Altimeter-derived surface circulation in the large scale NE Pacific Gyres. Part 1. seasonal variability

Transcription

1 Progress in Oceanography 53 (2002) Altimeter-derived surface circulation in the large scale NE Pacific Gyres. Part 1. seasonal variability P Ted Strub, Corinne James College of Oceanic and Atmospheric Sciences, Oregon State University, 104 Ocean Administration Building, Corvallis, OR , USA Abstract The seasonal variability of sea surface height (SSH) and currents are defined by analysis of altimeter data in the NE Pacific Ocean over the region from Central America to the Alaska Gyre. The results help to clarify questions about the timing of seasonal maxima in the boundary currents. As explained below, the long-term temporal mean of the SSH values must be removed at each spatial point to remove the temporally invariant (and large) signal caused by the marine geoid. We refer to the resulting SSH values, which contain all of the temporal variations, as the residual SSH. Our main findings are: 1. The maximum surface velocities around the boundaries of the cyclonic Alaska Gyre (the Alaska Current and the Alaska Stream) occur in winter, at the same time that the equatorward California Current is weakest or reversed (forming the poleward Davidson Current); the maximum surface velocities in the California Current occur in summer. These seasonal maxima are coincident with the large-scale atmospheric wind forcing over each region. 2. Most of the seasonal variability occurs as strong residuals in alongshore surface currents around the boundaries of the NE Pacific basin, directly connecting the boundaries of the subpolar gyre, the subtropical gyre and the Equatorial Current System. 3. Seasonal variability in the surface velocities of the eastward North Pacific Current (West Wind Drift) is weak in comparison to seasonal changes in the surface currents along the boundaries. 4. There is an initial appearance next to the coast and offshore migration of seasonal highs and lows in SSH, alongshore velocity and eddy kinetic energy (EKE) in the Alaska Gyre, similar to the previously-described seasonal offshore migration in the California Current. 5. The seasonal development of high SSH and poleward current residuals next to the coast appear first off Central America and mainland Mexico in May June, prior to their appearance in the southern part of the California Current in July August and their eventual spread around the entire basin in November December. Similarly, low SSH and equatorward transport residuals appear first off Central America and Mexico in January February before spreading farther north in spring and summer. 6. The maximum values of EKE occur when each of the boundary currents are maximum Elsevier Science Ltd. All rights reserved. Corresponding author. Fax: addresses: tstrub@oce.orst.edu (P.T. Strub); corinne@oce.orst.edu (C. James) /02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S (02)

2 164 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Contents 1. Introduction and background Data and methods Altimeter and tide gauge data Atmospheric forcing sea level pressure Statistical gridding Results Summary and discussion Alaska Gyre Connections around the boundaries of the subarctic and subtropical gyres Connections to the North Pacific Current Offshore propagation of the seasonal height and transport signals Connections to the equatorial current systems along the boundaries Introduction and background This is the first of a two-part analysis of temporal variability of the NE Pacific Ocean s surface circulation, as measured by satellite altimeters. Here we examine the seasonal variability. In Part 2 (Strub & James, 2002) we analyze the non-seasonal anomalies of the surface circulation over the period, during which the El Niño creates the largest signal. Formation of the seasonal cycles discussed here is the first step in creating the non-seasonal anomalies. The seasonal cycles themselves, however, provide new information on the response of the NE Pacific to strong seasonal forcing, on scales not previously addressed. This analysis quantifies the degree of connection, on seasonal time scales, between the boundary currents in the eastern subarctic and subtropical gyres, as well as the connection between the boundaries and the interior NE Pacific. It further shows a connection to the equatorial current system. Numerous papers describe aspects of the seasonal cycles for certain parameters in subregions of our larger domain. Chapters in Robinson and Brink (1998) review some of the past results from the coastal ocean in the regions between the Equator and the Alaska Gyre (Badan-Dangon, 1998; Hickey, 1998; Royer, 1998). Fig. 1 presents the climatological surface dynamic height field (relative to 500 m) in the NE Pacific, calculated from the long-term mean climatological temperature and salinity data of Levitus and Gelfeld (1992). The 500 m reference level is used to concentrate on the surface flow seen by altimeters. Although this climatology is overly smooth, it shows the major currents in the area. The broad, eastward North Pacific Current (also called the West Wind Drift) splits into the counterclockwise Alaska Gyre and the equatorward California Current. South of 20 N in summer, the California Current turns westward and flows into the North Equatorial Current, while in winter spring, part of it continues along the Mexican mainland before turning westward (Badan-Dangon, 1998; Fiedler, 1992, 2002). The long-term climatology shows both paths. The North Equatorial Countercurrent (NECC) flows eastward between 5 10 N to approximately 120 W, but is only weakly seen in the annual climatology from there to the cyclonic flow around the Costa Rica Dome near 8 N, 92 W. The NECC is a shallow current (found in the upper 200 m) and might appear more strongly if a shallower reference were used, but it is also seasonally intermittent. When the Intertropical Convergence Zone (ITCZ) is in its northern location near 10 N (summer), surface divergences and upwelling create a zonal trough in surface height, driving the NECC along the southern side of the trough. When the ITCZ moves south in winter, the NECC weakens or reverses.

