1. Introduction
Southern Delaware is situated on the “Delmarva” Peninsula, which is an irregularly shaped promontory that is surrounded by the Chesapeake Bay, the Delaware Bay, and the Atlantic Ocean. These bodies of water influence the mesoscale variability in the wind field, leading to complex low-level winds over the region, especially during the summertime, as a result of the influence of local circulation systems such as the sea breeze. Furthermore, coastal urban development and nearby ocean upwelling introduce variability in land and sea surface properties that influence the local wind regime. Delaware’s ocean coastline is fairly straight, with an average orientation of 175°, and the bay coastline is slightly convex, bowing inland with an average orientation of 330°.
The low-level local winds play an important role in agriculture, pollution transport, wind-energy potential, and tourism. The nearby presence of the Atlantic Ocean and Delaware Bay can influence the mesoscale winds, which often compete with the synoptic wind regime. Even though Delaware has a high-density network of meteorological observation stations, high-resolution modeling of these winds can provide insight into this dynamical balance where observations are scarce.
Agriculture is an important industry in Delaware, and it relies heavily on the distribution of precipitation, especially during the summer. Convection and convergence of low-level winds associated with the sea breeze can increase cloud production and precipitation across the region (e.g., Hill et al. 2010; Azorin-Molina et al. 2014). Abnormally high temperatures often influence the local and regional winds, which in return modify mesoscale circulations such as the sea breeze. This situation can have an impact on the occurrence and persistence of drought events, which have occurred 12 times in Delaware between 1948 and 2005 (Dowtin 2012).
High ozone concentrations are a continual problem throughout Delaware during the spring and summer months. In 2012, Sussex County, the southernmost county in the state, recorded 12 days on which ozone mixing ratios were in excess of 0.075 ppm for at least 8 h, which is in excess of the national ambient air quality standards set by the U.S. Environmental Protection Agency [data obtained online from the Delaware Department of Natural Resources and Environmental Control (DNREC) at http://apps.dnrec.delaware.gov/AirMonitoring/Reports/2012%201-hr%20and%208-hr%20exceedances.pdf]. The total amount of ozone in a region is dependent upon the proximity to point sources as well as the wind speed and direction (Husar et al. 2007). A recent study in Wilmington, Delaware, demonstrated that the type of particulate matter that was present in the air was dependent on the persisting wind direction (Reinard et al. 2007). They indicated that heavy metals such as iron and vanadium, which can be emitted from Delaware’s coal-fired power plants, can be transported over 10 km in strong winds.
Local winds have been shown to affect the currents and ecosystem of the Delaware Bay (Muscarella et al. 2011). Münchow and Garvine (1993) demonstrated that the wind forcing is often in competition with buoyancy forcing for control of the Delaware Coastal Current, an important dynamic for local fishermen. During the summer, larvae are transported southeastward by the Delaware Coastal Current and then northward because of dominant southerly winds (Epifanio et al. 1989). A study of below-source oxygen levels concluded that atmospheric conditions such as storm events can influence the dissolved oxygen content in these inland bays (Luther et al. 2004). Understanding the relationships among the winds, currents, and nutrient levels in the Delaware Bay and nearby coastal areas is an important component of disaster preparedness and coastal resource management.
The geographical variation in the low-level wind field influences the on- and offshore wind-energy potential of a region. Measuring and characterizing the prevailing winds can help to optimize the placement of wind turbines to maximize energy output and minimize wake effects. Characterizing the variability in the low-level winds also provides insight into how energy demands compare with the available wind resource on annual, seasonal, and diurnal scales. The U.S. Bureau of Ocean Energy Management has approved a wind-energy area for coastal Delaware that is just outside the mouth of the Delaware Bay, where a commercial wind-energy lease has been issued (Lennon 2010). The characteristics of the low-level wind resource for this area on the outer continental margin have been studied using reanalysis data and mesoscale modeling (Woods et al. 2013).
Several studies in the past 35 years have investigated general features of the wind climate over the Mid-Atlantic Bight including the Delaware Bay (Maurmeyer 1978; Moffatt and Nichol 2007; Garvine and Kempton 2008; Woods et al. 2013; Monaldo et al. 2014). Maurmeyer (1978) characterized the local winds at Wilmington Airport using data from the National Climatic Data Center (NCDC); her study indicated that the dominant wind speed and direction changed significantly as a function of season from the northwest in the winter to the southwest in the summer. This seasonal shift in the dominant wind direction was confirmed in a report prepared for the DNREC on coastal sedimentation using data from the Indian River, Dover, and Sussex airports obtained from the NCDC (Moffatt and Nichol 2007). Both of these studies were limited by the number of available surface stations.
Analysis of over 50 years of data from the Naval Air Station in Patuxent, Maryland, indicated a similar summertime wind regime (Scully 2010) and confirmed that persistent west winds over the Chesapeake Bay support hypoxic conditions. Atkinson et al. (2013) summarized the wind characteristics for several National Data Buoy Center (NDBC) stations off the mid-Atlantic coast. They found that B44009, located off Delaware’s coastline, recorded seasonal characteristics that are similar to those recorded by B44014, which is located well off the coast of Virginia. Furthermore, CHLV2, located at the Chesapeake Light Tower in Virginia, had a stronger southeastward wind component in the summer than did other nearby stations. Woods et al. (2013) and Monaldo et al. (2014) also observed variability in the offshore wind field using buoy and satellite data, respectively.
Other recent work on the regional wind climate indicated that there is a higher proportion of easterly wind at Indian River than at other inland stations during the summer because of its geographical location along the coastline. Garvine and Kempton (2008) used more than 10 surface stations from NDBC with record lengths of 1.7–19.6 yr to calculate the mean annual wind speed at 80 m above ground level (AGL) over Lewes, Delaware, Brandywine Shoal, and B44009 as a preliminary regional wind-resource assessment. The authors reported average 80-m wind speeds of 5.74, 7.70, and 7.93 m s−1, or approximately 4.68, 6.46, and 6.65 m s−1 at 10 m AGL using a logarithmic wind profile. These differences indicate that there is significant horizontal variability in the low-level wind field that is seen on a diurnal and seasonal basis.
The current study described in this article builds upon these prior results, increasing the spatial resolution of the observational data, investigating sources of variability in the wind field, and comparing these results with output from the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008).
