERRA) dataset. MERRA is a reanalysis 209 [31]. The fields for calculating wind speed and WPD 164 conducted by the Global Modeling and Assimilation 210 at hub height have been extracted from the MERRA 165 Office (GMAO) at NASA using the Goddard Earth 211 2D surface turbulent flux diagnostics available at a 166 Observing System Version 5 (GEOS-5). GEOS-5 is 212 single level at the top of the surface layer. Specifically, 167 a general circulation model (GCM) used within a 213 momentum roughness length (Z0M), friction velocity 168 data assimilation system where satellite and surface 214 (USTAR), surface air density (RHOA), and displace- 169 observations are utilized [30]. The dataset has a spatial 215 ment height (DISPH) are extracted from the dataset 170 resolution of 0.5° (latitude) × 0.67° (longitude) and 216 for the computation. Wind speed and WPD have 171 hourly output is available. For this study, a record 217 been computed at all hourly time-steps at 50 m, 80 m, 172 period from January 1, 1979 midnight (UTC) to 218 and 140 m corresponding to the typical hub heights 173 January 1, 2014 midnight (UTC) is chosen. This 219 of the three generations of wind turbines. Most prior 174 temporal range enables studying of wind variation 220 studies in the region did not consider terrain and 175 over different time scales from hours to decades. The 221 local conditions when adjusting wind speed to hub 176 spatial coverage of the dataset provides an opportunity 222 height (e.g.,[34]), where the effects of wind shear were 177 to understand the regional wind patterns, those that 223 demonstrated to be significant in power generation 178 coarse observations cannot address. 224 [35]. The WPD provides a proxy to wind power

179 Wind speed and WPD are two primary variables in 225 resource that is independent of a wind energy system’s 180 wind resource assessments. Wind speed at the turbine 226 speciﬁcation. It combines the contributions of both 181 hub height is calculated using the similarity theory for 227 wind speed and air density in illustrating the physical 182 the surface layer in which the turbine submerges [31]. 228 limit of wind power potential. 183 Wind speed is computed as

229 2.2. Metrics u∗ z − d u(z) = ln − ψ, (1) κ z0 230 Prior studies of the wind energy resource in the 231 Arabian Peninsula were mostly concerned with the 184 where z is the turbine hub height [31]. u∗ = 232 abundance of the wind resource in terms of average 0 0 2 0 0 2 1/4 185 [(u w ) + (v w ) ] is the friction velocity, defined 233 wind speed or annual energy production estimates. 186 with the surface kinematic momentum fluxes in x and 234 Variability and intermittency have not been investi- 187 y directions. κ = 0.41 is a standard accepted value of 235 gated comprehensively for this region. The understand- 188 the von Kármán constant [32]. d is the displacement 236 ing of these temporal characteristics is essential for grid 189 height that gives the vertical displacement of the entire 237 integration to pave the way for wind energy to become 190 flow regime over areas densely covered with obstacles. 238 part of a larger generation network contributing to 191 Roughness length z0 defines the height that the wind 239 the generation of base-load electricity. Therefore, 192 speed is assumed to vanish near the ground. The 240 we consider three essential metrics in evaluating the 193 parameter ψ depends on the stability of the boundary 241 wind energy resource: abundance, variability, and 194 layer. A neutral boundary layer is assumed, thus 242 intermittency. 3 −2 243 Abundance is the amount of wind energy available. 283 of 200 W m . The threshold has been chosen keeping −2 244 It is conventionally measured by time-averaged wind 284 in mind the 300 W m threshold (Wind Class 3) in 245 speed (u¯) or time-averaged WPD, which takes the 285 filtering sites for commercial scale power production 246 form of 286 in the Regional Energy Deployment System (ReEDS) 287 model of the National Renewable Energy Laboratory −2 1 Z T 288 (NREL) [36]. A lower threshold of 200 W m (Wind x¯ = x dt, (3) 289 Class 2) is chosen in light of recent developments in T 0 290 low-wind turbines which render conventionally low- 291 wind areas viable for energy applications. Availability 247 where x is the instantaneous wind speed or wind power 292 (A) is computed following [26], 248 density, and T is the length of the record period. Time- 249 averaged WPD and wind speed are used to compare n 250 with results from prior studies in the region. For each 1 X A = τ(pt), (6) 251 grid cell in the domain, the hourly wind power density n t=1 252 field is used to compute the average WPD over the 253 record period. −2 1, pt > 200 W m τ(pt) = 0, otherwise n 1 X p¯ = pt, (4) n 293 τ(pt) records the events when pt is greater than the t=1 294 threshold. Persistence reflects the steadiness over time

295 254 where pt is the average WPD at each hour (t) and n is in wind power generation above a given threshold. 296 255 the total number of time-steps in the record period. In Episode length is deﬁned as the duration of consecutive 297 p 256 addition, we compute the median WPD, pmed, which time steps when t is above the threshold.

257 is the value lying at the midpoint of a frequency 298 Persistence is measured by the median episode length 258 distribution of the time series at each grid point such 299 (MEL). The median length of wind episodes is the 259 that there is an equal probability to be above or below 300 median length of all recorded continuous periods when 260 it. The median is a metric robust to extreme samples 301 WPD is above the threshold [26], 261 and hence is a better characterization of central 262 tendency than the mean for skewed distributions such τ p . 263 as those of wind speed and WPD. The pmed is used to MEL = median[ ( t)] (7) 264 characterize the wind regimes in the further analysis: −2 302 265 regime I where pmed ≥ 67 W m , regime II where It should be noted that intermittency concerns the −2 −2 303 266 46 W m ≤ pmed < 67 W m , and regime III where statistics of threshold-crossing of the WPD while −2 304 267 pmed < 46 W m . The three regimes correspond to variability measures the ﬂuctuation of the WPD 305 268 values separated by the 33rd and 67th percentiles of in magnitude. This distinction was elaborated in 306 269 the median WPDs for all grid cells in the domain. Gunturu and Schlosser [26].

