Hydrographic responses at a coastal site in the northern Gulf of Alaska to seasonal and interannual...

ARTICLE IN PRESS

0967-0645/$ - se

doi:10.1016/j.ds

�CorrespondiE-mail addre

Deep-Sea Research II 52 (2005) 267–288

www.elsevier.com/locate/dsr2

Hydrographic responses at a coastal site in the northernGulf of Alaska to seasonal and interannual forcing

Thomas C. Royer�

Center for Coastal Physical Oceanography, Department of Ocean, Earth and Atmospheric Sciences, Old Dominion University,

Norfolk, VA 23529, USA

Accepted 26 September 2004

Available online 6 January 2005

Abstract

Three decades (1970–2000) of hydrographic (temperature–salinity–depth) sampling at the coastal site, GAK1, near

(601N, 1491W) in the northern Gulf of Alaska provides an opportunity to investigate the seasonal and interannual

variability within this water column. Over this relatively deep shelf (260m), the temperature and salinity are forced by

solar heating, coastal freshwater discharge, winds and El Nino-Southern Oscillation (ENSO) events. Seasonally, the

water temperatures at the surface and bottom change out of phase with one another. From about November until

April, there are temperature inversions with thermal stratification increasing from April to August. The upper-layer

(0–100m) salinity closely follows the seasonal freshwater discharge with the annual minimum in October and maximum

in March. However, this is in sharp contrast with the lower-layer (100–250m) salinity that has an April minimum and

October maximum. Water density cycles follow the salinity changes at this site, not the temperature changes. The lower-

layer salinity cycle is the sum of responses to buoyancy and wind forcing. Maximum freshwater discharge in autumn

should enhance the entrainment through the strengthening of both cross-shelf and alongshore pressure gradients,

causing a deep intrusion onto the shelf of relatively high-salinity offshore water. The contributions of the very weak,

summer upwelling winds to the increased lower-layer salinity are uncertain. The summer is a period when the

hydrography on the shelf relaxes, since the non-summer winds are the downwelling-type. They force less saline water

downward, diminishing the lower-layer salinity especially in late winter. This downwelling will force relatively warm

water downward until the temperature inversion occurs in November. After that, the downwelling will be forcing cooler

water downward. This leads to the maximum lower-layer temperature in November.

The interannual anomalies of temperature and salinity give insight into the potential forcing of this ecosystem.

Correlations between the forcing phenomena of local winds, freshwater discharge, Southern Oscillation Index or

patterns of sea-surface temperature (Pacific Decadal Oscillation (PDO)) suggest that interannual subsurface

temperature anomalies are linked to ENSO events with a propagation time from the equator to the Gulf of Alaska

of about 8–10 months. There are no significant interannual temperature variations in the surface layers (0–50m)

correlated with ENSO events. However, the interannual temperature variability throughout the water column does

respond to the interannual variability of coastal freshwater discharge and also follows the PDO. Additionally, the

e front matter r 2004 Elsevier Ltd. All rights reserved.

r2.2004.09.022

ng author. Fax: +1757 683 5550.

ss: [email protected] (T.C. Royer).

www.elsevier.com/locate/dsr2

ARTICLE IN PRESS

T.C. Royer / Deep-Sea Research II 52 (2005) 267–288268

water-column temperature anomalies are well correlated with local winds and regional winds up to about 1000 km

eastward, with delayed responses of 3–8 months.

The salinity anomalies in the upper layer (0–100m) correlate inversely with coastal freshwater discharge anomalies

with a 1-month delay. In contrast, the behavior of the salinities in the lower layer (150–250m) is opposite to the surface

layers. The deep interannual salinity anomalies increase with increasing freshwater runoff, reflecting a possible

strengthened cross-shelf circulation. The salinity anomalies do not follow PDO or ENSO. Winds over the eastern Gulf

of Alaska are well correlated with the salinity anomalies, though lags approach 5 years.

Interdecadal trends in these coastal temperatures and salinities are consistent with a general warming of the upper

layer (0–100m) of the water column, with temperatures increasing by about 0.9 1C since 1970 and 0.8 1C in the lower

layer (100–250m). During this same time period, the sea-surface salinity decreased by about 0.3 and the upper-layer

salinity decreased by 0.06, while the lower-layer salinity increased by about 0.04. Consequently, there is a tendency for

the stratification to increase. This has been accompanied by a tendency for less downwelling and increased freshwater

discharge. Both of these influences will tend to increase the coastal stratification.

r 2004 Elsevier Ltd. All rights reserved.

1. Introduction

Coastal hydrographic observations spanningnearly three decades at 59150.70N, 149128.00W(Gulf of Alaska, GAK1) in the northern NorthPacific Ocean (Fig. 1) allow the investigation ofhydrographic time scales that range from seasonalto interannual. Relatively large seasonal signals inthe winds and moisture fluxes are present whereasthe interannual time scales might respond to large-scale atmospheric forcing and remote equatorialforcing from El Nino-Southern Oscillation(ENSO) events. The seasonal atmospheric forcingof the northern North Pacific Ocean changes froma strong low-pressure system in winter to a weakhigh-pressure system in summer (Wilson andOverland, 1986). In winter, high-latitude stormsspawn in the western North Pacific Ocean as dry,cold air outbreaks. These storms propagate east-ward across the Pacific into the Gulf of Alaska,gaining heat and moisture from the ocean. Alongthe Pacific Northwest coastline, these stormsencounter an extensive barrier of coastal moun-tains. As they attempt to continue their passageeastward over the mountains, their relativelywarm, moist air masses are lifted adiabaticallycausing very high rates of precipitation, occasion-ally exceeding 800 cm year�1 (Wilson and Over-land, 1986). The steep coastal terrain and relativelynarrow coastal drainage area in Alaska does notallow the establishment of major river networks.Instead of entering the ocean as large river

discharges, the freshwater enters the coastal watersthrough a myriad of small coastal streams (Royer,1982). These sources sum to an annual average ofmore than 23,000m3 s�1 for the Alaskan coastlinealone, from its southern boundary with BritishColumbia to 1501W.In concert with the seasonal precipitation

variations, the coastal winds over the Gulf ofAlaska undergo large seasonal changes. Low-pressure domination in winter assures the presenceof downwelling winds and coastal convergences ofthe upper-layer waters at that time of the year.This winter convergence helps to maintain a bandof low-salinity water along the shore (Xiong andRoyer, 1984) that has been identified as the AlaskaCoastal Current (Schumacher and Reed, 1980;Royer, 1981). As summer approaches, the stormtracks move northward into the Bering Sea(Whittaker and Horn, 1982) and the strong winterdownwelling in coastal Gulf of Alaska is replacedby very weak upwelling.The relatively low water temperatures (aver-

aging less than 10 1C at the surface) and the rangeof temperatures enable the density to be more res-ponsive to salinity changes rather than tempera-ture changes. High rates of coastal freshwaterdischarge in combination with downwelling windsthroughout most of the year create nearshore hori-zontal and vertical coastal stratifications that drivethis alongshore flow cyclonically around the basin,averaging about 0.25 Sv (Royer, 1981; Schumacherand Reed, 1980) (1 Sv ¼ 1� 106m3 s�1). This

ARTICLE IN PRESS

150oW 147oW 144oW 141oW 138oW 135oW 56oN

57oN

58oN

59oN

60oN

61oN

62oN

200

200 200

200

200

20020

0400

400

400

400

1000

1000

1000

1000

60 N, 149 W60 N, 146 W

57 N, 137 W

GAK1

Alaska

Arctic Ocean

BeringSea

Gulf of Alaska

Study Area

Fig. 1. Gulf of Alaska with GAK1 and upwelling index sites indicated.

T.C. Royer / Deep-Sea Research II 52 (2005) 267–288 269

coastal current has a width comparable to theinternal Rossby radius of deformation, about10–20 km (Johnson and Royer, 1986). The hydro-graphic station, GAK1, is located within thiscoastal current. Shelf depths here are relativelydeep; usually greater than 100m within less than akilometer off the coast, increasing to severalhundred meters across the shelf. This is quiteunusual for a shelf to have such a rapidlyincreasing bathymetry; it is nearly a vertical wall.Farther offshore, bottom topography variationsmight exert significant control on the circulation,as onshore–offshore transports could be influ-enced by the numerous troughs and canyons onthis shelf.