2. Data and methods
a. Meteorological observations
The Delmarva Peninsula contains one of the densest networks of meteorological surface stations in the United States. The Delaware Environmental Observing System (DEOS) consists of over 40 meteorological stations that cover Delaware, with data collection beginning in 2005. In addition, NDBC has maintained marine meteorological stations in the region since 1984, several of which are located along or near Delaware’s coastline. This study includes eight surface meteorological stations from DEOS and NDBC that span southern Delaware, the Delaware Bay, and the nearby Atlantic Ocean. These stations are representative of their geographical locations, with minimal wind shadowing, and contain high-quality and relatively complete data records that vary in duration from 7 to 28 yr (Table 1). There are two stations located along Delaware’s coastline (DBNG and LWSD1), three over open water (SJSN4, BRND1, and B44009), and three inland (DLAU, DBRG, and DHAR), as shown in Fig. 1.
Meteorological stations in this study.
The innermost model domain and meteorological stations used in this project. The study area covers southern Delaware, the Delaware Bay, and the nearby oceanic region. (Map data copyright 2014 Google.)
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
b. Model description
Insight into the spatial variability of the local and regional winds can be provided from high-resolution mesoscale modeling (Jiménez et al. 2010; Colle and Novak 2010; Kain et al. 2010; Salamanca et al. 2011). WRF has been used extensively to examine many aspects of local and regional climates throughout the world. For example, Khan (2010) investigated the sea-breeze circulation in Auckland, New Zealand, for 2006–07 using WRF. Khan demonstrated that WRF is more skillful than The Air Pollution Model (Hurley et al. 2005) at simulating physical surface variables like temperature, relative humidity, and winds when compared with observations. Colle and Novak (2010) showed that WRF could reproduce the presence of a low-level jet over the New York Bight. Several other studies demonstrate that in some situations WRF overestimates the mean lower-level wind speeds by 1–2 m s−1 relative to observations (e.g., Shimada and Ohsawa. 2011; Carvalho et al. 2014), which is similar to biases found with the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5; Shimada et al. 2009; Khan 2010). Shimada and Ohsawa (2011) further showed that the positive bias in modeled offshore winds was the main source of error in wind-power-density estimations, even though the smallest difference between observed and WRF modeled wind is over water. Nawri et al. (2012) demonstrated that WRF is capable of capturing the seasonal shift in the dominant winds from offshore during the winter to onshore during the summer over Iceland using 16 meteorological stations for comparison. These studies indicate that WRF is capable of representing the climatological winds of a region with a high degree of skill.
WRF, version 3.5, is employed to simulate the low-level winds over the Delaware Bay and the surrounding landmass for a sampling of days between 2006 and 2012. This time frame was chosen because it coincides with the highest concentration of quality observational data across the region. Because it is very computationally expensive to run a high-resolution mesoscale atmospheric model for a 6-yr time period, we developed a set of runs for each season that approximate climatological norms while including a variety of realistic synoptic conditions. For each meteorological season [defined as December–February (DJF) for winter, March–May (MAM) for spring, June–August (JJA) for summer, and September–November (SON) for autumn] of each year, two periods of 5 consecutive days were randomly selected, resulting in a minimum of 70 days being simulated per season. There had to be at least 15 days between the starting dates of each run to prevent all of the sample days in one season from being represented by a persistent synoptic system. The representativeness of the resulting dataset was evaluated by comparing characteristic wind speed and direction for the entire period from 2006 to 2012 with that of the selected subset. As can be seen in Fig. 2, the differences between the full dataset and the subset are small, including a difference in the mean wind speed of less than 0.15 m s−1 between the sample set and the 2006–12 time frame, which is of the same order as instrument error and suggests that this selection method is reasonable.
Wind roses for B44009 comparing the following time frames: (a) 1984–2012, (b) 2006–12, and (c) 280 selected sample days from 2006 to 2012.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
Each 5-day simulation employs a 12-h spinup and is configured with three nests centered on coastal Delaware, with horizontal resolutions of 18, 6, and 2 km (Table 2). The innermost domain is depicted in Fig. 1 along with the meteorological stations analyzed in this study. The model time steps are 100, 33, and 11 s for the outer, middle, and inner domains, respectively, with instantaneous output saved at the start of every hour. The “Noah” land surface model is employed, which includes the effects of vegetation and diagnoses the surface skin temperature (Chen and Dudhia 2001). The lateral boundaries of the outer nest are forced by temperature and pressure data from the North American Regional Reanalysis (NARR) project (Mesinger et al. 2006). The data are ingested every 3 h with a resolution of 40 km, which is then downscaled by the inner domains. The Mellor–Yamada–Janjić (MYJ) planetary boundary layer scheme is used to calculate the vertical diffusion (Janjić 1994); the MYJ scheme is commonly used in high-resolution modeling runs (Jimenez et al. 2007). The Kain and Fritsch (1993) cumulus parameterization is used in the larger two nests. No cumulus parameterization was used in the inner domain because no scheme is designed for such high resolutions (see 2009 presentation slides by J. Dudhia on WRF physics options: http://www.mmm.ucar.edu/wrf/users/tutorial/200901/WRF_Physics_Dudhia.pdf). The Rapid Radiative Transfer Model (RRTM; Mlawer et al. 1997) was used for longwave radiation because it is an accurate and fast scheme that improves efficiency by using reference tables. For shortwave radiation we used the Dudhia (1989) scheme, which provides simple downward integration and is appropriate for high-resolution studies. Each run includes 28 vertical levels, with the 7 lowest levels within 1 km of the surface and a top pressure level of 100 hPa.
WRF configuration.
c. Method
Sampling 70 days per season over a 6-yr period will not capture the entire range of possible synoptic conditions, and therefore for this study we focused on the dominant synoptic types that occur in each season. The synoptic type is taken from a temporal synoptic index developed at the University of Delaware (D. Leathers 2013, personal communication) that is derived from a principal component analysis of meteorological variables such as temperature, dewpoint temperature, pressure, cloud cover, and wind magnitude, taken at a station in Philadelphia, Pennsylvania. This synoptic index accounts for the observed variance between synoptic events (Kalkstein and Corrigan 1986) and is used to classify each event according to the relative weight of each variable in the analysis, using a k-means clustering analysis (Kalkstein et al. 1987, 1990). The synoptic types are developed for each season on the basis of data from 1946 to 2012, with anywhere from 10 to 13 synoptic types identified per season; each day in the dataset is assigned a synoptic type. The temporal synoptic index has been used in a variety of fields such as improving the prediction of snowfall (Ellis and Leathers 1996; Leathers and Ellis 1996), analyzing the relationship between climate and wine production in Bordeaux (Jones and Davis 2000), assessing glacier mass balance (Fealy and Sweeney 2007), and predicting summertime mortality (Samet et al. 1998; Kyselý et al. 2010).