270 This classiﬁcation gives an indication of the relative 307 The present study aims at identifying regional features 271 spatial abundance of the wind resource in the region. 308 of wind resource while ﬁner local characteristics and

309 272 Variability characterizes the fluctuations of the wind circulations could not be represented. Our analysis 310 273 energy resource at a given location. It is measured by identifies regions for further down-scaling and micro- 311 274 the robust coefficient of variation (rCV) of WPD [26], siting assessment where models with finer resolutions 312 275 computed as follows for each cell: in space and time are appropriate. Dynamical down- 313 scaling in the mountainous regions of the Arabian 314 Peninsula would benefit from the use of Kalman filters, median(|p − pmed|) rCV = . (5) 315 as it has shown improvements in NWP wind speed pmed 316 forecast in complex terrain [37]. Sub-hourly data 317 would be needed to characterize wind intermittency at 276 The rCV provides a dimensionless measure of vari- 318 scales relevant to grid integration of individual wind 277 ability across the spatial domain. High rCV indicates 319 farms, as illustrated in the Wind Integration National 278 highly variable wind resource, thus less desirable for 320 Dataset (WIND) Toolkit by NREL [38]. Prevalence of 279 operation. 321 intense surface heating in the region requires proper 280 Intermittency is defined by availability and persistence 322 parameterization of the stability of the boundary layer 281 of the wind resource. Availability is the fraction of time 323 to better estimate the variation of wind speed with 282 over a time series when the WPD exceeds a threshold 324 height during day-time. It is also important to validate 4 325 model results with current in situ measurements at 326 different hub heights, such as the measurements taken 327 by K.A.CARE in Saudi Arabia [39].

Ar 30N Ma Aq R K 328 3. Results

W DM 199 25N Do A Y III Ar 100 Dl S II 30N Ma Aq R K J I 90 Latitude 20N

W DM 80 Do 25N Sa A Jz Y 70

Dl S 15N J 60 Latitude Ad 20N 50 35E 40E 45E 50E 55E

Jz Sa 40 Longitude

15N 30 Figure 2: Categorized median WPD (ﬁg.1) at 50 m AGL into −2 Ad three regimes: regime I where median WPD is above 67 W m , 11.5 regime II where median WPD is from 46 W m−2 to 67 W m−2, 35E 40E 45E 50E 55E and regime III where median WPD is below 46 W m−2. Eleva- Longitude tion contours at 400 m intervals are drawn using elevation data from MERRA. Selected locations of prior studies (table A.1) Figure 1: Median WPD (W m−2) computed at 50 m AGL using are shown. the wind ﬁelds reconstructed from the MERRA data. The color scale is uniform except for values beyond 100 W m−2 and below 30 W m−2. Large extreme values beyond the 98-percentile in the spatial domain are masked in grey. Selected locations of prior studies (table A.1) are shown.

532 329 3.1. Abundance

Ar 200 Ma 330 30N Near-surface wind resource over the Arabian Peninsula Aq R K 180 331 varies due to its diverse topographical features. We

332 discuss the wind resource over this vast area by W DM 160 Do 333 25N categorizing the spatial domain into three regimes A Y 140 334 based on the median WPD (ﬁg.1). The three wind Dl S 335 J 120

