A Coupled Global/Regional Circulation Model for Ecosystem Studies in the Coastal Gulf of Alaska
A. J. Hermann1, D. B. Haidvogel2, E. L. Dobbins1, P. J. Stabeno3 and P. S. Rand4
1Joint Institute for the Study of Atmosphere and Ocean (JISAO)
2Institute of Marine and Coastal Sciences
Rutgers University
New Brunswick, New Jersey 08903
4Department of Zoology
North Carolina State University
Raleigh, NC 27695
For submission to: Progress in Oceanography
ABSTRACT
To study the impact of interannual-to-decadal changes in circulation and hydrography on lower trophic level dynamics in the Coastal Gulf of Alaska (CGOA), we are developing a suite of nested physical and biological models as part of the West Coast U.S. GLOBEC program. Components of the multi-scale system include a variable-resolution, global circulation model; a higher-resolution, regional circulation model; a lower trophic level (NPZ) model; and an individual-based salmon model. Here we describe the attributes and coupled behavior of the first two of these components.
The global circulation model is a version of the Spectral Element Ocean Model implemented in layered, primitive equation form on an unstructured grid, and driven by global winds. The regional physical model consists of the S-Coordinate Rutgers University Model configured with approximately 22 km resolution in the CGOA and driven by runoff, heat fluxes and wind stresses appropriate to years 1976, 1995, 1996 and 1997. The coupled models develop appropriate boundary currents (the Alaskan Stream and the Alaska Coastal Current), and also spin up large (~200km) eddy features which appear to play a significant role in cross-shelf exchange.
Model-generated seasonal patterns and eddy structures are consistent with recent and historical data (hydrographic, drifter track, SSH, and SST). As part of our model validation for GLOBEC, we explore: 1) interannual differences in the broad regional circulation and temperature; 2) Eulerian and Lagrangian aspects of a large eddy feature near Sitka, Alaska; and 3) the impacts produced by forcing the regional model with barotropic data from the global model.
1. Background
1.1. Goals of this project
A core hypothesis of the U.S. GLOBEC Northeast Pacific program is that interannual to interdecadal variability in the circulation and hydrography of the Gulf of Alaska drives changes in productivity of zooplankton in the coastal zone, with consequent effects on the feeding success of salmonids and other species in the Gulf (U.S. GLOBEC, 1996). This hypothesis is partly based on strong evidence that a major shift in both the physical and biological character of the Gulf of Alaska occurred in 1976-1977 (Trenberth, 1990; Brodeur and Ware, 1992; Trenberth and Hurrel, 1994; Brodeur and Ware 1995; Brodeur et al., 1996). Trenberth (1990) documented a deepening and eastward shift of the Aleutian Low as part of that climate shift. However, a recent analysis by Lagerloef (1995) suggests that this deepening was in fact accompanied by weakened wind-driven cyclonic circulation in the Gulf. Polovina et al. (1995) showed that mixed layer temperatures rose in the Gulf after the climate shift, while the mixed layer shoaled. Brodeur and Ware (1992) demonstrated enhanced zooplankton levels after the shift, especially towards the coastal perimeter of the Gulf, which they attributed in part to greater Ekman drift of zooplankton out of the central Northern Pacific.
To address these issues for the CGOA, we have been developing with our biological colleagues a set of linked circulation models, coupled with a lower trophic level Nutrients-Phytoplankton-Zooplankton (NPZ) biological model, and an individual-based model (IBM) of salmon. Specific issues to be addressed by this set of models include the relative importance of surface Ekman flux, flows through submarine canyons, and mesoscale eddies on cross-shelf exchange, and the subsequent impacts of that exchange on plankton and fish (both through resupply of nutrients and transport of the organisms themselves).
As part of this larger effort for GLOBEC, we have thus far developed both global and regional circulation models, bathymetry and forcing datasets for each, and a method for passing information from the global to the regional model. Initial runs have yielded prominent spatial, seasonal and interannual differences. Here, our goals are: a) to describe the coupled global/regional physical modeling system; b) to assess the primary features of the physical circulation generated by the coupled physical models; c) to describe eddy dynamics of the regional model in the vicinity of the widely reported eddy near Sitka AK (the "Sitka Eddy" first reported by Tabata [1982]); and d) to use Lagrangian tracking as a zero-order method to infer how such eddy-scale variability and Ekman flux could affect plankton and salmon dynamics in the CGOA.
1.2. Overview of circulation in the GOA
We begin by reviewing present knowledge about the mean circulation and biology of lower trophic levels in the Gulf of Alaska, and their variability. The Gulf of Alaska contains two major current systems: the Alaskan Current (AC)- Alaskan Stream (AS) and the Alaska Coastal Current (ACC) (Fig. 1). The AS is the intensified northern boundary of the AC; both are part of the subarctic gyre forced by cyclonic winds in the northeastern Pacific. Reed and Schumacher (1986) summarized knowledge of the AC-AS and ACC; significant new data have been reported since their review.
The subarctic gyre of the Northeast Pacific is much broader in the east than in the western part of the GOA, where it forms the narrow, swift AS (Reed, 1984). This juxtaposition of eastern and western boundary currents is unique. The AS is constrained by a steep continental rise which parallels the Aleutian Island chain. It is generally steady on seasonal time scales, but varies interannually (Reed, 1984; Musgrave et al., 1992; Lagerloef, 1995). Chelton and Davis (1982) first suggested that the transport in the Alaska and California Current systems fluctuated out of phase on interannual time scales, both being fed by the West Wind Drift. Kelly et al. (1993) found such interannual fluctuations in altimeter data.
The annual mean flow of the AS is approximately 10 Sv above 500 m depth, and between 15-30 Sv integrated to the bottom, with significant interannual variability (Reed, 1984); this annual mean conforms to forcing by the annual mean integrated wind stress curl of the far northern Pacific (Musgrave et al., 1992). In AVHRR imagery, the AS typically shows up as a warm band of SST along the shelf break in the northern Gulf (Royer 1983). Lagerloef's (1995) analysis of wind climatology indicates a net southward Ekman transport of 1 Sv from the Gulf, and divergence ~300 km offshore at the head of the Gulf corresponding to an upwelling rate of 3x10-6 m/s. Interannual variability of sea surface height in the Gulf has been observed using altimeter data (Bhaskaran et al., 1993; Matthews et al., 1992; Strub and James, 2000a,b).
The intermittently formed Sitka eddy centered off Sitka, Alaska (Tabata, 1982), and meanders of the Alaskan Stream in the central and western Gulf (Musgrave et al., 1992; Reed and Stabeno, 1993; Stabeno and Reed, 1989; Thomson and Gower, 1998) are prominent mesoscale features with scales of ~200 km. These features have been measured with sea surface height (TOPEX), sea surface temperature (AVHRR), and hydrographic (T, S, O2) data. TOPEX data in particular has indicated lifespans of months to years (Crawford and Whitney, 1999). These large eddies typically drift at < 2 cm/s, are predominantly anticyclonic, are more commonly observed in spring. Modeling studies have suggested that such eddies near the shelf break are intensified by ENSO warm events (Melsom et al. 1999).
Royer (1989) found decadal variation in spatially averaged SST data for the Gulf, as did Polovina et al. (1995), who noted shoaling of mixed layer depths during 1976-1988. Lagerloef (1995), using EOF (empirical orthogonal function) analysis of XBT and CTD data spanning 1968-1990, found a consistent pattern for interdecadal variations in the Gulf. His time series for the first mode dynamic height EOF (referenced to 500db) exhibited a striking correlation with both: 1) the North Pacific anomaly, defined by Trenberth and Hurrel (1994) as an index of the strength and position of the Aleutian Low, and 2) a time series of observed SST anomalies for the Gulf. Evidently, when the Aleutian Low is strong (as in the period 1976-1989), the AC-AS circulation is weaker and the spatial mean SST of the Gulf rises. The weakened circulation appeared to result from anticyclonic anomalies in the wind stress curl pattern over the Gulf during periods of intense Aleutian Low. Concurrently in the central Pacific, mixed layer depth increased and SST declined (Polovina et al., 1995). During the early 1970's the Aleutian Low was weak, the anticyclonic circulation in the Gulf more intense and SST depressed below the long term mean value. Shorter-term fluctuations associated with El Nino also exhibit this spatial pattern (Lagerloef, pers. comm.).