3 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 1. Climatological dynamic height relative to 500 m in the NE Pacific. Based on the temperature and salinity climatology of Levitus and Gelfeld (1992). The most well-studied subregion of our larger domain is the California Current System, located between approximately N within 1000 km of the coast of North America. In addition to reviews by Hickey (1979, 1989, 1998), papers by Kelly, Caruso, and Austin (1993), Kelly et al. (1998), Brink et al. (2000) and Strub and James (2000) describe aspects of the seasonal circulation and eddy statistics in this region, using altimeter and surface drifter data, and provide reviews of previous work. Kelly et al. (1993), in particular, addresses both the Alaska Gyre and the California Current and comes the closest to covering the same range of territory and topics as in the present study. The seasonal circulation of the California Current System has been best defined off central and southern California, using the long CalCOFI data set (Chelton, 1984; Lynn & Simpson, 1987). Some of the early CalCOFI cruises also cover northern California and are shown by Hickey (1979), but were not repeated often enough to statistically define the seasonal cycle there. Hickey (1979) also presents fields derived from ship drift records. Other studies have inferred seasonal changes in the circulation over the shelf or farther offshore from shorter records of measured currents (Strub, Allen, Huyer, Smith, & Beardsley, 1987; Wickham, Bird, & Mooers, 1987) or from specific cruises covering one or more years (Kosro et al., 1991; Rienecker & Mooers, 1989). From these and other works, we expect an equatorward current that extends from Vancouver Island to the Southern California Bight in summer (Strub & James, 1995; Strub, Kosro, Huyer, & CTZ Collaborators, 1991), partially entering the cyclonic flow in the Bight and partially continuing along Baja California. An inshore poleward countercurrent develops next to the coast in mid-summer (Lynn & Simpson, 1987; Strub & James, 2000) and eventually extends to the Canadian border in winter, becoming the poleward Davidson Current. In spring, upwelling and equatorward flow begin in the south, next to the coast, and this also expands offshore and to the north to create the summertime California Current jet, completing the cycle (Strub & James, 2000). Thus, both equatorward and poleward phases of the seasonal currents begin next to the coast off southern California and migrate offshore and to the north, at least as far as Vancouver Island.

4 166 P.T. Strub, C. James / Progress in Oceanography 53 (2002) The seasonal wind forcing over the California Current has been well-defined by Bakun and Nelson (1991) at scales of km and more, using merchant ship data. They show the generally upwellingfavorable (equatorward) winds south of approximately 37 N, maximum in summer. North of 37 N, there are seasonally alternating poleward winds in winter and equatorward winds in summer associated with the expansion of the Aleutian Low (winter) and North Pacific High (summer) pressure systems. Other papers present similar seasonal cycles derived from measured and proxy coastal winds (Halliwell & Allen, 1987; Strub et al., 1987), model winds (Strub, James, Thomas, & Abbott, 1990), or scatterometer winds (Thomas, Carr, & Strub, 2001). There are far fewer historical studies of the seasonal circulation in the Alaska Gyre (Royer, 1998). Much of the past and recent work has focused on the Alaska Coastal Current (ACC), a narrow surface current found within km of the coast in the north and northwest part of the Gyre. It is driven by both fresh water inputs along the boundaries and cyclonic wind systems over the basin, which help retain the fresh water at the coast. The maximum transport is thought to occur in autumn, at the time of maximum fresh water input, rather than at the time of maximum wind forcing in winter (Royer, 1981a; Stabeno, Reed, & Schumacher, 1995). This would agree with the October November maximum in coastal tide gauge sea levels, which was first observed by Pattullo, Munk, Revelle, and Strong (1955). Velocities are maximum at the surface next to the coast at speeds of m s 1 or even greater (Johnson, Royer, & Luick, 1988; Stabeno et al., 1995). Mean transports are of order Sv and are correlated with the wind (Stabeno et al., 1995). The Alaska Current (AC) flows in the region farther offshore and is what we expect the altimeter to see. In the western half of the basin, it is called the Alaska Stream. Royer (1981b) uses a series of hydrographic sections along the western boundary of the Gyre to calculate the total transport (relative to 1500 m) in the AC as 9.2 Sv, with a seasonal variation of ±1.2 Sv. He finds the maximum transport to occur in late-winter or spring (March or even later), lagging the wind forcing. Bograd, Thomson, Rabinovich, and LeBlond (1999) report a somewhat different result, based on analysis of 5 years of WOCE drifter tracks in the Alaska Gyre. Along the shelfbreak in the northern gyre between W, there are twice as many high speed observations (velocities over 0.4 m s 1 ) in winter as in summer, with most of those coming from the very surface (drifters that had lost their drogue). They define winter as October March, which leaves a fair degree of uncertainty as to the time of the actual maximum. Bograd et al. (1999) also report a maximum in mean kinetic energy in winter (40 cm 2 s 2 ) in the North Pacific Current at 50 N, 150 W, compared to values less than 20 cm 2 s 2 in summer (a decrease of 40 50%). In the eastern Gyre, the measurements are even more sparse and seem more often to indicate a vigorous eddy field rather than a continuous flow (Crawford, Cherniawsky, & Foreman, 2000; Crawford & Whitney, 1999; Meyers & Basu, 1999; Royer, 1998; Tabata, 1982; Thomson & Gower, 1998). Royer (1998) characterizes the flow next to the coast off Vancouver Island and farther north as entirely poleward, even in summer, based on measurements of a bouyancy driven current in the 20 km next to the Vancouver coast in summer (Hickey, Thomson, Yih, & LeBlond, 1991). Hickey et al. (1991), however, show equatorward currents of up to 15 cm s 1 offshore of this current. Thus, the degree of connection between the eastern Alaska Gyre and the California Current in summer is not well-defined by earlier studies. Previous altimeter analyses of the Alaska Gyre have used Geosat data. Combining data from the Geodetic and Exact Repeat Mission (ERM) periods to form a nearly 4-year time series, Bhaskaran, Lagerloef, Born, Emery, and Leben (1993) identify an annual mode through the use of Empirical Orthogonal Function (EOF) analysis. This mode consists of a cell of low SSH values north of 55 N, maximum in December January. Other cells are found south of 55 N. This mode is correlated with the first EOF of atmospheric sea-level pressure (SLP), with no lag. Bhaskaran et al. (1993) suggest that this could either represent Ekman pumping of the cyclonic Gyre or errors in the atmospheric corrections to the altimeter data, which use the SLP. Kelly et al. (1993) use the 2.5 years of ERM data to form seasonal cycles of SSH by fitting annual and semiannual harmonics. The patterns presented (their Fig. 9) differ from Bhaskaran et al. (1993) and