For the selected 280 days of analysis, we verified that the relative frequency of occurrence for each synoptic type was similar to occurrences over the entire 66-yr period (see the appendix). Our subset underrepresents some of the less-frequent synoptic types, which is expected in this kind of sampling approach. We also compared frequency of occurrence with the 2006–12 synoptic-type frequency for the full dataset, since the frequency of occurrence of some of the synoptic types has evolved over the past six decades (D. Leathers 2014, personal communication). We focused on the four most frequently occurring synoptic types for each season (see the appendix for a complete list), which accounts for 60% of the total days in winter, 67% in spring, 90% in summer, and 73% in autumn. For each synoptic type, a minimum of 12 days is used. For the majority of these selected types, the original set of 280 days provided the required amount of sample days. For the remaining types, seven additional runs were performed between the years of 2006 and 2012 to bring the sample size up to 12 cases. Random start dates were generated for these runs, but only the ones that included the desired synoptic types were chosen. Data from these additional runs were only included to have enough samples in the analysis of the synoptic types as composites; the additional data were not included in the seasonal analysis to avoid oversampling the synoptic types relative to their presence in the climatic data.
General characteristics such as annual, seasonal, and monthly average wind speeds and directions were calculated and were compared with previous studies. The standard meteorological classification for each season (i.e., DJF, MAM, JJA, and SON) was used throughout our analysis. The mean wind direction is calculated by taking the mean of both the sine and cosine components of each sample wind direction and then calculating the angle of the resultant vector.
We constructed wind roses to assess the probability distribution of the wind speed and direction using 12 bins, each with a width of 30°. This visual and analytical tool lends insight into characteristics of the wind field and how this field evolves temporally and spatially. Feather plots were utilized to examine diurnal variability of each station. Diurnal wind direction variability is obtained by measuring the difference in the mean wind directions. The modeled wind fields were also analyzed, with focus on the afternoon hours to highlight the influence of the local sea-breeze circulation.
We computed the mean monthly wind speed, standard deviation, and extreme wind frequencies at B44009 to assess the interannual and longer-term variability in the wind field. The standard deviation was calculated on the basis of all available data points within a month. To assess the yearly trend, the linear regression was calculated for each month and the significance of the slope was assessed using a t test with a critical p level of 0.05. Only months with greater than 90% data availability were included in all analyses except for the extreme wind frequencies, which used all available data. Between 20 and 28 years of data fit the criteria for each month.
We compared modeled wind speed data with the geographically closest observed data at hourly intervals. Statistics such as absolute error in wind speed and direction were calculated to assess WRF’s skill in simulating the instantaneous wind speed at each meteorological station. For stations other than B44009, we employed a 13-sample hourly average of the wind speed and direction for comparison with model output.
3. Climatological wind description
a. Annual features
The annual wind speed and direction were characterized using wind roses compiled for all eight stations across the region, several of which are displayed in Fig. 3. Across inland stations, the dominant wind directions are from the south-southwest and north-northwest, consistent with prior studies. This feature is also in strong agreement with the dominant wind directions observed well offshore (B44009), suggesting that the overall wind climate is controlled by synoptic-scale atmospheric dynamics. Similar features have been observed at other nearby locations, including the Greater Wilmington Airport (Maurmeyer 1978) and the Georgetown Airport (Moffatt and Nichol 2007). Coastal stations across southern Delaware indicate a different annual wind regime, however, with no clear dominant wind direction, similar to what was observed at the Indian River Coast Guard Station, located on the barrier island between Dewey Beach and the Indian River inlet, between 1975 and 1984 (Moffatt and Nichol 2007).
Wind roses from select stations overlaid on their approximate locations for the entire dataset; B44009: 1984–2012, and other stations: 2005–12.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
Our results suggest further spatial variation in the wind field along the Delaware Bay. BRND1, an NDBC station mounted on a small island at the mouth of the bay, has dominant wind directions from both the south and north-northwest. This indicates a counterclockwise shift in the dominant southerly wind direction relative to both the offshore and inland regions. This counterclockwise shift of the dominant southerly flow is further observed at SNSJ4, located near the head of the Delaware Bay, with a south-southeast flow. The shifting of the wind direction to align with the axis of the Delaware Bay is similar to that observed for sea-breeze events at the New York Bight (Colle and Novak 2010) and introduces a significant easterly component to the surface winds on the Delaware side of the bay as seen from in situ observations in weather radar reflectivity data over and near the Delaware Bay (Hughes 2011; Gilchrist 2013). This shift seen in the annual wind roses is due to changes in the local pressure gradient related to the thermal gradient both along the bay coast and across the bay. The observed spatial variability of the winds is similar to that observed by Monaldo et al. (2014) using synthetic-aperture-radar data to retrieve ocean wind speed.
As expected, the mean annual wind speed at 10 m AGL is higher over open water than over coastal and inland regions (Table 3). For the water-based stations, there is a difference among the mean wind speeds of B44009 (6.8 m s−1), BRND1 (6.5 m s−1), and SJSN4 (5.7 m s−1) that is probably a function of the proximity of land to each station, with B44009 being in the coastal ocean and SJSN4 being at the head of the Delaware Bay. LWSD1 (4.8 m s−1), located in Lewes near the mouth of the Delaware Bay, has a higher mean wind speed than a nearby coastal station DBNG (3.5 m s−1). This is reasonable because the LWSD1 station is close to water bodies that span from the northwest direction clockwise to the southeast direction, and winds from those directions are mostly unhindered by friction with the land surface. The mean wind speed for the inland stations ranges from 3.0 to 3.9 m s−1, which indicates that the increased surface roughness of the land surface has a significant effect on the observed mean wind speed. The observed wind speeds agree with Garvine and Kempton’s (2008) results and suggest that wind speeds over the past 5 yr are representative of climatological values. Results from the modeled 280-day subset indicate that WRF overestimates the inland and coastal regions by 0.5 and 2.1 m s−1, respectively, whereas there is almost no bias in its representation of B44009.