2. Hydrographic forcing functions

2.1. Heat flux

The paucity of direct measurements of many ofthe atmospheric parameters over the Gulf ofAlaska requires the use of proxy data sets. Thesolar heat-flux variability is assumed to followthe seasonal changes in solar declination with the

maximum of about 250Wm�2 day�1 in late Juneat the latitude of GAK1 (Bryant, 1997). While theseasonal pattern of sensible heat flux generallyfollows the solar flux, in winter there can be coldair outbreaks over the region that can extract morethan 1000Wm�2 day�1 from the ocean surface(Namias, 1968). Neglecting the seasonal changesin cloudiness but considering heat storage, themaximum sea-surface temperatures (SSTs) shouldoccur in late summer 1–2 months after themaximum solar heat flux. A maximum surfacewater temperature in August is expected, similar tothat found at Ocean Station P (501N, 1451W) inthe Pacific (Pickard and Emery, 1990). Minimumsurface temperatures should occur approximately6 months later (i.e. February) according to asinusoidal seasonal signal based on solar declina-tion. This signal might be skewed due to verticalmixing since heating from below is possible, whichmight delay the surface minimum.

2.2. Wind stress

The absence of long-term direct wind measure-ments for the Gulf of Alaska requires that indirectestimates of the wind stress be used to describe the

ARTICLE IN PRESS


seasonal and interannual atmospheric forcing.Upwelling indices (www.pfeg.noaa.gov) are usedas a measure of the local and regional winds(Bakun, 1973). The upwelling index is an estimateof the onshore–offshore component (y-direction)of Ekman transport

My ¼ tx=f ; (1)

where tx is the wind stress in the alongshore(x-direction) and f is the Coriolis parameter,2O sinj; where O is the rate of rotation of theearth and j is the latitude.The wind stress is calculated from the geos-

trophic winds using sea-level pressure on a 3� 31grid. These winds are reduced by 30% and rotatedby 151 to the right to account for frictional effects.Bakun (1973) pointed out some difficulties withthe accuracy of these winds especially nearmountainous coastlines. Luick et al. (1987) in-vestigated the responses near to GAK1 usingmeasured winds and found that the calculatedwinds needed additional corrections, but that theirtemporal responses matched the actual winds atperiods greater than 1.16 days. For additional

Fig. 2. Monthly mean Bakun upwelling indices at three

discussions of the upwelling indices, see Schwinget al. (1996).The upwelling index at 601N, 1491W from 1946

to April 2001 (Fig. 2, upper panel) shows that,with few exceptions, the winds are downwellingfavorable (negative) with a distinct seasonal cycle(Fig. 3, upper panel). At other locations east andsouth along the coast (Fig. 3), the maximumseasonal downwelling always occurs in Januarythough the intensity varies around the perimeter ofthe gulf from 481N to 1601W (Table 1). The mostintense downwelling occurs seasonally in Januaryat 571N, 1371W (Fig. 3, lower panel) with themaximum seasonal onshore transport of192m3 s�1 along each 100m of coastline, ascompared with 120m3 s�1 along each 100m ofcoastline at 601N, 1491W. The maximum onshoreEkman transport decreases southward from 571Nto a minimum at 511N. In summer, the cyclonicatmospheric circulation weakens and the down-welling winds are replaced by upwelling winds,with much less strength than the downwellingwinds. Maximum upwelling (minimum downwel-ling) over the Gulf of Alaska occurs in July except

locations bordering the northern Gulf of Alaska.

http://www.pfeg.noaa.govhttp://

ARTICLE IN PRESS

Fig. 3. Average annual cycle of coastal upwelling indices from 1970 to 2000 from three locations bordering the northern Gulf of

Alaska as determined according to the method of Bakun (1973) with standard deviations.

Table 1

Upwelling indices for the Gulf of Alaska (after Bakun, 1973)

Location Maximum downwelling Maximum upwelling U.I.

correlation

coefficient

relative to

601N, 1491W

U.I.

correlation

coefficient/

discharge

Month Index S.D. Month Index S.D.

601N, 1491W 1 �120 79 7 5 4 1.0 0.16 (�25)

601N, 1461W 1 �155 102 8 2 5 0.74 �0.29 (1)

571N, 1371W 1 �192 106 7 0 8 0.57 �0.49 (1)

541N, 1341W 1 �99 82 7 1 12 0.48 �0.23 (1)

511N, 1311W 1 �65 74 7 17 19 0.19 �0.16 (48)

481N, 1251W 1 �95 76 7 38 21 0.23 �0.14 (48)

Note: Units are cubic meters per second per 100m of coastline. Negative indices are downwelling and positive indices are upwelling.

Correlations are for anomalies of upwelling indices and freshwater discharge with the phase shift in months in parentheses with

positive indicating that upwelling is leading. All correlations are 499% confidence level (CL).


at 601N, 1461W where it takes place in August.Upwelling amplitudes are very small from thenorthern gulf boundary to about 541N. There is nosignificant monthly averaged upwelling at 571N,1371W. Maximum monthly averaged upwelling

(38m3 s�1 along each 100m of coastline) along theBritish Columbia–Alaska coastline occurs at 481N,1251W. The seasonal variability of the upwellingindex at each site was subtracted to yield theanomalies (Fig. 4). Summer is characterized by a

ARTICLE IN PRESS

Fig. 4. Monthly mean Bakun upwelling index anomalies at three locations bordering the northern Gulf of Alaska.


lack of downwelling rather than active upwelling.Upwelling farther to the south was not incorpo-rated into this study since it is believed that southof this point will flow equatorward.The spatial coherence of the wind forcing was

determined using correlation techniques to relatethe upwelling index anomalies at 601N, 1491Wwith the indices along the coast (Table 1). There isa ‘break point’ at 511N where the correlationcoefficient drops below 0.20. Thus, the coastalconvergences along the coast of the Gulf of Alaskahave similar variability from 51N1 northward.The upwelling indices have large negative

anomalies beginning in about 1950 until the early1970s when a quiescent period was present untilabout 1979 (Fig. 4). A ‘normal’ 5-year periodfollowed. However, since about 1984, large down-welling anomaly events (4110m3 s�1/100m ofcoastline) have not occurred at 601N, 1491W and601N, 1461W whereas prior to 1985 such intenseevents took place every 2–3 years. Off SoutheastAlaska (571N, 1371W), there were no correspond-ing changes in the frequency of large downwellingevents.

2.3. Freshwater discharge

The freshwater discharge at the coast nearSeward, Alaska (Fig. 5) was determined using ahydrology model based on precipitation andtemperature that also incorporates snow and icemelt (Royer, 1982). The hydrology model uses themonthly mean temperature and precipitation forthe National Weather Service Southcoast andSoutheast divisions. The Southeast division runsfrom the southern Alaska border at BritishColumbia to a line south of Yakutat near 591 N(Fig. 1). The Southcoast division runs fromYakutat to Kodiak Island. The model allowsseasonal (winter) and interannual storage ofmoisture depending on the air temperatures. Itrequires that there be no net glacial accumulationor ablation. The lack of river discharge data for theentire period does not allow the incorporation ofthose flows into the model. It is estimated thatabout 10–15% of the flow is due to thosedischarges. It also excludes the influx of freshwaterfrom regions other than Southeast or SouthcoastAlaska, namely British Columbia andWashington.

ARTICLE IN PRESS

Fig. 6. Seasonal variability of coastal freshwater discharge for

the northern Gulf of Alaska estimated offshore of Seward,

Alaska (after Royer, 1982) with standard deviation.

Fig. 7. Anomalies of coastal freshwater discharge for the

northern Gulf of Alaska estimated offshore of Seward, Alaska

(after Royer, 1982). Heavy line is 5-year filtered (Butterworth)

discharge anomaly.

Fig. 5. Monthly coastal freshwater discharge for the northern

Gulf of Alaska estimated offshore of Seward, Alaska (after

Royer, 1982). Heavy line is 5-year filtered (Butterworth)

discharge.


