Performance of the WRF Model for Surface Wind Prediction around Qatar
1. Introduction
Wind fields around Qatar are important for many reasons. On the marine side, applications such as circulation and wave modeling, marine safety, coastal erosion, and contaminant transport all rely on suitable wind model predictions (e.g., Moeini et al. 2010; Liao and Kaihatu 2016). On the terrestrial side, emerging needs include studies pertaining to the installation of solar energy panels and the transport and settling of dust, which are becoming increasingly prominent (Shao 2008; Hamidi et al. 2014). Near-surface conditions for the Arabian Gulf may be obtained from “global” models run by various international agencies, such as the European Centre for Medium-Range Weather Forecasts (ECMWF), the U.S. National Center for Environmental Prediction (NCEP), NASA, and others; however, their resolution is coarse, typically on the order of 50 km × 50 km. This results in only a few grid points being located in Qatar. In fact, Moeini et al. (2010) have noted deficiencies in ECMWF wind fields for wave prediction in parts of the Arabian Gulf, such as the underestimation of wave heights resulting from their use. The underestimation is consistent with similar errors noted by Cavaleri and Bertotti (1997) in the context of global models applied to basins, such as the Mediterranean and the Adriatic.
Regional models that use fine-resolution grids are therefore often used; these are generally linked to the global models. For example, Bricheno et al. (2013) describe such efforts for an Irish Sea application and Olsson et al. (2013) for parts of the Gulf of Alaska. In regions such as Qatar, however, different weather patterns prevail. While the Arabian Gulf is dominated by northwesterly winds (the peaks of which are known as the shamal winds), the wind fields around Qatar can be complicated because of its size. Qatar is a narrow peninsula with an area of approximately 11 437 km2 extending northward from the southern coast of the Arabian Gulf (Fig. 1). With maximum diagonal lengths of approximately 160 and 88 km, it is surrounded by relatively shallow water, the average depth in the entire Arabian Gulf being approximately 35 m. The convex shape and narrowness of the peninsula renders it conducive to sea and land breezes from both sides, potentially leading to convergence and upward motions. The nature of such local spatial variations could be expected to depend on the strength of the synoptic forcing. For instance, shamal winds caused by strong synoptic forcing will tend to preclude small-scale thermal circulations, whereas strong sea and land breezes could be expected during periods of weak synoptic forcing.
Study area showing (left) topography and locations of 19 AWS and (right) two offshore buoys.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
To simulate near-surface atmospheric conditions, several factors must be represented adequately. These include topography, land use, soil properties, and the appropriate boundary layer parameterization (Shafran et al. 2000; Cheng and Steenburgh 2005). The topographic configuration in the Arabian Gulf (Fig. 1) affects the circulation patterns. The basinlike contours of the region, with the sharply rising Zagros Mountains of Iran to the north and northeast and a more gradual upsloping terrain to the west and southwest in Saudi Arabia, tend to direct the low-level airflow along a general northwest–southeast axis. This terrain complexity and variability may be an impediment to the currently available coarse-resolution global models. Also, land-use characteristics present a problem in terms of proper representation, owing to rapid urbanization and development in Qatar.
In terms of the PBL parameterizations, various schemes are available to represent the exchange of moisture, heat, and momentum in the atmospheric boundary layer occurring through turbulent eddies. These schemes affect the modeled kinematics and thermodynamic structure of the lower troposphere (Cohen et al. 2015). Several studies have compared the performances of different PBL schemes and concluded that no single scheme is suitable for all regions and conditions (Gunwani and Mohan 2017). Draxl et al. (2014) found that model performance strongly depends on atmospheric stability, while Pena Diaz et al. (2011) showed that PBL schemes performed quite differently for sites over sea and over land. The choice of the best model setup for a forecasting system for a particular region will thus depend on the typical distribution of atmospheric stability at the site (Draxl et al. 2014). Typically, PBL schemes may be classified as either local or nonlocal. In the case of local schemes, only those vertical levels that are directly adjacent to a given point directly affect variables at that point. On the other hand, in nonlocal schemes, even distant levels are used to determine variables at a given point. These differences may be expected to influence the simulations in a desertlike, highly convective environment. While, in general, certain disadvantages of local schemes have been noted by Stensrud (2007) and Cohen et al. (2015), Chaouch et al. (2017) have noted their superior performance in the United Arab Emirates (UAE) region for certain applications. Overall, there are few studies relating to the effect of various PBL schemes on modeled surface winds in the dry and arid regions of the Middle East.
In this context, we examine the performance of a regional version of the WRF Model applied to the area in and around Qatar, a region where the wind pattern changes on short spatial and temporal scales. This regional model is implemented through the dynamical downscaling of the global NCEP Final Analysis using the WRF Model (Skamarock et al. 2005), essentially following the approach of Olsson et al. (2013) and Carvalho et al. (2014). No fewer than 19 automated weather stations in Qatar and two offshore buoys (Fig. 1) are available for evaluation of model performance. These data sources comprise a relatively high density of measurements in the peninsula, providing a relatively rare scenario for model configuration and validation. In addition, we examine the ability of the model to suitably simulate the formation of land and sea breezes under different seasonal conditions in the region around Qatar. We also examine the difference in the modeled surface wind characteristics over land and over water. The fidelity of the simulated winds is then investigated by examining their effect on ocean waves through the use of a wave model and wave measurements. This detailed model implementation and validation study is intended to be the preliminary step prior to its use for forecasting.