The ACC is driven by a widely distributed coastal source of freshwater and downwelling favorable winds (Royer, 1981; Schumacher et al., 1990). Continuity of this current in the northern and western Gulf has been established (Stabeno et al., 1995a). Freshwater input is greatest in October and smallest in March (Royer, 1982), while downwelling-favorable winds peak in January for the northern Gulf (Royer, 1983; Wilson and Overland, 1986). Estimated runoff is well correlated with measured baroclinic transport in the northern Gulf (Royer, 1983); the transport is ~30 times larger than the runoff value. Salinity determines the density field for much of the year, but a seasonal thermocline forms in the summer (Royer, 1983; Stabeno et al., 1995a).
Significant bifurcation of the ACC occurs at several locations along the coast, with branches joining the AS. The first bifurcation is directly west of Kayak Island, near the eastern edge of Prince William Sound (Royer et al., 1979). The second is at Kennedy-Stevenson entrances east of Shelikof Strait, where ~25% of the transport flows along the south side of Kodiak Island, eventually joining with the AS. The third is at the western exit of Shelikof Strait, where -25% joins the AS. Finally, near the Shumagin Islands ~50% of the remaining ACC joins the AS (Stabeno et al., 1995a; Stabeno and Reed, 1989; Stabeno, unpublished data).
As with the AC-AS system, there are significant differences in the character of the ACC between the eastern and western Gulf. The shelf is much broader in the northern and western Gulf than further east, while bathymetric irregularities (banks and canyons) are found in all areas. Alongshore wind forcing is considerably weaker in the eastern Gulf (Wilson and Overland, 1986), with strongest downwelling-favorable winds in the north and west. Flow of the ACC is also weaker in the eastern GOA (Reed et al., 1981); presumably this is due mainly to the weaker local downwelling. It has been suggested that the baroclinic structure of the ACC at this upstream location is too weak to support baroclinic instability (Swaters and Mysak, 1985), which is prevalent in parts of the western Gulf (Mysak et al., 1981).
Interannual variability of the ACC has been documented by Stabeno et al. (1995a), and some decadal variability is evident in the long-term hydrographic series of Royer at approximately 150 degrees W (the GAK line). An 18.6 year signal was discovered in the GAK temperature series, which Royer (1993) hypothesized might be due to modulation by tidal forcing at that frequency. Positive temperature anomalies were also recorded at depth at the GAK line, 6-9 months after each of two El Nino events. In addition, Royer (pers. comm.) has noted decadal trends in his CGOA runoff time series, which could affect trends in the strength of the ACC and mixed layer depths of the Gulf.
There has been considerably more study of circulation in the Gulf west of 150 degrees W, largely as a result of the ongoing Fisheries Oceanography Coordinated Investigations (FOCI) program (Schumacher and Kendall, 1995) in that region. Moored current meter records, in conjunction with geostrophic calculations from CTD casts, indicate that the mean flux of the ACC through Shelikof Strait in the spring is ~0.6 x 106 m3/s (Reed and Schumacher, 1989; Reed and Bograd, 1995). However, data from cross-strait arrays of current meters convey intense variability, from weakly reversed transport to fluxes as strong as 3 x 106 m3/s, on a time scale of days (Schumacher et al., 1990). A sea valley connects deep waters of the Strait with the open Gulf (see Fig. 1). An estuarine type circulation pattern is observed in the sea valley, with deep water entering below 150 m, and southwestward surface outflow on the northwestern side (Schumacher et al., 1990).
Meanders and eddies are common features of the sea valley, resulting from baroclinic instability of the mean flow through the Strait (Mysak et al., 1981). Remotely-sensed sea surface temperature data (AVHRR), in conjunction with drifter tracks and CTD surveys, have clearly revealed the presence of eddies with approximately 25 km radius and lifetimes of several weeks (Vastano et al. 1992; Schumacher et al., 1993). These eddies extend below 100 m depth (Schumacher et al., 1993; Bograd et al, 1994), and translate at speeds much slower than the mean currents of the valley (Schumacher et al., 1993). A persistent eddy has also been revealed by hydrographic and drifter data just west of Kayak Island (Royer et al., 1979). It is presently unknown what role the nearby Copper River outflow may play in the maintenance of this eddy, although models of coastal buoyancy outflows (Chao and Boicourt, 1986; Kourafalou et al., 1996) suggest a possible link.
Tidal currents are strongest in the vicinity of Kodiak Island, and especially strong in Cook Inlet, where tidal currents in excess of 100 cm s-1 have been observed. Tidal models of Isaji and Spaulding (1986) and Liu and Leendertse (1986; 1990) show good agreement with these observations. Both models and the data of Schumacher and Reed (1980) indicate strong tides and resultant tidal mixing on Portlock Bank, just east of Kodiak Island.
2. Methods
2.1. Overview of CGOA models
Past physical modeling efforts for the CGOA relevant to the present study include: global eddy-resolving models which include the Gulf (Semtner and Chervin, 1992; Fu and Smith, 1996), North Pacific models which include the Gulf (Cummins and Mysak, 1988; Ingraham and Miyahara, 1988; Cummins, 1989; Hsieh and Lee, 1989; Heim et al., 1992; Hurlburt et al., 1992; Lee et al., 1992; Cummins and Freeland, 1993; Miller et al., 1994; Hurlburt et al., 1996), regional models of portions of the Gulf (Isaji and Spaulding, 1986; Liu and Leendertse, 1990; Hannah et al., 1991; Foreman et al., 1992; Waiters and Foreman, 1992; Foreman et a1., 1993) and very local models of specific estuaries and embayments. Both Cummins (1989) and Lee et al. (1992) noted the importance of bottom topography in limiting the variability and setting the vertical structure of flows in the AC-AS system.
As components of the Shelikof Strait FOCI program, Stabeno et al. (1995b) and Hermann and Stabeno (1996) have used a primitive equation model in topography-following coordinates (the S-coordinate Primitive Equation Model [SPEM] of Haidvogel et al., 1991) to explore interannual variability in the circulation between Kodiak Island and the Shumagin Islands in the northern and western GOA. The model is driven with FNOC winds and the runoff series of Royer (1982 and personal communication). Stabeno and Hermann (1996) demonstrate how the model compares favorably with observed currents near Shelikof Strait.
Large-scale wind forcing is obviously significant for the AC-AS, whereas for the ACC both wind and nearshore buoyancy forcing play a crucial role. Biologically relevant physics near the coast include the freshwater input, baroclinic instability at scales of tens of kilometers, tidal mixing fronts anchored to small-scale topographic features, and frontal activity near the shelf break due to the generation of internal tides (Huthnance, 1995). Quasigeostrophic and layer models cannot realistically incorporate the relevant coastal physics (runoff and tides), while full primitive equation level models are very expensive to run at very high resolution on basin scales. Simple 1-1/2 layer basin/coastal models allow high resolution and capture some wave dynamics, but exclude baroclinic instability and tidal mixing effects. The primitive equation model of Hermann and Stabeno (1996) for Shelikof Strait and downstream includes the baroclinic instability physics, but is limited in spatial coverage.
To satisfy the multiple aims and long time scales of GLOBEC, we have been developing a set of coupled global and regional models, with finest resolution in the CGOA. A suitable regional coastal model needs to have sufficient resolution to resolve at least some of the baroclinic instabilities of the flow. Vertical resolution needs to be sufficient to allow decoupling of the flows from topography under stratified conditions and development of appropriate shears when the flow is baroclinically unstable, and at least partially resolve boundary layers at the top and bottom of the water column, which meet in the shallow regions. Ideally the regional model should be informed at its boundaries with circulation and scalar fields appropriate to specific days and years. In the following sections we describe the global and regional models presently used to begin to achieve these intricate modeling objectives, and how global model output can be used to constrain the regional model simulations.