The seasonal discharge (1931–2000) (Fig. 6) isgreatest in October, coincident with fall stormsand the highest precipitation rates for bothSoutheast and Southcoast Alaska. The precipita-tion and discharge decrease throughout the winter,reaching the seasonal minimum in February–March. There is a relatively rapid increase in themonthly mean discharge from March to May from

the spring melting. Throughout the summer, therate is slightly below the annual mean(23,000m3 s�1) until August when it begins climb-ing to the seasonal maximum. The standarddeviations of the freshwater discharge are greatestduring the fall and winter months when theprecipitation is greatest and least in June and Julywhen the precipitation is least.The coastal freshwater discharge anomaly

(Fig. 7) has some low-frequency variations of14–20 years superimposed on longer period varia-tions of about 50 years, though the record length isinsufficient to resolve adequately these time scales.Beginning in 1970, there was an increasing trend inthe discharge, reaching a maximum in 1987, withsome declines in the 1990s followed by belownormal dip in 1996 and increasing discharge in1999 and continuing above normal in 2000.The freshwater discharge and upwelling index

anomalies (Table 1) indicate that only the upwel-ling index closest to GAK1 (601N, 1491W) andfreshwater discharge have a significant positivecorrelation (0.16 with 342 effective degrees offreedom 499% confidence level (CL)), with thefreshwater leading the upwelling index by 25months. The effective degrees of freedom are

ARTICLE IN PRESS


based on the autocovariance function (Emery andThomson, 1998, p. 262). There are no othercorrelations with 495% CL with other time shiftsfor these two parameters at this location. At otherlocations (Table 1), there are significant negativecorrelations with upwelling and freshwater nearlyin phase (upwelling anomaly leading freshwaterdischarge anomaly by 1 month). This is reasonablesince the hydrology model incorporates precipita-tion from Southeast and Southcoast Alaska withthe southeast discharge lagged by a month torepresent its transit time to Seward (Royer, 1982).High rates of freshwater follow enhanced down-welling by about 1 month. Storms in the coastalregions upstream of GAK1 should influence thefreshwater discharge along this coast. The dis-charge is most highly correlated with upwelling inthe eastern portion of the northern Gulf of Alaska.The highest correlation with freshwater discharge(�0.49) is with the upwelling index anomaly at571N, 1371W (located off Southeast Alaska).Additional significant correlations between upwel-ling index and freshwater discharge at lowerlatitudes have the upwelling index leading by 4years. Changes in large-scale atmospheric circula-tion systems could be responsible for thesecorrelations and these might be related to ENSOvariability since the upwelling index variabilities atlower latitudes have some linkages to ENSO.

Fig. 8. Seasonal variability of temperature at GAK1.

3. Seasonal variability of water temperature and

salinity

This description and explanation of coastaltemperature and salinity in the northern Gulf ofAlaska is an update of an earlier paper that dealswith the first look at the seasonal variability in theGulf of Alaska using less than a decade of data(Xiong and Royer, 1984). Water-column tempera-ture and salinity at GAK1 in 263m of water havebeen measured since 1970 at irregular temporalsampling intervals ranging from hours to months.Since 1990, the sampling has been more regular,approximately monthly. The initial sampling upuntil the mid-1970s utilized only Nansen bottlesto sample at standard depths. Later, instru-ments were employed that were capable of more

frequent sampling; first salinity–temperature–depths (STDs) were used followed by modernconductivity–temperature–depths (CTDs). Only T

and S at 10 standard depths from the surface to250m are used in this study to permit the longestpossible time history. For logistical reasons, thesampling site was originally selected because of itsproximity to research vessel berthing facilities inSeward, Alaska. Fortuitously, it was discoveredthe site is also located in the Alaska CoastalCurrent (Royer, 1981), ensuring good connectionwith the regional coastal circulation. The monthlymeans of salinity, temperature and density atstandard oceanographic depths have been deter-mined for the period from December 1970 toFebruary 2000 (Figs. 8–10).The seasonal cycles of the upper layers (0–50m)

monthly mean temperatures (Fig. 8) all have theirminimum in March. This is in sharp contrast totheir maxima that do not occur simultaneously.Instead, the maximum at the surface of more than13 1C occurs in August followed by lesser maximaat depth up to 3 months later. Though the 0–50mtemperatures decrease rapidly from October oninto winter, the 100 and 150m temperatures peakin November with the 75m temperatures intransition, holding steady in October–November.The downward propagation of the seasonal sur-face warming appears to be less important below75m. Insulation between the upper and lower

ARTICLE IN PRESS

Fig. 9. Seasonal variability of salinity at GAK1.

Fig. 10. Seasonal variability of density at GAK1.


layers is confirmed by the average seasonalpenetration of the mixed-layer depth being lessthan 150m (Sarkar, 2001). In contrast to thesurface waters, the maximum temperatures at 200and 250m occur in December and January,respectively, nearly 6 months out of phase withthe surface temperatures. The time interval be-tween the occurrence of the maximum andminimum temperature at depth is 5–7 months at

the surface (0–20m), decreasing to 3 months at250m. Except during July and August, thermalinversions with depth exist because the density isnot strongly influenced by temperature here. Thedepth dependencies of the seasonal maximum andminimum temperatures are of opposite sign; theseasonal minimum temperature increases withdepth while the maximum decreases with depth.The surface heat fluxes and water-column heatstorage are responsible for these depth dependen-cies. As noted earlier, the seasonal thermalvariability at 250m is not sinusoidal, implyingthat more than one factor is affecting thetemperature there. Seasonally, the minimum tem-perature at the sea surface (March) coincides(within 1 month) with the minimum near-bottomtemperature that occurs in April (5.4 1C). Thehighest temperature at 250m occurs in January(6.2 1C). These characteristics of the verticalthermal structure imply the mixing of heat frombelow in winter.The impact of the temperature inversions on

water-column stratification is mitigated by thevertical salinity stratification. Seasonal patterns ofmonthly mean salinity (Fig. 9) and densityvariability (Fig. 10) are similar because at lowseawater temperatures the influences of salinityvariations on density are larger than they are forwarmer water. This enhances the influence offreshwater on the ocean circulation at highlatitudes. At the sea surface, the minimum salinityoccurs in August, in phase with the maximumtemperature. Deeper in the water column, theminimum salinity occurs later in the year at:10–30m (October), 50m (November), 75m (De-cember), 100–150m (February) and 200–250m(March). Thus, the minimum salinity below 75mtakes place about 6 months out of phase with thesurface (Fig. 9). Similar depth dependent phasingis found in the maximum salinity. In the upper30m, the maximum salinity is in February–Marchand with increasing depth, it occurs later: 50m(June), 75m (July), 100–150m (August) and200–250m (October). The minimum vertical stra-tifications of salinity and density are in March withthe greatest stratification in August (Figs. 9 and10). Vertical inversions of salinity and density arenot present in these monthly mean data.

ARTICLE IN PRESS


3.1. Analysis as a two-layer system

Based on possible local surface forcing of theupper layers, the water column is considered as atwo-layer system, the seasonal progressions ofsalinity and temperature in the upper (0–100m)and lower layers (100–250m) are out of phase witheach other (Fig. 11). Maximum lower-layertemperatures do not coincide with maximumupper-layer temperatures, though as mentionedearlier, minimum bottom-layer temperatures(April) almost coincide with the minimum surfacetemperature (March) (Fig. 11, upper panel).Maximum salinities in the bottom layer coincidewith minimum surface salinities (August). Severalpossibilities exist for these contrasts. First, theupwelling is greatest in June, July and August(downwelling is least). This relaxation of thecoastal wind-driven convergence will allow high-salinity offshore water to intrude onto thecontinental shelf in the lower layer, eventuallyfinding its ways into the coast. Second, themaximum freshwater discharge occurs in October(Fig. 6) and the minimum in the upper-layersalinity occurs at that time. Accompanying an

Fig. 11. Seasonal variability temperature and s

enhanced alongshore flow, there also should be anenhanced offshore flow (i.e. ageostrophic flow). Asthe discharge enters the ocean along the coast,there will be an accumulation of freshwater as theflow progresses westward. This will create analongshore pressure gradient that will enhancethe offshore flow especially in the upper layers.This leads to speculation that an estuarine-typeflow could exist across the shelf. An upper-layeroffshore flow could, through entrainment (Tullyand Barber, 1960), induce an onshore flow in thelower layer, advecting relatively warm, high-salinity water across the shelf in fall, maintaininga relatively high salinity in the lower layer. Later,in November–December, the downwelling-typewinds increase and the coastal convergence in theupper layers will force lower-salinity water down-ward, causing the lower-layer salinity to decrease,reaching its seasonal minimum in February.Though this lack of downwelling has some of thecharacteristics of upwelling, it should not beconsidered the reverse procedure. The coastalconvergence with downwelling will accelerate thealongshore flow increasing the mixing due tofrictional processes. The high winds will also

alinity within the upper and lower layers.

ARTICLE IN PRESS

Fig. 12. Temperature anomalies versus time and depth at

GAK1.