regimes are illustrated in ﬁg.2. Latitude 20N 100

336 3.1.1. Regime I (relatively abundant) Jz Sa 80

15N 337 The Arabian Peninsula shows a varied abundance of 60

338 the near-surface wind resource spatially. Regime I Ad 40.3 339 shows the most abundant wind resource in the region 35E 40E 45E 50E 55E −2 340 with median WPD above 67 W m . Offshore regions Longitude 341 have the most abundant wind resource. Among the −2 342 onshore locations, the Hejaz Mountains east of Jeddah Figure 3: Average WPD (W m ) computed at 50 m AGL using the wind fields reconstructed from the MERRA data. The color 343 in Saudi Arabia, the southern coast of Oman, and scale is uniform except for values beyond 200 W m−2 and below 344 eastern Yemen show a relatively high abundance of 60 W m−2. Large extreme values beyond the 98-percentile in 345 wind resource. The mountainous coastline along the spatial domain are masked in grey. Selected locations of 346 the Gulf of Aqaba and central Jordan also have prior studies (table A.1) are shown. 347 moderately abundant wind resource. These regions 348 show a comparable spatial distribution of average and 349 median WPD and the average WPD is about twice the 5 −2 350 median, illustrating a positively skewed distribution 402 WPD is around 105 W m . In central Jordan, average −2 351 (fig. A.2). One exception is along the coast of the Gulf 403 WPD is found to be 176 W m . This finding seems 352 of Aqaba, where the average and median WPD are 404 to deviate from prior observations where the wind 353 close, leading to a less skewed distribution. 405 resource is more abundant in the southwestern area. 406 In particular, Fujaij shows an average WPD of around 354 The abundance of the wind resource is illustrated in −2 −1 407 91.6 W m (4.3 m s ), where the annual average 355 average WPD (fig.3). A map of average wind speed is −1 408 wind speed of 6.88 m s was reported [44]. 356 also included for reference and comparison with prior 357 studies where WPD is not available (fig. A.3). The 358 East of the Hejaz and Asir mountains in Saudi Arabia 409 3.1.2. Regime II (moderately abundant) −2 359 has average WPD of 143 W m . This is confirmed 410 Regime II shows areas with a moderately rich wind 360 by recent meteorological measurements from 1998 to −2 −2 411 resource of median WPD between 46 and 67 W m . 361 2002 at Dhulum with average WPD of 186 W m 412 It includes the northern coast of the Red Sea, the 362 at 40 m AGL [40]. Before the use of reanalysis, 413 coast of the Arabian Gulf, the areas west of Riyadh 363 wind resource assessments relied on data collected at 414 and along the borders of Saudi Arabia with Yemen 364 meteorological stations and compared among known 415 and Oman, and northwestern Kuwait. 365 locations. The consensus from earlier studies has 366 been that coastal areas possess more abundant wind 416 The northern Arabian coast of the Red Sea has been 367 resources (e.g., [41]). Our discovery of higher wind 417 popular for wind resource studies due to long-standing 368 resources away from the coast points to the need for 418 meteorological measurements at major settlements. 369 more in situ observations at locations indicated in this 419 Our results show that the surroundings of Yanbu −2 370 study. 420 have an average WPD around 106 W m , close to −2 421 the observed 134 W m [40]. Our results indicate 371 Bahrain and Qatar show average WPD (wind speed) −2 −1 −2 422 that a more abundant wind resource is present within 372 of around 119 W m (4.63 m s ) and 139 W m −1 423 100 km of the meteorological stations. Northwest of 373 (4.71 m s ) respectively. A prior study [24] shows −1 424 Yanbu shows a higher wind average WPD of around 374 an average wind speed of 8.65 m s at 60 m AGL −2 425 138 W m . Comparable wind abundance can also 375 extrapolated from hourly observations at 10 m AGL 426 be found in the mountains further north of Yanbu 376 at Bahrain International Airport from 2003 to 2005. 427 and along the coast of the Red Sea northwest of 377 These observations point to the need for further inves- 428 Yanbu. This analysis identifies coastal sites beyond 378 tigations to better understand the disparity between 429 those discussed and observed in prior studies (i.e., 379 the median and average WPD shown in Bahrain and 430 Rehman and Ahmad [21]). 380 Qatar where the ocean strongly influences circulation 381 over the island and the peninsula. 431 The coast of the Arabian Gulf illustrates a highly pos- 432 itively skewed distribution of WPD where the average 382 Along the Oman coast, our results show an average −2 433 value is about three times the median, indicating a 383 WPD of about 197 W m for southern coast and −2 434 potentially high variability in wind power previously 384 85.8 W m near Sur. Meteorological observations 435 not reported in the literature. The average WPD near 385 reported at Thumrait, Sur, Masirah, and Marmul, −2 −2 436 Dhahran is shown to be around 136 W m , close to 386 show average WPD of 230, 194, 165, and 109 W m −2 437 the observed 154 W m [40]. 387 at 10 m AGL [42]. Further investigations into the 388 long-term time-series would help reveal climatological 438 Kuwait also shows a highly positively skewed dis- 389 differences in the two sub-regions. 439 tribution in WPD as illustrated by the difference 440 between the median (fig.1) and the average WPD 390 In eastern Yemen around the Hadramaut Mountains, 441 (fig.3). Northwestern Kuwait has an average WPD 391 we find an average WPD (wind speed) of about −2 −2 442 of around 164 W m , where southern Kuwait shows 392 110 W m . The region surrounding Aden shows −2 −2 −1 443 an average WPD of around 147 W m . A prior study 393 an average WPD of about 105 W m (4.75 m s ), 444 shows an average WPD at 30 m AGL at Umm Omara 394 where the wind speed decreases further eastward. A −2 445 (northwest) and Al-Wafra (south) to be 271 W m 395 prior study [43] using five years of meteorological −2 446 and 273 W m respectively. These discrepancies 396 observations from the airport at Aden shows an average −1 447 could arise from the choice of a constant exponent 397 wind speed of 4.5 m s at 10 m AGL. The more 448 of 1/7 when adjusting the wind speed from 10 m AGL 398 abundant wind resource appears in the west of Aden, 449 observations to 30 m [45]. The 1/7 constant exponent 399 where future in situ observations would benefit from 450 has been a popular choice in the wind resource 400 further studies into the local resource characteristics. 451 assessment literature in extrapolating wind speed to 401 Along the coast of the Gulf of Aqaba, the average 452 hub height and deemed inappropriate for domains with 6 453 complex surface characteristics [46]. This choice of the 480 by the varying terrain. Offshore locations show lower 454 exponent in prior studies did not account for surface 481 increase than onshore sites. These observations are 455 characteristics, leading to an overestimate of the wind 482 consistent across average and median WPD. With 456 resource. 483 increased hub height, both the magnitude and the 484 frequency of higher wind power resource increase and 485 shift the distribution to higher WPD values, making 457 3.1.3. Regime III (least abundant) 486 it less skewed. The rate of increase is proportionally 458 Regime III shows areas with the comparatively least 487 consistent across the average and median WPD at the 459 abundant wind resource in the region where median 488 two altitudes. −2 460 WPD is below 46 W m . The area includes the

461 southern Arabian coast of the Red Sea, the southern 489 3.1.5. Comparison with Vestas’ wind map over Saudi 462 coastline of Yemen, the east coast of Oman, and Abu 490 Arabia at 100 m AGL 463 Dhabi in the UAE. 7.17

464 The southern Arabian coast of the Red Sea has an Ar 6.6 −2 30N Ma 465 R average WPD (wind speed) of around 70.4 W m Aq K −1 6 466 (3.57 m s ), which is conﬁrmed by observations at

467 Jizan, where monthly average wind speed is between 5.4 −1 −1 W D M 468 3.8 m s and 5.2 m s [21]. The southern Yemen 25N Do 4.8 469 A coast also shows comparable wind power density, with Y 4.2 470 a similar mountainous terrain along the coastline. Latitude Dl S

471 Northeastern Oman and the UAE show low average J 3.6 −2 472 WPD of 80.7 W m , consistent with observations 20N 3 473 [25, 42]. 2.4

Jz Sa 0 474 3.1.4. Effects of variation of hub height 35E 40E 45E 50E 55E Longitude 48% average 80 m average 140 m Figure 5: Average wind speed (m s−1) computed at 100 m AGL 30N 45% using the wind fields reconstructed from the MERRA data. The −1 25N color scale is uniform except for values beyond 6.6 m s and 40% below 2.4 m s−1. Selected locations of prior studies (table A.1) 20N are shown. 35% 15N 491 30% Figure5 shows the annual average wind speed at median 80 m median 140 m 492 100 m AGL by our reconstruction and fig.6 shows

Latitude 25% 493 its comparison to Vestas’ estimation. Vestas’ wind 494 speed map was calculated using the output from 20% 495 the Weather Research and Forecasting (WRF) model 496 with a spatial resolution of 3 km × 3 km and hourly 15% 497 temporal resolutions from 2000 to 2013 [47]. The