5 P.T. Strub, C. James / Progress in Oceanography 53 (2002) indicate SSH values that decrease toward the north of the Gyre in January, suggesting a decrease in the Gyre s strength in winter. The SSH residuals become positive in summer and autumn, implying an increase in the Gyre s strength. Kelly et al. (1993) interpret this to represent an autumn maximum in the Gyre s strength, as opposed to Bhaskaran et al. s (1993) winter and Royer s (1981b) spring maximum. One goal of this study is to clarify this timing, using the longer record of the more accurate TOPEX altimeter. In the southern part of our domain off Central America, Badan-Dangon (1998) summarizes the few direct measurements and Fiedler (1992, 2002) provides useful gridded maps of physical variables between 30 N 20 S. These include seasonal maps of ship-drift surface velocities, winds and thermocline depth (the depth of the 20 isotherm). Seasonal residuals (the deviation from the long-term mean) of ship-drift surface velocities and thermocline depth provide the closest comparison to the altimeter residual heights. These show a decrease and reversal of the NECC in winter-spring, after the ITCZ moves south, and an increase in NECC strength in summer autumn, after the ITCZ moves north, in agreement with other observations (Badan-Dangon, 1998; Tomczak & Godfrey, 1994). Current residuals next to the coast between 10 N 30 N are equatorward in winter and early spring (the Costa Rica Current is weak or reversed); poleward velocities are maximum in summer and autumn. Putting these various descriptions together, we expect the counterclockwise boundary currents in the Alaska Gyre to be strongest some time between autumn and spring (depending on the reference). The altimeter fields will not usually see the Alaska Coastal Current but should resolve the broader Alaska Current in the east and the Alaska Stream in the west. Drifter observations suggest a moderate winter maximum in the eastward North Pacific Current, part of which serves as the southern boundary of the Alaska Gyre. In the California Current, we expect the development and offshore/northward movement of a spring summer equatorward jet and winter poleward flow north of the Southern California Bight. Off Central America, we expect the eastward North Equatorial Countercurrent and the poleward Costa Rica Current to be strongest in summer and autumn. Thus, the goals of the examination of seasonal variability in the surface circulation in the NE Pacific are: (1) To determine the timing and structure of the seasonal maximum in transport in the Alaska Gyre; (2) To determine the extent and timing of direct transport between the subtropical and subarctic gyres along the boundary currents; (3) To determine the nature of the seasonal variability in the strength and position of the North Pacific Current and its contribution to the seasonal cycles of the boundary currents in the subarctic and subtropical gyres; (4) To look for connections between the equatorial currents and the boundary currents farther north. 2. Data and methods 2.1. Altimeter and tide gauge data The majority of altimeter studies use gridded fields of SSH, inferring the geostrophic circulation as occurring along constant heights. Here we also use geostrophic transports calculated between altimeter crossover points. The transports use the heights as recorded on the altimeter tracks directly, avoiding the effects of statistical gridding, although some alongtrack smoothing is applied. The height differences also reduce the effects of most environmental errors and seasonal steric changes in height, which have scales larger than the km differences in locations of the grid points. Over the open ocean, the SSH values come only from the TOPEX and POSEIDON altimeters on the TOPEX/POSEIDON (T/P) satellite. The TOPEX altimeter provides most of the data and has the highest signal to noise ratio of all altimeters. Next to the coast where high quality tide gauge data are available, altimeter data from crossovers km offshore are combined with SSH values from the coastal tide gauges (Fig. 2). This allows a calculation of the boundary transports with no gap between the altimeter data and the coast. This is done