Fig. 13. Salinity anomalies versus time and depth at GAK1.


enhance wind mixing. The lack of downwellingwill not enhance the flows nor will it mix the upperlayers. However, relaxation of the winds shouldallow the excursion of deep water onto the shelfhere.The seasonal temperature cycles in the upper

layer respond primarily to solar heat flux and windmixing. However, this mixing is also influenced bychanges in the vertical density stratification fromthe freshwater discharge. The upper-layer tem-peratures follow the solar cycle in a delayedfashion with a September maximum and a Marchminimum. The lower-layer temperatures are af-fected by a combination of freshwater discharge,winter cooling and downwelling. First, the summerrelaxation of winds will allow the advection ofrelatively warm, high-salinity water onto the shelfin the lower layer from the central Gulf of Alaskathroughout the summer and into the fall. How-ever, why does the maximum lower-layer tempera-ture occur in November? A possible explanation isthat as the downwelling conditions increase in fall,they will force the nearshore upper-layer watersdownward. This begins to reduce the salinity in thelower layer in fall (October–November). Thelower-layer temperature will continue to increasein fall by the downwelling. Downwelling will forcewarmer water downward since for most of the yearthe water column at GAK1 has a monotonicallydecreasing thermal structure with depth. However,with winter cooling and increasing salinities,temperature inversions begin in November–December (Fig. 8). After the temperature inversionis established, downwelling will cool the lowerlayer. The result is the maximum temperature inthe lower layer in November.

4. Temperature and salinity anomalies

To investigate further the relationships betweenthe potential forcing functions and the hydrogra-phy at GAK1, the monthly means were removedfrom the data to yield the anomalies of tempera-ture (Fig. 12) and salinity (Fig. 13). The tempera-ture anomalies versus time and depth (Fig. 12)have a pattern of negative (lower than normal)anomalies throughout the water column from 1970

until about 1977 when alternating warm and coldepisodes were established. The largest temperatureanomalies, especially the warm events, are foundat depth. For example, significant warm eventsoccurred between 50 and 150m in 1977, the earlyand mid-1980s, 1987 and 1992. The largest warmevent took place in 1998. Unlike earlier subsurfacewarmings prior to the 1998 warm event, it waspreceded by a surface warm event in summer of1997. Interspersed among these warm events are

ARTICLE IN PRESS


cold events though they are usually found atshallower depths than the warmings. Theseanomalies have fluctuations that do not havephase shifts with depth. This lack of slopingisopleths suggests that, either (1) they are advectedinto the region rather than being induced by localsurface evaporation, runoff, winds, heating andcooling or (2) the temporal sampling (Xmonthly)is so coarse that is does not allow such slopes to bemeasured. Some long-period (low-frequency)trends are apparent. Below normal water tempera-tures before 1975 were followed by a rapidwarming centered in 1977. Though there werebrief below normal excursions, most of the 1980swere above normal followed by a cool periodbeginning in 1989.As determined by least-squares sinusoidal curve

fitting, the greatest amount of the variance wasfound for periods of the upper-layer temperatureanomalies of 5 years (containing less than 3% ofthe total variance) and 9 years (3%), withconsiderable variance contained in the interdeca-dal periods. In the lower layers (150–200m), a5-year period fluctuation (10% of the variance) isaccompanied by interdecadal fluctuations with aperiod of 19–20 years that accounts for about 30%of the variance.The pattern of salinity anomalies versus depth

(Fig. 13) has a similar banded vertical structure asdoes the temperature. There are alternating freshand salty periods. The greatest salty period (43)

Table 2

Correlations of monthly GAK1 water temperature and monthly upw

Depth (m) 601N, 1491W 601N, 1461W 571N, 1371

0 �0.19 (6) �0.20 (6) �0.20 (6)

10 �0.20 (6) �0.19 (6) �0.20 (6)

20 �0.19 (6) �0.18 (7) �0.22 (7)

30 �0.17 (7) �0.20 (8) 0.22 (7)

50 �0.19 (5) �0.22 (8) �0.21 (8)

75 �0.23 (5) �0.21 (9) �0.20 (8)

100 �0.22 (5) �0.21 (5) �0.23 (5)

150 �0.21 (4) �0.24 (3) �0.27 (3)

200 �0.21 (3) �0.24 (4) �0.26 (3)

250 �0.20 (3) �0.22 (4) �0.25 (4)

Note: Effective degrees of freedom are calculated using coherence

Numbers in parentheses are lags in months with positive values re

Bold ¼ 499% CL, regular ¼ (99%4CL495%) and blank ¼ CLo9

coincided with the early 1970s cold period.Clearly, hydrographic conditions were unusual inthe Gulf of Alaska. At that time, there was also the168-year record low air temperature at Sitka,Alaska (Parker et al., 1995).The major periodicities of salinity at the surface

are 4 and 9 years but only about 5% and 3%,respectively, of the variance is explained byfluctuations at these periods. ENSO is oftenassociated with the 4-year period and the 9-yearperiodicities (Allan et al., 1996, p. 37). At 250m,the 4-year period accounts for more than 8% ofthe variance while the fluctuations at 14–15 yearscan explain more than 10%. The mechanisms ofthe interactions of ENSO and salinity here areunclear.Correlative methods are employed to investigate

possible relationships between the temperatureand salinity anomalies and forcing functions ofwind (upwelling), freshwater discharge and South-ern Oscillation Index (SOI) (Tables 2–4). Upwel-ling index (Bakun, 1973) is used as a proxy for thelocal winds. As discussed earlier, Luick et al.(1987) compared the index at 601N, 1491W withreasonable results. Large-scale indices such asNOIx (Schwing et al., 2002) were not used in thisanalyses since a significant amount of its varia-bility is correlated with SOI (0.77), North PacificIndex (0.57) and the Pacific Decadal Oscillation(PDO) (�0.67). Also, correlations between NOIxand the temperature and salinity anomalies did not

elling index anomalies around the Gulf of Alaska

W 541N, 1341W 511N, 1311W 481N, 1251W

0.19 (63) 0.15 (63) 0.14 (57)

�0.15 (7) 0.14 (�6) 0.14 (53)

�0.14 (2) 0.14 (57) 0.15 (57)

�0.14 (2) 0.16 (57) 0.19 (57)

0.14 (57)

�0.17 (4)

�0.18 (2) �0.13 (11)

�0.20 (2)

�0.18 (11)

�0.17 (12)

and autocorrelative techniques (Emery and Thomson, 1998).

presenting upwelling index leading the temperature anomaly.

5%.

ARTICLE IN PRESS

Table 3

Correlations of monthly GAK1 salinity and upwelling index anomalies

Depth (m) 601N, 1491W 601N, 1461W 571N, 1371W 541N, 1341W 511N, 1311W 481N, 1251W

0 0.15 (6) 0.19 (45) 0.19 (45) 0.19 (45) �0.20 (19) �0.19 (19)

10 0.15 (45) 0.16 (8) 0.19 (8) 0.14 (44) �0.19 (57) �0.20 (57)

20 0.16 (�51) 0.17 (8) 0.20 (8) 0.16 (�51) �0.19 (22) �0.18 (57)

30 0.18 (�52) 0.16 (�52) 0.16 (�51) 0.22 (�51) 0.18 (�52) �0.16 (19)

50 �0.19 (33) �0.17 (33) 0.17 (0) 0.18 (�50) 0.20 (�5) 0.17 (�51)

75 0.21 (�37) �0.20 (22) 0.20 (�2) 0.17 (�1) 0.20 (�27) 0.15 (32)

100 0.16 (3) 0.15 (3) 0.17 (11) 0.18 (11)

150 0.19 (2) 0.18 (2) 0.13 (2) 0.15 (1) 0.16 (�27) 0.19 (11)

200 0.26 (2) 0.25 (2) 0.17 (2) 0.22 (3) 0.16 (2) 0.16 (2)

250 0.25 (2) 0.28 (2) 0.22 (2) 0.29 (2) 0.21 (2) 0.19 (2)

Note: Positive lags in ( ) indicate that upwelling index leads salinity. Bold are more than 99% significant. Regular face are less than

99% CL but greater than 95% CL. Blank are less than 95% CL.

