TMCnet News
AN OCEAN OBSERVING AND PREDICTION EXPERIMENT IN PRINCE WILLIAM SOUND, ALASKA [Bulletin of the American Meteorological Society](Bulletin of the American Meteorological Society Via Acquire Media NewsEdge) Twenty years after the Exxon Valdez oil spill in Alaska a unique field experiment demonstrated an integrated ocean observing system with advanced technologies to enable weather, wave, and ocean circulation forecasting. S ystematic weather observations in North America have a long history dating to the eighteenth century and colonial times when the first country-wide weather organization was the U.S. Post Office and Benjamin Franklin was the Post Master General. In the nineteenth century, Matthew Maury of the U.S. Navy pioneered the collection and documentation of ocean weather and currents observed from ships so that mariners could use these data to shorten transoceanic voyages. The proliferation of the telegraph allowed terrestrial weather observations to be centralized, and newspapers distributed weather reports to the public. Technological innovations in the twentieth century, such as satellite imagery, telecommunications, powerful computers to drive numerical simulation models, and meteorological advancements, have provided a better mechanistic understanding of weather phenomena. Today there are thousands of weather stations reporting in near- real time and 10-day weather forecasts are routinely available from public and private sources. However, compared to terrestrial networks, observations from the oceans are limited and forecasts of winds, waves, and ocean currents are not as well developed. The National Oceanic and Atmospheric Administration (NOAA) Integrated Ocean Observing System (IOOS; information online at www.ioos.gov), through regional associations such as the Alaska Ocean Observing System (AOOS; see www.aoos.org), is developing an expansive infrastructure of networked observational platforms and forecast models. To demonstrate the utility of an ocean observing and forecasting system with diverse practical applications, such as oil spill response, search and rescue, fisheries, and risk management, a unique 5-yr project was conducted in Alaska. Prince William Sound (PWS) was chosen for the demonstration because of numerous historical studies conducted on oil spill- related impacts following the 1989 grounding of the Exxon Valdez tanker. Since that incident, a prototype ocean observing and forecasting system was developed focusing on oil spill trajectories and planning for different response scenarios. The overall objective was to demonstrate the ability of an ocean observing system to provide information that is critical for realtime decision making. A more specific objective was to quantitatively evaluate the performance of numerical models developed for PWS using an array of fixed and mobile observation platforms. Environmental setting. PWS is located in the northeast corner of the Gulf of Alaska (GOA) at about 60°N. The Trans-Alaska Pipeline carries oil south from the Arctic coast to the Port of Valdez (Fig. 1). The largest communities are Valdez (population 4,036) and Cordova (2,454). Chenega Bay (86) and Tatitlek (107) are Alaska Native villages, and Whittier (182) is mostly non- Native. Only Valdez and Whittier have highway access to the state's main road system. The economic base of PWS communities is almost entirely resource dependent. For example, the Cordova economy is based on commercial salmon fishing, and Valdez is supported primarily by the oil pipeline terminal. Cruise ships regularly transit PWS to take in the many natural wonders, including an intricate network of tidewater glaciers, rainforests, offshore islands, and marine life. PWS is surrounded by the Chugach Mountains, which reach 4,300 m, and contains the most extensive system of valley glaciers in North America. Most of the land area is in or adjacent to the Chugach National Forest, and much of it is designated or proposed wilderness area. With a shoreline length of about 6,900 km and a tidal range of 6-8 m, PWS has an enormously varied shoreline habitat of seastacks, reefs, rocky headlands, mud flats, eelgrass beds, wetlands, kelp forests, and cobble beaches. Meteorology and ocean circulation in the GOA and PWS. Atmospheric conditions over the GOA are primarily established by the interaction of storms associated with the Aleutian low and the surrounding coastal mountains (Wilson and Overland 1986). As a consequence of this interaction, the prevailing winds are cyclonic (counterclockwise), leading to positive wind stress curl over the ocean basin and downwelling-favorable wind stress over the continental shelf throughout most of the year. Upon encountering coastal mountains, moist storm air masses are elevated and adiabatically cooled, which leads to very high rates of coastal precipitation. Much