6 168 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 2. T/P Altimeter tracks and sea level locations (filled circles). The crossover points used in the transport calculations are indicated by filled squares. because the altimeter data become noisier on the parts of the altimeter tracks within km of the coast. Where no tide gauge data are available, the altimeter must be used for the most inshore height measurement, leaving a gap of order km next to the coast. Fig. 2 shows the T/P ground tracks, along which altimeter height data are available approximately every 6 km, repeating every 10 days. This provides approximately 3 height samples per month along each track, with 6 height samples each month at the crossover points. The tracks are separated by 316 km at the equator, which goes to zero at the inclination latitude of 66. The T/P data used to calculate transports are processed at JPL and made available by V. Zlotnicki. Standard environmental corrections are applied (inverted barometer, wet and dry troposphere, ionosphere and EM bias). Ocean tides are removed using the UT 3.0 tide model; poletides and solid earth tides are also removed. The data are regridded to a regular alongtrack grid with 6 km spacing. We form and remove a five-year ( ) long-term mean from each data point to remove the marine geoid and create a residual sea surface height (SSH) data set used in calculating transports. The crossover points used to calculate transport are indicated by filled squares in Fig. 2. To avoid missing features in the surface circulation due to the choice of the specific grid used here to calculate the transports, a different altimeter data set is used to calculate gridded 2-D fields of seasonal SSH for 2-month calendar periods. We use alongtrack SSH data from four satellite missions included in the NASA/NOAA Pathfinder data set: the T/P mission during the period October 1992 December 1997: the GEOSAT mission from November 1986 September 1989 (although data dropouts become numerous after autumn 1988), the ERS-1C mission from April 1992 December 1993, and the ERS-2 mission from April 1995 June The Geosat and ERS data provide higher spatial resolution and a longer data set to define the annual cycle of the 2-D SSH gradient fields. Their processing is similar to the processing of the primary T/P data set from JPL. See Strub and James (2000) for plots of the tracks for all three altimeters in the California Current region.

7 P.T. Strub, C. James / Progress in Oceanography 53 (2002) In order to combine the heights from different satellites, it is necessary to first remove any offsets in mean height between the orbits of the satellites. The approach we have found most successful is to remove the spatial mean from each complete data set for each specific period covered (here the two-month calendar periods for all years of data) before gridding. Besides eliminating height biases between the data sets, this procedure also removes the spatial average of the large-scale mean seasonal cycle in height, which is dominated by hemispheric steric height changes associated with seasonal heating (Leuliette & Wahr, 1999; Stammer, 1997). What remains are the two-month fields of height gradients, displayed as contours of height in a field with zero spatial mean. Before removing the mean, we also eliminate outliers using a 2-iteration scheme that examines the time series at each point and removes points that deviate by more than 4 and then 5 standard deviations from the mean (this primarily removes noisy points next to the coasts). Hourly tide gauge data were obtained from NOAA/NOS and the University of Hawaii. Tides are removed using a 46-hour half power filter and the hourly data are averaged to form a daily value. An inverse barometer correction is applied using sea level pressure data from the U.S. Navy atmospheric model (provided by the NOAA Pacific Fisheries Environmental Group). A 20-day filter is then applied to the corrected daily average sea level to reduce the effect of coastally trapped waves, which have a strong signal at the coast (in the tide gauge data) and much smaller signals km offshore in the altimeter signals. Temporal averages from the same 5-year period as used with the altimeter data are removed from the tide gauge time series. Finally, this daily time series is interpolated in time to the midpoint of each T/P cycle. The tide gauge data are from five locations along the California Current between N and at three locations around the Alaska Gyre (Fig. 2). Monthly geostrophic surface velocities over the large-scale NE Pacific are calculated from the height differences between selected crossovers and tide gauge locations. The velocities are multiplied by the distance between points to form surface transports. The reader can multiply by an assumed depth of the current to produce volume transports. Since the distance between crossovers decreases from approximately 250 km at the lower latitudes to 100 km at the highest latitudes, the scale arrow in some figures showing 5000 m 2 s 1 represents velocities of approximately 2 cm s 1 at low latitudes and 5 cm s 1 at high latitudes. The selection of the location of the points used here is the result of tests of different grids, with the goal of defining the boundary currents in the two gyres and the zonal currents in the interior North Pacific Current, in as simple a fashion as possible. The final choice of points defines the alongshore boundary currents around the basin, using 3 4 points oriented roughly perpendicular to the coast. The onshore offshore currents into this band around the basin margin are also included, to look for changes in the inflow to the boundary currents. Finally, four north south lines are included to monitor the zonal interior flow between W north of 46 N, a pair near 141 W and a pair near 146 W (Fig. 2). These lines allow us to see any connections between the zonal interior flow and the inflow to the margins. The geometry of adjacent pairs of north south altimeter lines produces east west transports that are staggered in space, increasing meridional resolution Atmospheric forcing sea level pressure Over the large-scale NE Pacific basin, surface momentum flux is due to winds associated with the atmospheric sea level pressure (SLP) fields. We use the pressure fields from the NCEP/NCAR Reanalysis atmospheric models to represent the basin-scale forcing, over the five complete years of JPL altimeter data ( ). Daily fields are used to form the mean fields presented here Statistical gridding Five-year means ( ) of SSH are formed at each along-track grid point and subtracted from the time series at each grid point. This removes the marine geoid (which is the largest SSH signal and must