The layout of the paper is as follows. Section 2 describes the sources of the wind and wave data available for the validation and the optimal choice of grid sizes and various parameters based on selected comparisons with data. Section 3 provides validation of simulated land and sea breezes, of selected shamal events, and of diurnal patterns over land and water (including the performance of PBL schemes to replicate these patterns). A consolidated comparison using data from all 19 automated weather stations (AWS) is also provided, along with validation using a wave model forced by the simulated winds and data from two offshore buoys. In section 4, the validated model results are used to estimate monthly mean wind characteristics for land and sea breezes for the Arabian Gulf. Conclusions are provided in section 5.
2. Data availability and model configuration
Observational data were obtained from 19 land-based automated weather stations and two offshore buoys deployed and maintained by the Qatar Meteorological Department (QMD). Figure 1 shows the locations of these measurement sites. Since historical data were not available, we downloaded current observations from the QMD website (http://qweather.gov.qa/EWSData.aspx) on an ongoing basis. Wind speed, wind direction, relative humidity, temperature, and significant wave height were archived for a period of 12 months from September 2015 to October 2016. For the automated weather stations, data available at 10-min intervals consist of wind speed and direction at the 10-m elevation, temperature and humidity at the 2-m elevation, and sea level pressure. The 10-min values were averaged on an hourly basis for convenience. For the offshore buoys, wind data were available at 30-min intervals at an elevation of 4 m; these were converted to corresponding values at the 10-m elevation by assuming neutral stability and a logarithmic profile. Significant wave heights were also available at 30-min intervals from the two buoys.
We used the WRF Model (version 3.7), developed by NCAR, for modeling mesoscale wind systems in the Arabian Gulf region. Two-way nested WRF simulations were performed with two grids (Fig. 1). The outer domain uses a 30-km grid resolution and the inner domain a 10-km resolution. We used the Dudhia scheme for shortwave radiation, the Rapid Radiative Transfer Model for longwave radiation, the WRF single-moment 3-class for microphysics, and the Noah scheme for land surface depiction. These options or other comparable ones are typically used for such applications (e.g., Viswanadhapalli et al. 2017). Both model domains utilize 39 vertical levels, with the model top situated at a pressure level of at 50 hPa and the lowest model level at approximately 30 m above the ground level. The initial and boundary conditions are obtained from the NCEP’s final operational analyses data at 1° × 1° resolution. The simulations were carried out for a 1-yr period starting from September 2015. The selection of this period was based on the availability of observational data for comparison.
Surface parameters like wind, temperature, and humidity can be influenced by the choice of the boundary layer parameterizations and model resolution, as well as the land surface representation. To examine the sensitivity of results to various boundary layer parameterizations, simulations were made with four commonly used schemes—Mellor–Yamada–Janjíc (MYJ), quasi-normal scale elimination (QNSE), Yonsei University (YSU), and the Asymmetric Convective Model, version 2 (ACM)—along with the surface layer schemes shown in Table 1. The MYJ and QNSE schemes are classified as turbulent kinetic energy closure schemes and allow only local transport. The YSU and ACM2 schemes, on the other hand, are widely used nonlocal schemes (details may be found in Hu et al. 2010).
Selected PBL and surface layer schemes.
For examining the sensitivity of the results to the different PBL parameterizations and for selecting a suitable one, a test period was randomly selected. The model was integrated for 54 h (the first 6 h as spinup) for simulating the diurnal evolution for 2 days (1800 UTC 11 May 2016–0000 UTC 14 May 2016). The measured wind speeds and directions at the different AWS locations (denoted as AWS 1, AWS 2, etc.) were compared with the simulations. For representative purposes, sample comparisons for AWS 1 and AWS 5 located in the middle of Qatar are shown in Figs. 2a,b, respectively. The plots show that, for the most part, the results of all four parameterizations are similar, although the MYJ and QNSE results may be viewed as slightly inferior relative to the measurements, especially during the daytime (the diurnal effect of the PBL schemes is discussed in detail later). Regarding the other two schemes, we resorted to the ACM, although YSU may produce similar results.
Wind speed comparison based on different PBL schemes at locations (a) AWS1 and (b) AWS5, and different grid resolutions at AWS1 for part of (c) May and (d) June 2016.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
To identify the appropriate resolution for long-term simulations, two grid sizes were tested: 10 km × 10 km (considerably smaller than the 50-km resolution used by global models) and 3.3 km × 3.3 km. Figures 2c,d show representative results for two events, namely, 11–14 May 2016 and 7–10 June 2016, respectively. These correspond to a synoptically weak period during the winter-to-summer transition (Fig. 2c) and to a synoptically strong shamal wind period (Fig. 2d). In both cases 54 h of continuous simulation was performed with one-way nesting. The plots indicate no significant difference in between the model results, and both cases appear to reproduce the salient features of the measurements. Therefore, we have resorted to the 10-km grid size for the remaining simulations in this paper.