10% 498 wind speed estimate is available at the Renewable 35E 40E 45E 50E 55E 499 Resource Atlas at King Abdullah City for Atomic Longitude 500 and Renewable Energy (K.A.CARE) in Saudi Arabia 501 [39]. The model was driven by the boundary con- Figure 4: Changes of median WPD (fig.1) and average WPD 502 ditions from the National Center for Environmental (fig.3) at 80 m and 140 m from those at 50 m AGL. Large 503 extreme values beyond the 98-percentile in the spatial domain Prediction (NCEP) Global Forecast System Analysis are masked in grey. The color scale is uniform except for values 504 of 1° × 1° spatial resolution and 6-hourly temporal beyond 45%. 505 resolution. Topography used in the Vestas model 506 was obtained from the Moderate Resolution Imaging 00 00 475 The rate of change of average and median WPD with 507 Spectroradiometer (MODIS) of 30 × 30 spatial 476 altitude depends on the roughness length (fig. A.4). 508 resolution. Figure5 indicates relatively high average 477 Roughness length defines the influence of surface 509 wind speed at the tip of the Gulf of Aqaba, north 478 characteristics on wind power resource. Figure4 shows 510 of Duba, west of Yanbu, south of Jeddah, and to 479 a general increase of WPD with altitude, much affected 511 the east of the Hejaz and Asir Mountains. Figure6 7 7.08 542 where surface-atmosphere interactions are accounted Ar 2.5 30N Ma 543 for. The 3TIER wind speed estimates were validated R Aq K 2 544 with 229 NCEP-ADP stations in the Middle East 1.5 545 and Africa, where only around 60 stations are in the 1 −1 W D M 546 Middle East [49]. The RMSE is 1.03 m s between 0.5 Do 25N 547 the annual average wind speed of station data and A Y 0 548 those from the model output in Africa and the Middle Latitude Dl -0.5 S549 East. However, the data for the 3TIER map are not J -1 550 retrievable. There is also a lack of publicly available -1.5 20N 551 documentation on the creation of the dataset. Hence, -2 552 the following discussion is not illustrated. -2.5 Jz Sa -2.9 553 Our wind speed reconstruction and 3TIER’s estimates 35E 40E 45E 50E 55E 554 are qualitatively similar. On the coast of the Red Sea, Longitude 555 the two maps agree on the lower wind speed between 556 the southern Red Sea and the mountains. Prominent Figure 6: Differences in the average wind speed (m s−1) at 100 m AGL between our calculation and that from estimation 557 wind patterns agree over the north, the east, and the of Vestas (Vestas - MERRA) are illustrated with a color scale 558 south of Saudi Arabia, the UAE, Qatar, Kuwait, the −1 that is uniform except for values beyond 2.5 m s and below 559 eastern and the southern coasts of Oman, and Yemen. −2.5 m s−1. Selected locations of prior studies (table A.1) are 560 shown. The MERRA dataset is regridded to 1 km resolution by Known locations at Yanbu, Jeddah, and Al Qahma nearest neighbor for illustration. 561 appear to show higher wind speed in the 3TIER map 562 than in the MERRA reconstruction.

512 shows that the wind speed patterns of the two maps 563 3.2. Variability 513 are qualitatively similar, with root-mean-square error −1 514 (RMSE) of 1.15 m s . However, at these identiﬁed 1.47 515 high wind locations the Vestas’ map holds higher

516 values than from our reconstructions using MERRA, Ar 1.45 Ma −1 30N R 517 with the Vestas’ average wind speed being 0.424 m s Aq K 1.4 518 higher on average. The area in the north near Arar D 519 and Rafha, and in the east near the coast of the W M 25N Do 1.35 A 520 Arabian Gulf show similar spatial patterns in wind Y 521 speed with the Vestas’ map, but higher values in Dl S 1.3 J 522 general. This qualitative assessment reflects the higher Latitude 20N 1.25 523 spatial variability that is shown in the model with 524 higher spatial resolution. A converse pattern is shown 1.2 Jz Sa 525 on the Vestas’ map where there is a higher wind speed 15N 1.15 526 east of Najran than in the Empty Quarter. There is Ad 527 a higher wind speed shown to the west of the Hejaz 0.994 528 Mountains where fine topographical features exist in 35E 40E 45E 50E 55E 529 the Vestas’ map. It remains to investigate further Longitude 530 these discrepancies and their relation to the models Figure 7: Robust coefficient of variation computed at 50 m AGL 531 and assimilation schemes. using the wind fields reconstructed from the MERRA data. The color scale is uniform except for values beyond 1.45 and below 1.15. Selected locations of prior studies (table A.1) are shown. 532 3.1.6. Comparison with 3TIER’s wind speed map at 533 80 m AGL 564 The robust coefficient of variation (fig.7) measures 534 The average wind speed at 80 m AGL from our 565 the median deviation as a fraction of median WPD. 535 reconstruction using the MERRA dataset is compared 566 A higher rCV indicates relatively higher variability 536 with 3TIER’s average wind speed map. The 3TIER 567 at a location in WPD. Low variability reduces the 537 wind map is available with 5 km spatial resolution 568 output volatility and is favorable for power generation 538 hosted at the Global Atlas for Renewable Energy of the 569 Figure7 indicates that the sea-ward side of elevated 539 International Renewable Energy Agency (IRENA) [48]. 570 areas shows greater variability than other sites in the 540 The 3TIER’s map uses over ten years of hourly data 571 spatial domain. Within the Regime I area, the inland 541 generated from statistical and dynamical downscaling 572 face of the Hejaz and Asir Mountains shows relatively 8 573 low variability. Similarly, the Hadramaut Mountains 601 reduced effect of boundary layer friction on the wind 574 in Yemen and the Wahiba Sands desert in eastern 602 by the surface. Increases in availability with hub 575 Oman show relatively low variability in WPD. Along 603 height are most significant in regions with low wind 576 the Gulf of Aqaba, the mountainous region also shows 604 abundance, except southern coastal Yemen. 577 relatively low variability. Within Regime II, both the 83% 578 80 m 140 m coastlines of the Red Sea and the Arabian Gulf show 60% 579 higher variability than inland locations. However, the 30N 50% 580 surroundings of Yanbu show only moderate variability. 40% 25N 35% 581 Changes in hub height lead to insignificant changes in 30% 582 variability (fig. A.5). Latitude 20N 25% 20% 15N 15% 583 3.3. Intermittency 4% 35E 40E 45E 50E 55E 584 Intermittency is characterized by availability and Longitude 585 persistence. Figure 9: Changes of availability (fig.8) at 80 m and 140 m from those at 50 m AGL. Large extreme values beyond the 98- 586 3.3.1. Availability percentile in the spatial domain are masked in grey. The color scale is uniform except for values beyond 40% and below 15%. 0.5