8 170 P.T. Strub, C. James / Progress in Oceanography 53 (2002) be removed) and also the temporal mean dynamic SSH and circulation field. The resulting residual SSH data sets contain all of the temporal variability, including the seasonal cycles. These are used to construct the time series of 2-D horizontal fields of residual SSH with 10-day spacing that are used in the EOF analysis (centered on each time step using 35-days of data, in order to include all of the ERS tracks). The gridding method is that of successive corrections (Bratseth, 1986; Vazquez, Zlotnicki, & Fu, 1990). In this method, heights are estimated at each point on a regular grid (0.5 is used here). A weighted mean of the data surrounding each grid point forms a first guess of the height at the grid point, which is modified during three iterations, reducing the spatial scale of the quadratic weighting function from 1.25 (first guess) to 1.0, 0.75, and 0.5 (Vazquez et al., 1990). The resulting field is smoothed with a final Laplacian filter, which reduces the amplitude of features with scales less than approximately km. See the Appendix of Strub, Chereskin, Niiler, James, and Levine (1997) for a discussion of the statistical aspects of this gridding. The same data set is used to form the seasonal means of residual SSH, also using successive correction, but combining all data from a given season from all altimeters. 3. Results Fig. 3 shows the 3-month seasonal residuals of surface geostrophic transports between all chosen points, based on five complete years, Winter consists of January March, etc. Transports of 2000 m 2 s 1 are darkened to emphasize the dominant circulation patterns. Similar 3-month seasonal means formed from only the first three and four years of data produce similar fields. When the 5-year mean altimeter height is subtracted to remove the marine geoid, the mean circulation is also removed. To restore the mean Fig. 3. Seasonal residuals of transport between selected T/P crossovers and tide gauges. Winter is JFM, spring is AMJ, summer is JAS and autumn is OND. Based on 5 years, , after removing the 5-year mean from all heights. Transports greater than 2000 m 2 s 2 are darkly shaded to highlight the main pattern of flow.

9 P.T. Strub, C. James / Progress in Oceanography 53 (2002) circulation, one should mentally add back a more intense version of the circulation shown in Fig. 1: cyclonic flow around the boundaries of the Alaska Gyre, equatorward flow in the California Current to around 20 N, poleward flow in the Costa Rica Current between 5 15 N, and eastward flow in the North Pacific Current between N and the NECC between 5 10 N. The winter transport residuals show a strong counter-clockwise Alaska Gyre and a weak or reversed California Current (the poleward Davidson Current) as far south as Baja California. Strengthened eastward return flow occurs along the southern flank of the Alaska Gyre in the interior ocean at scattered locations between N. Equatorward flow is found in the offshore regions of the California Current. In spring, equatorward transport residuals appear next to the California coast, strongest south of 40 N, as the poleward flow moves offshore. By summer, equatorward transport residuals next to the coast are strongest north of 40 N and extend around the Alaska Gyre. Since the long-term mean is removed, this represents a weakening (or reversal) of the normal counterclockwise flow in the Alaska Gyre and a strong California Current, especially north of 40 N. The seasonally weakened Alaska Gyre produces westward transport residuals on its southern border in the interior at some locations between N. Poleward seasonal residuals begin to appear next to the California coast south of 40 N even in summer, representing the inshore countercurrent described by Lynn and Simpson (1987) and Strub and James (2000). By autumn, poleward residuals begin to dominate, as the equatorward California Current moves offshore. The residual transports become consistently poleward around the basin in winter. The magnitude of the largest transports in both summer and winter are equivalent to velocities of order 5 15 cm s 1. These are the correct order for geostrophic climatological currents in the California Current (Chelton, 1984; Lynn & Simpson, 1987). One striking aspect of the seasonal circulation residuals in Fig. 3 is the degree to which the strongest and most coherent transports are concentrated around the boundaries. This makes the whole system look more like a single basin-scale gyre, rather than the two-gyre system seen in the mean flow of Fig. 1. The primary connection to the interior zonal flow across the four north south lines is in the Alaska Gyre between N. Thus the northern portion of the Alaska Gyre appears to strengthen and weaken without a corresponding change in the interior flow in the subtropical gyre. If the North Pacific Current simply changed its seasonal location, moving north in winter to preferentially feed the Alaska Gyre and south in summer to feed the California Current, the expected residual pattern in winter would consist of eastward transport residuals north of westward residuals. The opposite pattern (westward north of eastward residuals) would occur in summer. The interior north south lines primarily show only the part of this pattern associated with the Alaska Gyre. Extending these north south lines farther south does not change this result. Five-year ( ) seasonal means of surface forcing, as represented by the SLP fields, are shown in Fig. 4. Note that the long-term mean is not removed from these fields. The forcing is consistent with the circulation being primarily wind-driven. The Aleutian Low is strong in winter, representing the net result of cyclonic storms that drive the counterclockwise Alaska Gyre and oppose the equatorward flow in the northern half of the California Current System. In spring, the strengthened North Pacific High creates equatorward wind stress in the California Current. By summer the pressure gradient causes eastward wind stress in the northern Coastal Gulf of Alaska and equatorward wind stress along the entire eastern margin. This reverses in the autumn north of central California, beginning the return to counterclockwise seasonal wind forcing around the basin in winter. Note that the summer and winter patterns are more basinwide, while the spring and autumn patterns are more like the mean double-gyre pattern expected in the ocean. The region where the seasonal circulation residuals (Fig. 3) flow counter to the winds most consistently is in the southern half of the California Current System, where poleward transport residuals are found next to the coast in summer and autumn, despite strong equatorward winds. The tendency for poleward flow next to the coast in summer off California is documented and referred to as the Inshore Countercurrent (IC) by Lynn and Simpson (1987) and is also seen in higher resolution seasonal height fields from the combined Geosat, ERS and TOPEX altimeters, presented by Strub and James (2000). The in situ data of Lynn and Simpson (1987) indicate that it may be concentrated closer to the coast than depicted by Fig.