To obtain a suitable model description of land cover, we rely on the USGS land-use characterization (Fig. 3) based on Loveland et al. (2000). This characterization indicates some “shrubland” areas in the northwest surrounded by “barren” regions elsewhere. Model results were compared with measurements at the 19 automated weather stations. Since time series comparisons can be influenced by random fluctuations, we examined systematic errors using the mean diurnal cycle for a 1-yr period (September 2015–October 2016). Sample results are shown in Fig. 4, which indicate that the simulations reproduced measurements well at several stations (e.g., 1, 3, 5, 6, 8, and 16). However, at the remaining stations (2, 12, 15, and 19) overestimation was seen. The source of these discrepancies was explored. The default land-cover characterization was based on AVHRR data spanning April 1992–March 1993. The MODIS IGBP 20-category land-use characterization incorporates more recent observations (circa 2000); however, our examinations showed the two datasets to be comparable in the study area. Recent years have witnessed considerable changes in the landscape in Qatar, including extensive urbanization and expansion of industrial facilities. For example, between 1997 and 2010, land use/land cover near the capital Doha in Qatar appears to have increased by 280% in the built-up areas and by 426% in recreational areas, according to Hashem and Balakrishnan (2015), who also projected a 20% increase by 2020. We therefore attempted to manually make some adjustments based on Google Earth images. After several trials covering grids in various parts of the country, the region near Doha was specified as “urban.” A similar characterization was assigned to some grids in the northeast, in the vicinity of Ras Laffan Industrial City. The modified land use is shown in Fig. 3 (bottom right). Effectively, the roughness length in the model was altered from 0.01 m (for sparsely vegetated areas) to 0.5 m (for urban areas). The results showed considerable improvement at urban stations (e.g., at 2, 12, 15, and 19). Elsewhere, this land-use modification does not have a significant effect. In the remaining simulations, we therefore adopt this modified land-use description. The sensitivity of the local wind simulation to the land-use/land-cover description (seen in Fig. 4 for some stations) suggests that accurate representation is necessary for reproducing reliable results.
(top) Default land use for model domain 2. (bottom) Land use zoomed around Qatar: (left) default and (right) modified, where A is Doha and B is Ras Laffan Industrial City.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Composite (annual) diurnal wind speed comparison.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
3. Analysis and discussion
We consider three scenarios: 1) simulations that help elucidate the formation of land and sea breezes, 2) simulations of specific high-intensity Shamal wind events, and 3) simulations to examine contrasts in the diurnal patterns over land and water. In each scenario we have attempted to qualitatively and quantitatively assess model performance. We also obtain bulk statistics, such as root-mean-square error, bias, and correlation coefficients based on a yearlong comparison with data from the 19 AWS and two offshore buoy locations. Further, we provide indirect validation of the modeled winds by examining their ability to reproduce observed wave conditions through the use of a wave model.
a. Land and sea breezes
As stated earlier, land and sea breezes are dependent on the strength of thermal gradients and synoptic forcing. High thermal gradients are typical of the summer months. Figure 5 shows a sample simulation for this season at a time of weak synoptic forcing for the morning and afternoon hours. The pattern of land and sea breezes is apparent for many parts of the gulf region in Figs. 5a,b. On a regional scale too, the wind pattern around Qatar, illustrated in Fig. 5c, shows the sea-breeze dynamics clearly. The modeled wind vectors are comparable to the measurements provided by the automated weather stations, which inspires confidence in the predictions. In this case the sea breeze develops from both sides of the coast and converges in the middle of the peninsula. This can lead to the development of a zone of upward motions in the middle of Qatar, which can contribute to cloud formation and local convective rains under favorable conditions. This is consistent with the observations of Al Mamoon and Rahman (2017), who noted the propensity for rainfall in the middle of the country during the summer months. When the synoptic forcing is stronger—for example, when winds with a magnitude of 5–7 m s−1 from the northwest are present—a sea breeze penetrating inland from the east coast cannot fully develop because of the strong opposing winds. The condition on 3 June 2016 is an example of such a case (Fig. 6). The wind reversal (i.e., sea breeze) occurs only along the southeast coast of Qatar, which is somewhat sheltered from the prevailing winds. This leads to the convergence zone shifting eastward (compare with Fig. 5c). When the synoptic forcing is strongest, namely, during the frequent summer shamal events, the high wind speeds preclude the development of clearly identifiable land and sea breezes, as can be seen in Fig. 7. Thus, the model appears to be appropriately capturing the effects of variable synoptic forcing in a qualitative sense; quantitatively, too, the results in Figs. 6, 7 also are mostly confirmed by the measurements (red vectors).
Modeled winds (m s−1) for 18 Sep 2015 (a) at 0300 UTC, showing a land breeze; (b) at 0900 UTC, showing a sea breeze; and (c) an enlarged version of (b) showing detail around Qatar (AWS, red; WRF, black).
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Wind speeds (m s−1) at 1200 UTC 3 Jun 2016 (AWS, red; WRF, black).
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Simulations (m s−1) showing relatively constant wind direction around Qatar at (left) 0300 and (right) 1400 UTC (AWS, red; WRF, black).
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
b. Shamal wind events
As noted by Rao et al. (2003), the dominant northwesterly shamal winds have distinct features in the winter and the summer. The winter shamals (October–March) are related to midlatitude disturbances and frontal systems, and are characterized by abrupt and intense winds often accompanied by thunderstorms and high seas. Summer shamals (May–August) occur as a result of the pressure difference between the monsoon low pressure system centered over northern India and the stationary high pressure system centered over the eastern Mediterranean. These pressure differences are responsible not only for strong winds but also for dust storms that occur during the summer months (see Yu et al. 2016 for details).
