Multidecadal trends in North American Atlantic salmon ... · Multidecadal trends in North American...

Multidecadal trends in North American Atlanticsalmon (Salmo salar) stocks and climate trendsrelevant to juvenile survival

K.D. Friedland, D.G. Reddin, J.R. McMenemy, and K.F. Drinkwater

Abstract: Landings of North American Atlantic salmon (Salmo salar) over the past century show multidecadalpatterns, which most recently characterize unprecedented declines in abundance. Stock size is compared with seasurface temperature (SST) data in the marine nurseries of post-smolt Atlantic salmon. A previously described correla-tion between stock abundance and winter SST conditions was again documented; however, of more relevance to thesurvival of salmon post-smolts, a correlation was also observed between abundance and spring SST in the Gulf ofSt. Lawrence. The relevance of the winter SST correlation was further investigated by considering winter conditions inthe freshwater nurseries as a factor causing elevated overwintering mortality of pre-migrant parr. The salmon abundancetime series was compared with air temperature and rainfall trends averaged over time and space. Air temperature andrainfall do not appear to be significant environmental variables in shaping salmon recruitment. The timing of smoltruns appears to be out of synchronization with ocean conditions in the post-smolt nursery areas. The relationshipbetween marine and freshwater impacts may change with changing climate conditions. Persistent positive phase forcingin the North Atlantic Oscillation raises the concern that recent declines in Atlantic salmon are, in part, due to globalclimate change.

Résumé : Les débarquements de saumons de l’Atlantique (Salmo salar) en Amérique du Nord au cours du siècle der-nier suivent des patterns qui s’étendent sur plusieurs décennies et qui tout récemment affichent des déclins del’abondance jamais encore enregistrés. La taille des stocks est examinée en fonction de la température de surface de lamer (SST) dans les zones d’engraissement des saumons de l’Atlantique (ayant terminé le stade de saumoneau). Unerelation décrite antérieurement entre l’abondance du stock et les conditions de la SST en hiver est confirmée; cepen-dant, une corrélation plus pertinente pour la survie des saumons après le stade de saumoneau a été observée entrel’abondance et la SST du printemps dans le golfe du Saint-Laurent. La pertinence de la corrélation impliquant la SSTd’hiver a été étudiée plus à fond en considérant les conditions hivernales dans les zones d’engraissement d’eau doucecomme facteur de mortalité accrue des tacons en hiver avant leur migration. La série chronologique des abondances desaumons a été mise en relation avec les tendances moyennes dans le temps et l’espace de la température de l’air et desprécipitations. Les températures de l’air et les précipitations ne semblent pas être des variables significatives qui affec-tent le recrutement du saumon. La phénologie de la descente des saumoneaux semble être déphasée par rapport auxconditions océaniques dans les zones d’engraissement des saumons qui ont terminé le stade de saumoneau. Les rela-tions entre les impacts marins et d’eau douce risquent de changer avec la modification des conditions climatiques. Leforçage de phase positif qui se maintient dans l’oscillation de l’Atlantique nord fait craindre que les déclins récentsdans l’abondance du saumon de l’Atlantique soient dus, en partie, aux changement climatique global.

[Traduit par la Rédaction] Friedland et al. 583

Introduction

Questions about the recruitment of Atlantic salmon (Salmosalar) in North America remain enigmatic, especially in re-gard to the relationship between events in fresh water and inthe sea. Although there are concerns about the impact that

climate can ultimately have on recruitment, there are fewdata to link recruitment of these stocks to broad-scale cli-mate forcing. Thus, the dramatic decline in stocks reproduc-ing at the southern end of the range in North America, asevidenced by the recent endangered listings of salmon popu-lations in Maine, U.S.A., and the Bay of Fundy, Canada, re-

Can. J. Fish. Aquat. Sci. 60: 563–583 (2003) doi: 10.1139/F03-047 © 2003 NRC Canada

563

Received 3 March 2003. Accepted 31 March 2003. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on17 June 2003.J17371

K.D. Friedland.1 UMass/NOAA CMER Program, Blaisdell House, University of Massachusetts, Amherst, MA 01003, U.S.A.D.G. Reddin. Science Branch, Department of Fisheries and Oceans, P.O. Box 5667, St. John’s, NL A1C 5X1, Canada.J.R. McMenemy. Vermont Department of Fish and Wildlife, 100 Mineral Street, Suite 302, Springfield, VT 05156-3168, U.S.A.K.F. Drinkwater. Department of Fisheries and Oceans, Ocean Sciences Division, Bedford Institute of Oceanography,P.O. Box 1006, Dartmouth, NS B2Y 4A2, Canada.