10 172 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 4. Seasonal SLP fields, based on NCEP fields. Seasons are as in Fig. 3 and the 5-year mean is not removed. 3. A tendency for poleward flow in summer off central and southern California has also been noted in previous studies using in situ data, as discussed further below. Fig. 5 shows the first two Empirical Orthogonal Functions (EOFs) of the time series of transports over the entire basin. These represent a total of 33% of the total variance and more compactly summarize the principal features of the seasonal surface circulation seen in Fig. 3, with higher (monthly) temporal resolution. They also give some indication of the year-to-year variability in the seasonal cycles. The spatial pattern of the first EOF (representing 23.5% of the variance) consists of poleward/counterclockwise boundary currents at the 1 2 locations next to the coast. There are no regions of strong transports in the interior or into the margins. The time series for the first EOF forms a clear seasonal cycle, peaking in December February in each year. This represents the strongest period of the circulation in the Alaska Gyre and the weak or reversed California Current. The negative summer minimum of the time series occurs between June August and corresponds to the strongest period of equatorward flow in the California Current and the weakest period of poleward flow in the Alaska Gyre. Looking forward to the analysis of interannual variability in Part 2 (Strub & James, 2002), we note that the summer minimum is weakest in mid-1997 and that the winter maximum is strong and longer in duration during the El Niño. In the California Current, the spatial pattern for the second EOF (representing 10% of the variance) is equatorward for the transports next to the coast south of 37 N, poleward farther offshore between N. In the Gulf of Alaska, the pattern strengthens the counterclockwise rotation in the northwest part of the Alaska Gyre next to the coast and weakens it slightly farther offshore. The time series is usually positive in spring and negative during late-summer and autumn. Thus the second EOF is uncorrelated with the first, by virtue of a three month phase shift in the time series (in approximate quadrature). The second EOF corresponds to the earlier (spring) beginning of equatorward flow next to the coast in the southern half of the California Current System, as the winter poleward flow moves offshore. In late summer and autumn, the pattern represents the early poleward flow next to the coast and the offshore migration of the

11 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 5. EOFs of transport residuals: (a) first EOF (23.5% of the variance); (b) second EOF (9.7% of the variance). In the spatial pattern, transports with EOF weightings greater than 1 m 2 s 2 are darkly shaded to highlight the main pattern of flow. equatorward California Current in the same region. In the Gulf of Alaska, the second EOF continues the westward flow next to the coast in the northwest during spring, prior to the reversal of the seasonal residuals in summer. To check whether the fairly specific grid of transports used above misses aspects of the seasonal flow, more complete spatial maps derived from all available altimeter surface heights are used. Two-month seasonal height residuals in Fig. 6, based on over seven years of altimeter data, cover the very large-scale NE Pacific, from the equator to 61 N and from W. As stated above, the spatial mean is removed from each two-month field, eliminating the expected large-scale steric rise and fall of hemispheric sea level with seasonal heating and cooling. The remaining seasonal gradients in height are represented by the twocentimeter contour spacing. The magnitude of the transports are inversely proportional to the distance between height contours at a given latitude, while arrows on the height contours indicate the direction in which geostrophic currents flow. The height residuals confirm that most of the strong, coherent variability north of 20 N is confined to the boundaries. There is only the weakest indication that the strength of the zonal North Pacific Current varies systematically on seasonal time scales. Strong variations in its strength or position would appear as zonally oriented changes in these seasonal residuals. This type of variation is apparent south of 20 N, due to seasonal changes in the Equatorial Current System, especially in the zonal NECC near 5 10 N (strongest in September December and absent or reversed in March June). No strong zonal patterns such as these appear between N. However, if one averages the fields from November April (autumn winter), the resulting field is smoother but similar to the November December residual in Fig. 6f, with a slight tendency for an eastward flow from 150 W to 130 W between N, due to a north south gradient in SSH caused by generally lower values to the north. Similarly, the average of the May October (spring summer) fields produces slightly higher SSH values in the north, creating a weak tendency for a westward spring summer transport residual in this region (somewhat similar to the May June residual in Fig. 6c,

12 174 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 6. Two-month seasonal SSH residuals based on Geosat, T/P, ERS-1 and ERS-2. Arrows indicate the direction of geostrophic flow and geostrophic surface transports are inversely proportional to the contour spacing. The contour interval is 2 cm. but smoother). Thus, there is a slight increase in surface transport in the North Pacific Current in winter, as found in the drifter velocities by Bograd et al. (1999), but the seasonal changes are much weaker than changes in the boundary currents. The strongest pattern that relates the boundaries to the interior in California Current and Alaska Gyre is the general offshore migration of the alongshore bands of height (and the associated bands of alongshore