Statistics (mean ± std dev; m s−1) for observed mean wind speed at 10 m AGL for the entire data record.
To investigate further the long-term wind speed variability, we utilized wind speed and direction data from B44009. NDBC has been collecting meteorological data at this location for 28 yr. The analysis in this study includes only years with at least 90% availability, which provides an average of 23.4 valid years for each month. It is important to note that, within the dataset, there are several gaps of 2 weeks or longer. The monthly-mean wind speed over the entire dataset indicates a clear seasonal cycle (Fig. 4a). In general, the mean wind speed for this location is 6.3 m s−1, with the highest wind speeds seen in the winter (7.8 m s−1) and the lowest windspeeds seen in the summer (5.0 m s−1).
(a) Climatological monthly wind speed and (b) yearly trend in monthly-mean wind speed at B44009 (1984–2012). The standard deviation is indicated in the figure by error bars (see also Table 3). The dark-shaded bars in (b) indicate trends that are statistically significant (p < 0.05). Only months that are missing less than 10% of the wind speed data were included in this analysis.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
The average annual wind speed for B44009 during the first five years of the analysis (1984–88) was 6.2 m s−1 at anemometer height (5 m). This value increased slightly to 6.3 m s−1 during the last five years of the record (2008–12). This compares well to a global study that suggests that the mean wind speed near the mid-Atlantic coastline has shown no statistically significant increase between 1991 and 2008 as based on satellite data (Young et al. 2011). Schofield et al. (2008) observed a wintertime increase in wind speed at B44009 between the two periods of 1996–2007 and 1987–95, however, and attributed it to changes in the Atlantic multidecadal oscillation. When the entire time series from B44009 (1984–2012) is analyzed on a monthly basis, excluding months that are missing more than 10% of the data, a statistically significant increase (p = 0.006) is seen in the mean wind speed from 6.4 to 7.0 m s−1 in October (0.04 m s−1 yr−1) and a statistically significant decrease (p = 0.026) from 6.0 to 5.5 m s−1 is seen in May (0.03 m s−1 yr−1), as shown in Fig. 4b. There are smaller positive trends for the other months in the autumn and winter seasons, along with smaller decreases during the summer months. These changes in the wind speed may be attributed to natural climatic variability or may be related to changes in atmospheric circulation resulting from climate change. We repeated the analysis, allowing for minimum monthly data coverage of 75%, with similar results.
The maximum sustained wind speed observed at B44009, which occurred on both 27 September 1985 and 18 August 2004 during Hurricanes Gloria (1985) and Charley (2004), was 25 m s−1. Wind speeds of roughly 24–28 m s−1 that are due to hurricanes are expected to occur roughly once per decade according to work done using historical hurricane tracks (MMI Engineering and Atmospheric and Environmental Research 2012]. A peak wind gust of 33 m s−1 was recorded during Hurricane Gloria (1985). Wind speeds in excess of 10.9 and 15.2 m s−1 occur with frequencies of approximately 10% and 1%, respectively, for this location. These frequencies do not include times during which the buoy was not recording data; such outages may be more likely to occur during major storm events.
b. Seasonal characteristics
Seasonal wind speed statistics were computed for each observation station for the years from 2005 through 2012 (Table 3). The highest seasonal wind speeds are observed in the winter at all three locations, with inland, coastal, and open-water wintertime mean speeds ranging from 3.3 to 4.6, 3.9 to 5.8, and 6.3 to 8.2 m s−1, respectively. The seasonal winds are weakest during the summer, with values ranging from 2.3 to 5.4 m s−1. Wind speeds over inland stations are approximately 0.7 m s−1 stronger during spring than during autumn, but the mean wind speed values for coastal stations during autumn were very similar to spring values. B44009 measured higher wind speeds in autumn (7.1 m s−1) than in spring (6.7 m s−1). These differences may be caused by strong synoptically driven east winds that occur in autumn. The seasonal variability in wind speed in this dataset is similar to that obtained by Garvine and Kempton (2008). This indicates that the seasonal cycle in wind speed between 2008 and 2012 is comparable to previous years (1984–2007 for B44009 and 2006–07 for BRND1, LWSD1, and SJSN4).
From observations, the randomly selected subset of 280 days shows seasonal characteristics that are similar to those described above. For the 280 days selected for this analysis, WRF generally underestimates the winds during the winter and overestimates the winds for the remaining seasons (Table 4). The positive WRF Model wind speed bias has been found in several other studies (Shimada et al. 2009; Khan 2010; Shimada and Ohsawa 2011). Carvalho et al. (2014) observed both positive and negative wind speed biases relative to observations, although they do not explore the seasonal nature of these biases. Some of these biases may be due to the model boundary and surface layer parameterizations employed. For example, Otkin and Greenwald (2008) found that the Mellor–Yamada–Janjić boundary layer parameterization (Janjić 1994) may better simulate the lower boundary layer in high-resolution runs than does the Yonsei parameterization. In our own sensitivity studies, we found that the modeled low-level winds were more sensitive to the input sea surface temperature than to the boundary layer parameterization (Hughes 2011).
Comparison of modeled and observed data for all 280 days. The directional difference and speed difference are both the absolute mean difference between modeled and observed instantaneous hourly data.
In this study, the station with the largest bias (DBNG) is a land-based station that is situated on land in the model but neighbors an open-water point, whereas in actuality the station is 1 km inland from the coast. The mean wind speeds of coastal stations were proportionally higher than inland stations. Therefore both modeled and observed analyses indicate that the land surface is playing an important role in its wind speed even in the presence of strong synoptic conditions.
At each meteorological station, we analyzed the seasonal variation of the mean wind direction. Winter winds demonstrated the least spatial variation, with prevailing winds from the west and northwest at all stations. In the spring, the mean winds shift to the southwest (counterclockwise) by approximately 30°. The mean winds shift farther to the south in summer by about 15° and show greater coherence than in spring. In autumn, the mean wind direction shifts back to the northwest as the increasing variability in direction indicates another transition season. Dominant (summer/winter) and transition (spring/autumn) seasonal wind patterns were also observed in several buoys off the mid-Atlantic coast (Atkinson et al. 2013).