of this precipitation is presumed to enter the ocean relatively rapidly because of the steep terrain, except in the winter season, when it is stored in mountain snowpacks. Downwelling-favorable winds are weakest in summer, build rapidly through fall to a winter maximum, and decrease through spring. In contrast, coastal freshwater discharge is a maximum in fall, minimal in winter (when precipitation is stored as snow), and increases gradually through spring and summer because of melting. Alongshore winds and freshwater discharge from coastal rivers and glaciers generate the Alaska Coastal Current (ACC) over the continental shelf. The ACC originates on the British Columbian shelf, flows cyclonically around the GOA, and then enters the Bering Sea through the Aleutian Islands (Royer 1998). Seasonal variations in wind and freshwater buoyancy forcing give rise to large changes in the strength and density structure of the ACC (Johnson et al. 1988). The ACC is narrow (<10 km), swift (30-100 cm s-1), and deep (~150 m) in winter; and broad (~40 km), relatively sluggish (10 cm s-1), and shallow (<50 m) in summer (Weingartner et al. 2005). October is a transition month, during which time the ACC evolves from its summer to its winter structure as winds intensify and runoff increases. Nevertheless, maximum near-surface currents are typically observed in late fall resulting from the strong baroclinic nature of the current at this time (Stabeno et al. 2004). Two major entrances connect the GOA with the central basin of PWS-Hinchinbrook Entrance (east) and Montague Strait (west). Two bathymetric troughs extend from the northern edge of the central basin. A 300-m-deep trough tends to the northeast and terminates in Valdez Arm. A second trough curves to the northwest, and broadens to form a smaller basin where maximum depth exceeds 700 m (Fig. 1). As the westward-flowing ACC encounters the Hinchinbrook Entrance, a substantial fraction of it turns northward into PWS. The remainder of this current continues across the mouth of Hinchinbrook Entrance, and then southwestward along Montague Island and westward again after rounding the southern tip of the island. Once in PWS, the upper-layer flow proceeds cyclonically around the central basin, with some of the flow feeding the waters exiting through Montague Strait (and also perhaps along the western side of Hinchinbrook Entrance) and some of it continuing into the northern PWS (Gay and Vaughan 2001; Vaughan et al. 2001). The surface circulation pattern varies seasonally in accordance with the cycle of winds and runoff and appears to be strongest in late fall and winter and weakest in summer. As much as 40% of the volume of PWS above 100-m depth is exchanged in summer (May-September) and 200% is exchanged in winter (October-April; see Niebauer et al. 1994). Although these estimates are uncertain, they nevertheless suggest that exchange between the shelf and PWS is substantial and efficient, and should therefore profoundly inf luence circulation and ecosystem processes in PWS. Moreover, it is conceivable that the timing and magnitude of the exchange of water between PWS and the ACC is somewhat episodic and may vary considerably from year to year. PWS OBSERVING SYSTEM. Following the Exxon Valdez oil spill, the Prince William Sound Science Center and the Oil Spill Recovery Institute (OSRI) implemented a prototype nowcast-forecast project to aid in the oil spill response. The system combined near-real-time and conventional ocean and meteorological observations with a numerical ocean circulation model to simulate (nowcast) and predict (forecast) ocean currents, temperature, and salinity. The ultimate objective was to couple the atmospheric and oceanic models with biological data and develop an ecosystem forecast model. Improvements in forecasting regional weather and ocean circulation would benefit local oil spill responders as well as state and federal fishery managers. In 2004 the observing system included coastal weather stations, a high-frequency (HF) radar array used to image surface currents in the central basin, a 4-km-grid regional atmospheric model, and a 1-kmgrid ocean model. The ocean circulation modeling system for PWS, called the PWS Nowcast/Forecast System, was based on the Princeton Ocean Model (POM) and was forced by winds, tides, heat flux, and freshwater buoyancy (Mooers et al. 2009, and references therein). Tidal heights and currents were computed from tidal harmonics (amplitudes and phases) interpolated from a northeast Pacific tide model. Wind stress was computed by the Regional Atmospheric Modeling System (RAMS), which is a mesoscale-resolving atmospheric model operated by the Alaska Experimental Forecast Facility (AEFF). Heating and cooling were given by the climatological monthly heat flux from the Comprehensive Ocean and Atmospheric Data Set (COADS). Freshwater runoff was derived from a hydrological model (Wang et al. 2001) and applied at surface grid points next to the land. Since 2005 the PWS observing system rapidly evolved to integrate with AOOS, and 1 of the 11 regional associations formed within the national infrastructure of IOOS. Observational components that were added to the existing array include eight terrestrial weather stations (three are collocated with real-time web cams), for model validation and improvement of estimates of precipitation; acoustic Doppler current profilers (ADCP) and surface ocean temperature and salinity sensors mounted on two existing National Data Buoy Center (NDBC) weather buoys; a NDBC wave buoy with an ADCP that was deployed in Montague Strait; a new weather buoy in the GOA off Montague Island; a vessel-mounted thermosalinograph, for monthly synoptic surveys; two instrumented subsurface moorings in Hinchinbrook Entrance and two in Montague Strait to improve estimates of water exchange between the GOA and PWS; a stream discharge gauge on the Copper River; and an upgraded surface current radar system. A new tide gauge was also installed in Whittier to supplement two existing gauges in Valdez and Cordova. AOOS designed the PWS observing system to demonstrate the utility of a high time-space density of environmental observations directly to the public through the Internet to enable accurate forecasts by regional-scale weather and wave models, and for data assimilation by an ocean circulation model. These models include the Weather Research and Forecasting (WRF) model, Simulating Waves in the Nearshore (SWAN) model, and Regional Ocean Modeling System (ROMS). The observational infrastructure in PWS and forecast model development took 5 yr to complete, brought together over 65 scientists from eight states, and cost approximately $5 million (Table 1). SOUND PREDICTIONS 2009. Quantitative evaluations of forecast model skill were based on 1) retrospective analyses of historical observations and model hindcasts, 2) comparisons with baseline performance during a field experiment in 2004, and 3) observational data collected during the intensive 2-week predictions period of the Prince William Sound Field Experiment (PWS FE) spanning spring and neap tides in July and August 2009. During the PWS FE, the fixed array of observing system instruments was augmented by thermosalinograph surveys (see Okkonen and Bélanger 2008), and additional vessel-based measurements of pressure (depth), conductivity (salinity), temperature, chlorophyll fluorescence, turbidity, and nutrients along latitudinal and longitudinal transects in the central basin (Fig. 2). Nearly continuous measurements of temperature and salinity were also collected using a Slocum glider and a REMUS-100 autonomous underwater vehicle (AUV). Four types of drifting buoys were used to track water velocities at different depths. Argospheres (Metocean Data Systems) are 28-cm-diameter spherical buoys that are designed to track oil floating on the surface. These drifters use an onboard GPS receiver and the Argos satellite system for location and tracking. Microstar drifters (Pacific gyre) are designed to track the mean surface current at a depth of about 1 m. A surface float contains the GPS receiver, telemetry system, antenna, batteries, and sensors. The U.S. Coast Guard (USCG) Self-Locating Data Marker Buoy (SLDMB) made by Metocean is designed specifically for deployment from a vessel or aircraft and for unattended operation during a 30-day lifetime. The SLDMB is accompanied by an onboard electronics package, which includes GPS positioning. The data are transmitted to a secure USCG website for use by trained search-and-rescue personnel. The USCG uses these drifters to construct current vectors from sequential SLDMB positions. Surface Velocity Program (SVP) drif ters (Pacif ic gyre) are 38-cm-diameter spherical buoys to which a drogue is attached, and they are expected to drift with the water at the depth of the center of the drogue. The drogue is a 2.5-m-long fabric tube suspended at a depth of 10 m. SVP drifters also use an onboard GPS receiver and the Argos satellite system for location and tracking. A total of 44 drifting buoys were repeatedly deployed, retrieved, and redeployed during the 16-day period. Model validation of surface and deeper currents in the central basin were emphasized, and the majority of drifter deployments occurred within the field of view of radar surface current measurements. Additional deployments occurred around the perimeter of PWS to validate the velocity of surface currents forced mostly by freshwater runoff from perimeter snow fields and glaciers. Weather forecasting. The WRF atmospheric circulation model is run twice daily at AEFF (http://aeff. uaa.alaska.edu) to enable 48-h ocean and 36-h wave forecasts that are more