Typically, a winter shamal system is initiated in the northwestern part of the gulf and then spreads southeast behind an advancing cold front. Moderate to strong winds can suspend dust and reduce visibility. Storm surges generated by the shamal winds, coupled with tidal effects, can lead to several meters of sea level rise (El-Sabh and Murty 1989). One such event occurred 2–5 January 2016. Figure 8a shows that the model was able to capture the rapid initial increase of the wind speed (from about 5 to 14 m s−1) starting on 3 June and the subsequent slow decline observed in the north buoy measurements. The preceding and subsequent smaller events are also reproduced well. A well-defined summer shamal event occurred in June 2016. Modeled wind speeds for this event, shown in Figs. 8b,c, bear considerable fidelity to the measurements. These two simulations demonstrate the models’ capability to appropriately simulate shamal events.
Wind speed comparison (a) during a winter shamal period at the north buoy, (b) during a summer shamal period at the north buoy, and (c) during a summer shamal period at the south buoy.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
c. Diurnal patterns over land and water
Comparisons of model results with data from the land-based automated weather stations and ocean-based buoys often showed differences in the diurnal patterns. Figure 9 shows one such comparison for 9–18 October 2015. Over both land and water, the model appears to capture the salient features of the measurements reasonably well. However, at the land-based station, diurnal fluctuations on the order of 6 m s−1 are seen. These are due to thermal mixing of the atmospheric boundary layer, as described later. Over the ocean, though, the fluctuations occur only at low wind speeds (right part of Fig. 9, bottom frame). At higher wind speeds (left part of Fig. 9, bottom frame), the synoptic features dominate, overwhelming the diurnal fluctuations. The general rise and fall of the wind speed from 10 to 15 October observed over water are seen over land as well, as indicated by the dashed lines connecting the diurnal peaks. However, this pattern is subsumed by the diurnal fluctuations.
Wind speed comparison at (top) AWS 1 and (bottom) the north buoy.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
A clearer representation of the hourly fluctuations may be obtained by estimating yearly averages for each hour. Toward this end, the model was run for the entire year. The results, shown in Fig. 10, indicate that the diurnal signal is predominant on land. At the location of AWS 1 (in the middle of Qatar), the observed maximum of 5.5 m s−1 occurs approximately at 1400 UTC and the minimum of approximately 3 m s−1 at 0400 UTC. Just after midday, solar heating and hence the surface temperature reach a peak, leading to an unstable surface layer, the mixing of momentum, and consequently higher surface wind speeds. During the night, as a result of longwave cooling, a stable layer forms near the surface, which precludes mixing, leading to low surface winds. The magnitude of the diurnal oscillation is about 2.5 m s−1. Over water, in contrast, the diurnal fluctuations are considerably smaller. The model captures these patterns in both cases, although with a slight overestimation. These differences between land and water locations also manifest themselves in the vertical wind profiles. For the purpose of illustrating the diurnal fluctuations, we choose one day, 3 January 2016, by way of example. Figure 11 (top) shows the modeled wind profiles for time instants corresponding typically to day and night. The unstable surface layer and momentum mixing lead to vertical homogeneity for both day and night over water, but only during the daytime over land. The middle plots in Fig. 11 show the model-derived eddy diffusivity profiles. Over land, the diffusivities are much smaller at night than during the day, leading to greater vertical variations in the wind speeds, as shown in the top plots. Over water the eddy diffusivities during day and night are essentially the same and are somewhat similar to those over land during the daytime. This may be attributed to higher sea surface temperatures than to the overlaying air temperature, leading to an upward sensible heat flux (as shown in Fig. 11, bottom), convective conditions, and vertical mixing, and hence homogeneous wind profiles. Over land at night, on other hand, Fig. 11 (bottom) shows a downward heat flux because the skin temperatures are lower than the air temperatures; this precludes vertical mixing and leads to nonhomogeneous profiles. Simulations were also performed using different PBL schemes. While there were some differences (not shown), the overall pattern produced by all schemes was similar to that shown in Fig. 11; the absence of vertical profile data prevents a more detailed investigation.
Diurnal cycle of wind over land (red) and water (blue).
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Simulated (top) vertical wind profiles and (middle) eddy diffusivity for 3 Jan 2016 night (0000 UTC, black) and day (1200 UTC, red) at AWS 1 (over land) and at the north buoy (over water). (bottom) Modeled sensible heat flux at AWS1 and the north buoy.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
d. Validation based on bulk parameters
It is difficult to provide detailed comparisons (such as those shown in the previous section) for all wind events that occurred during the year. Therefore, to assess model performance over an extended period, bulk parameters were used. First, we consider wind speed histograms, that is, the percent of time that wind speeds of a specified magnitude occur. This was estimated using model results and measured data at half-hour intervals at the location of the two offshore buoys and the 19 land-based stations. The histograms are plotted for the north and south buoys in Figs. 12a,b, respectively. The frequency of low wind speed occurrences are slightly underpredicted by the model. The histogram for the north buoy (in the middle of the Arabian Gulf) shows a slightly bimodal distribution (relative to the south buoy), which is a typical feature of channeled flows (Jaramillo and Borja 2004). Similar histograms are plotted in Fig. 12c for the 19 AWS locations. Though the overall variations are generally well captured, the frequencies of low wind speed occurrences (<3 m s−1) and high wind speed occurrences (>13 m s−1) are underestimated. On the other hand, the occurrences in the interval 4–12 m s−1 appear more frequently in the model than in the observations.