Modeled wind directions for the subset of 280 days followed a seasonal wind shift that was similar to that of the observed data (Table 4). The mean absolute error in the wind direction varied from 22° to 38° and was typically higher in the summer when mesoscale effects are stronger, similar to the findings of Carvalho et al. (2014). Some of this error may be attributed to the model having a counterclockwise bias of approximately 15°–30° that persisted for each season. The wind direction error is greatest at the coastal locations, similar to what Colby (2004) found in high-resolution (4 km) MM5 runs of the New England sea breeze. Colby observed that the coastal stations showed greater variability in wind speed and direction absolute error than did inland locations. This result suggests that winds at the coastline are sensitive to the resolved details of the land–sea boundary.
Within each season we constructed wind roses to compare the modeled and observed data for the four dominant synoptic types for B44009 (Figs. 5 and 6; appendix Table A1). The model captures most of the wind speed and direction characteristics for the majority of the synoptic types. For example, observed values for type 2010 (spring) had two dominant directions (southwest and east). The model accurately recreated the direction and relative frequency of both modes. The model generally performed the worst when the frequency of the dominant wind directions was relatively small, such as in type 1032, which is wintertime southwest flow, and type 4037, which is strong southwest flow in autumn. For type 4037, the model did not represent the frequency of the east and northeast wind component that was present in the observations. The model overestimated strong northwest summertime winds except for extreme conditions such as in the presence of thunderstorms.
Observed 10-m wind roses for B44009 for the most frequent synoptic types per season. Each wind rose is plotted on the same axes as in Fig. 1 and is composed of at least 12 days between 2006 and 2012.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
As in Fig. 5, but for modeled wind speed and direction.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
To further compare the spatial variation of the wind field across the region between the model and observations, we investigated how the wind regimes of a coastal grid cell (representing DBNG) and an inland grid cell (representing DLAU) compare for the four dominant synoptic types in the summertime (Fig. 7) when there is the greatest variability in the observed winds. The composite sea level pressure for the dominant types indicates the following synoptic winds: 3031—southwest (weak), 3032—southwest (strong), 3033—northwest, and 3035—east. In the model runs, types 3031 and 3032 show a 30° shift in the dominant wind direction from south-southwest inland to south at the coastal location. At the location of this shift, the coastal winds are noticeably stronger, showing significantly variability in the wind fields—a feature that is not observed for types 3033 and 3035. Similar differences were observed in the spring and autumn seasons but were not as prevalent as in the summer.
Summertime, model-derived wind roses for DLAU (inland) and DBNG (coastal), arranged by synoptic type. Composites of the sea level pressure are based on 12 modeled days for each type.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
c. Diurnal features
The diurnal variability within the regional wind speeds and directions is greatly influenced by both the season and the proximity to the coastline. To understand this better, the hourly wind speeds and directions averaged by season are compared for an inland (DLAU) station and a coastal (DBNG) station (Fig. 8). The winter season has the least wind direction variability, with persistent northwest winds across the region, and the most significant variability is observed in the summer. Across the inland region during the summer the mean wind flows from the southwest during the early-morning hours, in alignment with the synoptic flow. The mean wind flow then shifts to the west during the afternoon before shifting back to the south in the evening. There are several possible causes for this feature, including down-sloping winds from the Appalachian Mountains, strengthening of the Bermuda high, and interaction with the Chesapeake Bay breeze. At the coast, the early-morning wind regime is similar across the entire study region, coming from the southwest and controlled by the synoptic-scale pressure gradient, similar to that shown in Woods et al. (2013). In the late-morning hours, however, the mean coastal wind shifts to the east with the onset of the sea breeze driven by the surface temperature gradient. Throughout the afternoon and into the evening, the winds continue to rotate clockwise because of the influence of the Coriolis force and eventually realign with the mean synoptic wind. This is similar to the evolution of the wind direction seen by Colle and Novak (2010) for the New York Bight during evenings with lower-level jets.
Feather plot of the mean hourly wind vectors for each season at an inland station (DLAU; black) and a coastal station (DBNG; gray). The size of each vector is proportional to the mean wind speed for that hour.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
The large-scale flow is predominantly driven by the pressure gradient resulting from the semipermanent high pressure off the East Coast. Days with southwesterly flow have the highest occurrence of sea breezes (Gilchrist 2013). The largest difference between the average winds at the inland and coastal stations occurs during the afternoon hours, mostly because of the presence of the sea breeze. This characteristic wind speed and direction shift is observed at both BRND1, which is inside the Delaware Bay, and B44009, which is outside the bay mouth. For both of these open water locations, however, the wind shift starts later than at the coastal stations and is not as pronounced, covering a smaller range of directions. This is probably because of the delayed impact of the sea-breeze circulation, which tends to develop at or near the coast. At SJSN4, which is much farther up the Delaware Bay, the mean wind shifts slightly to the south-southwest during the evening hours. The spring and autumn seasons follow a diurnal pattern that is similar to the summertime trends for all regions.
The hourly diurnal mean wind speeds at each station indicate considerable temporal and spatial differences throughout the day (Fig. 9). The mean hourly wind speeds at B44009 fluctuate by only 0.5 m s−1, with the lowest values occurring during the late afternoon. The largest fluctuation is observed at DLAU with a 2.5 m s−1 amplitude. Contrasting with B44009, the mean wind speeds at DLAU are strongest during the afternoon hours. Throughout the day, hourly mean wind speeds remain constant at BRND1 except in the early-afternoon hours where there is a pronounced drop of over 1 m s−1, which may be caused by the passage of a sea-breeze front moving up the bay. This was not seen in the analysis of Garvine and Kempton (2008). In contrast, winds at LWSD1 are strongest during the afternoon hours. In general, hourly changes in the mean wind speed are largest during the afternoon hours at all locations, and these changes are dependent on the land surface type in and around each location.