Table 4

Correlations of monthly temperature (T) and salinity (S) anomalies with freshwater discharge at Seward (F), SOI and PDO

Depth (m) T=F T/SOI T/PDO S=F S/SOI S/PDO

0 0.30 (3) None 0.44 (2) �0.22 (1) 0.20 (46) None

�0.24 (22)

10 0.30 (1) None 0.45 (3) �0.30 (1) 0.23 (�27) �0.27 (2)

�0.23 (22) �0.27 (25)

�0.19 (32)

20 0.31 (1) None 0.47 (3) �0.34 (1) 0.24 (�17) �0.23 (�22)

�0.21 (23) �0.23 (2)

30 0.32 (1) �0.27 (7) 0.43 (3) �0.35 (1) �0.21 (39) �0.23 (3)

�0.18 (22)

50 0.34 (1) �0.30 (7) 0.35 (2) �0.34 (1) �0.24 (37) None

�0.28 (51)

75 0.33 (2) �0.36 (10) 0.43 (3) 0.21 (�11) 0.20 (6) �0.22 (�40)

�0.28 (1) 0.23 (18)

100 0.34 (3) �0.37 (8) 0.49 (2) 0.25 (�12) 0.24 (7) None

— �0.28 (1)

150 0.37 (2) �0.37 (8) 0.52 (2) 0.20 (�13) �0.21 (18) �0.22 (�1)

0.18 (1)

0.20 (30)

200 0.38 (3) �0.36 (8) 0.53 (2) 0.23 (�34) None �0.22 (�2)

0.24 (�19) 0.24 (28)

0.23 (30)

250 0.39 (3) �0.32 (10) 0.47 (4) 0.23 (�18) None �0.26 (0)

�0.18 (�2) 0.26 (29)

0.25 (30)

Note: Numbers in parentheses are lags in months with positive values representing freshwater discharge anomaly, SOI or PDO leading

the temperature or salinity anomaly. Bold ¼ 499% CL, regular ¼ (99%4CL495%) and blank ¼ CLo95%.


produce results that differed significantly fromthose presented in this paper. Correlations be-tween the temperature and salinity anomalies and

PDO are also included (Table 3), though PDO isnot a forcing mechanism but is instead a pattern ofSST in the North Pacific (Mantua et al., 1997).

ARTICLE IN PRESS


4.1. Upwelling index responses

Temperature anomalies are inversely well corre-lated (CL499%) with upwelling anomalies at601N, 1491W in the northern Gulf of Alaska at alldepths (except 30m) (Table 2). The effectivedegrees of freedom are determined using coherenceand autocorrelation properties of the data sets(Emery and Thomson, 1998; Chelton, 1983). Thelags in months of these maximum correlations areshown in parentheses with positive lags, indicatingthat upwelling index anomalies lead the tempera-ture anomalies. The negative correlation in thenorthern Gulf of Alaska can be attributed toincreased downwelling (negative upwelling), lead-ing to higher water temperatures throughout thewater column. Of course, this relationship is notfollowed during that portion of the year withtemperature inversions as discussed previously.The segregation of data with those conditions willbe the subject of a future paper. For the upper75m, ocean temperature anomalies lag the upwel-ling index anomalies with a 99% CL as far southas 571N, 1371W by 5–8 months. The delay couldbe due to changes in the alongshore advection.Increased downwelling with intense winter cyclo-nic activity might increase the advection of warmerwater from the south. This water would arriveduring the following summer or fall, 5–8 monthslater. South of 571N, the correlation drops and thesign at the surface is reversed. Farther south, thecorrelations are only significant at the 95% CL forthe upper 50m and their signs are reversed (withone exception). This is probably related to the 5year periodicity found in the upper-layer watertemperature anomalies in the northern Gulf ofAlaska.Below 75m, the linkages between tempera-

ture and upwelling index anomalies are as strongas those in the surface layers but their responsesare more rapid, 3–5 months (Table 2). Addition-ally, the temperature anomalies in the lowerlayer (100–250m) at GAK1 are most highlycorrelated with the upwelling indices at 571N,1371W, near Sitka, Alaska (Fig. 1), and the timelag is less. Either the deep layers acceleratemore rapidly than the surface layers or thealongshore thermal gradients are larger in the

deep layers or other processes are present in theupper layer.The relationships between salinity and upwelling

index anomalies add another dimension to thisanalysis since the direction of flow of the upwelled/downwelled water is generally parallel to the maxi-mum salinity gradients (i.e. offshore/onshore).Generally, salinity anomalies are significantlycorrelated with upwelling anomalies (Table 3). Inthe upper layers, the salinity leads or lags theupwelling index by 6–57, months making theirinterpretation tenuous. However, for the deeplayers (100–250m) the salinity anomaly lags theupwelling index anomaly by 2–3 months, verysimilar to the temperature responses. Concep-tually, increased downwelling leads to decreasedsalinity through the downward movement of low-salinity upper-layer water, though the thermalresponse to downwelling is more complex, asdiscussed earlier. Additionally, a depression ofthe isohaline (isopycnal) surfaces will acceleratethe alongshore baroclinic flow, bringing highertemperature water northward and westwardaround the perimeter of the Gulf of Alaska.Conversely, increased upwelling (decreased down-welling) will allow the excursion of higher-salinitywater onto the shelf and will elevate the nearshoreisohaline surfaces decreasing the baroclinic flow inthe upper layers. The upper 0–50m might beresponding to decreased downwelling (less stormi-ness) with an increased salinity through lesscontainment of low-salinity water at the coastand less runoff. South of 541N, increased upwel-ling off British Columbia (more anticyclonicactivity) accompanies decreased surface salinityin the northern Gulf of Alaska about 19 monthslater. The salinity anomalies in the deepest layers(200–250m) increase with increased upwelling,very similar to the response in the northern partof the Gulf of Alaska.

4.2. Freshwater responses

The variations of freshwater discharge with T

and S are examined (Table 4) along with SOI andPDO. As expected, increased freshwater dischargedecreases the upper-layer (0–100m) salinity anom-aly at GAK1 (Table 4). Changes in freshwater

ARTICLE IN PRESS


discharge anomaly lead the salinity changes by1 month, though there are significant correlations(499% CL) with freshwater leading salinity by22–23 months. In the deep layers (150–250m), thesalinity responds in the opposite sense, that is, thecorrelation coefficient is positive. Higher rates offreshwater discharge correlate with higher sali-nities in the deep waters about 30 months later.This time lag at 499% CL is present from 150mto the bottom. There are also highly signifi-cant correlations with salinity leading freshwater(negative lags) at 13, 18 and 19 months. Thecombination of the significant correlations at –18and +30 months suggests a time scale of about48 months, similar to the time scales that separatethe upwelling indices in southern Gulf of Alaskafrom the freshwater discharge responses in thenorthern gulf near Seward (Table 1).Surprisingly, the responses of the temperature

anomalies to freshwater discharge anomalies aremore definitive than those of salinity. Significantpositive correlations exist throughout the watercolumn, with the temperature lagging the fresh-water by 1–3 months (Table 4). With the exceptionof the surface temperature, the longer lags arefound in the deeper layers (75–250m). Elevatedtemperatures can accompany increased freshwaterdischarge through an accelerated alongshore flow,either with increased barotropic flows with in-creased sea-level gradient or increased baroclinicflows with changes in the density structure.What alongshore accelerations within the ACC

are necessary to produce the observed temperaturefluctuations? To determine this, the alongshorethermal gradients must be determined. Alongshoresynoptic data for the Gulf of Alaska are rare, butusing nearshore measurements over seven sectionseast of GAK1 to Yakutat, Alaska, an alongshorethermal gradient of 0.026 1Ckm�1 (increasingeastward) was determined on the 1025.8 isopycnallayer. This layer was selected because it was belowthose depths with significant seasonal temperaturesignals at GAK1. The hydrographic sections usedin this gradient determination were from July andOctober 1974, February and June 1975. Theselected isopycnal of 1025.8 was found at 200mat GAK1 and shoaled eastward. A shallowerdepth would have intersected with the seasonal

mixed layers and concurrent upstream densitydata at the coast do not exist for 250m. Salinitychanges on this density surface were an order ofmagnitude less than the temperature changes.Using this thermal gradient and assuming asteady-state thermal field and thermal gradient,to increase advectively, the temperature at GAK1below 100m by 1 1C (e.g. 1977), an increased flowof 0.05m s�1 westward is required.