Observed and modeled histogram at the (a) north buoy, (b) south buoy, and (c) 19 AWS locations.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
To illustrate the directional variability of the wind, the AWS observations are shown as wind rose plots (Fig. 13); the same is done for WRF simulations as well. The frequency distribution seen in the model as well as in the observations suggests that the predominant wind direction is from the north-northwest. While observations show greater scatter in the wind direction compared to the model, the northwesterly dominance is clearly reproduced.
Wind rose plot (wind speed, m s−1) based on the (left) 19 AWS data and (right) WRF Model.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Statistical measures are used to compare the modeled and observed wind speeds for the entire period of one year for all 19 AWS locations and the two offshore locations. These are shown in Table 2. The values shown are comparable to those reported elsewhere (e.g., Langodan et al. 2016). The results shown in the table also suggest better model performance at the offshore buoy locations than at the land-based AWS locations. This is not surprising, since model parameters (the complicated land-use descriptors, the associated nonhomogeneous friction coefficients, etc.) are difficult to accurately represent over land compared to the ocean. As seen from the bias, an overestimation of wind occurs over both land and water; however, the bias is much lower over water compared to that over land. The diurnal variation in the bias is plotted in Fig. 14 for AWS location A1 (as an example of overland model performance) and for the north buoy (over water). The dependence of the bias on the choice of the PBL schemes is also shown. It is obvious that all schemes lead to overprediction. Overprediction of this kind has by reported by other researchers as well (e.g., Zhang et al. 2013; Hariprasad et al. 2014; Madala et al. 2016). The overprediction is significant during the daytime (when convective condition are expected to prevail over land) while using the local schemes. This can be attributed to the deficiencies of local PBL schemes in highly convective conditions because, during the daytime, large eddies prevail and local schemes rely on only the characteristics relating to the immediate proximity of a given point. A scheme relying on nonlocal mixing (which better represents the large eddies) would be expected to yield better performance, as confirmed by Fig. 14. During the nighttime all schemes yield similar results and somewhat less overprediction. At the location of the north buoy (i.e., over water), all the schemes show similar performance. During the nighttime the model marginally overpredicts the wind speeds and during the daytime underpredicts them. The pattern of bias at this location is different from that over land. This diurnal difference in bias is indicative of the performance of the local and nonlocal PBL schemes’ performance relative to atmospheric stability.
WRF performance for surface winds.
Diurnal cycle of wind bias for different PBL schemes (top) over land and (bottom) over water.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
e. Wind validation using wave simulation
The wave conditions needed for many practical applications, such as maritime safety, offshore design, wave energy studies, etc., are generally estimated through the use of wave prediction models forced by winds [see Etemad-Shahidi et al. (2011) and Liao and Kaihatu (2016) for such studies in coastal regions of the Arabian Gulf]. It is generally recognized that errors in wind fields tend to be a major cause of errors in predicted wave fields (e.g., Signell et al. 2005; Bhaskaran et al. 2013). The availability of wave data at two offshore buoys therefore serves as an opportunity for indirect validation of the reliability of the wind simulations for practical uses. A wave simulation of 1-yr duration was performed using the Simulating Waves Nearshore (SWAN) model (Booij et al. 1999). SWAN is a third-generation wave model specifically designed for simulating waves in shallow waters and accommodates wind-induced wave generation, energy transfer caused by quadruplet and triad wave–wave interaction, and dissipation caused by breaking and bottom friction. The model is widely used for various applications, such as siting aquaculture operations (Panchang et al. 2008) and offshore oil platforms (Panchang et al. 2013), wave forecasting (Singhal et al. 2010), etc.
For running the wave model, the Arabian Gulf region was discretized into 315 × 315 grids with 2′ × 2′ resolution, and the computational time step was selected to be 30 min. The model was forced with WRF winds (see section 2). Significant wave height (SWH) data from the two offshore buoys have been used to evaluate model performance qualitatively and quantitatively. For demonstration, three periods characterized by high wind speeds are chosen. The first two periods correspond to winter shamal conditions (Figs. 15a–d, 15e–h, respectively) that cover a period of approximately 10 days, and the third period to a summer shamal (Figs. 15i–l) that covers a whole month. In all cases, the patterns in the wave field appear to follow the patterns in the wind fields. However, the wave heights at the south buoy are approximately half those at the north buoy, in spite of the wind speed being about the same at both locations. The comparisons shown in Fig. 15 indicate that both the wind model and the wave model yield satisfactory simulations.
Comparison of (right) wind speed and (left) SWH for (a)–(d) 7–17 Oct 2015, (e)–(h) 3–13 Feb 2016, and (i)–(l) 17 May–7 Jun 2016.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
Modeled and measured significant wave heights are compared in Table 3, which suggests that there is a good match with the model results and observations at the north buoy. The high correlation coefficient, the low mean absolute deviation, the low bias, and the low root-mean-square difference suggest that the model performance is reliable for most practical applications. A detailed investigation revealed that the maximum underestimation of about 0.46 m occurs in the 99th percentile (i.e., the underestimation is largely at the peaks). The deviation between the modeled and observed wave heights at the south buoy is somewhat higher than at the north buoy (Table 3), but in both cases the model appears to capture the salient features of the measurements reasonably well. The mean wave height at the north and the south buoys are 0.77 and 0.49 m, respectively. The lower wave climate near the south buoy may be attributed to fetch limitations. The effect of fetch-limited wind-wave generation on the eastern side of Qatar is also reported by Liao and Kaihatu (2016).The wave field near the south buoy is influenced by more rapid fluctuations in the winds—that is, land and sea breezes—which are more prominent than in the vicinity of the north buoy (where synoptic features tend to prevail).
