The Alaska Coastal Current (ACC) is driven by multiple sources of coastal freshwater discharge and propagates alongshore over hundreds of kilometers. The ACC is also subject to downwelling-favorable winds associated with cyclonic wind systems. Spatially-uniform, downwelling winds trap the buoyant ACC waters nearshore. However, we demonstrate with numerical experiments that spatial wind variations, due either to a stationary or translating cyclone, can enhance the offshore transport of buoyant coastal waters in comparison to no-wind conditions. A stationary atmospheric cyclone induces a strong convergence in the coastal current at the upstream periphery of the cyclone. This convergence generates an offshore filament of buoyant water, which evolves into detached anticyclone. A transient atmospheric cyclone enhances the offshore spreading of freshwater by intensifying mesoscale variability of the coastal current. Thus, the spatial structure of the wind field represents a potentially important mechanism for cross-shelf freshwater transport.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Alaska Coastal Current (ACC) is driven by multiple sources of fresh, buoyant water associated with numerous rivers, high precipitation rates in the surrounding coastal mountains, and melting glaciers during the warm season. The ∼2000 km long ACC circumscribes the Gulf of Alaska (GOA) shelf with the coast on its right facing downstream [Williams et al., 2007]. This flow direction induces a downwelling transverse circulation on the shelf, which is augmented by the prevailing cyclonic (coastal downwelling-favorable) wind field over the Northeast Pacific. The resulting jet is narrow, partially baroclinic; its total transport can reach or even exceed 1 Sv, and the transport is well correlated with the winds. The strongest density anomaly is in fall (following peak discharge) while the strongest flow is in winter due to strongest wind forcing [Stabeno et al., 2004; Weingartner et al., 2005].
 Previous studies found that downwelling-favorable winds trap buoyant water near the coast, steepen the isopycnals, and enhance/arrest the downstream/offshore freshwater transport [e.g., Garvine, 1996; Hetland and Signell, 2005; Williams et al., 2007]. However, in these studies a spatially-uniform wind forcing was applied. Spatially uniform wind is justified for a buoyant plume originating from a single freshwater source, since the spatial scale of the plume is typically smaller than the wind patterns. However, the alongshore extent of the ACC exceeds the spatial scales of meso- and even synoptic-scale atmospheric systems, so spatial variations in the wind field cannot be ignored. In this study we present numerical experiments that mimic the ACC setting: a coastal current, driven by multiple sources of coastal freshwater discharge, and by either a stationary or transient cyclonic wind field. Such an atmospheric forcing can in fact promote the offshore transport of freshwater beyond its offshore extension in the absence of wind.
2. Model Experiments
 In numerical simulations of the idealized buoyancy driven coastal current we utilize the Regional Coastal Modeling System (ROMS) [e.g., Song and Haidvogel, 1994; Shchepetkin and McWilliams, 2005] in the same configuration as described by Rogers-Cotrone et al. . The model solves nonlinear momentum and mass balance equations on an f-plane under hydrostatic and Boussinesq approximations. The Coriolis parameter f is 10−4 s−1. The vertical turbulent viscosity and diffusivity are parameterized by the nonlocal KPP closure scheme [Large et al., 1994] with the background friction and mixing coefficients both set to 5 × 10−6 m2 s−1.
 The model domain is a meridional periodic channel 80 km wide and 500 km long (Figure 1) where x and y represent across-shelf and alongshore coordinates, respectively. The model cross-shelf depth profile resembles the GOA shelf: depth linearly increases from 5 to 200 m over a distance of 10 km (6.875 ≤ x ≤ 16.875 km) and remains constant farther offshore. The horizontal resolution is 1.25 km along the x-coordinate and 2.5 km along the y-coordinate. The vertical s-coordinate contains 25 grid cells with the finest resolution occurring near the surface and bottom boundaries. Five westernmost grid cells are masked as a strip of coastal land cut through by ten zonal inlets delivering buoyant water into the domain (Figure 1). The time steps are 6 seconds for a barotropic mode and 120 seconds for baroclinic modes.
 The modeled flow is a subject to the following boundary conditions: a quadratic friction with a drag coefficient of 3 × 10−3 is applied at the bottom, while eastern and western walls are slippery. The freshwater enters the domain through ten inlets (0 ≤ x ≤ 6.875) spaced 25 km apart in the middle of the coastal strip. The zero salinity discharge is introduced at the head of each inlet (x = 0) at a constant rate of 200 m3 s−1. Our model domain represents a fraction of the total length of the ACC. Hence, we use periodic boundary conditions in order to introduce a continuous inflow of buoyant water from the upstream. Further details of the model configuration and setup are given by Rogers-Cotrone et al. .
 At the surface, wind forcing is applied either as a spatially-uniform, downwelling-favorable (northerly) wind stress of 0.025 Pa, or as a mesoscale cyclone. The cyclone, previously used by Yankovsky , is prescribed analytically as follows: we assume a Gaussian-shaped, radially-symmetric atmospheric pressure perturbation with a minimum value of −5 hPa in the cyclone center (its coordinates are defined as xs and ys) The azimuthal wind is determined from the gradient wind approximation (Figure 1, insert) and converted to the wind stress following Large and Pond . The wind pattern is centered around xs, ys, and its movement is determined by the translation speed of the cyclone center.