Ar 0.45 30N Ma Aq R K 605 3.3.2. Persistence 0.4 15 W DM 0.35 50 m 80 m Do 25N Ar Ar A 30N Ma Ma 10 Y 0.3 Aq R K Aq R K

Dl S W DM W DM 25N Do Do 9 J 0.25 Y A Y A Latitude S S 20N J J 0.2 20N 8 Jz Sa Jz Sa Jz Sa 0.15 15N 7 Ad Ad 15N 0.1 140 m 6

Latitude Ar Ad Ma 0.024 Aq R K 5 35E 40E 45E 50E 55E W DM Do Longitude Y A S 4 J

Figure 8: Availability of wind resource computed at 50 m AGL Jz Sa 3 using the wind fields reconstructed from the MERRA data. The color scale is uniform except for values beyond 0.45 and Ad 2 below 0.1. Large extreme values beyond the 98-percentile in the 35E 40E 45E 50E 55E spatial domain are masked in grey. Selected locations of prior Longitude studies (table A.1) are shown. Figure 10: Median length of wind episode (h) computed at 50 m, 587 Availability (fig.8) measures the fraction of time when 80 m, and 140 m AGL using the wind fields reconstructed from 588 the wind power resource is above a given threshold the MERRA data. The color scale is uniform except for values −2 beyond 10 h. Selected locations of prior studies (table A.1) are 589 (i.e., 200 W m ). It represents the amount of time shown. 590 with meaningful power generation by a turbine. Most 591 onshore locations have availability between 10% to 606 Median episode length (fig. 10) measures the persis- 592 30%, except for the low wind abundance areas in 607 tence of the wind resource, which indicates persistent 593 Yemen, northeast Oman, and along the southern 608 up-time of energy production at a specific location. In 594 coast of the Red Sea. Areas with abundant wind 609 Regime I, MEL is eight hours in the western mountains 595 resource appear to have high availability. The western 610 of Saudi Arabia, central Jordan, and the southern coast 596 mountains in Saudi Arabia, the southern coast of 611 of Oman. It indicates that the wind resource at these 597 Oman, the tip of the Gulf of Aden, and Kuwait show 612 locations is persistent for at least eight consecutive 598 availability of around 25%. 613 hours during half of the record period. At the tip 599 Figure9 shows that an increase in hub height 614 of Yemen at the Gulf of Aden, the MEL is 14 hours. 600 contributes to increased availability onshore due to 615 Moderate MEL above five hours by the coast of the 9 616 Gulf of Aqaba is noticed. Areas with moderate wind 666 [11, 52–54]. Specifically, wind power production in 617 speed near the Arabian Gulf such as Kuwait and 667 this region can be modeled in a way similar to those 618 eastern Saudi Arabia show a higher persistence of 668 performed in Northern Ireland [55], Great Britain [56], 619 the wind with the MEL of greater than eight hours. 669 and Sweden [57]. In light of an estimated 25 years 620 Low wind abundance regions exhibit persistence with 670 turbine lifetime [58], sustainable deployment of wind 621 the MEL of below five hours. The MEL tends to 671 energy systems in the region requires assessments of 622 increase with hub height with almost no exceptions 672 the effects of climate change on the regional wind 623 (fig. A.6). 673 resource using developed approach [59]. A measure- 674 correlate-predict (MCP) approach can be used to 675 estimate the long-term wind resources at a target

676 624 4. Conclusions site using our reconstructed wind fields in conjunction 677 with short-term wind measurement campaigns [60]. 678 An investigation into the best pattern for wind power 625 This study provides the first regional assessment of 679 aggregation through region-wide interconnection to 626 the abundance, variability, and intermittency of wind 680 mitigate intermittency will be a timely and valuable 627 resource over the Arabian Peninsula. Employing 681 next step towards an optimal integration of large-scale 628 the MERRA dataset, the wind field at different hub 682 wind energy systems in the Arabian Peninsula [61]. 629 heights is reconstructed applying similarity theory 630 using the roughness length, friction velocity, and

631 displacement height. The reconstructed wind field 683 Acknowledgement 632 spans over three decades with consistent spatial and 633 temporal resolution. The wind power density field 684 Research reported in this publication was supported by 634 enables analysis of various aspects of the large-scale 685 the King Abdullah University of Science and Technol- 635 features of the wind energy resources in the Arabian 686 ogy (KAUST) and the Saudi Basic Industries Corpo- 636 Peninsula. The wind resource is also characterized 687 ration (SABIC) under grant number RGC/3/1815-01. 637 using metrics of wind variability and persistence. This 688 For computer time, this research used the resources 638 work improves upon the earlier comprehensive studies 689 of the Supercomputing Laboratory at King Abdullah 639 in providing an areal overview of the wind resource 690 University of Science and Technology (KAUST) in 640 with higher spatial resolution and metrics of temporal 691 Thuwal, Saudi Arabia. MERRA data used in this 641 variations in the wind. Previously unrecognized 692 study have been provided by the Global Modeling and 642 locations of interest with high wind abundance and low 693 Assimilation Office (GMAO) at NASA Goddard Space 643 variability and intermittency have been identified in 694 Flight Center through the NASA GES DISC online 644 this study and confirmed by recent on-site observations 695 archive. We thank the two anonymous reviewers for 645 [40]. In particular, the western mountains of Saudi 696 their careful reading of our manuscript and their many 646 Arabia experience more abundant wind resource than 697 insightful comments and suggestions. 647 most Red Sea coastal areas. The wind resource is 648 more variable in coastal areas along the Arabian Gulf 649 than their Red Sea counterparts at a similar latitude. 698 A. Appendix 650 More persistent wind is also found along the coast of 651 the Arabian Gulf.