Diurnal mean wind speed at 10 m AGL. Values were computed using all available years for each station.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
Using model output, the mean wind speed and direction are plotted for each grid cell for the four dominant summer synoptic types at 1400 local time (Fig. 10). This time was chosen because it highlights the diurnal differences in the wind magnitudes that are seen across the region. Both synoptic types 3031 and 3032 indicate the clear presence of a corkscrew-originating sea-breeze circulation (Miller et al. 2003). The sea-breeze signature is better defined and larger for type 3031, and this is probably attributable to the weaker synoptic winds. Strong offshore synoptic winds can prevent the development of a sea breeze or prevent it from advancing landward. Type 3033 shows the signature of a backdoor sea breeze in the presence of a strong northwest synoptic wind. The development of such a sea breeze is more difficult than for the corkscrew sea breeze because the Coriolis force is working against it. The wind regime in type 3035 is onshore, and there is no clear sea-breeze signature present, although its presence cannot be completely ruled out. It is clear that the strength and structure of the simulated sea-breeze circulation are affected by the synoptic pressure regime and resulting wind field.
Composite of the U 10-m wind component (west wind positive) at 1400 local daylight time for the dominant summertime synoptic types. The mean surface pressure for each synoptic type is shown in Fig. 7.
Citation: Journal of Applied Meteorology and Climatology 54, 1; 10.1175/JAMC-D-14-0011.1
4. Summary
The winds over southern Delaware and its coastal areas are subject to strong interannual, seasonal, and diurnal variations as well as considerable spatial variability. Overall, the mean wind speed in the region has changed little in the past 30 years. Our analysis of observations from B44009, located offshore the mouth of the Delaware Bay, indicates that there is a statistically significant increase in the mean wind speed in the past three decades in October and a similar magnitude decrease in May, however. As seen in previous work, northwesterly winds predominate in the winter and southerly winds persist in the summer, with wintertime wind speeds being significantly higher than summertime speeds. The mean summertime wind is observed to shift from the south, offshore and well inland, to the southeast over the Delaware Bay, which is related to the occurrence of the sea breeze. Spring and autumn are transition seasons with a broader spread in the dominant wind directions. As expected, wind speeds are strongest over open water, weaker at coastal stations, and lightest across inland stations.
Diurnal variability in wind direction persists over all regions and all seasons, with the most variable winds occurring during the afternoon in the summer. Coastal stations experience a clockwise shift from the afternoon to overnight hours, which may be driven by local thermal circulations such as the sea or land breeze. The largest spatial diurnal variability in wind direction occurs during summer afternoons with an onshore component at coastal stations and an offshore component both inland and offshore. This effect can be seen in the transition seasons as well.
WRF generally overestimates the mean wind speed by up to 1 m s−1 over the entire region for the nonwinter seasons, which is consistent with other studies and typical of mesoscale model bias in coastal regions. The model accurately simulates dominant wind directions and wind speeds both seasonally and for the most frequently occurring synoptic types; there is a persistent counterclockwise shift in the mean wind direction, however. Model simulations for the four dominant synoptic types in the summer clearly capture the large differences in the local wind regimes. Of the four dominant types, three clearly demonstrate the signature of a sea-breeze circulation while the other dominant type shows minimal spatial variation within the wind field.
Our analysis indicates that the wind regime throughout southern Delaware is complex, especially along the coastline. Mesoscale events such as the sea breeze clearly introduce large differences in wind speed and direction along the coastline in the spring, summer, and autumn seasons. Coastal air and sea surface temperatures are sensitive to changing wind patterns. Therefore, we conclude that the inland, coastal, bay, and ocean wind regimes are unique during the nonwinter seasons. These contrasting wind regimes affect the temperature and humidity across the region and may have an impact on many aspects of Delaware’s economy, including tourism, agriculture, energy, and air quality.
Acknowledgments
This work was generously supported by the Delaware Sea Grant foundation, Project R/ETE-12. We acknowledge the Delaware Environmental Observing System, National Data Buoy Center, and National Centers for Environmental Prediction for providing observational and model-derived data. We extend our thanks to Dan Leathers of the University of Delaware, who generously provided us with the temporal synoptic index data for the Philadelphia region, and to Rich Pawlowicz of the University of British Columbia who provided a MATLAB toolbox (M_MAP) that was used extensively in several of our figures. In addition, we thank the anonymous reviewers for their insightful and helpful comments.
APPENDIX
Description of the Dominant Synoptic Types Used in this Study
In each season, the four most frequent synoptic types were explored using both observations and WRF Model simulations to investigate the typical variability found in the region. For the winters, these types were (in order of frequency) strong northwest flow (1031), southwest flow (1032), anomalous northwest flow (1033), and weak high pressure (1034). The distribution of recorded winds speeds and directions by season at B44009 is shown in Fig. 5 for the selected days of the study. It is clear that each synoptic type is distinct from the others. For example, 1031 and 1033 both have dominant northwest winds, but the mean wind speed is 2.5 m s−1 greater in 1031. Type 1031 is also colder with a lower mean surface pressure, which is indicative of an Arctic air mass.
For the summertime, the four dominant synoptic types seen in the 60-yr time series are southerly flow with approaching front (3032), southwest flow (3031), southern trough aloft (3035), and cold-frontal passage (3033). Of these types, three of the four have strong southerly or southwesterly flow, with 3035 showing a distinct easterly flow. The wind roses for the spring and autumn show significant differences among the dominant synoptic type, as expect for these transition seasons. The spring dominant types are southwest flow (2031), weak pattern (2010), northwest flow (2032), and cold front off the coast (2034). For autumn, the most frequent synoptic types are high pressure (4032), northwest flow (4036), strong southwest flow (4037), and weak southwest flow (4038). Table A1 shows a description of these dominant types and the corresponding wind speeds and directions. It is interesting to note that southwest flow is a dominant type in all four seasons.
Dominant synoptic types and conditions (B44009) by season for 2006–12.
REFERENCES
Atkinson, L. P., J. L. Blanco, and R. Chekuru, 2013: Final report for task 2.0 extreme wind events and design parameters. U.S. Department of Energy Office of Energy Efficiency and Renewable Energy Wind and Water Power Program Final Rep., 34 pp. [Available online at http://www.ceoe.udel.edu/windpower/resources/ODU-final-report-28-June-2013.pdf.]
Azorin-Molina, C., S. Tijm, E. E. Ebert, S. M. Vicente-Serrano, and M. J. Estrela, 2014: Sea breeze thunderstorms in the eastern Iberian Peninsula. Neighborhood verification of HIRLAM and HARMONIE precipitation forecasts. Atmos. Res., 139, 101–115, doi:10.1016/j.atmosres.2014.01.010.