1Corresponding author (e-mail: [email protected]).

J:\cjfas\cjfas60\cjfas6005\F03-047.vpJune 12, 2003 6:57:10 AM

Color profile: Generic CMYK printer profileComposite Default screen

mains largely unexplained (Anderson et al. 2000). And thoughnot as severe, there is also evidence of a decline in salmonin the core distribution areas of the North American stockcomplex, which is also largely unexplained and which com-plicates the imposition of management restrictions (Chaputet al. 1998). Where climate-impact studies are typically asearch for prediction models of the survival rates withinpopulations, analyzing the impact of climate on Atlanticsalmon is evolving into a search for survival rates amongpopulations.

Climate studies on the marine phase of Atlantic salmonhave focused on ocean conditions and their consequences forpost-smolt survival. Run reconstructions of large stock com-plexes and experiments with individual stocks show thatmortality is very high during the first year at sea. Post-smoltmortality has been shown to exert a more profound effect onthe pattern of recruitment than mortality in fresh water(Chadwick 1987). Studies done thus far on marine condi-tions and North American stocks have focused on winter cli-mate signals. Reddin (1988) and Ritter (1989) found thatsurvival rates could be related to thermal conditions duringthe winter of the post-smolt year. The winter correlationswere hypothesized to be an indication that post-smolt migra-tion patterns were changing in response to temperature, andas a result, predation pressure varied (Ritter 1989). Reddinand Friedland (1993) computed thermal habitat for winter byrelating catch rate at temperature to the temperature fields ofthe Northwest Atlantic and found that the winter thermalhabitat index and North American stock size were positivelycorrelated. Friedland et al. (1993) systematically searchedfor correlates between a stock index based on landings andthermal habitat for the months and temperature ranges rele-vant to post-smolts and reported that thermal habitat varia-tion in winter was the strongest correlate. However, usingstock size as a proxy for survival, it is not obvious what sortof survival mechanisms are at work. We are faced with thedilemma that Atlantic salmon are quite large by their firstwinter and have outgrown vulnerability to most predatorsthat affect them earlier in the post-smolt year (Allen et al.1972). The absence of a spring correlate, indicative of theocean conditions when the fish first enter the marine envi-ronment, is puzzling.

The concept of a “critical period” for salmonids is mostcommonly applied to the factors affecting juveniles whenthey first enter the marine environment. The fish are chal-lenged by a wide array of potential sources of mortality;however, most fish are believed to be lost to predation dur-ing the first weeks in the ocean. There is a rich body of liter-ature supporting this contention from the Pacific (Fisher andPearcy 1988; Holtby et al. 1990) and Baltic (Eriksson 1994;Salminen et al. 1995). Climate analyses with Atlantic salmonin European waters suggest that Atlantic salmon populationsbehave much the same way. Friedland et al. (1998a) ob-served that thermal regimes during the first weeks at sea arecorrelated with survival patterns of two index stocks, onemigrating from southern Norway, the other from westernScotland. In years when warm thermal regimes invaded theNorth Sea and southern Norwegian coast in a time frame co-incident with the post-smolt migration, survival was good.Alternatively, when cold conditions persisted in the regionduring the migration period, survival was poor. Climate me-

diation of survival was suggested by the environmental cor-relate; however, the nature of the survival mechanism wasfurther elucidated with growth data for one of the two indexstocks. Friedland et al. (2000) reported that post-smolt growthfollowed a pattern similar to the time series of survival andclimate variation. When warm sea surface temperatures (SSTs)were present as post-smolts began their ocean migrations,survival was higher and post-smolt growth was also corre-lated. Thus, post-smolt survival of European Atlantic salmonappears to be influenced by the same mechanisms hypothe-sized for a wide range of marine species, better growth dur-ing a critical period is associated with the ability to escapepredation (Anderson 1988; Pepin 1991).

Post-smolt growth data for North American salmon isequivocal in its support of the hypothesis that growth andsize influence post-smolt mortality. Studies on hatchery fishshow that differences in survival rates between stocks can berelated to post-smolt growth (Friedland et al. 1996); how-ever, no evidence has yet been presented to show that inter-annual variation in survival is related to post-smolt growthfor North American stocks (Friedland 1998). This may berelated to the selection of the stocks analyzed thus far; per-haps the analysis of wild stocks over a wide geographicrange will shed further light on this issue.