Wave model performances based on wave height comparison.
4. Mean wind conditions in and around Qatar
The assessment of model performance demonstrated in the previous section inspires confidence in the simulations that can then be used for various applications. The results are therefore used to determine monthly mean wind features that can serve, for example, as input to contaminant transport models (e.g., Elhakeem et al. 2015). For the entire Arabian Gulf, Zhu and Atkinson (2004) have observed sea-breeze and land breeze circulations to be a perennial feature with vertical and horizontal extent varying seasonally and diurnally. On the local scale, a systematic climatological analysis of sea-breeze and land breeze circulation was done by Eager et al. (2008) for the UAE region. In Qatar, using observations from one coastal location, Singha and Sadr (2012) examined the diurnal variations for one month (June 2011) from the perspective of atmospheric turbulence; however, studies with greater spatial coverage are lacking and are hence addressed below.
As stated by Zhu and Atkinson (2004), the Arabian Gulf experiences four main periods with distinct tropospheric circulations. The winter season (November–February) is characterized by northwesterly winds. The summer season (June–September) is influenced by an extension of the Indian monsoon low. The March–May period and October represent the transition periods without strong circulations. From the full-year simulations performed, here we choose October 2015, January 2016, April 2016, and July 2016 to show the mean characteristics as representative of the four periods.
The overall monthly averages (for all 24 h) as well as the monthly averages for the early morning (0300 UTC) and afternoon (1400 UTC) were calculated. These are displayed in Fig. 16 in and around Qatar. It can be seen that during October, the mean condition is calm and not characterized by any dominant circulation pattern. The figure shows a flow reversal along the coast during the day (i.e., the land /sea-breeze effect is pronounced). Along the coast of Qatar, there is a band of very low wind speeds in the morning. This is due to the land breeze opposing the mean flow pattern. In the afternoon hours, a sea breeze is clearly dominant, particularly on the east coast.
Monthly mean (m s−1) of hourly WRF wind simulations.
Citation: Journal of Atmospheric and Oceanic Technology 35, 3; 10.1175/JTECH-D-17-0125.1
In January, the mean flow is considerably stronger (5 m s−1) than in October and in fact it is higher than in the other months. This precludes the development of meaningful land breezes, especially on the west coast. The afternoon hours see a shift in the winds toward the south. Relative to January, April sees a decline in the northwesterly wind flow. Wind reversal between night and day is apparent, with a significant sea-breeze effect on the southern Arabian Gulf coast. In the morning hour, the western side of Qatar is calmer than the eastern side. In the afternoon, both coasts of Qatar experience sea breezes and there is a tendency toward convergence in the southern part of country. In July (summer) southerly winds dominate in terms of the mean. The morning and afternoon hours experience conditions similar to the winter, but there is a greater propensity for sea breezes to develop.
In general, the Arabian Gulf and the peninsula of Qatar experience land/sea breezes throughout the year. Sea breezes are more predominant than land breezes as result of the high land/sea temperature contrast during the daytime (vide supra). The strength of sea breeze is greater during the summer than during the winter. The results for the complete model domain (not shown) indicated that the southern coast of the Arabian Gulf is more conducive to sea breezes throughout the year. In Qatar, in the summer, under favorable conditions, sea breezes develop on both coasts and wind convergence occurs in the middle of the peninsula.
5. Summary and conclusions
To provide wind information for various applications in and around Qatar, the WRF Model has been used after testing appropriate grid sizes, boundary layer parameterizations, and land-use/land-cover specifications. While a 1-yr simulation confirms that land and sea breezes and the shamal winds dominate the wind pattern on this peninsula, the configurations used have resulted in capturing reliably the spatial and temporal variability. Several individual land breeze/sea-breeze cases were identified, and a comparison using data at the 19 AWS locations showed realistic model performance, including convergence in the middle of the country when the large-scale winds are favorable. Similarly, significant shamal wind events were also reliably reproduced, based on data from offshore buoys. The seasonal characteristics of land and sea breezes are identified and described. The differences in the diurnal evolution of the wind at inland and ocean locations were examined. The sensitivity of PBL parameterization to surface wind bias was examined. Over land, nonlocal schemes were proven to be superior to local schemes, especially during the daytime when highly convective condition prevail. The land locations showed high diurnal variation compared to offshore locations. This may be attributed to a thermally unstable boundary layer leading to vertical homogeneity, and it was affirmed by the model (Fig. 11). The yearlong surface winds at the 19 AWS locations and the two buoy locations show RMSEs of 2.6 m s−1 on land and 1.7 m s−1 over water; these are similar to other studies (e.g., Viswanadhapalli et al. 2017). The usefulness of the modeled winds has been demonstrated by using them to force a coastal wave model; the similarity of the resulting wave heights to observations serves as an indirect verification of wind simulations. The results lend confidence to the expected use of this model configuration for ocean forecasting purposes (for modeling contaminant transport, for maritime safety, etc.), as well as terrestrial forecasting (e.g., for modeling dust transport, which is important in this region).
Acknowledgments
This work was supported by the Qatar National Research Fund (a member of the Qatar Foundation) under Grant NPRP 7-660-1-124.
REFERENCES
Al Mamoon, A., and A. Rahman, 2017: Rainfall in Qatar: Is it changing? Nat. Hazards, 85, 453–470, https://doi.org/10.1007/s11069-016-2576-6.