Carvalho, D., A. Rocha, M. Gómez-Gesteira, and C. Silva 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, doi:10.1016/j.apenergy.2013.12.001.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569–585, doi:10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Colby, F. P., Jr., 2004: Simulation of the New England sea breeze: The effect of grid spacing. Wea. Forecasting, 19, 277–285, doi:10.1175/1520-0434(2004)019<0277:SOTNES>2.0.CO;2.
Colle, B., and D. Novak, 2010: The New York Bight jet: Climatology and dynamical evolution. Mon. Wea. Rev., 138, 2385–2404, doi:10.1175/2009MWR3231.1.
Dowtin, A. L., 2012: Developing a drought climatology for Delaware (1948–2005). M.S. thesis, Dept. of Geography, University of Delaware, 103 pp. [Available online at http://sites.udel.edu/delaware-water-watch/files/2014/06/Dowtin2012_Developing-a-Drought-Climatology-for-Delaware-1948-2012-vvwxrr.pdf.]
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Dvorak, M. J., C. L. Archer, and M. J. Jacobson, 2010: California offshore wind energy potential. Renewable Energy, 35, 1244–1254, doi:10.1016/j.renene.2009.11.022.
Ellis, A., and D. Leathers, 1996: A synoptic climatological approach to the analysis of lake-effect snowfall: Potential forecasting applications. Wea. Forecasting, 11, 216–229, doi:10.1175/1520-0434(1996)011<0216:ASCATT>2.0.CO;2.
Epifanio, C. E., A. K. Masse, and R. W. Garvine, 1989: Transport of blue crab larvae by surface currents off Delaware Bay, USA. Mar. Ecol. Prog. Ser., 54, 35–41, doi:10.3354/meps054035.
Fealy, R., and J. Sweeney, 2007: Identification of frequency changes in synoptic circulation types and consequences for glacier mass balance in Norway. Norw. J. Geogr., 61, 76–91, doi:10.1080/00291950701374328.
Garvine, R. W., and W. Kempton, 2008: Assessing the wind field over the continental shelf as a resource for electric power. J. Mar. Res., 66, 751–773, doi:10.1357/002224008788064540.
Gilchrist, J., 2013: A radar analysis of the Delaware sea/bay breeze. M.S. thesis, School of Marine Science and Policy, University of Delaware, 89 pp. [Available online at http://udspace.udel.edu/handle/19716/13023.]
Hill, C. M., P. J. Fitzpatrick, C. H. Corbin, Y. H. Lau, and S. K. Bhate, 2010: Summertime precipitation regimes associated with the sea breeze and land breeze in southern Mississippi and eastern Louisiana. Wea. Forecasting, 25, 1755–1779, doi:10.1175/2010WAF2222340.1.
Hughes, C. P., 2011: The climatology of the Delaware Bay/sea breeze. M.S. thesis, School of Marine Science and Policy, University of Delaware, 129 pp. [Available online at http://udspace.udel.edu/handle/19716/10659.]
Hurley, P. J., W. L. Physick, and A. K. Luhar, 2005: TAPM: A practical approach to prognostic meteorological and air pollution modelling. Environ. Modell. Software, 20, 737–752, doi:10.1016/j.envsoft.2004.04.006.
Husar, R. B., R. L. Poirot, and N. Frank, 2007: Exceptional air pollution events: Regional and intercontinental dust and smoke. 2007 Fall Meeting, San Francisco, CA, Amer. Geophys. Union, Abstract A53C-1355.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, doi:10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Jimenez, B., F. Durante, B. Lange, T. Kreutzer, and J. Tambke, 2007: Offshore wind resource assessment with WAsP and MM5: Comparative study for the German Bight. Wind Energy, 10, 121–134, doi:10.1002/we.212.
Jiménez, P. A., J. F. González-Rouco, E. García-Bustamante, J. Navarro, J. P. Montávez, J. Vilà-Guerau de Arellano, J. Dudhia, and A. Muñoz-Roldan, 2010: Surface wind regionalization over complex terrain: Evaluation and analysis of a high-resolution WRF simulation. J. Appl. Meteor. Climatol., 49, 268–287, doi:10.1175/2009JAMC2175.1.
Jones, G. V., and R. E. Davis, 2000: Using a synoptic climatological approach to understand climate–viticulture relationships. Int. J. Climatol., 20, 813–837, doi:10.1002/1097-0088(20000630)20:8<813::AID-JOC495>3.0.CO;2-W.
Kain, J. S., and M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 24, Amer. Meteor. Soc., 165–170.
Kain, J. S., S. R. Dembek, S. J. Weiss, J. L. Case, J. J. Levit, and R. A. Sobash, 2010: Extracting unique information from high-resolution forecast models: Monitoring selected fields and phenomena every time step. Wea. Forecasting, 25, 1536–1542, doi:10.1175/2010WAF2222430.1.
Kalkstein, L. S., and P. Corrigan, 1986: A synoptic climatological approach for geographic analysis: Assessment of sulfur dioxide concentrations. Ann. Assoc. Amer. Geogr., 76, 381–395, doi:10.1111/j.1467-8306.1986.tb00126.x.
Kalkstein, L. S., G. Tan, and J. A. Skindlov, 1987: An evaluation of three clustering procedures for use in synoptic climatological classification. J. Climate Appl. Meteor., 26, 717–730, doi:10.1175/1520-0450(1987)026<0717:AEOTCP>2.0.CO;2.
Kalkstein, L. S., P. C. Dunne, and R. S. Vose, 1990: Detection of climate change in the western North American Arctic using a synoptic climatological approach. J. Climate, 3, 1153–1167, doi:10.1175/1520-0442(1990)003<1153:DOCCIT>2.0.CO;2.
Khan, B. A., 2010: Sea breeze circulation in the Auckland region: Observational data analysis and numerical modeling. Ph.D. thesis, University of Canterbury, 262 pp. [Available online at http://ir.canterbury.ac.nz/handle/10092/4476.]
Kyselý, J., R. Huth, and J. Kim, 2010: Evaluating heat-related mortality in Korea by objective classifications of ‘air masses.’ Int. J. Climatol., 30, 1484–1501, doi:10.1002/joc.1994.
Leathers, D., and A. Ellis, 1996: Synoptic mechanisms associated with snowfall increase to the lee of Lakes Erie and Ontario. Int. J. Climatol., 16, 1117–1135, doi:10.1002/(SICI)1097-0088(199610)16:10<1117::AID-JOC80>3.0.CO;2-4.