accurate than those provided by earlier regional atmospheric models (Liu et al. 2007). The U.S. National Weather Service forecasts with 12-km grid spacing are used to provide the lateral boundary for the 4-km WRF forecasts. Weather observations used to validate the WRF forecasts are provided by land-based weather stations and NDBC buoy-mounted stations. The simulations are performed in a nested grid fashion. The outer domain (grid 1) has 90 × 72 points with a 12-km ?x, y, and the nested inner grid (grid 2) has 141 × 78 points with a 4-km ?x, y. The model has 42 vertical levels. This configuration is referred to as PWS-WRF. The intent of the 4-km PWS-WRF grid 2 is to resolve some of the more localized surface flow features. The model is initialized using forecasts from the operational North American Mesoscale Model (NAM) run at the National Centers for Environmental Prediction. While NAM is run on a 12-km grid, the PWS-WRF is initialized using data interpolated to a 40-km grid [World Meteorological Organization (WMO) grid 216], the highest-resolution grid available at the time of the PWS FE. Most of the analysis and verification effort during the PWS FE was focused on surface (10 m) winds as input to the ocean and wave models (Fig. 3). The surface meteorology in the northern GOA in the summer can be characterized as fairly quiescent, punctuated by periods of storminess as frontal systems associated with low pressure systems either pass through or move to the south of the region. From the standpoint of the ocean model input, it is most instructive to focus the comparison with data from NDBC buoy 46060 in the central basin of PWS because interference from terrestrial features is relatively minimal at this site (Fig. 4). Perhaps most striking was the very consistently observed east-southeast wind direction during the PWS FE. The model shows a clear bias toward more southerly winds, averaging about 15° throughout the period. The observed wind speeds were typically below 10 m s-1 but three storm events occurred when winds exceeded 10 m s-1. For the most part (with some notable exceptions) the simulated wind speeds tracked the observed winds, with a mean bias of around 1 m s-1. However, there were episodes when the model went astray, both underpredicting (21, 22, and 25 July) and overpredicting (20 and 26 July) wind speed by several meters per second. Similar results occurred at other weather stations. The causes of error in the surface wind predictions, apart from systematic model bias, arose from two main factors. The first was poor initialization by the operational NAM, which resulted in inaccurate forecasts, and hence inaccurate initialization fields for the PWS-WRF runs. Even a 50-km error in the initial location of a low in the northern GOA can result in significant errors in wind direction throughout the duration of the run. Also, when the pressure gradient was very weak over the region, significant wind errors occurred. The observed wind direction would then become quite variable, while the model tended to produce persistent winds of a few meters pre second from the southeast, suggesting the need for more weather observations in this area of complex topography. Wave forecasting. Wind-forced waves constitute an extremely energetic component of the physical oceanography affecting coastal Alaska. From a practical standpoint, information about the wave conditions in the Alaskan coastal oceans is needed to assess the fate of oil spills, related recovery efforts, and safe vessel operations. The grid-based SWAN model was established in PWS because buoy and satellite altimeter measurements of waves in coastal waters suffer from spatial and/or temporal sampling limitations. This effort was based at the Texas A&M University Maritime Systems Engineering Department (see www.tamug.edu/mase). The wave model is forced by the PWS-WRF winds, while NOAA's global wave forecasts (through the thirdgeneration wave model WAVEWATCH III) provide the offshore boundary conditions. The model runs every 24 h to predict significant wave height (SWH) out to 36 h. The wave model uses data collected from four NDBC buoys for ongoing validation in PWS, as well as the Cape Suckling (buoy 46082) and Cape Cleare (buoy 46076) buoys to validate GOA waves (Singhal et al. 2010). During the PWS FE, the waves were generally small with SWHs around 0.5 m, except during a storm event on 22 July when the SWH exceeded 2 m in the central sound (Fig. 5). In general, the spatial variability in the measured SWHs is reasonably captured by the model. Based on a detailed analysis of the model results at the location of NDBC buoy 46060, however, the model overpredicted the SWHs during the PWS FE (Fig. 6). Nevertheless, the RMS difference for the 12-h forecast was smaller than its 36-h counterpart (RMS differences of ~0.19 and ~0.29 m for the 12- and 