2.2. The Global Model
A large-scale context for our regional studies is provided by simulations with the Spectral Element Ocean Model (SEOM; Haidvogel and Beckmann, 1999). SEOM has been developed for the purpose of high-resolution basin-scale modeling on unstructured global grids (Iskandarani et al., 1994). The governing equations are the 3-D, Reynolds-averaged Navier-Stokes equations with Boussinesq and hydrostatic assumptions. Lateral subgridscale mixing of momentum is parameterized using the shear- and mesh-size-dependent formulation of Smagorinsky (1963),
which has proven to be highly effective on these horizontally heterogeneous grids. Vertical transfer of momentum is represented with weak (linear) interfacial drag and (nonlinear) sress laws. The resulting class of large-scale circulation models has several significant virtues over those using more traditional approaches, including complete geometric flexibility, regionally selective horizontal resolution, and the ability to avoid open boundary conditions by use of global grid refinement.
The spectral element circulation model has now been applied in its reduced gravity form to a variety of test problems and global oceanic/atmospheric applications. When applied to a now-standard suite of shallow water test problems on the sphere, the SEOM model is shown to be highly competitive with other numerical models, including those based on spherical harmonic methods (Taylor et al., 1996). Oceanic applications on global, non-uniform grids show that these favorable properties are maintained in the presence of continental geometry and highly unstructured elemental meshes (Haidvogel et al., 1996).
Here, for economy, SEOM has been implemented on a global grid in layered form with a total of five isopycnal layers. Following Hurlburt el al. (1996), outcropping of the layers is avoided by mass sharing between layers as a minimum "entrainment thickness", here taken to be 40 meters, is reached. Table 1 gives the relevant parameters (resting layer thicknesses and reduced gravities) used in the global simulation. A significant limitation of the non-outcropping layered model is that topographic variations must be contained within the lowermost layer. We have done so here by clipping topography (obtained from ETOP05) at 200 meters (minimum) and 5000 meters (maximum), and then multiplying the resulting topographic excursions above 5000 meters by 0.85. Some of the effects of this topographic shrinkage are noted below.
As a first test of behavior, the global layered model has been forced with a repeating cycle of NCEP winds, corresponding to the period of NCSAT wind availability (August 1996 through July 1997). A comparable simulation using NSCAT winds has also been prepared, as a basis for a future sensitivity study. No explicit thermodynamic forcing is included; therefore, the resulting simulations can at most represent the wind-driven component of the large-scale, low-frequency circulation. The SEOM model is highly scalable on parallel computing platforms (Curchitser et al., 1998). The simulations reported below have been obtained on a Beowulf-type cluster of Sun Ultra-5 workstations maintained at the Institute of Marine and Coastal Sciences, Rutgers University. On this system, a year's simulation requires approximately six cpu days when run on 12 processors.
2.3. The Regional Model
To capture regional circulation in the CGOA, we employed the S-Coordinate Rutgers University model (SCRUM) of Song and Haidvogel (1994). This free surface, primitive equation model uses curvilinear-orthogonal coordinates in the horizontal, while the stretched, bottom-following "s-coordinate" allows for flexible spacing of vertical grid points. The latter feature is especially useful in resolving boundary layers at the top of the water column (important for wind mixing) and near the bottom (important for tidal mixing). Initial experiments with a coast-following versus rectilinear coordinate system established the latter as the more economical choice for the highly curved CGOA coastline. Ultimately we implemented the model on a rectilinear telescoped grid, oriented at 38 degrees to true north. In SCRUM, land areas are "masked out" after the calculation of each timestep, but still entail some computational overhead. Our rotated grid is designed to efficiently cover coastal and basin areas of the GOA, while minimizing coverage of land areas to enhance computational efficiency.
A telescoped horizontal grid was employed to further reduce computational overhead associated with horizontal boundary conditions. The grid has 145 by 113 horizontal gridpoints with 17 telescoped gridpoints on the southern and western boundaries (Fig. 2). The finely resolved area of the model domain reaches from the northeast corner to Queen Charlotte Island in the south and Unimak Pass in the west. The telescoped region continues to the south end of Vancouver Island, and to Amukta Pass in the west. Grid resolution varies from 22 km in the finely resolved area, to 200 km near the western and southern walls. As described in a subsequent section, the telescoped regions here serve primarily to recirculate flows into and out of the area of interest.
To allow proper resolution of top and bottom boundary layers, we employed 20 vertical levels. We utilized the s-coordinate feature of SCRUM to achieve quasi-uniform spacing near the surface, with limits of 2.5 m in the shallowest areas and 3.6 m in the deepest areas. This quasi-uniform spacing will be of especial benefit for planned coupling with the biological models. Our CGOA implementation of SCRUM is forced by winds, coastal runoff and atmospheric heat flux. Details of the forcing, bathymetry, and boundary conditions are given in the following sections. SCRUM is written in highly vectorized code, and most of the simulations were obtained on vector architectures (CRAY J932) at the Arctic Region Supercomputing Center. However, recent simulations have been obtained on an equally fast, single-processor workstation at PMEL. On the CRAY platform a year's simulation requires approximately 7 CPU days; on the local workstation, 5 CPU days are required.
2.3.1. Bathymetry
Model bathymetry was interpolated from a specially developed 5-minute bathymetric map of the CGOA, based on ETOP05 and other sources. While ETOPO5 has the advantage of broad coverage, it is well known to be inaccurate in many coastal areas. More detailed and accurate bathymetric data used to improve ETOP05 was obtained from two different sources: 1) Nearshore data from the National Ocean Service (NOS) Hydrographic Data Base, error checked and gridded to 30 seconds by National Geophysical Data Center (NGDC) and distributed as the TerrainBase data collection. These data are focused on specific coastal areas such as Cook Inlet. 2) Offshore data from Smith and Sandwell (1997), who collected and verified coastline and marine ship track data from many sources, and distributed that data as part of their 2 minute measured and estimated digital topographic map. Though their estimated bathymetry (based partly on gravity anomalies) contains too much noise to be useful on the continental shelf, the measured bathymetry was easily extracted and used to improve ETOPO5 values offshore. A complete collection of descriptions of available bathymetry and topographic data sets has been compiled by Robert A. Kamphaus (http://newport.pmel.noaa.gov/~kamphaus/time/data.html).
Data from these two detailed sources were combined and interpolated to a 5-minute grid using Global Mapping Tools (GMT). To reduce computation effort, the interpolations were done for 10 deg. by 10 deg. areas that overlap by .25 deg. The interpolated grid points match those of ETOPO5 so that when they were combined, ETOPO5 seamlessly supplies data in areas where the detailed bathymetry data set is lacking.
The final data set is particularly accurate in areas of high data resolution, such as along the shelf break. It constitutes a major improvement over ETOPO5, especially in shallow shelf areas such as the Trinity Banks southwest of Kodiak Island. After interpolation to the SCRUM grid, bathymetry was cropped to 50 m minimum and 4000 m maximum, and filtered with six passes of a Shapiro filter, for numerical stability of the simulation. . The result is shown in Fig. 4. Even after filtering, the result is considerably more accurate than any bathymetry obtainable with ETOP05 alone.
2.3.2. Heat Flux and Wind Stress
Both wind and heat flux contribute substantially to the near-surface dynamics of the CGOA. Suitable values of wind and heat flux in specific years were obtained from the NCEP/NCAR Global Reanalysis Project. NCEP products include a global data set of atmospheric variables, obtained by combining a global spectral model with historical data. Their model has been run for the years 1958-present, and the output is available online. Resolution of these data is roughly 2 degrees. Temporal resolution is 6 hours, but we have chosen to use daily averages as input to our regional circulation model.
Daily average NCEP/NCAR values for latent and sensible heat net flux, and net longwave and shortwave radiation were summed to provide total heat flux from the ocean. Daily average U-wind and V-wind at 10m height above the ocean surface were converted to wind stress using the simple formula:
t = ra* Cd * U10 * |U10|
where t is the wind stress in N/m2 , U10 is the vector of wind speed in m/s, ra is the air density and Cd = .0012 .