4.3. Pacific Decadal Oscillation response

The PDO Index, the leading principal compo-nent of North Pacific monthly SST variability, hasthe strongest correlation with temperaturethroughout the water column (Table 4). TheGAK1 temperatures lag the PDO by 2–4 months,reflecting a delayed high-latitude response. It issurprising that the entire water column respondsto this pattern. Therefore, the influences of PDOare not restricted just to the sea surface in thiscoastal region, but they extend throughout thewater column. The correlations of salinity varia-bility with PDO are not as significant as thetemperature variability. Generally, the correla-tions are negative, suggesting that increases inPDO (higher SST) are associated with decreases insalinity. This is consistent with increased precipita-tion and runoff during warm periods. However,the sea-surface salinity does not have a significantcorrelation with PDO. At 10 and 20m, the PDOleads the decrease in salinity by 2 months. Thestrong PDO-water column provides evidence thatthe PDO reflects not only just the SST variabilitybut also the upper water-column heat content inthe northern Gulf of Alaska.

4.4. El Nino-Southern Oscillation response

As mentioned earlier, the water temperaturesmight be responding ENSO events of 1977,1982–3, 1987, 1992 and 1997–8 and a 5-yearperiodicity in ENSO is apparent. Correlationsbetween temperature and salinity at GAK1 andthe SOI (http://www.cpc.ncep.noaa.gov/data/in-dices/index.html) indicate that the upper-layer(0–20m) temperature anomalies are not signifi-cantly correlated with SOI (Table 4). The most

http://www.cpc.ncep.noaa.gov/data/indices/index.html

http://www.cpc.ncep.noaa.gov/data/indices/index.html

ARTICLE IN PRESS


significant correlations were for the 75–150mtemperature anomalies when they lagged SOI by8–10 months. The correlation coefficients arenegative, which is consistent with an ENSO event(negative SOI) relating to a positive tempera-ture anomaly. A pattern of 7–10-month lags wasfound throughout the water column except in the0–20m layer. Based on the transit time afterthe ENSO event, the speed of propagation fromthe equator would be 0.54–0.41m s�1. The salinityanomalies are not as well correlated with SOI andtheir lags do not have a consistent pattern andhave no significant correlations at 200 and 250m(Table 4).The elevated temperatures at GAK1 following

an ENSO event could be propagated througheither the ocean or atmosphere or both. Atmo-spheric forcing should be greatest in the upperlayers or in those portions of the water columnthat contain the thermocline since upwelling anddownwelling might play an important role. Anearlier study (Royer, 1982) emphasized the lack ofresponse of either the SST or 150m temperatureanomalies to local wind stress or wind-stress, curlbut that study failed to include upstream responsesto conditions in Southeast Alaska. It appears thatthe atmospheric teleconnections between the winds(upwelling indices) over the eastern Gulf ofAlaska, freshwater discharge, SOI and PDO ofthe eastern Gulf of Alaska are as significantas those atmospheric conditions near Seward(Table 5). Approximately 1 month prior to ENSO,there is likely to be increased downwelling(negative upwelling) at 571N and southward(Table 5). This is supported by Emery andHamilton (1985), who suggested that weak winter

Table 5

Correlations of upwelling index anomalies with various parameters

Location Freshwater discharge

601N, 1491W 0.16 (�25)

601N, 1461W �0.30 (1)

571N, 1371W �0.48 (1)

541N, 1341W �0.22 (1)

511N, 1311W 0.15 (12)

481N, 1251W �0.14 (22)

Note: Bold are greater than 99% CL. Positive lags indicate upwelling

atmospheric circulation prior to major ENSOevents is followed by a rapid strengthening of theatmospheric circulation, i.e. increased downwel-ling. Using the 150m temperature as a proxy forthe deep portion of the GAK1 water column, thisdownwelling could lead to elevated GAK1 150-mtemperature anomalies 4 months later (Table 2).However, since the sum of these phase shifts isonly 3 months rather than the most significantlycorrelated lag of 8 months (Table 4), evidence for alocal, atmospherically forced ENSO signal atGAK1 is lacking. Therefore, attention is directedto baroclinic coastal oceanic processes.Focusing on the temperature anomalies at 150m

(Fig. 14, upper panel), they contain the early 1970cold event followed by several warm events thatexceeded 1 S.D. (0.715 1C) (Fig. 14, dashed line) in1977, 1982–3, 1987, 1992–3 and 1997–8. Thisdepth was selected because it contained the largeamplitude temperature anomalies (warm) and it iswell removed from any local sea-surface warming.SOI during this period (Fig. 14, lower panel)contains minima that are associated with ENSOevents. These ENSO events coincide with tem-perature maxima at 150m approximately 8months later. The 1972 ENSO event is not evidentand it has been noted that it did not propagate tohigh latitudes (Enfield and Allen, 1980) and theevent for late 1978 could have been missed due toinadequate temporal hydrographic sampling. Thelast warm event (1997–8) was more than 2 S.D.above normal. It reached its maximum (1.60 1C) inFebruary 1998 and was back below 1 S.D. (theENSO threshold) by May. A cautionary note onthe temporal sampling is that since September1990, sampling has been approximately monthly

SOI PDO

None None

None None

0.15 (1) �0.27 (1)

0.18 (2) �0.27 (1)

0.21 (2) �0.31 (1)

0.19 (2) �0.25 (2)

index anomalies lead parameter.

ARTICLE IN PRESS

Fig. 14. Temperature anomalies at 150m (upper panel) with SOI (lower panel). One standard deviation is indicated with dashed line.

ENSO events are noted with vertical lines between panels.


rather than irregularly as was the case with the‘‘ship of opportunity’’ sampling from 1970 to1990. Therefore, ENSO events are better resolvedafter 1990.It is uncertain as to whether the ENSO signal

propagates only in the subsurface layers orwhether it propagates throughout the watercolumn but is masked by other forcing in theupper layers. However, an internal Kelvin wave isa possible mechanism for the propagation of thisthermal signal. ENSO signal propagation by aKelvin wave mechanism along the west coast toOregon was initially reported by Enfield and Allen(1980). More recently, Johnson and O’Brien (1990)discussed the movement of the 1982–3 ENSOevent from the equatorial region to about 501N.Their average Kelvin wave velocity was about0.4m s�1 that agrees with a travel time of about 10months. Additionally, this Kelvin wave speedagrees with a wave propagation of 0.40 m s�1

observed along the Oregon coast in 1982 (Huyerand Smith, 1985).

Applying the theory of nonlinear Kelvin wavedynamics after Allen and Hsieh (1997) to theKelvin wave bore, it is assumed that there are twolayers (100–150m) and (150–200m) with theGAK1 average respective densities of 1025.28and 1025.61 kgm�3. The calculated baroclinicwave speed is 0.40m s�1. Of course this is a localspeed and does not necessarily apply for the entireKelvin wave transit from the equator that dependson the stratification and water depth. The currentspeed associated with this bore is dependent on theamplitude of the interface displacement betweenthe two layers. If a 10m displacement is assumed,the current speed increase should be 0.056m s�1.As mentioned previously, an alongshore tempera-ture gradient of 26.3� 10�4 1Ckm�1 exists on the1025.8 kgm�3, decreasing westward. An increaseof 1.3 1C in the water temperature anomaly over0.4245 year as was observed in 1997–8 at 200m,would require a displacement of isotherms of494 km. This translates to an increase in thecoastal current of 0.037m s�1. Thus, the internal

ARTICLE IN PRESS


Kelvin wave speed and the required horizontaldisplacement are very similar. A more precisedetermination is impossible since the exact hor-izontal isopycnal temperature gradient is un-known, but this does suggest that thetemperature changes could be a result of theKelvin bore. This is consistent with Melsom et al.(1999), who found enhanced eddy generation inthe Gulf of Alaska under ENSO conditions due toincreased alongshore flows. In summary, thearrival time of the temperature signal at GAK1is consistent with Kelvin wave dynamics and thetemperature elevations are consistent with aKelvin wave bore. The 1977–8 arrival of thesubsurface ENSO signal is contrary to the Sub-botina et al. (2001), who reported that the pole-ward oceanic propagation of that ENSO signal itstopped near central California.