699 References 652 This analysis points to the areas previously not recog-

653 nized. Studies at finer resolutions for these identified 700 [1] Marvel K, Kravitz B, Caldeira K. Geophysical limits to 654 areas are necessary to resolve spatial features and local 701 global wind power. Nature Climate Change 2013;3(2):118– 702 21. doi:10.1038/nclimate1683. 655 circulations relevant to wind power generation. Our 703 [2] Barthelmie RJ, Pryor SC. Potential contribution of wind 656 reconstructed wind field will enable investigations on 704 energy to climate change mitigation. Nature Climate 657 the impact of large-scale circulations on regional wind 705 Change 2014;4(8):684–8. doi:10.1038/nclimate2269. 658 resources. Effects of the El Niño Southern Oscillation 706 [3] McElroy MB, Lu X, Nielsen CP, Wang Y. Poten- 707 tial for wind-generated electricity in china. Science 659 (ENSO) [50] and the North Atlantic Oscillation (NAO) 708 2009;325(5946):1378–80. doi:10.1126/science.1175706. 660 [51] at various hub heights can be assessed using 709 [4] Bryden J, Riahi L, Zissler R. MENA renewables 661 the reconstructed wind fields at different hub heights. 710 status report. Tech. Rep.; 2013. URL: http: 662 The economic viability of wind energy applications 711 //www.ren21.net/Portals/0/documents/activities/ 712 Regional%20Reports/MENA_2013_highres.pdf. 663 at a regional scale can be conducted with higher 713 [5] El-Katiri L, Fattouh B, Oxford Institute for Energy Studies 664 spatial and temporal resolution using this dataset 714 . A roadmap for renewable energy in the Middle East and 665 to provide greater consistency than in prior studies 715 North Africa. 2014. ISBN 9781907555909 1907555900. 10 1.25

30 Countries 1.00 Bahrain Jordan

25 Kuwait 0.75 Hub height Oman 140m AGL Qatar Latitude 80m AGL 20 Saudi Arabia United Arab Emirates 0.50 Yemen Fractional change in WPD 15 0.25

35 40 45 50 55 60 Longitude 0.001 0.100 Median roughness length (m in log10-scale) Figure A.1: Country map in the Middle East Figure A.4: Fractional change in WPD at 80 m and 140 m from 50 m over temporal median of roughness lengths

90000

600

60000 count 400 Hub height 140m AGL

30000 count 80m AGL

200

0 250 500 750 1000 WPD (W m−2) 0

Figure A.2: Illustration of a skewed distribution of WPD at -0.004 0.000 0.004 0.008 Yanbu Fractional changes in rCV

Figure A.5: Distribution of fractional changes in rCV at 80 m

8.5 and 140 m from 50 m out of 1638 grid cells

Ar 8 30N Ma Aq R K 7.5

W DM 6.5 900 25N Do A Y 6 Dl S J 5.5 Latitude 600 Hub height 20N 5 140m AGL

count 80m AGL 4.5 Jz Sa 4 300 15N 3.5

Ad 3 35E 40E 45E 50E 55E 0 Longitude -2 -1 0 1 2 3 Changes in Median Episode Length (hour) Figure A.3: Average wind speed (m s−1) is computed at 50 m AGL using the wind ﬁelds reconstructed from the MERRA Figure A.6: Distribution of diﬀerences in MEL at 80 m and data. Selected locations of prior studies (table A.1) are shown. 140 m from 50 m out of 1638 grid cells