Climate studies on the freshwater phase of Atlantic salmonhave mainly focused on river conditions that control themortality and growth of parr over the course of a number ofgrowing seasons. Ghent and Hanna (1999) revisited Elson’shypothesis relating rainfall and water levels with bird preda-tion in freshwater nurseries. Minns et al. (1995) consideredthe effect of temperature change on the extent of availableparr habitat and the ramification for juvenile production,predicting changes in smolt age distributions. Arndt et al.(2002) recently examined the impact of summer precipita-tion on the feeding and growth of parr and found popula-tions were resilient to rainfall events. However, during therecent decline in stock abundance of North American salmon,conservation measures have increased the population sizesof salmon juveniles in fresh water, lessening our concernsover climate factors in freshwater environments affecting re-cruitment (Anonymous 2002; Swansburg et al. 2002). Ofgreater interest is the impact of ice stability on juvenile habi-tats and mortality of pre-migrant parr the winter before theysmoltify and go to sea (Cunjak et al. 1998; Whalen et al.1999a; Prowse and Beltaos 2002). Freshwater censuses ofjuvenile populations are often conducted before winter andthus before any winter mortality on pre-migrant parr. Deteri-oration of conditions in the freshwater nursery during thewinter before migration could result in lower smolt produc-tion than expected from the census numbers the summer be-fore. Thus, climate impacts on parr populations during wintercould potentially control recruitment; furthermore, a climateeffect in fresh water during winter would establish a linkageto the previously observed correlates relating abundance toclimate conditions in the sea during winter.

The migration of smolts from freshwater nurseries is gov-erned by physiology and environmental cues. Freshwater parrare primed by photoperiod to undergo physiological changesto transform into marine adapted smolts; the migration isinitiated by common seasonal cues (Whalen et al. 1999b).Obviously, photoperiod is relatively constant year to year,

© 2003 NRC Canada

564 Can. J. Fish. Aquat. Sci. Vol. 60, 2003



thus our focus on environmental timing of migration mustbe on other factors. Whalen et al. (1999b) concluded thattemperature is the primary trigger for smolt migration andhas been most widely associated with a 10°C threshold. Wa-ter flow has been associated with the initiation of somesmolt migrations, but these examples are often in environ-mental settings lacking temperature contrast (Hesthagen andGarnås 1986). As Gargett et al. (2001) suggests, the mostuseful mechanistic hypotheses underlying climate-forcedpatterns in salmonid recruitment may be related to the tim-ing of salmon entry into the marine environment. It is notclear whether the smolt migration has remained in synchro-

nization with early ocean conditions for North Americanstocks of Atlantic salmon.

Our goal with this investigation is to examine three sur-vival control points: overwintering mortality of parr the win-ter before migration to sea, the synchronization of smoltmigration and marine conditions, and conditions related tothe first weeks to months at sea for post-smolts. We will ap-ply the newly released enhanced version of the Comprehen-sive Ocean–Atmosphere Data Set (COADS) to see if furtherinsights on marine conditions and post-smolt survival can beachieved. Additionally, data sets associated with freshwaterrearing areas, specifically air temperature and rainfall, will

© 2003 NRC Canada

Friedland et al. 565

Fig. 1. Map showing five marine habitat or sea surface temperature index areas for Atlantic salmon (Salmo salar).

Fig. 2. Map showing location, indicated by plus sign (+), of air temperature and precipitation observing stations. Number associatedwith “+” is the station number.



© 2003 NRC Canada


Mea

nm

onth

lyai

rte

mpe

ratu

reD

aily

high

air

tem

pera

ture

Mea

nm

onth

lyra

infa

ll

Num

ber

Nam

eL

ongi

tude

,°W

Lat

itud

e,°N

Cou

ntry

Pro

vinc

eor

stat

eS

tart

End

Sta

rtE

ndS

tart

End

1A

roos

took

–67.

7246

.80

Can

ada

New

Bru

nsw

ick

1914

1996

2B

agot

vill

e–7

1.00

48.3

3C

anad

aN

ova

Sco

tia

1908

1996

3C

artw

righ

t–5

7.03

53.7

0C

anad

aN

ewfo

undl

and

1935

1996

1935

1996

4C

harl

otte

tow

n–6

3.13

46.2

8C

anad

aP

.E.I

.19

0819

9619

0819

925

Cha

tham

–65.