In the present study, daily wind stress and heat flux for years1976, 1995, 1996 and 1997 have been utilized to force the regional model. These years were chosen to span a range of interannual variability in winds and freshwater input, and to overlap with the repeating cycle of NCEP winds used to force the global model. A comparison of computed low-pass filtered wind stress near Sitka, AK for the four years is presented in Fig. 5a. Note how the 1976 wind stress was especially strong and northwestward, relative to the other years.
2.3.3. Freshwater Input
Freshwater from distributed sources is a major source of buoyancy to the CGOA. Time series of "line-source" freshwater input are shown in Fig. 4 for 1995. These monthly values of integrated freshwater runoff along segments of the coastline were derived from snowpack, precipitation and temperature data by Royer (1982 and pers. comm.). The data represent runoff from areas seaward of the coastal mountain range. The "Southeast" region begins at the southern border of Alaska and extends northward to a location between Glacier Bay and Yakutat Bay. The "Southcoast" extends northward from there to include the southern side of the Kenai Peninsula. These regions are roughly equivalent to 130W-141W and 141W-152W longitude.
These line-source estimates were supplemented by river discharge data for significant inland areas draining into the CGOA. The river discharge data were obtained from USGS sources. The Susitna and Copper Rivers were among the few rivers that were gauged, and there were not many years of data for either. Therefore, a monthly climatology was computed with all the available data, and this is used for every model year. Peak mean values for river discharge occur in July. However, little discharge is provided by these large rivers (~10 percent of the total), relative to the line sources (~90 percent of the total). In Fig. 5b, we compare the total discharge from Southeast plus Southcoast regions plus river input for the four years noted in Fig. 5a. While the seasonal runoff pattern is similar between years, significant variability does exist, with especially wet months in September 1995 and November 1996, and an especially wet spring for 1997.
2.3.4. Mixing
Vertical mixing is parameterized as a function of local shear and stability, using Mellor-Yamada level 2.0 closure (Mellor and Yamada, 1974). Horizontal mixing of both scalars and momentum is calculated with a biharmonic operator, and scaled by the local grid spacing as described in Hermann and Stabeno (1996). Mixing is computed along geopotential surfaces, rather than along sigma surfaces (Haidvogel and Beckmann, 1999).
2.3.5. Horizontal Boundary conditions
The philosophy of our approach to regional modeling is similar to that of Hermann and Stabeno (1996), as follows. True open boundary conditions are difficult to achieve in eddy-resolving models, and especially so if tidal motions must be permitted simultaneously. Open boundaries are notoriously unstable to vigorous mesoscale signals. One sensible approach is to avoid open boundaries entirely through the use of a telescoped grid and a simple closed box (Fig. 2). Within this box, adjacent to the finely resolved area of interest, are placed bands where flow and scalar values are nudged towards desired values, but not so strongly as to prevent the escape of any out-going, internally generated mesoscale signals. The desired boundary values may be based on data or an externally run large scale model, or some combination of the two. The area between nudging bands and the solid wall is intended to function as a free recirculation zone, satisfying continuity and possibly absorbing any mesoscale features which escape from the interior. When both tidal and subtidal forcing is desired, a useful approach is to separately apply the tidal signal on sea surface elevation in a nudging band adjacent to the solid wall. If nudging constants for the subtidal nudging band are chosen properly, the tidal signal will pass cleanly through that band without interference. In the present work, we will show only results from the subtidally nudged model. Future work will add simultaneous tidal forcing, as has been accomplished for the Bering Sea (Hermann and Haidvogel, 2000)
2.3.6. Float Tracking
Ultimately our physical modeling system will feed information to a three-dimensional lower trophic level (NPZ) model, and an individual-based model of juvenile salmon. The NPZ model is designed to include preferred prey of juvenile salmon (e.g. euphausiids), and is currently running in one-dimensional (depth-time) form (Sarah Hinckley, personal communication). As an intermediate step on the path to fully coupled models, we are tracking representative fish (here, passive numerical floats) using surface currents generated by the regional physical model.
Hartt and Dell (1986) reported high catches of juvenile sockeye salmon in July centered at ~55 N 133 W, just northwest of the Queen Charlotte Islands and within 40 km of the coast. Guided by this observation, and in order to address how purely physical advection of salmon (and plankton generally) might compare with their observed histories, our floats are seeded in northwest-southeast lines at 0,100,200,300 and 400km from the coast between the Queen Charlotte Islands and Sitka. Subsequent tracks are strongly influenced by the Sitka Eddy, as described in the results.
3. Model experiments
We have set up several model experiments to test features of the coupling scheme and performance of the regional model. The experiments conducted are as follows:
1) Run of the global model. The global model is spun up with 5 years of a repeating cycle of daily NCEP winds, spanning the period of NCSAT wind availability (Aug 1996-July 1997). Resulting depth-integrated velocities were low-pass filtered and stored on a daily basis, for use by the regional model.
2) Free runs of regional model. We spin up the regional model from a state of rest, starting on January 15 of each simulated year. Temperature and salinity fields are initialized with Levitus climatology for January, and driven with winds, heat flux, and runoff as described in the Methods section. Runs were executed for calendar years 1976, 1995 and 1997.
3) Nested (boundary nudging) runs of the regional model. Here we utilize monthly climatologies (Levitus) for T and S, and daily barotropic velocities from the global model, as candidate fields to be applied to the nudging bands in the regional model. Wind, heat flux and coastal buoyancy forcing is applied for the period August-December 1996, when applicable SEOM results are available.
3a) In the first assimilation experiment, the regional model is nudged during July 27-Aug 1 with daily barotropic velocities (from SEOM), and monthly T,S fields (from Levitus climatology). A uniform nudging coefficient is applied in the finely resolved domain, and ramped to zero approaching the outer edges of the (telescoped) domain. This pattern of assimilation allows the interior to assume the SEOM velocity values, without forcing any flows through the solid walls. Subsequent to the five-day spinup, the model is nudged only in the telescoped domain, with nudging coefficient values ramped to zero at the edge of the finely resolved domain and at the outer walls of the telescoped area. The initial spinup everywhere with SEOM provides the essential broad patterns of the flow field, including some larger eddies. Subsequently the regional model is free to develop finer-scale circulation in its interior, under the influence of local wind and buoyancy forcing and with the SEOM velocities and Levitus T,S applied as a horizontal boundary condition.
3b) In the second assimilation experiment, the model is spun up as in case 3a). Subsequently, however, we remove all influence of the SEOM velocities, and assimilate only the climatological T,S as a horizontal boundary condition. By contrasting runs 3b) and 3a), we clarify the influence of global barotropic velocities on the interior regional solution.
4. Results
4.1. Global model results
Here we present barotropic velocities and free surface height fields from SEOM on a regular lat-long grid, for mid-July (Fig. 6) and early November (Fig. 7). Coastal and shelf-break currents are evident, as are meanders and eddies in the deep basin. A significant portion of the shelf-break current in the northwestern GOA (the Alaskan Stream) appears to come from the deep basin in these simulations, in addition to shelf-break currents upstream. In early November, a spatially continuous flow proceeds westward (counterclockwise) around the GOA. Results for mid-July exhibit a significant reversal of that near-coastal flow in the eastern GOA. Seasonal weakening/reversal of the coastal currents in the eastern GOA has been noted by Royer (1998) and is suggested by the analyses of seasonal altimetric patterns of Strub (2000). In both frames, the Alaskan Stream is wider and weaker than is typically observed, probably due to the limited spatial resolution and smoothed bathymetry of the model.
4.2. Regional model free run results.
4.2.1 Prominent features of the circulation and SST
Regional model results reproduce many of the major observed features in the CGOA (Figs. 8-10). In this section we examine 130-day runs initialized with Levitus climatological January T and S and driven with winds, heat flux and runoff appropriate to 1995 and 1997, respectively. Both runs exhibit a prominent AC-AS system and a weaker ACC; a weak ACC is appropriate for this time of year, as buoyancy fluxes are at their seasonal minimum. As in the global model results, the AS is weaker and somewhat wider than is typically observed; even finer than 22 km resolution would be required to improve this result. A general warming is observed in model results from early March through mid-May 1995 (Figs. 8 and 9), with some persistent cold areas near the coast. A tongue of warm water penetrates west along the shelf break in model results, as has been noted in SST images and in hydrographic data. A comparison of the simulations for 1995 versus 1997 (Fig. 10) reveals warmer temperatures in 1997. A similar degree of large eddy activity was observed in the May results for those two years.