36-h forecasts, respectively), indicating that the forecast accuracy is better for shorter lead times. Overall, the SWAN model errors, especially for the smaller waves, could be attributed to the corresponding errors in the PWS-WRF winds during the PWS FE. Ocean circulation nowcasting and forecasting. A real-time nowcast-forecast system for PWS circulation was developed at the NASA Jet Propulsion Laboratory (http://ourocean.jpl.nasa.gov/PWS) based on the ROMS, a community model widely used for regional and coastal ocean modeling studies (Shchepetkin and McWilliams 2005). The PWS- ROMS consists of a three-domain nested configuration covering the northeastern Pacific Ocean at 10-km resolution, the northeastern GOA at 3-km resolution, and PWS at 1-km resolution. The PWS- ROMS is forced by the PWS-WRF regional atmospheric model. Freshwater input is computed from a modified hydrological model of Wang et al. (2004) using a digital elevation model and the air temperature and precipitation data from the PWS-WRF. For all three domains, a three-dimensional variational data assimilation (3DVAR) methodolog y (Li et al. 2008) was used to enable nowcast (analysis) and forecast. This ROMS 3DVAR forecasting system has been used successfully in several other field experiments, including the 2003 Autonomous Ocean Sampling Network (www.mbari.org /aosn) experiment in Monterey Bay (Chao et al. 2009). During the PWS FE, nowcasts were produced every 6 h and a 48-h forecast was made once per day. In addition, a 16-member ensemble of forecasts was executed in order to estimate the uncertainties in the simulated drifter trajectories. Data were assimilated from the glider and AUV, buoy-mounted ADCPs, and radar surface current mappers. Quantitative analyses of these data are in progress and will be presented in a forthcoming special issue. The PWS-ROMS surface circulation nowcast was evaluated first with the radar surface current data used in the assimilation and then with independent drifter trajectory data that were not used for assimilation. As noted above, during most of the PWS FE, there were strong east-southeast winds. This wind forcing resulted in strong north-northwestward surface flow in much of central PWS. Observations and the PWS-ROMS nowcasts show this general flow pattern, which was also confirmed by drifter trajectories in the region. The time mean (16 July-3 August 2009) radar surface currents in central PWS were compared with the PWS-ROMS surface current nowcasts (Fig. 7), and the variation in time of the root-mean-square difference of the radar observations were compared with collocated values from the PWS-ROMS surface currents (Fig. 8). The RMS differences are between 5 and 10 cm s-1, though some are significantly larger, particularly just before the radars stopped working (e.g., days 202 and 207). The RMS differences for the run without radar surface currents (red triangles) shows a clear and substantial impact of the radar data assimilation because the RMS differences are generally twice as large (or more), ranging from 10 to 20 cm s-1 in most cases. In addition, there is somewhat greater day-to-day variability in the RMS values in this run. Next, the PWS-ROMS forecast results are compared with independent data (i.e., observations not used for assimilation), specifically, the Microstar and USCG SLDMB drifter track data. The distance between the simulated drifter locations and the observed drifter locations as a function of time from deployment is examined (Fig. 9). Starting from the same drifter release locations, the simulated drifter tracks were computed using hourly forecasted surface currents over a 48-h period. There is a gradual increase in separation up to 10-15 km through the first 20 h. Note that during the second half of the PWS FE, a cyclonic eddy, similar to that observed in 2004, dominated the central sound. Most drifters were released in the central sound to avoid hitting land, and these generally followed the cyclonic eddy. The asymptote feature (Fig. 9) of the drifter track comparison likely reflects the curvature of the observed and/or simulated drifter trajectories. A more careful study is needed to further understand this asymptote feature around 10-15 km or ideally a drifter release experiment in a larger unbounded ocean would be needed to quantify the increase of error in forecasting the drifter trajectory. Education and outreach. AOOS and the Center for Ocean Sciences Education Excellence (COSEE) Alaska (www.coseealaska.net) and COSEE Networked Ocean World (NOW) combined resources to use the PWS FE as a major opportunity for outreach and ocean education to Alaskans and the nation. This collaboration also involved the Alaska Sea Grant Program and the University of Alaska Fairbanks School of Fisheries and Ocean Sciences. The PWS FE was utilized as a tool for outreach about