 Initially, the model domain is filled with a quiescent water of constant density (salinity of 32). Both freshwater discharge and spatially-uniform wind stress are ramped to their prescribed values over the 1st day of model run and then are held constant through day 30 of the integration. The wind forcing is weak and is required to prevent an excessive offshore spreading of buoyant water at the early stage of the adjustment when the coastal buoyancy driven current has not “returned” to the upstream end of the channel (y = 500 km) and the continuous coastal flow has not yet formed. Freshwater discharges form individual buoyant plumes which merge into continuous coastal current propagating with the coast on its right. A plan view of the buoyant water distribution at the surface on day 30 is presented in Figure 1; hereafter, we identify the offshore edge of the buoyant layer as the position of the 31.5 isohaline.
 After 30 days, the wind subsides to zero over a 1-day period, and three different scenarios are utilized in the three model runs discussed below. In model run 1 only the freshwater discharge is retained. The coastal current relaxes offshore as the wind ceases and by day 35 its offshore edge is at x ≈ 20 km. Also, mesoscale features start developing along the front, especially where the buoyant water is discharged. The instabilities continue to grow through day 50 and the offshore spreading of buoyant water occurs predominantly through the mesoscale dynamics, reaching x ≈ 40 km by the end of the model run (day 50; Figure 1).
 In model run 2, the coastal buoyancy-driven current is forced by a stationary cyclonic wind field. The cyclone starts translating at 3 m s−1 westward on day 30 from outside the model domain (xs = 240 km, ys = 250 km) until its center reaches x = 40 km by day 30.77. Thereafter it remains at this location (marked with dot in Figure 1) through the end of the integration (day 50). The buoyant layer is shifted closer to the coast at 150 ≤ y ≤ 300 km under the influence of moderately strong downwelling-favorable wind. The offshore edge of the coastal current does not exhibit substantial change with time (as in model run 1), except between 350 ≤ y ≤ 450 km, along the upstream periphery of the cyclone (Figure 1). Here, buoyant water continually spreads offshore and reaches x ≈ 70 km by day 50, almost twice its offshore extension under the no-wind condition in model run 1.
 In model run 3 the cyclone is transient: it starts from the same location and at the same time as in model run 2 and travels westward at a constant translation speed of 3 m s−1 until it exits the model domain. The cyclone center crosses the western boundary on day 30.93 and the wind subsides completely by day 31.69. Thus, the coastal current is affected by the passing cyclone for about 1.5 days. However, the impact of this forcing event lasts longer. The mesoscale variability of the coastal current is stronger than in model run 1, as is especially evident on days 45–50 (two weeks after cyclone passage; Figure 1). The buoyant layer extends farther offshore than in model run 1, the mesoscale features are larger, and there are offshore parcels of buoyant water detached from the coastal current (characterized by closed isohalines).
 The net freshwater fluxes through alongshore transects at x = 20 km are shown in Figure 2a. In model runs 1 and 3, the transect is placed in the central part of model domain (100 ≤ y ≤ 400 km) which is affected by both the freshwater inflow and the atmospheric cyclone. In model run 2, the transect is shifted upstream (325 ≤ y ≤ 475 km), where the offshore filament is formed. The offshore freshwater flux continuously grows in model run 1 due to accumulation of freshwater in the coastal current and the resulting lateral expansion. However, its net value remains relatively low between day 38 and 43 ranging within 250–350 m3 s−1. The predominantly eddy-driven freshwater flux in model run 3 peaks during the same time with maximum values exceeding 1000 m3 s−1 (that is, thrice the value in model run 1 and more than a week after the cyclone's passage). But the strongest offshore freshwater flux develops in model run 2: it peaks on day 42 at ∼2800 m3 s−1 (order of magnitude higher than in model run 1 on the same day). Subsequently, the enhanced freshwater fluxes in model runs 2 and 3 subside to the background level occurring in model run 1.
 By day 45, the freshwater content in the whole domain increases offshore and decreases inshore in model runs with cyclonic wind forcing relative to model run 1 (Figure 2b). In the case of a transient atmospheric cyclone, the freshwater anomaly is positive from x = 20 ÷ 45 km peaking at x = 31 km and turning negative inshore of x = 20 km. In model run 2 the positive freshwater anomaly extends farther offshore albeit at a lower maximum value than in model run 3. A stationary atmospheric cyclone shifts the buoyant layer closer to the coast in the central part of domain so that the negative freshwater anomaly reaches x = 25 km but reverses to positive values at x = 13–15 km.