5. Linear trends in water properties

Did significant changes in the water propertiesat GAK1 occur during the period of theseobservations? To simplify this analysis, the two-layer water column (0–100 and 100–250m) isrevisited. Linear fits to the temperature andsalinity anomalies and the freshwater discharge

Table 6

Linear fits for 1970–2000 to anomalies of sea surface water properties (

freshwater and upwelling indices with F-statistic test

Parameter Slope

TA0 +0.0327 1Cyear�1

SA0 �0.0107 year�1

Freshwater +182m3 s�1 year�1

Upper layer T +0.0315 1Cyear�1

Lower layer T +0.0262 1Cyear�1

Upper layer S �0.00219 year�1

Lower layer S +0.00127 year�1

Upwelling indices

601N, 1491W +0.300m3 s�1 (100m)�1 year�1

601N, 1461W +0.665m3 s�1 (1 0 0)m�1 year�1

571N, 1371W �0.225m3 s�1 (100m)�1 year�1

541N, 1341W �0.173m3 s�1 (100m)�1 year�1

511N, 1311W �0.188m3 s�1 (100m)�1 year�1

481N, 1251W �0.441m3 s�1 (100m)�1 year�1

and upwelling indices (Table 6) reveal highlysignificant changes in the temperature and fresh-water anomalies. The upper-layer (0–100m) lineartemperature anomaly has a slope 0.032 1Cyear�1,that accounts for 14% of the variance with thetotal linear response over the period of thestudy of +0.94 1C. The lower layer has a slopeof 0.026 1Cyear�1, accounting for 15% of thevariance or about +0.78 1C for the study.As noted earlier here and by others elsewhere(Hare and Mantua, 2000), the pre-1978 tempera-ture behavior appears to be different thanthe 1978–2000 changes. Indeed, the slopes ofthe early portion of the record for the upper andlower layers were +0.269 and +0.249 1Cyear�1,respectively, an order of magnitude greater thanthe overall average. These linear trends areresponsible for more than 42% of the varianceand were 499% significant. Since 1978, thetemperature anomaly slopes have been 0.0182and 0.0112 1Cyear�1 for the upper and lowerlayers, respectively, accounting for 3–4% ofthe variance, though once again 499% signifi-cant. The average temperature anomaly changepre-1978 of about 1.5 1C was about five timesthe change since 1978. Thus, an analysis ofwater temperature changes in the Gulf ofAlaska using the 8 years of measurements from

TA0 and SA0), upper- and lower-layer temperature and salinity,

F Significance (%)

34.10 499

2.22 486

14.05 499

57.63 499

59.20 499

1.57 478

1.50 477

1.85 482

6.36 498

0.59 455

0.39 446

0.65 457

3.03 491

ARTICLE IN PRESS


1970 to 1978 would not reflect the more ‘‘typical’’variations in the temperature since that time.The salinity anomaly linear trends are in contrast

to the monotonic depth structure of the tempera-ture anomaly trends. The upper layer (0–100m) hasdiminishing salinities (�0.00219 year�1) while thelower-layer salinities are increasing at a slightlylower (+0.00127 year�1). The surface fresheningrate of –0.011 year�1 is very close to the rate ofsurface freshening along the British Columbia coastof about �0.009 year�1 (Freeland et al., 1997).These linear salt changes account for less than0.5% of variance and have lower significance levels(477%). However, it does suggest that since thesalinity trends in the upper and lower layers are ofopposite sign, so the stratification of the watercolumn will be increasing. Such changes instratification most likely occur primarily at thecoast so that the cross-shelf density gradient willincrease. The concept of an acceleration of thealongshore current is also supported by thesignificant (499%) increase in the freshwaterdischarge (182m3 s�1) during this period. Thisacceleration of the alongshore current will bringwarmer water from the south, producing higherwater temperatures. This warming supports thefeedback mechanism developed in Royer et al.(2001) that involves an increased freshwater dis-charge accelerating the alongshore flow and in-creasing the cyclonic activity in the Gulf of Alaska.These low-frequency temporal variations of

temperature and salinity are consistent with (1)increased freshwater discharge, (2) increasedalongshore circulation advecting warmer water atall depths and (3) enhanced estuarine-like flow thatdecreases the upper-layer salinity and increases thelower-layer salinity (offshore upper layer/onshorelower layer).The upwelling indices from 1971 to 2001 have a

trend of decreased downwelling along the northernboundary of the Gulf of Alaska, especially at 601N,1461W. However, off Southeast Alaska, the ten-dency is for increased downwelling (though statis-tically not significant). The decreased downwellingimmediately upstream of GAK1 will allow in-creased onshore flow of deep, relatively warm andsaline water, which is consistent with the watermass changes for the two layers outlined in Table 6.

6. Conclusions

Water-column temperature and salinity in thenorthern Gulf of Alaska respond to seasonalchanges of heat flux, wind and freshwater dis-charge. The responses of the water column to thesechanges are complicated by the high rates offreshwater discharge and relatively low watertemperatures. The maximum freshwater dischargein fall precedes the maximum downwelling windsin winter. Thus, the maximum vertical densitystructure is created about 3 months prior to thepeak alongshore wind stress. They combine todrive an intense coastal current. The freshwaterupper layer creates a system that can be consideredan estuarine analogy for this coastal system, aswas suggested by Tully and Barber (1960). Sincethe heat flux, wind and freshwater all contribute tothe density structure and their amplitudes andphases are not necessarily dependent on eachother, changes in their relative phasing arepossible. This could alter the timing and intensityof the vertical density distribution. This, in turn,could affect the primary production in the upperlayers through changes in the nutrient flux frombelow and the retention of phytoplankton in theeuphotic zone.Very significant changes in temperature and

salinity at GAK1 occur on a decadal basis. Thehydrographic record began during a relativelycold, dry period in the early 1970s. Rapid increasesin water temperature accompanied the 1977–8‘regime’ change that accompanied the 1976–7ENSO. This was immediately followed by a seriesof ENSO events in the 1980s. ENSO correlatesbest with temperatures of those portions of thewater removed from the upper layers, and theseresponses lag ENSO by 7–10 months. Deep-water(150m) temperature anomalies at GAK1 reflecteach of the ENSO events since 1977. These ENSOresponses appear to be an alongshore displace-ment of the coastal temperature gradient. Thesefairly brief responses, on the order of months, aresuperimposed on other long-term thermal changes.The relatively brief duration (on the order ofmonths) of the ENSO events and their GAK1responses are phenomena that require morefrequent sampling than was provided in most of

ARTICLE IN PRESS


this study. However, the 1997–8 ENSO event iswell resolved though the spring sampling intervalwas greater than 1 month. Since 2000, a fixedmooring has sampled at GAK1 so that thetemporal sampling interval has been muchreduced and future ENSO events should beresolved better. Additionally, the cross-shelfhydrographic responses to ENSO events fromOctober 1997 until December 2004 will bemeasured as part of the Northeast PacificGLOBEC program. The program consists of sixcruises per year with nutrient, primary productionand zooplankton sampling concurrent with thehydrographic sampling across the shelf to beyondthe shelf break.A weak link in the analysis of the wind forcing is

the uncertainty of the winds over this shelf.Chelton et al. (2004) have reported small-scalevariability in the global wind-stress curl since 1999using satellite scatterometer data and show thatthe Seward Line (GAK1) has a very strong curl.Comparisons with the NCEP reanalysis windsindicate a failure to resolve the small-scalevariability. Also, Ladd and Bond (2002) suggestthat the NCEP winds are inadequate near bound-aries. So, the upcoming GLOBEC analysis will usethe winds either from the scatterometer or themesoscale model five (MM5) winds (N. Bond,Pacific Marine Environmental Laboratory) in theatmospheric forcing study.While the similarity of the GAK1 surface

temperature and the PDO was expected, itwas surprising to find high correlation betweenPDO and temperature throughout the entirewater column. This suggests that the PDO canbe used to represent water column changeshere, not just the SST. Since the PDO-typewarm water at GAK1 also is found below theseasonal thermocline, it is likely that increasedalongshore coastal advection brings this warmwater to GAK1 from the east where it has beenbrought northward along the southeastern coast ofAlaska.Over the 30-year period of these observations,

there have been statistically significant increases inthe temperatures of both the upper and lowerlayers, along with an increase in the freshwaterdischarge. The increased freshwater is consistent

with enhanced alongshore advection. There hasbeen a less significant decrease in the upper-layer salinity and increase in the lower-layer salinity. These salinity changes are consistentwith the increased freshwater discharge. They alsoare consistent with an increase in upwelling(decreased downwelling) at a site 300 km east ofGAK1.The freshwater discharge model (Royer, 1982)

used in this analysis did not allow any net glacialablation or accumulation from 1931 to 1981.However, Arendt et al. (2002) estimate that forsome of the glacial fields in coastal Gulf of Alaska,there is a contribution of more than 50 km3 year�1

to the runoff from ablation. In the past decade,this ablation has increased to more than90 km3 year�1. New computations of thiscoastal freshwater discharge should be incorpo-rated along with the missing componentsfrom gauged rivers and the influxes from BritishColumbia and the coasts farther south. Theimpacts of these long-term changes on the hydro-graphic structure, circulation, nutrient fluxes andbiological productivity need to be investigatedfurther.