Lennon, G., 2010: Presentation to Delaware Renewable Energy Task Force. NRG Energy, Inc., Presentation Slides, 13 pp. [Available online at http://www.dnrec.delaware.gov/energy/information/Lists/Delaware%20Renewable%20Energy%20Task%20Force%20Calendar/Attachments/5/NRG%20Presentation%20on%20Utility%20Scale%20Solar.pdf.]
Luther, G. W., S. Ma, R. Trouwborst, B. Glazer, M. Blickley, R. W. Scarborough, and M. G. Mensinger, 2004: The roles of anoxia, H2S, and storm events in fish kills of dead-end canals of Delaware inland bays. Estuaries, 27, 551–560, doi:10.1007/BF02803546.
Maurmeyer, E. M., 1978: Geomorphology and evolution of transgressive estuarine washover barriers along the western shore of Delaware Bay. Ph.D. thesis, University of Delaware, 546 pp.
Mesinger, F., G. DiMego, and E. Kalnay, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/BAMS-87-3-343.
Miller, S. T. K., B. D. Keim, R. W. Talbot, and H. Mao, 2003: Sea breeze: Structure, forecasting, and impacts. Rev. Geophys., 41, 1011, doi:10.1029/2003RG000124.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the long-wave. J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237.
MMI Engineering, and Atmospheric and Environmental Research, 2012: Prediction of wind energy resources on the outer continental shelf with weather models. Bureau of Ocean Energy Management Task 2 Final Rep. M11PC00017, 57 pp. [Available online at http://www.boem.gov/uploadedFiles/BOEM/Environmental_Stewardship/Environmental_Studies/Renewable_Energy/M11PC00017_Task2_FinalReport.pdf.]
Moffatt, J., and F. Nichol, 2007: Sediment management plan: Rehoboth Bay, Sussex County, Delaware. DNREC Final Rep., 132 pp. [Available online at http://www.dnrec.delaware.gov/swc/Shoreline/Documents/Regional%20Sediment%20Management%20Plan%20-%20Rehoboth%20Bay.pdf.]
Monaldo, F. M., X. Lui, W. Pichel, and C. Jackson, 2014: Ocean wind speed climatology from spaceborne SAR imagery. Bull. Amer. Meteor. Soc., 95, 565–569, doi:10.1175/BAMS-D-12-00165.1.
Münchow, A., and R. Garvine, 1993: Buoyancy and wind forcing of a coastal current. J. Mar. Res., 51, 293–322, doi:10.1357/0022240933223747.
Muscarella, P., N. Barton, B. L. Lipphardt Jr., D. E. Veron, and A. D. Kirwan Jr., 2011: Surface currents and winds at the Delaware Bay mouth. Cont. Shelf Res., 31, 1282–1293, doi:10.1016/j.csr.2011.05.003.
Nawri, N., H. Björnsson, K. Jónasson, and G. N. Petersen, 2012: Evaluation of WRF mesoscale model simulations of surface wind over Iceland. Icelandic Meteorological Office Rep. VÍ 2012-010, 44 pp. [Available online at http://en.vedur.is/media/2012_010_web.pdf.]
Otkin, J. A., and T. J. Greenwald, 2008: Comparison or WRF model-simulated and MODIS-derived cloud data. Mon. Wea. Rev., 136, 1957–1970, doi:10.1175/2007MWR2293.1.
Reinard, M. S., K. Adou, J. M. Martini, and M. V. Johnston, 2007: Source characterization and identification by real-time single particle mass spectrometry. Atmos. Environ., 41, 9397–9409, doi:10.1016/j.atmosenv.2007.09.001.
Salamanca, F., A. Martilli, M. Tewari, and F. Chen, 2011: A study of the urban boundary layer using different urban parameterizations and high-resolution urban canopy parameters with WRF. J. Appl. Meteor. Climatol., 50, 1107–1128, doi:10.1175/2010JAMC2538.1.
Samet, J., S. Zeger, J. Kelsall, J. Xu, and L. Kalkstein, 1998: Does weather confound or modify the association of particulate air pollution with mortality? An analysis of the Philadelphia data, 1973–1980. Environ. Res., 77, 9–19, doi:10.1006/enrs.1997.3821.
Sathe, A., and W. Bierbooms, 2007: Influence of different wind profiles due to varying atmospheric stability on the fatigue life of wind turbines. J. Phys. Conf. Ser., 75, 012056, doi:10.1088/1742-6596/75/1/012056.
Schofield, O., and Coauthors, 2008: The decadal view of the Mid-Atlantic Bight for the COOL room: Is our coastal system changing? Oceanography, 21, 108–117, doi:10.5670/oceanog.2008.08.
Scully, M. E., 2010: The importance of climate variability to wind-driven modulation of hypoxia in Chesapeake Bay. J. Phys. Oceanogr., 40, 1435–1440, doi:10.1175/2010JPO4321.1.
Shimada, S., and T. Ohsawa, 2011: Accuracy and characteristics of offshore wind speeds simulated by WRF. SOLA, 7, 21–24, doi:10.2151/sola.2011-006.
Shimada, S., T. Ohsawa, S. Serizawa, and I. Yoneda, 2009: Characteristics of offshore winds at Shirahama Oceanographic Observatory. Proc. ISOPE 2009, Osaka, Japan, ISOPE, 424–428. [Available online at http://e-book.lib.sjtu.edu.cn/isope2009/data/papers/2009-TPC-407.pdf.]
Silva, J., C. Ribeiro, and R. Guedes, 2007: Roughness length classification of Corine Land Cover classes. European Wind Energy Conf., Milan, Italy, EWEA, 10 pp. [Available online at http://www.ewea.org/ewec2007/allfiles2/545_Ewec2007fullpaper.pdf.]
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3.pdf.]
Woods, B. K., T. Nehrkorn, and J. M. Henderson, 2013: A downscaled wind climatology on the outer continental shelf. J. Appl. Meteor. Climatol., 52, 1878–1890, doi:10.1175/JAMC-D-12-0216.1.
Young, I. R., S. Zieger, and A. V. Babanin, 2011: Global trends in wind speed and wave height. Science, 332, 451–455, doi:10.1126/science.1197219.