AOOS, ocean observing and forecasting, oceanographic technologies, and the value of ocean ecosystems. This was accomplished through the use of podcasts (http://coseenow.net/podcast), blogs (http:// princewilliam2009.blogspot.com), the AOOS website, and a series of public seminars and events in Anchorage, Cordova, and Valdez. DISCUSSION AND SUMMARY. The 5-yr AOOS demonstration of an ocean observing system in PWS was a success in terms of routinely collecting field observations and providing forecasts of weather, wave, and ocean conditions in near-real time to the public through the Internet. All three forecasting models show significant skill in predicting 36-48 h into the future, and a quantitative assessment of these simulations will be presented in a special journal publication as the data are more fully analyzed. The preliminary results also identified a number of weaknesses, including the following: 1) Weather forecasting by WRF: i) Wind direction was difficult to predict in light and variable conditions (< 5 m s-1) due to topographical deflection in some mountainous areas of PWS. ii) Poor initialization from the NAM resulted in inaccurate regional forecasts. 2) Wave forecasting by SWAN: i) Wave height forecasts (in general) were overpredicted for the PWS FE. ii) Forecast errors, especially for smaller waves, could be attributed to corresponding errors in the input winds. 3) Ocean circulation nowcasting and forecasting by ROMS: i) There is a rapid growth of errors in forecasting the drifter trajectories within the first 24 h, suggesting challenges and perhaps limitations of using surface current measurements derived from HF radar and relatively limited subsurface observations for data assimilation. ii) Although not specifically addressed in this paper, the lack of either any real-time hydrological measurements within PWS or estimates of freshwater input from surface and subsurface glacial melt will need to be resolved to quantify the likely significant effect of buoyancy forcing on surface currents. This ocean observing demonstration utilized some of the most sophisticated technology available and the expertise of a team of scientists from across Alaska and the nation. The organizational and logistical obstacles encountered during the PWS FE were formidable mostly due to the remoteness of the study area and the distributed nature of the resources and assets focused on PWS for the 2-week period. An HF radar hardware failure during an unusually severe summer storm was an example. Parts had to be located and flown to Cordova and then, because the seas were too rough for a surf landing by skiff, a helicopter was used to transport technicians to the site for hardware repairs. Other examples are described in a daily blog (http://princewilliam2009. blogspot.com). Planning for the experiment took over 2 yr and monthly telephone conferences for all the participants began 6 months before and continued twice daily during the PWS FE. Observational sampling methods yield data that were often limited in space and time, but models enable nowcasts and forecasts over relatively large domains. Alaska's oceans continue to be undersampled and numerical simulation models, especially using model ensembles, will be an important tool for filling data and information gaps. The ocean observing system in PWS demonstrated that we can make routine ocean circulation forecasts with performance similar to weather forecasts that are widely accepted, even though they are often acknowledged as being inaccurate (especially for longer-term forecasts). Even with their uncertainty and limitations, real-time ocean circulation data and forecasts are used by federal and state agencies and industry for oil spill response, and by the USCG for the new Search and Rescue Optimal Planning System (SAROPS) for planning search-and-rescue missions. The challenge for the PWS observing system is how to assess the costs and benefits and determine which components, along with the most relevant scales of space and time, are most responsive to the needs of stakeholders. Interestingly, the most popular products from the AOOS web site are the relatively low-cost weather stations observations and images from web cameras. The wind, wave, and ocean circulation observations and forecasts are probably the most critical component for the community of oil spill and search-and-rescue responders, but generally only when an event occurs. Paying for a system in advance of an anticipated need has proven to be difficult to justify. For these reasons developing an operational ocean observing system, especially in a remote setting, continues to be a significant challenge. ACKNOWLEDGMENTS. Funding was provided by the Alaska Ocean Observing System and the Prince William Sound Oil Spill Recovery Institute. Additional funding was provided by the National Aeronautics