11 Mark Location 756 newable and Sustainable Energy Reviews 2011;15(3):1545– Ma Maan 757 55. doi:10.1016/j.rser.2010.11.055. Y Yanbu 758 [17] Kirby B, Ela E, Milligan M. Chapter 7 - analyzing the 759 impact of variable energy resources on power system re- Ar Arar 760 serves. In: Jones LE, editor. Renewable Energy Integration. R Rafha 761 Boston: Academic Press. ISBN 978-0-12-407910-6; 2014, D Dhahran 762 p. 83–99. 763 [18] Kwon SD. Uncertainty analysis of wind energy potential Aq Aqaba 764 assessment. Applied Energy 2010;87(3):856–65. doi:10. Sa Salalah 765 1016/j.apenergy.2009.08.038. W Al Wajh 766 [19] Lin J, Sun Yz, Cheng L, Gao Wz. Assessment of the power Jz Jizan 767 reduction of wind farms under extreme wind condition 768 by a high resolution simulation model. Applied Energy S Sur 769 2012;96:21–32. doi:10.1016/j.apenergy.2011.10.028. Do Doha 770 [20] Jung S, Arda Vanli O, Kwon SD. Wind energy potential M Manama 771 assessment considering the uncertainties due to limited A Abu Dhabi 772 data. Applied Energy 2013;102:1492–503. doi:10.1016/j. 773 apenergy.2012.09.011. Ad Aden 774 [21] Rehman S, Ahmad A. Assessment of wind energy potential J Jeddah 775 for coastal locations of the kingdom of saudi arabia. Energy K Kuwait 776 2004;29(8):1105–15. doi:10.1016/j.energy.2004.02.026. Dl Dhulum 777 [22] Rehman S, Halawani TO, Husain T. Weibull parameters 778 for wind speed distribution in saudi arabia. Solar Energy 779 1994;53(6):473–9. doi:10.1016/0038-092X(94)90126-M. Table A.1: Selected locations of prior studies in the region as 780 [23] Rehman S, Mahbub Alam AM, Meyer JP, Al-Hadhrami indicated on the maps. 781 LM. Wind speed characteristics and resource assessment 782 using weibull parameters. International Journal of Green 783 Energy 2012;9(8):800–14. doi:10.1080/15435075.2011. 784 641700. 716 [6] Smith A, Lott N, Vose R. The integrated surface database: 785 [24] Jowder FAL. Wind power analysis and site matching of 717 Recent developments and partnerships. Bulletin of the 786 wind turbine generators in kingdom of bahrain. Applied 718 American Meteorological Society 2011;92(6):704–8. doi:10. 787 Energy 2009;86(4):538–45. doi:10.1016/j.apenergy.2008. 719 1175/2011BAMS3015.1. 788 08.006. 720 [7] Ansari J, Madni IK, Bakhsh H. Saudi arabian wind energy 789 [25] Ouarda TBMJ, Charron C, Shin JY, Marpu PR, Al- 721 atlas. Riyadh, Saudi Arabia: KACST 1986;:1–27. 790 Mandoos AH, Al-Tamimi MH, et al. Probability distribu- 722 [8] Rehman S, Halawani TO. Statistical characteristics of 791 tions of wind speed in the UAE. Energy Conversion and 723 wind in saudi arabia. Renewable Energy 1994;4(8):949–56. 792 Management 2015;93:414–34. doi:10.1016/j.enconman. 724 doi:10.1016/0960-1481(94)90229-1. 793 2015.01.036. 725 [9] Ucar A, Balo F. Investigation of wind characteristics and 794 [26] Gunturu UB, Schlosser CA. Characterization of wind 726 assessment of wind-generation potentiality in uludağ-bursa, 795 power resource in the united states. Atmos Chem Phys 727 turkey. Applied Energy 2009;86(3):333–9. doi:10.1016/j. 796 2012;12(20):9687–702. doi:10.5194/acp-12-9687-2012. 728 apenergy.2008.05.001. 797 [27] Cosseron A, Gunturu UB, Schlosser CA. Characterization 729 [10] Fyrippis I, Axaopoulos PJ, Panayiotou G. Wind energy 798 of the wind power resource in europe and its intermittency. 730 potential assessment in naxos island, greece. Applied 799 Energy Procedia 2013;40:58–66. doi:10.1016/j.egypro. 731 Energy 2010;87(2):577–86. doi:10.1016/j.apenergy.2009. 800 2013.08.008. 732 05.031. 801 [28] Fant C, Gunturu B, Schlosser A. Characterizing wind power 733 [11] Akdağ SA, Bagiorgas HS, Mihalakakou G. Use of two- 802 resource reliability in southern africa. Applied Energy 734 component weibull mixtures in the analysis of wind 803 ????;doi:10.1016/j.apenergy.2015.08.069. 735 speed in the eastern mediterranean. Applied Energy 804 [29] Hallgren W, Gunturu UB, Schlosser A. The potential wind 736 2010;87(8):2566–73. doi:10.1016/j.apenergy.2010.02. 805 power resource in australia: A new perspective. PLoS ONE 737 033. 806 2014;9(7):e99608. doi:10.1371/journal.pone.0099608. 738 [12] Ohunakin O, Adaramola M, Oyewola O. Wind energy eval- 807 [30] Rienecker MM, Suarez MJ, Gelaro R, Todling R, Bacmeis- 739 uation for electricity generation using WECS in seven se- 808 ter J, Liu E, et al. MERRA: NASA’s modern-era 740 lected locations in nigeria. Applied Energy 2011;88(9):3197– 809 retrospective analysis for research and applications. Journal 741 206. doi:10.1016/j.apenergy.2011.03.022. 810 of Climate 2011;24(14):3624–48. doi:10.1175/JCLI-D-11- 742 [13] Jervase JA, Al-Lawati AM. Wind energy potential 811 00015.1. 743 assessment for the sultanate of oman. Renewable and 812 [31] Emeis S. Wind Energy Meteorology Atmospheric Physics 744 Sustainable Energy Reviews 2012;16(3):1496–507. doi:10. 813 for Wind Power Generation. Berlin, Heidelberg: Springer; 745 1016/j.rser.2011.12.011. 814 2013. ISBN 9783642305238 3642305237. 746 [14] Al-Yahyai S, Charabi Y, Al-Badi A, Gastli A. Nested 815 [32] Dyer AJ, Hicks BB. Flux-gradient relationships in the 747 ensemble NWP approach for wind energy assessment. 816 constant flux layer. Quarterly Journal of the Royal 748 Renewable Energy 2012;37(1):150–60. doi:10.1016/j. 817 Meteorological Society 1970;96(410):715–21. doi:10.1002/ 749 renene.2011.06.014. 818 qj.49709641012. 750 [15] Al-Yahyai S, Charabi Y. Assessment of large-scale wind 819 [33] Archer CL. Spatial and temporal distributions of u.s. 751 energy potential in the emerging city of duqm (oman). 820 winds and wind power at 80 m derived from measurements. 752 Renewable and Sustainable Energy Reviews 2015;47:438– 821 Journal of Geophysical Research 2003;108(D9). doi:10. 