4547

.02

Can

ada

New

Bru

nsw

ick

1908

1996

1908

1996

1908

1996

6D

eer

Lak

e–5

7.40

49.2

2C

anad

aN

ewfo

undl

and

1934

1996

1934

1996

7E

astp

ort

–67.

0044

.92

U.S

.M

aine

1908

1996

1908

1996

8F

rede

rict

on–6

6.53

45.8

7C

anad

aN

ewB

runs

wic

k19

0819

9619

0819

9619

1419

969

Gas

pé–6

4.48

48.7

7C

anad

aQ

uebe

c19

1619

9619

1619

9610

Goo

seB

ay–6

0.42

53.3

2C

anad

aL

abra

dor

1942

1996

1942

1996

11L

aT

uque

–72.

7847

.40

Can

ada

Que

bec

1912

1996

1912

1996

12M

onct

on–6

4.68

46.1

2C

anad

aN

ewB

runs

wic

k19

0819

9619

0819

9619

0819

9613

Mt.

Joli

–68.

2248

.60

Can

ada

Que

bec

1908

1996

14N

appa

n–6

4.25

45.7

7C

anad

aN

ova

Sco

tia

1913

1996

1913

1996

15N

atas

hqua

n–6

1.82

50.1

8C

anad

aQ

uebe

c19

1519

9619

1519

9616

Por

tlan

d–7

0.30

43.6

6U

.S.

Mai

ne19

0819

9617

Pre

sque

Isle

–68.

0046

.66

U.S

.M

aine

1909

1996

18Q

uébe

c–7

1.38

46.8

0C

anad

aQ

uebe

c19

0819

9519

0819

9519

Sab

leIs

land

–60.

0243

.93

Can

ada

Nov

aS

coti

a19

0819

9519

0819

9520

Sai

ntJo

hn–6

5.88

45.3

2C

anad

aN

ewB

runs

wic

k19

0819

9619

0819

9621

St.

Ant

hony

–55.

6351

.37

Can

ada

New

foun

dlan

d19

0819

9519

0819

9522

St.

John

’s–5

2.73

47.6

2C

anad

aN

ewfo

undl

and

1908

1996

1935

1996

23S

usse

x–6

5.53

45.7

2C

anad

aN

ewB

runs

wic

k19

0819

9619

0819

9624

Syd

ney

–60.

0546

.17

Can

ada

Nov

aS

coti

a19

0819

9619

0819

9625

Yar

mou

th–6

6.08

43.8

3C

anad

aN

ova

Sco

tia

1908

1996

1908

1996

Not

e:N

umbe

ris

used

onm

apan

dfi

gure

sto

iden

tify

stat

ions

.T

hest

art

and

end

year

ofav

aila

ble

data

are

show

nfo

rai

rte

mpe

ratu

re,

prec

ipita

tion,

and

rain

fall

data

.

Tab

le1.

Lan

d-ba

sed

obse

rvin

gst

atio

nsin

Nor

thA

mer

ica

used

inan

alys

isof

salm

onpo

pula

tion

tren

ds.



be examined in regard to overwintering conditions and theinitiation of smolt migrations.

Methods

Our primary method is that of exploratory correlation anal-ysis. It is predicated on the use of proxy data to representNorth American Atlantic salmon stock complex size and en-vironmental conditions in the freshwater nursery during win-ter and spring and in the ocean nursery for post-smolts. Wewill be using an index based on landings of salmon through-out North America spanning a relatively long time period,viz. fishery years 1910–1997; regrettably, it is not possibleto apply the methods of salmon run reconstruction used byassessment working groups because of the limitations of the

historical data (Anonymous 2002). Therefore, we are relianton the landings time series as a proxy variable for stock sizeand survival at key life-history stages. We will, however,draw on the assessment estimates of stock abundance in acomparative way to evaluate the landings index. Likewise,relevant environmental conditions in salmon nursery zoneshave not been specifically measured, thus we use SST datato represent conditions in marine nursery habitats and air tem-perature and rainfall data from land observing stations to repre-sent environmental conditions in freshwater nursery habitats.

Stock size of the North American Atlantic salmon stockcomplex

The North American stock complex was represented bythe time series of Canadian and U.S. landings from 1910 to

© 2003 NRC Canada


Fig. 3. Mean monthly temperature by latitude for each observing station averaged over the study period 1910–1996.

SST index area

1 2 3 4 5

Aroostook Bagotville Charlottetown Sydney St. AnthonyChatham Eastport Nappan St. John’sFredericton LaTuqueGaspé PortlandMoncton QuébecMt. Joli Saint JohnNatashquan YarmouthPresque IsleSussex

Table 2. Observing stations associated with sea surface temperature (SST) index areas used inthe synchronization of smolt migration and ocean conditions analysis.