and Space Administration (NASA) Earth Science. We are especially grateful for the support from NASA Public Health program managers John Haynes and Sue Estes. Support was also provided by the Prince William Sound Science Center and the Prince William Sound Regional Citizens' Advisory Council. The research for Y. Chao was carried out, in part, at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. The demonstration project and field experiment investigators include A. Allen, C. Bèlanger, M. Burdette, R. Campbell, F. Chai, J. Ewald, M. Halverson, E. Howlett, M. Johnson, P. Li, Z. Li, R. McClure, M. Moline, J. C. McWilliams, C. Ohlmann, S. Okkonen, V. Panchang, S. Pegau, and T. Weingartner. We thank the three anonymous reviewers for suggestions that greatly improved an earlier version of this manuscript. REFERENCES Chao, Y., and Coauthors, 2009: Development, implementation and evaluation of a data-assimilative ocean forecasting system off the central California coast. Deep-Sea Res. II, 56, 100-126, doi:10.1016/j. dsr2.2008.08.011. Gay, S. M., and S. L. Vaughan, 2001: Seasonal hydrography and tidal currents of bays and fjords in Prince William Sound, Alaska. Fish. Oceanogr., 10, 159-193. Johnson, W. R., T. C. Royer, and J. L. Luick, 1988: On the seasonal variability of the ACC. J. Geophys. Res., 93, 12 423-12 437. Li, Z., Y. Chao, J. C. McWilliams, and K. Ide, 2008: A three-dimensional variational data assimilation scheme for the Regional Ocean Modeling System. J. Atmos. Oceanic Technol., 25, 2074-2090. Liu, H., P. Q. Olsson, and K. Volz, 2007: SAR observation and modeling of gap winds in the Prince William Sound of Alaska. Sensors 2008, 8, 4894-4914, doi:10.3390/s8084894. Mooers, C. N. K., X. Wu, and I. Bang, 2009: Performance of a nowcast/forecast system for Prince William Sound, Alaska. Cont. Shelf Res., 29, 42-60, doi:10.1016/j.csr.2007.09.008. Niebauer, H. J., T. C. Royer, and T. J. Weingartner, 1994: Circulation of Prince William Sound, Alaska. J. Geophys. Res., 99C (14), 113-114. Okkonen, S. R., and C. Bélanger, 2008: Annual period temperature and salinity signals of surface waters in Prince William Sound, Alaska. Geophys. Res. Lett., 35, L14604, doi:10.1029/2008GL034456. Royer, T. C., 1998: Coastal processes in the northern North Pacific. The Sea, A. R. Robinson and K. H. Brink, Eds., The Global Coastal Ocean: Regional Studies and Syntheses, Vol. 11, J. Wiley and Sons, 395-414. Shchepetkin, A. F., and J. C. McWilliams, 2005: A threedimensional variational data assimilation scheme for the Regional Ocean Modeling System: Implementation and experiments. Ocean Modell., 9, 347-404. Singhal, G., V. G. Panchang, and J. L. Lillibridge, 2010: Reliability assessment for an operational wave forecasting system in Prince William Sound, Alaska. J. Waterw. Port. Coastal Ocean Eng., 136, 337-349. Stabeno, P. J., N. A. Bond, A. J. Hermann, N. B. Kachel, C. W. Mordy, and J. E. Overland, 2004: Meteorology and oceanography of the Northern Gulf of Alaska. Cont. Shelf Res., 24, 859-897, doi:10.1016/j. csr.2004.02.007. Vaughan, S. L., C. N. Mooers, and S. M. Gay, 2001: Physical variability in Prince William Sound during the SEA Study (1994-98). Fish. Oceanogr., 10, 58-80. Wang, J., M. Jin, V. Patrick, J. R. Allen, D. L. Eslinger, and R. T. Cooney, 2001: Numerical simulations of the seasonal circulation patterns and thermohaline structures of Prince William Sound, Alaska. Fish. Oceanogr., 10, 132-148. _____, _____, D. L. Musgrave, and M. Ikeda, 2004: A hydrological digital elevation model for freshwater discharge into the Gulf of Alaska. J. Geophys. Res., 109, C07009, doi:10.1029/2002JC001430. Weingartner, T. J., S. L. Danielson, and T. C. Royer, 2005: Freshwater variability and predictability in the Alaska Coastal Current. Deep-Sea Res. II, 52, 169-191, doi:10.1016/j.dsr2.2004.09.030. Wilson, J. G., and J. E. Overland, 1986: Meteorology. The Gulf of Alaska, Physical Environment and Biological Resources, D. W. Hood and S. T. Zimmerman, Eds., DOC/NOAA, 31-54. AFFILIATIONS: Schoch-Alaska Ocean Observing System, and Coastwise Services, Anchorage, Alaska; Chao-Jet Propulsion Laboratory and California Institute of Technology, Pasadena, California, and University of California, Los Angeles, Los Angeles, California; Colas and Farrara-University of California, Los Angeles, Los Angeles, California; McCamm on-Alaska Ocean Observing System, Anchorage, Alaska; Olss on-Alaska Experimental Forecast Facility, University of Alaska, Anchorage, Anchorage, Alaska; Singhal-Texas A&M University, College Station, Texas CORRESPONDING AUTHOR: Carl Schoch, 1199 Bay Ave., Homer, AK 99603 E-mail: [email protected] The abstract for this article can be found in this issue, following the table of contents. DOI:10.1175/2011BAMS3023.1 In final form 10 March 2011 ©2011 American Meteorological Society (c) 2011 American Meteorological Society |