Bhaskaran, P. K., S. Nayak, S. R. Bonthu, P. N. Murty, and D. Sen, 2013: Performance and validation of a coupled parallel ADCIRC–SWAN model for THANE cyclone in the Bay of Bengal. Environ. Fluid Mech., 13, 601–623, https://doi.org/10.1007/s10652-013-9284-5.
Booij, N., R. C. Ris, and L. H. Holthuijsen, 1999: A third‐generation wave model for coastal regions: 1. Model description and validation. J. Geophys. Res., 104, 7649–7666, https://doi.org/10.1029/98JC02622.
Bricheno, L. M., A. Soret, J. Wolf, O. Jorba, and J. M. Baldasano, 2013: Effect of high-resolution meteorological forcing on nearshore wave and current model performance. J. Atmos. Oceanic Technol., 30, 1021–1037, https://doi.org/10.1175/JTECH-D-12-00087.1.
Carvalho, D., A. Rocha, M. Gómez-Gesteira, and C. S. Santos, 2014: WRF wind simulation and wind energy production estimates forced by different reanalyses: Comparison with observed data for Portugal. Appl. Energy, 117, 116–126, https://doi.org/10.1016/j.apenergy.2013.12.001.
Cavaleri, L., and L. Bertotti, 1997: In search of the correct wind and wave fields in a minor basin. Mon. Wea. Rev., 125, 1964–1975, https://doi.org/10.1175/1520-0493(1997)125<1964:ISOTCW>2.0.CO;2.
Chaouch, N., M. Temimi, M. Weston, and H. Ghedira, 2017: Sensitivity of the meteorological model WRF-ARW to planetary boundary layer schemes during fog conditions in a coastal arid region. Atmos. Res., 187, 106–127, https://doi.org/10.1016/j.atmosres.2016.12.009.
Cheng, W. Y., and W. J. Steenburgh, 2005: Evaluation of surface sensible weather forecasts by the WRF and the Eta models over the western United States. Wea. Forecasting, 20, 812–821, https://doi.org/10.1175/WAF885.1.
Cohen, A. E., S. M. Cavallo, M. C. Coniglio, and H. E. Brooks, 2015: A review of planetary boundary layer parameterization schemes and their sensitivity in simulating southeastern U.S. cold season severe weather environments. Wea. Forecasting, 30, 591–612, https://doi.org/10.1175/WAF-D-14-00105.1.
Draxl, C., A. N. Hahmann, A. Peña, and G. Giebel, 2014: Evaluating winds and vertical wind shear from Weather Research and Forecasting model forecasts using seven planetary boundary layer schemes. Wind Energy, 17, 39–55, https://doi.org/10.1002/we.1555.
Eager, R. E., S. Raman, A. Wootten, D. L. Westphal, J. S. Reid, and A. Al Mandoos, 2008: A climatological study of the sea and land breezes in the Arabian Gulf region. J. Geophys. Res., 113, https://doi.org/10.1029/2007JD009710.
Elhakeem, A., W. Elshorbagy, and T. Bleninger, 2015: Long-term hydrodynamic modeling of the Arabian Gulf. Mar. Pollut. Bull., 94, 19–36, https://doi.org/10.1016/j.marpolbul.2015.03.020.
El-Sabh, M. I., and T. S. Murty, 1989: Storm surges in the Arabian Gulf. Nat. Hazards, 1, 371–385, https://doi.org/10.1007/BF00134834.
Etemad-Shahidi, A., B. Kamranzad, and V. Chegini, 2011: Wave energy estimation in the Persian Gulf. Proc. Int. Conf. on Environmental Pollution and Remediation (ICEPR 2011), Ottawa, Ontario, Canada, International ASET Inc., 223, https://doi.org/10.13140/RG.2.1.1089.6805.
Gunwani, P., and M. Mohan, 2017: Sensitivity of WRF model estimates to various PBL parameterizations in different climatic zones over India. Atmos. Res., 194, 43–65, https://doi.org/10.1016/j.atmosres.2017.04.026.
Hamidi, M., M. R. Kavianpour, and Y. Shao, 2014: Numerical simulation of dust events in the Middle East. Aeolian Res., 13, 59–70, https://doi.org/10.1016/j.aeolia.2014.02.002.
Hariprasad, K. B. R. R., C. V. Srinivas, A. B. Singh, S. V. B. Rao, R. Baskaran, and B. Venkatraman, 2014: Numerical simulation and intercomparison of boundary layer structure with different PBL schemes in WRF using experimental observations at a tropical site. Atmos. Res., 145–146, 27–44, https://doi.org/10.1016/j.atmosres.2014.03.023.
Hashem, N., and P. Balakrishnan, 2015: Change analysis of land use/land cover and modelling urban growth in Greater Doha, Qatar. Ann. GIS, 21, 233–247, https://doi.org/10.1080/19475683.2014.992369.
Hu, X. M., J. W. Nielsen-Gammon, and F. Zhang, 2010: Evaluation of three planetary boundary layer schemes in the WRF model. J. Appl. Meteor. Climatol., 49, 1831–1844, https://doi.org/10.1175/2010JAMC2432.1.
Jaramillo, O. A., and M. A. Borja, 2004: Wind speed analysis in La Ventosa, Mexico: A bimodal probability distribution case. Renewable Energy, 29, 1613–1630, https://doi.org/10.1016/j.renene.2004.02.001.