753 47. doi:10.1016/j.rser.2015.03.024. 822 1029/2002JD002076. 754 [16] Charabi Y, Al-Yahyai S, Gastli A. Evaluation of NWP per- 823 [34] Rehman S. Wind energy resources assessment for yanbo, 755 formance for wind energy resource assessment in oman. Re- 12 824 saudi arabia. Energy Conversion and Management 892 seta.2013.01.002. 825 2004;45(13–14):2019–32. doi:10.1016/j.enconman.2003. 893 [53] Rehman S, Halawani TO, Mohandes M. Wind power cost 826 11.009. 894 assessment at twenty locations in the kingdom of saudi 827 [35] Fırtın E, Güler Ö, Akdağ SA. Investigation of wind 895 arabia. Renewable Energy 2003;28(4):573–83. doi:10.1016/ 828 shear coefficients and their effect on electrical energy 896 S0960-1481(02)00063-0. 829 generation. Applied Energy 2011;88(11):4097–105. doi:10. 897 [54] Marafia AH, Ashour HA. Economics of off-shore/on- 830 1016/j.apenergy.2011.05.025. 898 shore wind energy systems in qatar. Renewable Energy 831 [36] Short W, Sullivan P. Regional energy deployment system 899 2003;28(12):1953–63. doi:10.1016/S0960-1481(03)00060- 832 (ReEDS). Tech. Rep. NREL/TP-6A20-46534; National 900 0. 833 Renewable Energy Laboratory; 2011. 901 [55] Kubik ML, Brayshaw DJ, Coker PJ, Barlow JF. Exploring 834 [37] Cassola F, Burlando M. Wind speed and wind energy 902 the role of reanalysis data in simulating regional wind 835 forecast through kalman filtering of numerical weather 903 generation variability over northern ireland. Renewable 836 prediction model output. Applied Energy 2012;99(0):154– 904 Energy 2013;57:558–61. doi:10.1016/j.renene.2013.02. 837 66. doi:10.1016/j.apenergy.2012.03.054. 905 012. 838 [38] Draxl C, Clifton A, Hodge BM, McCaa J. The wind 906 [56] Cannon DJ, Brayshaw DJ, Methven J, Coker PJ, Lenaghan 839 integration national dataset (WIND) toolkit. Applied 907 D. Using reanalysis data to quantify extreme wind power 840 Energy 2015;151:355–66. doi:10.1016/j.apenergy.2015. 908 generation statistics: A 33 year case study in great britain. 841 03.121. 909 Renewable Energy 2015;75:767–78. doi:10.1016/j.renene. 842 [39] King Abdullah City for Atomic and Renewable Energy 910 2014.10.024. 843 (K.A.CARE) . Renewable resource atlas. 2015. URL: 911 [57] Olauson J, Bergkvist M. Modelling the swedish wind power 844 https://rratlas.kacare.gov.sa. 912 production using MERRA reanalysis data. Renewable 845 [40] Al-Abbadi NM. Wind energy resource assessment 913 Energy 2015;76:717–25. doi:10.1016/j.renene.2014.11. 846 for five locations in saudi arabia. Renewable Energy 914 085. 847 2005;30(10):1489–99. doi:10.1016/j.renene.2004.11.013. 915 [58] Staffell I, Green R. How does wind farm performance 848 [41] Rehman S. Prospects of wind farm development in saudi 916 decline with age? Renewable Energy 2014;66:775–86. 849 arabia. Renewable Energy 2005;30(3):447–63. doi:10.1016/ 917 doi:10.1016/j.renene.2013.10.041. 850 j.renene.2004.04.008. 918 [59] Fant C, Adam Schlosser C, Strzepek K. The impact of 851 [42] AL-Yahyai S, Charabi Y, Gastli A, Al-Alawi S. Assessment 919 climate change on wind and solar resources in southern 852 of wind energy potential locations in oman using data 920 africa. Applied Energy 2015;doi:10.1016/j.apenergy. 853 from existing weather stations. Renewable and Sustainable 921 2015.03.042. 854 Energy Reviews 2010;14(5):1428–36. doi:10.1016/j.rser. 922 [60] Carta JA, Cabrera P, Matías JM, Castellano F. Compari- 855 2010.01.008. 923 son of feature selection methods using ANNs in MCP-wind 856 [43] Algifri AH. Wind energy potential in aden-yemen. Re- 924 speed methods. a case study. Applied Energy 2015;158:490– 857 newable Energy 1998;13(2):255–60. doi:10.1016/S0960- 925 507. doi:10.1016/j.apenergy.2015.08.102. 858 1481(97)00069-4. 926 [61] Gunturu UB, Schlosser CA. Behavior of the aggregate wind 859 [44] Bataineh KM, Dalalah D. Assessment of wind energy 927 resource in the ISO regions in the united states. Applied 860 potential for selected areas in jordan. Renewable Energy 928 Energy 2015;144:175–81. doi:10.1016/j.apenergy.2015. 861 2013;59:75–81. doi:10.1016/j.renene.2013.03.034. 929 02.013. 862 [45] Al-Nassar W, Alhajraf S, Al-Enizi A, Al-Awadhi L. Po- 863 tential wind power generation in the state of kuwait. 864 Renewable Energy 2005;30(14):2149–61. doi:10.1016/j. 865 renene.2005.01.002. 866 [46] Schwartz MN, Elliott DL. Wind shear characteristics at 867 central plains tall towers. Tech. Rep.; National Renewable 868 Energy Laboratory; 2006. URL: http://www.mapcruzin. 869 com/wind-power-publications/wind-issues/40019.pdf. 870 [47] Vestas . Mesoscale model description. Tech. Rep.; 2013. 871 URL: http://rratlas.kacare.gov.sa/RRMMDataPortal/ 872 Content/pdf/121019mazag_MesoscaleModelDescription. 873 pdf. 874 [48] International Renewable Energy Agency . Global atlas for 875 renewable energy 2.0. 2015. URL: http://irena.masdar.ac. 876 ae. 877 [49] 3TIER Inc. . 3tier global wind dataset annual mean 878 validation. Tech. Rep.; 3TIER Inc.; 2009. URL: http:// 879 www.3tier.com/static/ttcms/us/documents/publications/ 880 validations/3TIER_WP_FL_gbl_wind_validation.pdf. 881 [50] Klink K. Atmospheric circulation effects on wind speed 882 variability at turbine height. Journal of Applied Meteo- 883 rology and Climatology 2007;46(4):445–56. doi:10.1175/ 884 JAM2466.1. 885 [51] Cullen HM, Kaplan A, Arkin PA, deMenocal PB. Impact of 886 the north atlantic oscillation on middle eastern climate and 887 streamflow. Climatic Change 2002;55(3):315–38. doi:10. 888 1023/A:1020518305517. 889 [52] Shawon MJ, El Chaar L, Lamont LA. Overview of wind 890 energy and its cost in the middle east. Sustainable Energy 891 Technologies and Assessments 2013;2:1–11. doi:10.1016/j.