Acknowledgements

Appreciation is extended to all the scientists andcrew who have assisted in making the observationsthat comprise this data set, but especially thescientists and crew of R./V. ACONA and R./V.ALPHA HELIX. The personnel aboard theseships gathered the vast majority of the measure-ments and have made the occupation of thishydrographic station part of the seagoing ritual atthe University of Alaska. Captains Ken Turner,Mike Demchenko and Bill Rook are singled outfor their dedication to this sampling program.Geoffrey Pierson assisted with the programmingand data processing. C.E. Grosch assisted with hisdiscussions, some of the statistical comparisonsand his review comments. Thomas Weingartnerprovided valuable input and review. Commentsfrom the two reviewers and editor are appreciatedand served to clarify and greatly improve thispaper. These hydrographic data are available

ARTICLE IN PRESS


on-line at http://www.ims.alaska.edu:8000/GAK1/GAK1.dat. This program was supported throughthe Northeast Pacific program US GLOBEC(Contribution Number 479), through NOAA/CI-FAR NA67RJ0147 and NSF Grant OCE0100973.

References

Allan, R., Lindesay, J., Parker, D., 1996. El Nino-Southern

Oscillation & Climate Variability. CSIRO Publishing,

Collingwood, Vic., Australia (405pp).

Allen, S.E., Hsieh, W.W., 1997. How does the El Nino

generated coastal current propagate past the Mendocino

escarpment? Journal of Geophysical Research 102,

24,977–24,985.

Arendt, A.A., Echelmeyer, K.A., Harrison, W.D., Lingle, C.S.,

Valentine, V.B., 2002. Rapid wastage of Alaska glaciers and

their contribution to rising sea level. Science 297, 382–386.

Bakun, A., 1973. Coastal upwelling indices, west coast of North

America, 1946–71. US Department of Commerce, NOAA

Technical Report NMFS SSRF-671 (103pp).

Bryant, E., 1997. Climate Process and Change. Cambridge

University Press, Cambridge, UK (209pp).

Chelton, D.B., 1983. Effects of sampling error in statistical

estimation. Deep-Sea Research 30, 1083–1101.

Chelton, D.B., Schlax, M.G., Freilich, M.H., Milliff, R.F.,

2004. Satellite measurements reveal persistent small-scale

features in ocean winds. Science 303, 978–983.

Emery, W.J., Hamilton, K., 1985. Atmospheric forcing of

interannual variability in the northeast Pacific Ocean:

connections with El Nino. Journal of Geophysical Research

90, 857–868.

Emery, W.J., Thomson, R.E., 1998. Data Analysis Methods in

Physical Oceanography. Pergamon Press, Exeter, Great

Britain (634pp).

Enfield, D.B., Allen, J.S., 1980. On the structure and dynamics

of monthly mean sea level anomalies along the Pacific coast

of North and South America. Journal of Physical Oceano-

graphy 10, 557–578.

Freeland, H., Denman, K., Wong, C.S., Whitney, F., Jacques,

R., 1997. Evidence of change in the winter mixed layer in

the Northeast Pacific Ocean. Deep-Sea Research I 44,

2117–2129.

Hare, S.R., Mantua, N.J., 2000. Empirical evidence for North

Pacific regime shifts in 1977 and 1989. Progress in

Oceanography 47 (2–4), 103–146.

Huyer, A., Smith, R.L., 1985. The signature of El Nino off

Oregon, 1982–83. Journal of Geophysical Research 90,

7133–7142.

Johnson, M.A., O’Brien, J.J., 1990. The northeast Pacific Ocean

response to the 1982–1983 El Nino. Journal of Geophysical

Research-Oceans 95 (C5), 7155–7166.

Johnson, W.R., Royer, T.C., 1986. A comparison of two

current meters on a surface mooring. Deep-Sea Research 33,

1127–1138.

Ladd, C., Bond, N.A., 2002. Evaluation of the NCEP/NCAR

reanalysis in the NE Pacific and Bering Sea. Journal of

Geophysical Research 107, 3158.

Luick, J.L., Royer, T.C., Johnson, W.R., 1987. Coastal

atmospheric forcing in the Gulf of Alaska. Journal of

Geophysical Research 92, 3841–3848.

Mantua, N.J., Hare, S.R., Zhang, Y., Wallace, J.M., Francis,

R.C., 1997. A Pacific interdecadal climate oscillation with

impacts on salmon production. Bulletin of the American

Meteorological Society 78, 1069–1079.

Melsom, A., Meyers, S.D., Hurlburt, H.E., Metzger, E.J.,

O’Brien, J.J., 1999. ENSO effects on Gulf of Alaska eddies.

Earth Interactions, 3. Available online at: http://Earth-

Interactions.org.

Namias, J., 1968. The labile Gulf of Alaska cyclone—key to

large-scale weather modification elsewhere. In: Proceedings

of the International Conference on Cloud Physics, August

26–30, 1968, Toronto (11pp).

Parker, K.S., Royer, T.C., Deriso, R.B., 1995. High latitude

climate forcing and tidal mixing by the 18.6-year lunar

nodal cycle and low frequency recruitment trends in Pacific

halibut (Hippoglossas stenolepis). Canadian Special Publica-

tion on Fisheries and Aquatic Sciences 121, 447–459.

Pickard, G.L., Emery, W.J., 1990. Descriptive Physical Oceano-

graphy, An Introduction. Pergamon Press, Terrytown,

New York (320pp).

Royer, T.C., 1981. Baroclinic transport in the Gulf of Alaska.

Part II. Fresh water driven Alaska Coastal Current. Journal

of Marine Research 39, 251–266.

Royer, T.C., 1982. Coastal fresh water discharge in the

Northeast Pacific. Journal of Geophysical Research 87,

2017–2021.

Royer, T.C., Grosch, C.E., Mysak, L.A., 2001. Interdecadal

variability of northeast Pacific coastal freshwater and its

implications on biological productivity. Progress in Oceano-

graphy 49, 95–111.

Schumacher, J.D., Reed, R.K., 1980. Coastal flow in the

northwest Gulf of Alaska: the Kenai Current. Journal of


Sarkar, N., 2001. Hydrographic variability on decadal and

interdecadal scales in the northern Gulf of Alaska. Masters

Thesis, Old Dominion University (56pp.).

Schwing, F.B., O’Farrell, M., Steger, J., Baltz, K., 1996.

Coastal upwelling indices, west coast of North America

1946–1995. US Department of Commerce, NOAA-TM-

NMFS-SWFSC-231 (207pp).

Schwing, F.B., Murphee, T., Green, P.M., 2002. The northern

oscillation index: a new climate index for the Northeast

Pacific. Progress in Oceanography 53, 115–139.

Subbotina, M.M., Thomson, R.E., Rabinovich, A.B., 2001.

Spectral characteristics of sea level variability along the west

coast of North America during the 1982–83 and 1997–98 El

Nino events. Progress in Oceanography 49, 353–372.

Tully, J.P., Barber, F.G., 1960. An estuarine analogy in the

subarctic Pacific Ocean. Journal of the Fisheries Research

Board of Canada 17, 91–112.

http://www.ims.alaska.edu:8000/GAK1/GAK1.dat

http://www.ims.alaska.edu:8000/GAK1/GAK1.dat

http://EarthInteractions.org

http://EarthInteractions.org

ARTICLE IN PRESS


Whittaker, L.M., Horn, L.H., 1982. Atlas of Northern Hemi-

sphere Extratropical Cyclone Activity, 1958–1977. Depart-

ment of Meteorology, University of Wisconsin, Madison,

WI (65pp).

Wilson, J.G., Overland, J.E., 1986. Meteorology. In: Hood,

D.W., Zimmerman, S.T. (Eds.), The Gulf of Alaska

Physical Environment and Biological Resources. OCS

Study MMS86–0095. Mineral Management Service, Spring-

field, VA, pp. 31–54.

Xiong, Q., Royer, T.C., 1984. Coastal temperature and salinity

in the northern Gulf of Alaska, 1970–1983. Journal of


Hydrographic responses at a coastal site in the northern Gulf of Alaska to seasonal and interannual...

Documents

Transcript of Hydrographic responses at a coastal site in the northern Gulf of Alaska to seasonal and interannual...