Extensive mesoscale circulation features are observed in model output, on the continental shelf, at the shelf break, and in the deep basin. In the deep basin, many of these (modeled) features are locked to prominent seamounts. Animated results exhibit a clockwise propagation of disturbances around the seamounts. On the shelf, small eddies with 50-100km diameter are produced, but appear less dynamic than has been observed in drogued drifter studies (Stabeno and Hermann, 1996). During the first week of March, 1995, the regional model exhibits an intense, rapidly evolving, 200 km-scale eddy field along the shelf-break (Fig. 7). The eddy sizes and locations generally correspond to those which have been observed in AVHRR imagery for this period by Thompson and Gower (1998) (temperature signals are less evident than velocity signals, however, in the model result).
Eddy activity at the 200 km scale is most prominent in the vicinity of Sitka, AK. In the spring of each simulated year, its location is slightly north of the traditionally reported Sitka eddy (e.g. Tabata, 1982), moving slowly offshore (that is, southwest) from spring to fall (Fig. 11). This seasonal migration has been observed in the layered model studies of Melsom et al. (1999). The Sitka eddy has been reported in other studies as topographically generated (Swaters and Mysak, 1985), and has average surface currents of 15 cm/s, with a maximum of 110 cm/s as measured by drifters. Our regional model-generated surface drifter tracks for spring 1997 in this area (Fig. 12) compare favorably with observed drifter tracks reported in Tabata et al (1982), but do tend to exhibit weaker velocities than observed. Specifically, observed drifters transit the Sitka eddy in about 11 days, while the model's drifters require about 30 days to complete one circuit. Velocities shown in Fig. 11 also show a jet of offshore flow, feeding the eddy with coastal waters from the south.
4.2.2. Numerical Drifters
Numerical drifter tracks for July-Dec 1976 (Fig. 13 a-f) underscore the significance of the large eddy activity, and the Sitka eddy in particular, to salmon dynamics in the Gulf. Floats were initialized in a matrix parallel to the coast south of Sitka AK (indicated by asterisk in Fig. 13a), and subsequently tracked for July-Dec 1976. Half of the drifters released near the coast (Fig. 13b) become trapped in the Sitka eddy, with penetration further north only for tracks very close to the coastline. Drifters released 100km offshore (Fig. 13c) exhibit similar entrapment in the eddy feature. At 200km (Fig. 13d) offshore most of the drifters move around the western rim of the eddies; these either merge into the Alaskan Stream, traveling far to the west, or run aground between Sitka and Prince William Sound due to downwelling circulation (Fig. 5). Releases at 300 and 400km offshore (Fig. 13e,f) typically run aground somewhere between Yakutat Bay (east of Prince William Sound) and Cook Inlet (east of Kodiak Island).
4.3. Nested run results
Here we compare: the global model (SEOM) barotropic velocity results (experiment 1); a regional model run with boundary nudging of both global barotropic velocities (from SEOM output) and T,S climatology (experiment 3a); a regional model run with boundary nudging of T,S climatology only (experiment 3b). Results of each case are presented on the grid used by the regional model (in the case of the global model, results from the unstructured grid were interpolated onto this regular grid). Each of the regional model simulations was started on August 1, initialized with the results of the 5 day spinup period previously described. T,S climatology appropriate to that date. After 90 days, the velocities associated with the SEOM fields have penetrated far along the coastal waveguide and, to a lesser degree, into the interior of the basin. This is consistent with the fast propagation of coastally trapped signals.
First, consider the global model results for early November 1996. As noted in previous figures, the velocities in the global model are generally weaker than those in the regional model. The Alaskan Stream is less well developed than in the regional model, and the Sitka region exhibits far less mesoscale structure than in the regional model. These results are expected; the layered model employs a coarser grid, smoother bathymetry and a less realistic coastline than the regional model in the CGOA, and has no buoyancy forcing. As such, it cannot be expected to generate as much mesoscale activity, or as vigorous currents, as the regional model.
By contrast, the regional model for the corresponding day exhibits a highly developed near-coastal eddy field associated with the Alaska Coastal Current to the east of Kodiak Island. As noted in Fig. 5, buoyancy forcing is typically strongest in the fall, hence eddy generation via baroclinic instability (abetted by a temporary weakening of the downwelling-favorable winds) is a likely source of these smaller eddies. Larger meanders near the shelf break are also in evidence, but the smaller-scale variability dominates.
A closeup of the Sitka region exhibits an eddy feature just southwest of Sitka, as in other runs. Near-coastal velocities on this day (DOY 305) are southeastward, in contrast to the northwestward velocities of the global model (see Fig 14). This may be a reflection of the more accurate, shallower bathymetry of the regional model, with local daily winds exerting a stronger influence on the depth-integrated currents. A time average of velocities for the previous 10 days (not shown) in fact exhibits the more typical northwestward (that is, cyclonic around the GOA) current pattern. Such current reversals have in fact been observed in Shelikof Strait and in the eastern CGOA during periods of upwelling-favorable winds, but not in all of the areas (e.g. Gore Point, just upstream of Kodiak Island) indicated by the model (Stabeno et al., 1995a).
Boundary velocity information provided by the global model can be expected to influence the regional model most strongly near that boundary. Coastal-trapped waves, Rossby waves and simple advection can be expected to carry boundary information into the interior, as well. Indeed, one strong motivation for using global model results (or real data) at the boundary of the regional model is to provide just such information about remotely forced coastal-trapped waves to the regional model interior. Here we calculated the RMS difference in velocities between a run with boundary nudging of global model barotropic velocities and climatological T and S, versus a run with boundary nudging of climatological T and S only. The difference was normalized by the RMS velocity of the model run with boundary nudging of climatological T and S only (Figs. 17 and 18). This normalized difference is intended as an approximate metric for how much velocity signal is added to different regions of the interior through nudging at the boundary, relative to background velocity values. The largest normalized differences were observed along the shelf break, with much smaller values along the coast. In addition, a local maximum is observed in the vicinity of the Sitka eddy. The relatively weak influence on near-coastal flows may reflect the fact that only barotropic information is provided by the global model, whereas the true coastal current (that is, the ACC) has a strong baroclinic signature.
5. Discussion and Conclusions
Our initial experiments with various configurations of two circulation models for the Gulf of Alaska suggest that this area can be realistically modeled using a regional scale model, but
results are strongly affected by the inclusion of boundary information from a global model.
These results are an important step towards our goal of exploring the effect that interannual and decadal changes in GOA circulation have on lower trophic level dynamics and fishery populations.
Our regional model with ~20 km horizontal resolution and 20 vertical s-coordinate levels captures much of the spatial and seasonal patterns GOA circulation. Both the AC-AS and ACC current systems are easily identified in model output, though with somewhat reduced strength than is observed in the environment. Seasonally, a weakening/reversal of the coastal current in the eastern CGOA is observed. Interannual differences are also evident between the different years simulated; for instance, spring 1997 has generally higher SST that spring 1995. This is important because it implies that using this model, we will be able to investigate effects of the physical regime shift that has been documented in the GOA system.
Barotropic information from the global model can be effectively incorporated into this regional model as a boundary condition, through the use of suitably designed nudging bands located outside the region of interest. Spurious effects of rigid wall dynamics are avoided by allowing
free compensatory flow in the telescoped region outside the nudging band. As presently constructed, this approach yields greatest impact, relative to background velocities, in the AS-AC current system along the shelf break. Far weaker influence is felt in the mainly baroclinic, near-coastal ACC.
We believe these results can be further improved by including the influence of baroclinic information from the global model. Specifically, we plan to generate multiyear simulations with the global layered model, and project information about the first baroclinic mode from those results onto the first baroclinic mode of the regional model in the nudging band. Tidal forcing will be added as well, to better capture the spatial variability of vertical mixing in the GOA. Intensification of current speeds to more closely match observed current magnitudes and spatial focus of the AS-AC and ACC may require higher (for example, 10 km) horizontal resolution in the regional model. New regional model codes, which take advantage of parallel computer architectures, will hopefully allow these higher resolutions to be implemented for multiyear runs.