 Of the three model runs presented the farthest offshore spreading of buoyant water occurs when downwelling wind associated with a stationary cyclone forces the coastal current. Figure 3 elucidates this process. A stationary cyclonic wind stress generates an Ekman transport divergence with the low sea level anomaly formed in the center of the atmospheric cyclone. A geostrophic adjustment to this low pressure anomaly produces a cyclonic flow in the model domain that follows the wind pattern. The cyclonic current interacts with the coastal boundary and the adjacent coastal buoyancy-driven current. In the upstream segment of a cyclone, this interaction causes strong flow convergence, also added to by a radial Ekman drift with the upstream component at this location. The convergence can be identified as a high pressure/sea level anomaly near the coast, at y ∼ 300–350 km, and initiates an offshore flow that advects buoyant water seaward. The sea level increase at the upstream periphery of the cyclone continues through day 40 so that a high sea level anomaly extends 60 km offshore as an anticyclonic bulge. By day 45 the anticyclone detaches from the coastal current, which is evident in the discontinuity of buoyant water with salinity <31, as well the abrupt reduction in the offshore freshwater flux (Figure 2). The anticyclonic flow continues to develop in the detached bulge forming a well-pronounced eddy by day 50 (see the 31.5 isohaline in Figure 3). The high sea level anomaly associated with the anticylone's center continues to move offshore and slightly upstream, it is more than 20 km away from the coastal current by day 50. Meanwhile, a new bulge starts forming nearshore at the upstream periphery of an atmospheric cyclone.
 Alongshore transects in Figure 4 demonstrate that the offshore buoyant flow is deep (density anomaly reaches ∼100-m depth on day 40 at its near-maximum development), strong (up to 30 cm s−1) and develops in a geostrophic manner (that is, centered at the upstream flank of a density anomaly). In contrast, the buoyant layer is shallower elsewhere with the halocline between ∼10–40 m depth. The offshore flow subsides and the pycnocline relaxes by day 45, after the eddy detachment.
3. Discussion and Conclusion
 It is commonly accepted that downwelling-favorable winds along a shelf enhance downstream transport of buoyant water and reduce its offshore transport. While this perception is certainly true for spatially-uniform, moderate-to-strong winds, the situation can dramatically change when the wind field has spatial variations. Recently, Rogers-Cotrone et al.  found that light and spatially-variable downwelling-favorable winds can generate stronger offshore transport of buoyant water than under no-wind conditions in a coastal current driven by multiple sources of freshwater (such as the Alaska Coastal Current). This transport is eddy-driven and is produced through the enhanced mesoscale variability of the coastal current.
 The present communication further explores this possibility. We showed that the mesoscale cyclonic wind system, either stationary or translating, can induce offshore spreading of buoyant coastal water seaward of the “natural” offshore coastal current boundary in the absence of wind forcing, and at much higher rate.
 The cyclonic wind field applied in these model experiments has a diameter of ∼300 km and clearly represents a mesoscale atmospheric system. Such systems are very energetic over the Northeastern Pacific and GOA in particular, especially during the cold season. The examples include the formation of polar lows [e.g., Bond and Shapiro, 1991; Douglas et al., 1991], which are mesoscale cyclonic vortices of ∼300 km horizontal scale embedded in low pressure synoptic scale systems. The evolution of polar lows is often transient, but they can persist for several days near the same location making a looping motion with little net translation [Douglas et al., 1991]. They are characterized by strong winds in excess of 20 m s−1 near the center. Other examples of mesoscale atmospheric dynamics in the GOA are reported by Businger and Walter  and Ralph . While this study focuses on the mesoscale atmospheric forcing, it is quite possible that larger, synoptic-scale atmospheric cyclones can generate similar offshore flow patterns on the GOA shelf. Indeed, the coastal mountains surrounding the GOA often cause lows to stagnate as they approach the coast mimicking the situation in model run 2.
 The ACC exhibits strong mesoscale variability in the form of eddies, meanders and filaments [e.g., Royer et al., 1979; Bograd et al., 1994]. On the GOA shelf the anticyclones are far more frequent than the cyclones [Bograd et al., 1994; Henson and Thomas, 2008]. They are likely to have different generation mechanisms, including instability and topographic forcing, but the offshore advection of light, buoyant coastal water appears to play an important role in their formation [Henson and Thomas, 2008]. Here we demonstrate that the cyclonic wind field can directly generate an anticyclone detached from the coastal current. The presence of coastal and topographic irregularities could further amplify this process.
 The GOA shelf supports a highly productive ecosystem in spite of the prevailing downwelling wind stress and a large coastal freshwater discharge that is low in nitrate [Childers et al., 2005] but enriched in iron [Wu et al., 2009]. Within the euphotic zone, these authors find that nitrate (iron) concentrations increase (decrease) seaward of the ACC. Strom et al.  suggest that the mid- and outer portions of the shelf may be iron-limited at least in spring and summer. Hence the offshore transport of freshwater on this shelf may regulate primary production. While a multitude of processes can affect the cross-shelf spread of ACC waters, our results, and those of Rogers-Cotrone et al.  suggest that spatial variations in the shelf wind field may be important.
 A.Y. was supported by NSF grant OCE-0752059. G.M. was supported by NSF grant OCE-0752059 (REU supplement) and by a grant from the University of South Carolina Magellan Scholar Program. We are indebted to anonymous reviewers for their insightful comments.