1997 amended by the West Greenland catch of North Amer-ican Atlantic salmon (May and Lear 1971; Anonymous2002) using the approach outlined in Friedland et al. (1993).We will refer to this as the North American Atlantic salmonlandings index. Most of the catch at West Greenland is ofone sea-winter (1SW) salmon and is partitioned by continentof origin (Reddin and Friedland 1999) and sea age. Beforethe development of the Greenland fishery in the 1960s, thesefish would have returned to home waters the following yearas 2SW fish and would have either been caught in fisheriesor returned to fresh water and spawned. Though there are3SW salmon and repeat spawners in the North Americanstock complex, most fish are caught or spawn at either 1SWor 2SW ages. To create equivalency over the time series, weused the same methodology as Friedland et al. (1993) andadded the Greenland catch to the time series as an estimateof 2SW catch likely to have been taken in Canada in lieu ofthe 1SW catch in Greenland.

Thermal conditions in the marine nurseryTime series of SST were derived from the COADS data

set. These data are assembled from ship reports, research ves-sels, buoys, and other devices and are summarized asmonthly means averaged by 2° × 2° square latitude–longi-tude boxes (Woodruff et al. 1998). Their compilation pro-vides a depiction of SST worldwide. We used the data fromrelease 1a/1b/1c, also referred to as the enhanced version.Mean monthly SSTs were calculated for five index areas

© 2003 NRC Canada


Fig. 4. (a) North American Atlantic salmon landings index; thebold line represents five-point moving average filter smoothing.(b) North American Atlantic salmon landings index (solid line)and the estimate of one and two sea-winter pre-fishery stockabundance (PFA; broken line) for North American stocks.

Are

a1

Are

a2

Are

a3

Are

a4

Are

a5

Lag

1L

ag2

Lag

1L

ag2

Lag

1L

ag2

Lag

1L

ag2

Lag

1L

ag2

Mon

thr

Nr

Nr

Nr

Nr

Nr

Nr

Nr

Nr

Nr

NJa

n.–0

.14

57–0

.14

58–0

.34

87–0

.29

87–0

.23

87–0

.27

87–0

.25

87–0

.27

87–0

.09

84–0

.08

85F

eb.

–0.0

943

–0.1

944

–0.4

687

*–0

.49

87*

–0.2

388

–0.3

088

–0.1

587

–0.1

687

–0.0

582

–0.0

683

Mar

.0.

1948

0.20

49–0

.33

87–0

.31

87–0

.22

88–0

.23

88–0

.02

85–0

.03

85–0

.13

79–0

.08

80A

pril

–0.2

574

–0.1

675

–0.2

488

–0.2

188

–0.1

687

–0.1

587

–0.1

987

–0.1

987

–0.1

082

–0.1

083

May

–0.4

084

*–0

.34

84–0

.14

88–0

.12

88–0

.10

88–0

.09

88–0

.24

88–0

.20

88–0

.04

870.

0088

June

–0.2

883

–0.1

783

–0.0

588

–0.0

788

0.10

880.

1088

–0.1

187

–0.0

688

0.05

880.

1188

July

–0.2

484

–0.1

984

–0.0

188

–0.0

388

0.06

880.

0588

–0.0

688

–0.0

588

0.07

880.

1288

Aug

.–0

.27

86–0

.23

860.

0187

0.01

87–0

.15

88–0

.18

88–0

.04

88–0

.07

88–0

.06

880.

0388

Sep

t.–0

.36

85–0

.31

86–0

.03

880.

0488

–0.1

088

–0.1

288

–0.0

888

–0.1

188

0.03

870.

0787

Oct

.–0

.25

87–0

.12

87–0

.15

88–0

.12

88–0

.21

88–0

.18

88–0

.06

880.

0388

0.02

880.

1388

Nov

.–0

.38

83–0

.26

84–0

.42

88–0

.34

88–0

.27

88–0

.14

88–0

.24

88–0

.11

88–0

.02

870.

0987

Dec

.0.

0068

–0.0

169

–0.4

186

–0.3

486

–0.2

287

–0.1

887

–0.1

888

–0.1

388

0.09

880.

1188

Not

e:Si

gnif

ican

ce:

*,p

=0.

05.

Tab

le3.