The model also captures the dominant 200 km scale variability observed in the Gulf. It displays especially strong mesoscale dynamics in the vicinity of the Sitka eddy. The locus of this activity migrates slowly offshore from spring through winter. The strength of this feature is significantly affected by the upstream assimilation of global model results.
Resolution of this eddy is particularly relevant to salmon migration. The shelf in this area is very narrow, so passing salmon would be particularly vulnerable here to offshore advection. Modeled surface particles released directly upstream (that is, southeast) of the Sitka eddy are typically
captured or deflected by it for periods exceeding one month. However, very close to shore, some particles move northwestward along the coast, avoiding its influence, and the influence of the offshore jet which is sometimes associated with the eddy in the model results.
Hartt and Dell (1986) present a summary of sockeye salmon migration patterns in the Northeast Pacific derived from a detailed examination of catch data from coastal and high seas cruises. This work stands as an important synthesis of the available data on salmon distribution and migration patterns during the early phase of the salmon marine life history. The coastal migration patterns of salmon are thought to involve movement along a narrow band over the continental shelf (Hartt and Dell 1986, Groot and Cooke 1987). A directed northwest swim bearing during this time has been inferred based on higher net catches in seine sets open in the direction of migration (Hartt and Dell 1986). These studies have concluded that fish tend to exhibit ground speeds of approximately 18.5 km/d (Hartt and Dell 1986). Sockeye salmon migrate from coastal waters adjacent to Sitka, AK to the Kodiak Island region over a six month period during July to December. The timing and position of offshore migration remains unclear, but is thought to occur in late winter and early spring.
Results from our simulations suggest passive drifters greatly underestimate the net travel rate of salmon along the coastal Gulf of Alaska. The most appropriate float track to compare to observations of salmon migration are the ones released nearshore (Fig. 13b). While salmon appear to make forward progress at a rate of approximately 18.5 km/d, our surface floats in the simulation moved at a significantly slower rate (~6 km/d). While the model itself may underestimate the true current speeds, these fish in nature appear to exhibit relatively strong compass bearing and directed swimming, and hence the discrepancy could be explained by the fishes' innate ability to direct their movements while migrating. The swim speed that would be necessary to account for this discrepancy would be on the order of 12 km/d (or 13 cm/s, approximately 1.3 body length/s), which is biologically plausible. Theoretical predictions of optimal swim speed (speed that minimizes energy cost of locomotion per unit distance traveled) for this size class of fish are in the range of 14-27 km/d (or 25-30 cm/s, approximately 2.5-3 body lengths/s, Webb 1995). This theoretical prediction represents a more instantaneous rate, and thus is not directly comparable to our measures of net travel rate that can include non-direction movements along the migration path.
While several float tracks released nearshore became entrained in the Sitka eddy, to our knowledge no one has identified whether this oceanographic feature can entrain juvenile salmon. Depending on the intensity of the currents, this eddy may have an influence on migration timing, growth and survival of juvenile salmon. Further modeling efforts are required to determine what impact this entrainment might have on migration patterns, and if salmon swimming behavior can influence the risk of becoming trapped in an eddy.
Acknowledgments
This is contribution 2174 from NOAA/ Pacific Marine Environmental Laboratory, contribution FOCI-xxxx to NOAA's Fisheries Oceanography Coordinated Investigations, and contribution xxx from the Joint Institute for the study of the Atmosphere and the Oceans under cooperative agreement NA90RAH00073. GLOBEC is sponsored by the National Science Foundation and the Coastal Ocean Program of NOAA. The second author gratefully acknowledges the hospitality and financial support of the Miller Institute for Basic Research in Science received during a sabbatical visit to the University of California, Berkeley. Support from the Arctic Region Supercomputing Center is also gratefully acknowledged.
Table 1. Mean interface depths (m) and reduced gravity at interfaces (m/s2) for the layered implementation of the Spectral Element Ocean Model.
Interface depth, m Gravity at interface m/s2
0. 9.81
References
Bhaskaran, S., G.S.E. Lagerloef, G.H. Born, W.J. Emery and R.R. Leben. 1993. Variability in the Gulf of Alaska from Geosat altimetry data. J. Geophys. Res. 98: 16311-16330.
Bograd, S.J., P.J. Stabeno and J.D. Schumacher. 1994. A census of mesoscale eddies in Shelikof Strait, Alaska, during 1989. J. Geophys. Res. 99: 18243-18254.
Brodeur, R. D., B. W. Frost, S. R. Hare, R. C. Francis, and W. J. Ingraham, Jr. 1996. Interannual variations in zooplankton biomass in the Gulf of Alaska and covariation with California Current zooplankton. CaICOFI Reports, 37: 80-99.
Brodeur, R. D. and D. M. Ware. 1992. Long-term variability in zooplankton biomass in the subarctic Pacific Ocean. Fish. Oceanogr. 1: 32-38.
Brodeur, R. D., and D. M. Ware. 1995. Interdecadal variability in distribution and catch rates of epipelagic nekton in the Northeast Pacific Ocean. pp. 329-356 in Beamish, R. J. (ed.), Climate Change and Northern Fish Populations, Canadian Special Publication of Fisheries and Aquatic Sciences, 121, Ottawa, Canada.
Chao, S-Y. and W.C. Boicourt. 1986. Onset of estuarine plumes. J. Phys. Oceanogr. 16: 21372149.
Chelton, D. B. and R. E. Davis. 1982. Monthly mean sea-level variability along the west coast of North America. J. Phys. Oceanogr. 12: 757-784.
Crawford, W. R. and F. A. Whitney. 1999. Mesoscale eddy aswirl with data in Gulf of Alaska. EOS Transactions of the American Geophysical Union 80:365-370.
Curchitser, E.N., M. Iskandarani and D.B. Haidvogel. A spectral element solution of the shallow-water equations on multiprocessor computers. J. Atm. Oceanic Tech. 15:510-521.
Cummins, P. F. 1989. A quasi-geostrophic circulation model of the Northeast Pacific. Part 2: Effects of topography and seasonal forcing. J. Phys. Oceanogr. 19: 1649-1668.
Cummins, P. F. and H. J. Freeland. 1993. Observations and modeling of wind-driven currents in the Northeast Pacific. J. Phys. Oceanogr. 23: 488-502.
Cummins, P. F. and L. A. Mysak. 1988. A quasi-geostrophic circulation model of the Northeast Pacific. Part I: A preliminary numerical experiment. J. Phys. Oceanogr. 18: 1179-1189.
Foreman, M. G. G., A. M. Baptista and R. A. Waiters. 1992. Tidal model studies of particle trajectories around a shallow coastal bank. Atrmos.-Ocean 30: 43-69
Foreman, M. G. G, R. F. Henry, R. A. Waiters and V. A. Ballantyne. 1993. A finite element model for tides and resonance along the north coast of British Columbia. J. Geophys. Res. 98: 2509-2531
Fu, L-L. and R.D. Smith. 1996. Global ocean circulation from satellite altimetry and highresolution computer simulation. Bull. Am. Met. Soc. 77: 2625-2636.
Haidvogel, D. B., J. L. Wilkin, and R. Young. 1991a. A semi-spectral primitive equation ocean circulation model using vertical sigma and orthogonal curvilinear horizontal coordinates. J. Comput. Phys. 94: 151-185.
Haidvogel, D.B., E. Curchitser, M. Iskandarani, R. Hughes and M. Taylor. 1996. Global modeling of the ocean and atmosphere using the spectral element method. Atmosphere-Ocean, vol. XXXV, No. 1, 505--531.
Haidvogel, D.B. and A. Beckmann, 1999: Numerical Ocean Circulation Modeling. Imperial College Press, 318 pp.