13 P.T. Strub, C. James / Progress in Oceanography 53 (2002) transports along height contours). Thus, in winter (January February) there are high values of SSH next to the coast, associated with poleward seasonal transport residuals extending from southern California to the Alaska Gyre. At the same time, there is an offshore band of low SSH, associated with equatorward transport on its western side. This band of low SSH is the expression of the band of low SSH found next to the coast in July August, which has moved offshore. The band of high SSH next to the coast in January February, in turn, moves offshore by July August. At times there is a weak third band, farther offshore, of the same sign as found next to the coast, representing the previous year s coastal signal. Previous studies of the California Current using altimeter data and surface drifters have also documented the offshore seasonal movement of the eddy kinetic energy (EKE) generated in the vicinity of the seasonal jet (Kelly et al., 1998). Strub and James (2000) use intraseasonal variances of crosstrack geostrophic velocity, calculated from the alongtrack gradients of altimeter SSH, to demonstrate this offshore movement of EKE in the California Current. In Fig. 7, we present similar maps of intraseasonal variance of crosstrack geostrophic velocity in the Alaska Gyre during winter and summer. Since these variances only include one component of velocity, they are only equivalent to the complete EKE field if the velocities are isotropic. Still, they show the general pattern of the offshore movement. To create these maps, the mean crosstrack velocity at each point along the T/P tracks is removed for each season. This leaves the variability caused by changes within each season and by changes between years (representing both intraseasonal and interannual Fig. 7. Intraseasonal variances of geostrophic cross-track velocities, based on 5 years of T/P data. Only the winter (January March) and summer (July September) periods are shown. Variances are formed by subtracting the 5-year seasonal mean from each observation. Thus, these variances include both intraseasonal and interannual variability.

14 176 P.T. Strub, C. James / Progress in Oceanography 53 (2002) variability, although we call these intraseasonal variances for simplicity). In the California Current, this proxy for EKE is greatest in summer, surrounding the equatorward jet (Strub & James, 2000). In the Alaska Gyre, Fig. 7 shows that the proxy EKE is greatest in winter in the km next to the coast, where the Alaska Current also has a winter maximum. In spring and summer, the variances move farther offshore and diminish in magnitude, similar to the fate of the EKE in the California Current in autumn and winter. This is clearest in the north and northeast regions of the Alaska Gyre, less apparent in the Alaska Stream (where the data is often flagged as suspect, due to the wide shelf). The net result is that EKE is greatest near the strongest flow in both the Alaska Gyre and the California Current, suggesting dynamical instabilities as a possible source of the EKE. In both cases the EKE diminishes and moves offshore, similar to the movement of the SSH signals. Besides the offshore migration of SSH, alongshore currents and EKE in both the Alaska Gyre and California Current, the larger picture provided by Fig. 6 also suggests a connection between height and transport residuals in the southern part of the California Current and those in the eastern tropical Pacific farther south. The seasonal height residuals in January February (Fig. 6a) indicate low sea level and weaker poleward flow (or actual equatorward flow) along the Mexican mainland and Central America. These extend as far north as southern Baja California in January February. The equatorward residuals further strengthen off Central America and Mexico in March April. This winter and early spring period is the time when the normal eastward flow in the NECC and the northward flow along Central America (the Costa Rica Current) weaken or reverse in ship-drift currents shown by Fiedler (1992, 2002). By March April, the residual flow is strongly equatorward from Central America to Baja California and weakly equatorward as far north as northern California. Equatorward residuals off Central America next to the coast reverse in May June and by July August strong poleward residuals occur as far north as the tip of Baja California, with weaker poleward residuals as far north as central California. The strongest poleward residuals move to Baja California in September October, then extend around the Gulf of Alaska in November December, while they weaken off Mexico. Thus, the observed earlier poleward (and equatorward) flow in the southern part of the California Current appears to be connected to similar, stronger flow off Mexico and Central America. The poleward progression of the seasonal signals is most clearly defined by the timing of the annual maximum SSH at locations km from the coast, as a function of latitude (reconstructed from annual and semiannual harmonics). The combination of these two harmonics identify the steady progression of peak SSH (not the first arrival of the SSH signal, but the peak) from Central America in June to mainland Mexico (July August), Baja California (September November), southern California (December) and then around the Alaska Gyre in January March. A similar analysis of the month of minimum SSH also starts off Central America (in February March) and progresses to Oregon British Columbia by July September. This pattern of northward seasonal progression of coastal SSH from Central America to the Alaska Gyre also appears in the seasonal cycles of coastal tide gauge records presented by Enfield and Allen (1980). 4. Summary and discussion With respect to the four objectives stated at the end of the Introduction, we summarize the most certain of our results, then discuss them further. 1. The maximum surface currents in the Alaska Current and Alaska Stream occur in winter (December February), as demonstrated in Figs. 3, 5 and 6. This is in phase (approximately) with large-scale surface forcing (Fig. 4). 2. The primary seasonal variability in surface circulation and transport in the NE Pacific north of 20 N is in the boundary currents stretching from mainland Mexico around the Alaska Gyre (Figs. 3 and 6).