Langodan, S., Y. Viswanadhapalli, and I. Hoteit, 2016: The impact of atmospheric data assimilation on wave simulations in the Red Sea. Ocean Eng., 116, 200–215, https://doi.org/10.1016/j.oceaneng.2016.02.020.
Liao, Y. P., and J. M. Kaihatu, 2016: Numerical investigation of wind waves in the Persian Gulf: Bathymetry effects. J. Atmos. Oceanic Technol., 33, 17–31, https://doi.org/10.1175/JTECH-D-15-0066.1.
Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, Z. Zhu, L. W. J. Yang, and J. W. Merchant, 2000: Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sens., 21, 1303–1330, https://doi.org/10.1080/014311600210191.
Madala, S., K. H. Prasad, C. V. Srinivas, and A. N. V. Satyanarayana, 2016: Air quality simulation of NOX over the tropical coastal city Chennai in southern India with FLEXPART-WRF. Atmos. Environ., 128, 65–81, https://doi.org/10.1016/j.atmosenv.2015.12.052.
Moeini, M. H., A. Etemad-Shahidi, and V. Chegini, 2010: Wave modeling and extreme value analysis off the northern coast of the Persian Gulf. Appl. Ocean Res., 32, 209–218, https://doi.org/10.1016/j.apor.2009.10.005.
Olsson, P. Q., K. P. Volz, and H. Liu, 2013: Forecasting near-surface weather conditions and precipitation in Alaska’s Prince William Sound with the PWS-WRF modeling system. Cont. Shelf Res., 63, S2–S12, https://doi.org/10.1016/j.csr.2011.12.012.
Panchang, V. G., C. K. Jeong, and D. Li, 2008: Wave climatology in coastal Maine for aquaculture and other applications. Estuaries Coasts, 31, 289–299, https://doi.org/10.1007/s12237-007-9016-5.
Panchang, V. G., C. K. Jeong, and Z. Demirbilek, 2013: Analyses of extreme wave heights in the Gulf of Mexico for offshore engineering applications. J. Offshore Mech. Arctic Eng., 135, 031104, https://doi.org/10.1115/1.4023205.
Pena Diaz, A., A. N. Hahmann, C. B. Hasager, F. Bingöl, I. Karagali, J. Badger, M. Badger, and N.-E. Clausen, 2011: South Baltic wind atlas: South Baltic offshore wind energy regions project. Denmark Forskningscenter Risø Risø-R-1775, Danmarks Tekniske Universitet Risø Nationallaboratoriet for Bæredygtig Energi, 66 pp.
Rao, P. G., H. R. Hatwar, M. H. Al‐Sulaiti, and A. H. Al‐Mulla, 2003: Summer shamals over the Arabian Gulf. Weather, 58, 471–478, https://doi.org/10.1002/wea.6080581207.
Shafran, P. C., N. L. Seaman, and G. A. Gayno, 2000: Evaluation of numerical predictions of boundary layer structure during the Lake Michigan Ozone Study. J. Appl. Meteor., 39, 412–426, https://doi.org/10.1175/1520-0450(2000)039<0412:EONPOB>2.0.CO;2.
Shao, Y., 2008: Physics and Modelling of Wind Erosion. Atmospheric and Oceanographic Sciences Library, Vol. 37, Springer Science & Business Media, 456 pp., https://doi.org/10.1007/978-1-4020-8895-7.
Signell, R. P., S. Carniel, L. Cavaleri, J. Chiggiato, J. D. Doyle, J. Pullen, and M. Sclavo, 2005: Assessment of wind quality for oceanographic modelling in semi-enclosed basins. J. Mar. Syst., 53, 217–233, https://doi.org/10.1016/j.jmarsys.2004.03.006.
Singha, A., and R. Sadr, 2012: Characteristics of surface layer turbulence in coastal area of Qatar. Environ. Fluid Mech., 12, 515–531, https://doi.org/10.1007/s10652-012-9242-7.
Singhal, G., V. G. Panchang, and J. L. Lillibridge, 2010: Reliability assessment for operational wave forecasting system in Prince William Sound, Alaska. J. Waterw. Port Coastal Ocean Eng., 136, 337–349, https://doi.org/10.1061/(ASCE)WW.1943-5460.0000056.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers, 2005: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp., http://dx.doi.org/10.5065/D6DZ069T.
Stensrud, D. J., 2007: Parameterization Schemes: Keys to Understanding Numerical Weather Prediction Models. Cambridge University Press, 459 pp.
Viswanadhapalli, Y., H. P. Dasari, S. Langodan, V. S. Challa, and I. Hoteit, 2017: Climatic features of the Red Sea from a regional assimilative model. Int. J. Climatol., 37, 2563–2581, https://doi.org/10.1002/joc.4865.
Yu, Y., M. Notaro, O. V. Kalashnikova, and M. J. Garay, 2016: Climatology of summer Shamal wind in the Middle East. J. Geophys. Res. Atmos., 121, 289–305, https://doi.org/10.1002/2015JD024063.
Zhang, H., Z. Pu, and X. Zhang, 2013: Examination of errors in near-surface temperature and wind from WRF numerical simulations in regions of complex terrain. Wea. Forecasting, 28, 893–914, https://doi.org/10.1175/WAF-D-12-00109.1.
Zhu, M., and B. W. Atkinson, 2004: Observed and modeled climatology of the land–sea breeze circulation over the Persian Gulf. Int. J. Climatol., 24, 883–905, https://doi.org/10.1002/joc.1045.