Hannah, C.G., P.H. Leblond, W. R. Crawford and W. R. Budgell. 1991. Wind-driven depthaveraged circulation in Queen Charlotte Sound and Hecate Strait. Atmos.-Ocean 29: 712-736
Heim, P. K. II, M. A. Johnson, J. J. O'Brien. 1992. The influence of the Alaskan Gyre on the coastal circulation in the Gulf of Alaska. J. Geophys. Res. 97: 17765-17775
Hermann, A. J. and D. B. Haidvogel. 2000. Simultaneous modeling of tidal and subtidal dynamics in the Bering Sea. Proceedings of the International conference on coastal ocean and semi-enclosed seas: circulation and ecology modeling and monitoring, Moscow, Russia (8-12 September 1998).
Hermann, A. J. and P. J. Stabeno. 1996. An eddy resolving model of circulation on the western Gulf of Alaska shelf. I. Model development and sensitivity analyses. J. Geophys. Res. 101: 11291149.
Hsieh, W.W. And W. G. Lee. 1989. A numerical model for the variability of the Northern Pacific ocean. In R.J. Beamish and G.A. MacFarlane (eds.), Effects of Ocean Variability on Recruitment and an Evaluation of Parameters Used in Stock Assessment Models, Can. Spec. Publ. Fish. Aquat. Sci. no. 108, pp.247-254.
Hurlburt, H. E., A. J. Wallcraft, Z. Sirkes and E. J. Metzger. 1992. Modeling the global and Pacific Oceans: on the path to eddy-resolving ocean prediction. Oceanography 5: 9-18.
Hurlburt, H.E., A.J. Wallcraft, W.J. Schmitz, Jr., P.J. Hogan, and E.J. Metzger. 1996. Dynamics of the Kuroshio/Oyashio current system using eddy-resolving models of the North Pacific Ocean. J. Geophys. Res. 101:941-976.
Huthnance, J.M. 1995. Circulation, exchange and water masses at the ocean margin: the role of physical processes at the shelf edge. Prog. Oceanogr. 35: 353-431.
Ingraham, W.J. and R.K. Miyahara. 1988. Ocean surface current simulations in the North Pacific Ocean and Bering Sea. U.S. Dep. Commer., NOAA Tech. Memo NMFS-F/NWC-130, 155pp.
Isaji, T. and M.L. Spaulding. 1986. A numerical model of the M2 and K1 tide in the Northwestern Gulf of Alaska. J. Phys. Oceanogr. 17: 698-704.
Iskandarani, M., D.B. Haidvogel and J.P. Boyd, 1994. A staggered spectral finite element model for the shallow water equations. International Journal of Numerical Methods in Fluids, 20, 393--414
Kelley, K.A., M.J. Caruso and J.A. Austin. 1993. Wind-forced variations in sea surface height in the Northeast Pacific Ocean. J. Phys. Oceanogr. 23: 2392-2411.
Kourafalou, V.H., L.-Y. Oey, J.D. Wang and T.N. Lee. 1996. The fate of river discharge on the continental shelf 1. Modeling the river plume and the inner shelf coastal current. J. Geophys. Res. 101: 3415-3434.
Lagerloef, G.S.E. 1995. Interdecadal variations in the Alaskan Gyre. J. Phys. Oceanogr. 25: 2242-2258.
Lee, W.G., G. Holloway, W. W. Hsieh. 1992. Importance of bottom topography in the seasonal cycle of the North Pacific subarctic gyre. Atmos.-Ocean 30: 140-145.
Liu, S-K. and J.J. Leendertse. 1986. A three dimensional model of the Gulf of Alaska. pp. 26062618 in Twentieth Coastal Engineering Conference, Proceedings of the International Conference, Vol. 3, November 9-14, 1986.
Liu, S.-K. and J. J. Leendertse. 1990. Modeling the Alaskan Continental Shelf Waters. Final Report, Outer Continental Shelf Environmental Assessment Program (OCSEAP), Minerals Management Service, Vol. 70, 626pp.
Matthews, P.E., M.A. Johnson and J.J. O'brien. 1992. Observation of mesoscale ocean features in the Northeast Pacific using Geosat radar altimetry data. J. Geophys. Res. 97:17829-17840.
Mellor, G.L. and T. Yamada. 1974. A hierarchy of turbulence closure models for planetary boundary layers. J. Atm. Sci. 31: 1791-1806.
Melsom, A., S.D. Meyers, H.E. Hurlburt, E.J. Metzger and J.J. O'Brien. 1999. ENSO Effects on Gulf of Alaska Eddies, Earth Interactions, 3. [Available online at http://EarthInteractions.org]
Miller, A.J., D.R. Cayan, T.P. Barnett, N.E. Graham and J.M. Oberhuber. 1994. Interdecadal variability of the Pacific Ocean: model response to observed heat flux and wind stress anomalies. Clim. Dyn. 9: 287-302.
Musgrave, D. L., T. J. Weingartner, and T. C. Royer. 1992. Circulation and hydrography in the northwestern Gulf of Alaska. Deep-Sea Res. 39: 1499-1519.
Mysak, L. A., R. D. Muench, and J. D. Schumacher. 1981. Baroclinic instability in a downstream varying channel Shelikof Strait, Alaska. J. Phys. Oceanogr. 11: 950-969.
Polovina, J. J., G.T. Mitchum and G.T.Evans. 1995. Decadal and basin-scale variation in mixed layer depth and the impact on biological production in the Central and North Pacific, 1960-1988. Deep-Sea Res. 42: 1701-1716.
Reed, R. K. 1984. Flow of the Alaskan Stream and its variations. Deep-Sea Res. 31: 369-386.
Reed, R. K. and S. Bograd. 1995. Transport in Shelikof Strait, Alaska: an update. Cont. Shelf Res. 15:213-218.
Reed, R. K. and J. D. Schumacher. 1986. Physical Oceanography, pp. 57-75 in Hood, D.W. and S. Zimmerman. (eds.), The Gulf of Alaska: Physical Environment and Biological Resources. Alaska Office, Ocean Assessments Division, NOAA, U.S. Dept. Commerce.
Reed, R. K. and J. D. Schumacher. 1989. Transport and physical properties in central Shelikof Strait, Alaska. Cont. Shelf Res. 9: 261-168.
Reed, R.K. and P.J. Stabeno. 1993. The recent return of the Alaskan Stream to Near Strait. J. Mar. Res. 51: 515-527.
Reed, R.K., J.D. Schumacher and C. Wright. 1981. On coastal flow in the Northeast Gulf of Alaska near Yakutat. Atmos.-Ocean 19: 47-53.
Royer, T. C. 1981. Baroclinic transport in the Gulf of Alaska, Part II. A fresh water driven Coastal Current., J. Mar. Res. 39: 251-266.
Royer, T.C. 1982. Coastal freshwater discharge in the northeast Pacific. J. Geophys. Res. 87: 2017-2021.
Royer, T.C. 1983. Observations of the Alaska Coastal Current. pp. 9-30 in H.G. Gade, A. Edwards and H. Svendsen (eds.), Coastal Oceanography, Plenum Press.
Royer, T.C. 1985. Coastal temperature and salinity anomalies in the northern Gulf of Alaska, 1970-1984. In Wooster, W.S. and Fluharty, D.L. (eds.) El Nino north: El Nino effects in the eastern subarctic Pacific Ocean. Wash. Sea Grant No. WSG-WO 85-3, Univ. Wash., Seattle, WA, pp. 268-281.
Royer, T. C. 1989. Upper ocean temperature variability in the Northeast Pacific Ocean: is it an indicator of global warming? J. Fish. Res. Bd. Can. 17: 221-233.
Royer, T. C. 1993. High-latitude oceanic variability associated with the 18.6 year nodal tide. J. Geophys. Res. 98: 4639-4644.
Royer, T. C., D. V. Hansen, and D. J. Pashinski. 1979. Coastal flow in the northern Gulf of Alaska as observed by dynamic topography and satellite-tracked drogued drift buoys. J. Phys. Oceanogr. 9: 785-801.
Royer, T.C. 1998. Coastal Processes in the Northern North Pacific. Pp. 395-414 in A.R. Robinson and K.H. Brink (eds.), The Sea, Volume 11, John Wiley and Sons, Inc.