15 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Seasonal residuals of surface transport are equatorward in summer around the Alaska Gyre and in the California Current, except for an inshore poleward countercurrent south of approximately 40 N after July. In winter, the seasonal residuals flow poleward from Baja California around the Alaska Gyre. 3. Seasonal residuals in transport between the interior of the NE Pacific (the North Pacific Current) and the boundaries of the subarctic and subtropical gyres are very weak compared to the residuals along the boundaries (Fig. 6). The most consistent pattern around the basin is the seasonal offshore movement of bands of high and low SSH, with corresponding cells of alongshore geostrophic current and EKE. The offshore scale of these cells is approximately km. 4. Seasonal transport residuals along Central America and the Mexican mainland lead those in the California Current off central California by approximately six months (Fig. 6). This creates an out-of-phase relationship between SSH in the low latitudes and SSH in the mid- and high-latitudes. Transport residuals are equatorward off Central America in January April, while residuals are mostly poleward in the California Current and Alaska Gyre. Residuals are poleward off Central America in May October, while residuals in the California Current are mostly equatorward, except in the inshore countercurrent Alaska Gyre Our finding of a winter seasonal maximum in surface transport in the Alaska Gyre confirms the tentative conclusion of Bhaskaran et al. (1993) and is consistent with the general finding of an October March maximum in the Alaska Stream from surface drifters (Bograd et al., 1999). It is somewhat at odds with the March or later date of Royer (1981b) and the autumn date of Kelly et al. (1993). The transport fields in Figs. 3 and 5 are based on five years of consistently processed data from the high quality TOPEX altimeter. The height fields in Fig. 6 are based on over seven years of data from four altimeters of varying quality and differing orbits, but still with more regular sampling than the hydrography available to Royer (1981b) and a much longer record than available to Bhaskaran et al. (1993) and Kelly et al. (1993). Bhaskaran et al. (1993) based their analysis on an EOF decomposition of nearly four years of Geosat data, which found an annual cycle in one mode. This corresponded to a low SSH cell north of 55 N, lowest in winter, similar to the cell we find in two-month means from seven years of data in Fig. 6. Kelly et al. (1993) based their seasonal cycle on the fit of annual and semiannual harmonics to 2.5 years of Geosat data. Their SSH field for January consists primarily of low residuals, sloping down to the north, which may partly result from seasonal steric cooling. Examination of their fields (their Fig. 9), however, also shows that they were forced to eliminate data within approximately 2 3 of the coast, precisely where the combined altimeter data find the high SSH residuals associated with the strengthened gyre in January February. Likewise their fields of high SSH in July and October miss the low SSH band next to the coast to the north. It seems likely that the analysis of Kelly et al. (1993) found only the southern half of the winter low and summer high cells, offshore of the strong boundary signals in the NE Alaska Gyre in Fig. 6. Thus, the results of both analyses of Geosat data are consistent with our results, with somewhat different interpretations. With respect to the nature of the Alaska Gyre seasonal residuals, in Fig. 6 the offshore migration of the summer coastal low and winter coastal high produces an elongated NW SE gyre, alternating in transport direction between winter and summer. This is in contrast to the location of the center of the climatological gyre (Fig. 1), which is at 54 N, 150 W. Thus, the seasonal variability of the Alaska Gyre is not in the overall gyre, but in the structure along the boundary and in the interior in the northeastern portion of the basin. To place these seasonal residuals into perspective, it is useful to combine them with the mean dynamic height field shown in Fig. 1. In Fig. 8 we do this for the January February and July August periods from Fig. 6. This makes it clear that the effect of the seasonal residuals in the NE corner is to expand the central low of the Gyre farther into the NE corner in winter and strengthen the gradient next to the coast. In summer, the central low contracts to its climatological mean position and gradients next

16 178 P.T. Strub, C. James / Progress in Oceanography 53 (2002) Fig. 8. Two-month seasonal height residuals (from Fig. 6) added to dynamic height climatology (from Fig. 1). Only winter and summer are shown. The contour interval is 2 cm. to the coast become weaker than in the climatology. The location and strength of the center of the gyre, however, is not greatly affected (Fig. 6), although it may be difficult to determine that by looking at the combined fields of residual plus long-term mean (Fig. 8). We have chosen to present the seasonal residual fields alone in previous figures, since this isolates the temporally varying signal, which is all that the altimeter can resolve Connections around the boundaries of the subarctic and subtropical gyres In the western and northern Alaska Gyre, it is doubtful that the currents actually reverse in summer along the boundaries, despite the appearance of eastward coastal flow in Fig. 8. The climatology is too smooth to portray the actual strength of the mean boundary currents in those regions and so is overpowered by the altimeter seasonal residuals, which are not as smooth. Royer (1998) cites measurements in the Alaska Coastal Current at the apex of the Gyre, showing that velocities drop from cm s 1 (or more) in September March to cm s 1 in April June but do not reverse (Johnson et al., 1988). More recent measurements of the ACC farther west in April September 1991 also show no seasonal reversal (Stabeno et al., 1995). In the eastern Alaska Gyre, there is a greater likelihood of seasonal reversals, offshore of a narrow, poleward, buoyancy-driven jet. This poleward jet has been measured at the northern end of Vancouver Island (Hickey et al., 1991) but its behavior north of Vancouver Island is not well known. Hickey et al. (1991) also show a broader equatorward flow of over 15 cm s 1 offshore of the poleward jet. The altimeter transport values and height differences correspond to velocities of 5 10 cm s 1 in the summer north of Vancouver Island, similar to those measured by Hickey et al. (1991). In addition, Thomson and Gower (1998) and Thomson, Hickey, and LeBlond (1989) depict summer currents over the slope as equatorward north of and along Vancouver Island. Thus, we postulate that the surface velocities are equatorward over