Schumacher, J. D. and A. W. Kendall, Jr. 1995. An example of fisheries oceanography: walleye pollock in Alaskan waters. Rev. Geophys. Suppl.: 1153-1163.
Schumacher, J. D., P. J. Stabeno, and S. J. Bograd. 1993. Characteristics of an eddy over a continental shelf Shelikof Strait, Alaska. J. Geophys. Res. 98: 8395-8404.
Schumacher, J.D. and R.K. Reed. 1980. Coastal flow in the northwest Gulf of Alaska: the Kenai Current. J . Geophys. Res. 85: 6680-6688.
Schumacher, J. D., P. J. Stabeno, and A. T. Roach. 1990. Volume transport in the Alaska Coastal Current. Cont. Shelf Res. 9: 1071-1083.
Semtner, A. J., Jr. and R. M. Chervin. 1992. Ocean general circulation from a global eddy resolving model. J. Geophys. Res. 97: 5493-5550
Song, Y. and D. B. Haidvogel. 1994. A semi-implicit ocean circulation model using a generalized topography-following coordinate system. J. Comput. Res. 115: 228-244.
Smagorinsky, J., 1963: General circulation experiments with the primitive equations. I. The basic experiment. Mon. Wea. Rev. 91:99-164.
Smith, W. H. F., and D. T. Sandwell, Global sea floor topography from satellite altimetry and ship depth soundings. Science 277:1956-1962.
Stabeno, P. J., R. K. Reed and J. D. Schumacher. 1995a. The Alaskan Coastal Current, Continuity of Transport and Forcing. J. Geophys. Res. 100: 2477-2485.
Stabeno, P. J., A. J. Hermann, N. A. Bond and S. J. Bograd. 1995b. Modeling the possible impact of climate change on the survival of larval walleye pollock (Theragra chalcogramma) in the Gulf of Alaska. Can. Spec. Pub. Fish. Aquat. Sci. 121: 719-727.
Stabeno, P.J. and R.K. Reed. 1989. Recent observations of variability in the path and vertical structure of the Alaskan Stream. J. Phys. Oceahogr. 19: 1634-1642.
Stabeno, P. J. and A.J. Hermann. 1996. An eddy resolving model of circulation in the western Gulf of Alaska II. Comparison of results to oceanographic observations. J. Geophys. Res. 101: 1151-1161.
Strub, P. T. and C. James. 2000a. Altimeter-derived surface circulation in the large-scale NE Pacific gyres: Part 1. Annual variability. This volume.
Strub, P. T. and C. James. 2000b. Altimeter-derived surface circulation in the large-scale NE Pacific gyres: Part 2. 1997-1998 El Nino anomalies. This volume.
Swaters, G.E. and L.A. Mysak. 1985. Topographically-induced baroclinic eddies near a coastline, with application to the Northeast Pacific. J. Phys. Oceahogr. 15: 1470-1485.
Tabata, S. 1982. The anticyclonic, baroclinic eddy off Sitka, Alaska, in the Northeast Pacific Ocean. J. Phys. Oceanogr. 12: 1260-1282.
Taylor, M., J. Tribbia, and M. Iskandarani. 1997. The spectral element method for the shallow water equations on the sphere. Journal of Computational Physics 130: 92.
Thompson, R. E. and J. F. R. Gower. 1998 A basin-scale oceanic instability event in the Gulf of Alaska. J. Geophys. Res. 103:3033-3040.
Trenberth, K. E. 1990. Recent observed interdecadal climate changes in the Northern Hemisphere. Bull. Amer. Met. Soc. 71: 988-993.
Trenberth, K. E., and J. W. Hurrell. 1994. Decadal atmosphere-ocean variations in the Pacific. Clim. Dyn. 9: 303-319.
U.S. GLOBEC. 1996. U.S. GLOBEC Northeast Pacific Implementation Plan. U.S. GLOBEC Report No. 17. University of California, Berkeley, CA.
Vastano, A. C., L. S. Incze, and J. D. Schumacher. 1992. Observation and analysis of fishery processes larval pollock at Shelikof Strait, Alaska. Fish. Oceanogr. 1:20-31.
Walters, R. A. and M. G. G. Foreman. 1992. A 3D, finite element model for baroclinic circulation on the Vancouver Island continental shelf. J. Mar. Syst. 3: 507-5186.
Wilson, J.G. and J.E. Overland. 1986. Meteorology. pp. 31-54 in Hood, D.W. and S. Zimmerman. (eds.), The Gulf of Alaska: Physical Environment and Biological Resources. Alaska Office, Ocean Assessments Division, NOAA, U.S. Dept. Commerce.
Figure Captions
Fig. 1. Overview of circulation in the Gulf of Alaska.
Fig. 2. Layout of quadrilateral elements for our layered implementation of the Spectral Element Ocean Model (SEOM). Structure within each quadrilateral is represented with a polynomial basis set of order eight. The resulting average "grid spacing" is approximately 25 km around the periphery of the North Pacific Basin, and increases to about 100 km elsewhere.
Fig. 3. Layout of the telescoped rectilinear grid for our regional implementation of the S-Coordinate Rutgers University Model (SCRUM).
Fig. 4. Smoothed bathymetry used for the regional model simulations. In this and subsequent regional model results figures, the axes are aligned with those of the model (that is, rotated 38 degrees relative to true north), and units are model gridpoints in the two coordinate directions ("xi" and "eta"). Distance between successive gridpoints is approximately 22 km.
Fig. 5.a) Low-pass filtered wind stress near Sitka, AK (N/m2), computed from NCEP reanalyses. b) Interannual comparison of total monthly runoff from line-sources and river discharge, integrated along the length of the CGOA. "Southeast" values were lagged by one month relative to "Southcoast" values to obtain these totals, as in Royer (1982).
Fig 6. Global model SSH (shaded, m) and barotropic velocity (m/s) for DOY 155.
Fig 7. Global model SSH (shaded, m) and barotropic velocity (m/s) for DOY 304.
Fig. 8. Regional model SST (shaded, degrees C) and barotropic velocity (m/s) for DOY 64.5 1995.
Fig 9. Regional model SST (shaded, degrees C) and barotropic velocity (m/s) for DOY 144.5 1995.
Fig 10. Difference in regional model SST for 1997 vs. 1995 at DOY 144.5. Note the warmer temperatures in 1997, especially at the location of the Sitka eddy.
Fig. 11. Regional model results for 1976: SSH relative to areal mean (shaded, m) and surface velocities (m/s; for clarity only every other vector is plotted) in the vicinity of Sitka, AK for March (spring), June (summer), September (fall) and December (winter). Note the gradual migration of the "Sitka eddy" offshore from Spring through Winter 1976.
Fig. 12. Trajectories of surface drifters in the vicinity of Sitka, AK, based on regional model output. Tracks begin at large red asterisks on April 4, 1997. Small asterisks mark daily intervals and end on May 24, 1997. Surface velocity on April 4, 1997 is marked with blue arrows (m/s). Masked land area is shown in green.
Fig 13. a) Initial positions in float tracking experiments using regional model surface velocities. Floats are released to the south of Sitka AK (indicated by asterisk in figure), and subsequently tracked for July-Dec 1976, with positions marked at half-daily intervals. b-f) Float tracks subsequent to release nearshore (b), and at 100km (c), 200km (d), 300km (e), and 400km (f) offshore.
Fig. 14. Global model (SEOM) barotropic velocities (m/s) and speeds (shaded, m/s) for DOY 303 1996.
Fig 15. Regional model barotropic velocities (m/s) and speeds (m/s, shaded) for DOY 303.5 1996. Global model barotropic velocities and climatological T and S were used to spin up the regional model for DOY 210, and were subsequently assimilated as a horizontal boundary condition for DOY 210-304.
Fig. 16. Closeup of regional model velocities and speeds in the vicinity of Sitka AK.
Fig. 17. Normalized RMS difference in regional model barotropic velocities, averaged over DOY 214-244, for two different boundary nudging schemes. In the first case, boundary nudging includes both global model barotropic velocities and climatological T and S. In the second case only climatological T and S were included.
Fig. 18. Closeup of RMS difference between the two cases in the vicinity of Sitka AK.