Cor

rela

tion

betw

een

Nor

thA

mer

ican

Atl

anti

csa

lmon

(Sal

mo

sala

r)la

ndin

gsin

dex

and

sea

surf

ace

tem

pera

ture

(SS

T)

for

five

SS

Tin

dex

area

s,by

mon

th,

at1

year

befo

reca

tch

year

(lag

1)an

d2

year

sbe

fore

catc

hye

ar(l

ag2)

.



(Fig. 1) meant to represent the various habitats that salmonpost-smolts utilize in the marine environment. Though ourprimary interest is the time period in which post-smolt areusing these habitats, i.e., spring to fall, or from emigrationfrom fresh water to emigration to the overwintering areas,we calculated monthly mean temperatures for the full calen-dar year. The time series of mean SSTs were correlated withthe North American salmon landings index at two time lagsusing Pearson’s product-moment correlation. Because mostof the landings were of 1SW and 2SW fish, the landings in-dex was compared with the two years that the respectivesmolt cohorts would have been post-smolts; year i landingswere compared with the environmental conditions (SST) inyear i – 1 with respect to the 1SW component (lag 1) and toyear i – 2 with respect to the 2SW component (lag 2).

Thermal and rainfall conditions in the freshwater nurseryMonthly means of air temperature and rainfall, together

with daily maximum air temperatures, were assembled for arepresentative group of observing stations across the range offreshwater salmon habitat in North America (Fig. 2). The 25stations included 22 from Canada and three from the UnitedStates (Table 1). With the exception of Labrador, all stationshad temperature data for the overwintering seasons and smoltmigration years 1908–1996. The analysis of daily air tem-perature was restricted to four stations with complete dataand representative of the primary salmon producing areas.Rainfall data were only available for the Canadian stations.The numbering system in Table 1 is simply to facilitate la-beling in the station map.

Air temperatures for Canadian stations were extracted froma database of homogenized, long-term temperature time se-ries. Missing values were estimated using highly correlatedneighboring stations and by joining short-term station seg-ments to create long-term series. Using a recently developedtechnique based on regression models, Vincent (1998) iden-tified inhomogeneities in the temperature series, which areoften nonclimatic in origin and are due to station alterations,including changes in site exposure, location, instrumenta-tion, observer, observing program, or a combination of theabove. Monthly adjustments were derived from the regres-sion models, and adjustments were applied to bring each ho-mogeneous segment into agreement with the most recenthomogenous part of the series (Vincent and Gullett 1999).

A similar effort was undertaken to create a long-term, ho-mogenous database of precipitation data for Canadian sta-tions (Mekis and Hogg 1999). Adjustments were applied tothe daily level for rain and snow separately. For each rain-gauge type, corrections to account for wind undercatch andevaporation were implemented. Gauge-specific wetting losscorrections were also applied for each rainfall event. Over-lapping periods were used to minimize possible inhomo-geneities.

Air temperature and precipitation for stations in the UnitedStates were derived from the U.S. Historical ClimatologyNetwork (Karl et al. 1990), which is a high-quality data setof monthly averaged maximum, minimum, and mean tem-perature and total monthly precipitation developed to assistin the detection of regional climate change.

The analyses of air temperature and rainfall were limitedto winter months as identified by the distribution of mean

monthly air temperatures over the study period. Meanmonthly air temperatures are below freezing for nearly allobserving stations for the months of December throughMarch (Fig. 3). The only exceptions are December andMarch means for the most southerly stations. November andApril monthly means were below freezing for only the north-ern most stations in Newfoundland and Labrador. Becausethe main salmon producing areas are largely contained be-tween 44 and 50°N, it was decided that December throughMarch constituted the winter and subsequent analyses willbe restricted to these months.

Air temperature and rainfall trends were compared withthe North American Atlantic salmon landings index by bothindividual observing stations and by spatially and temporallycombined data within a winter season. The time series ofstation air temperatures and rainfall were correlated with theNorth American salmon landings index at two time lags(same rationale as used with the SST data) using Pearson’sproduct–moment correlation. December data were lagged 2and 3 years to make them contiguous with the correct winterseason. Spatial combinations of the data were guided by aprincipal component analysis of the winter data. Once spa-tial groups of stations were decided, means and first princi-pal component factor scores were correlated with the NorthAmerican Atlantic salmon landings index at the previouslydescribed time lags using Pearson’s product–moment corre-

© 2003 NRC Canada


Fig. 5. (a) North American Atlantic salmon landings index,(b) February mean sea surface temperature (SST) in area 2, and(c) May SST in area 1. Bold lines represent five-point movingaverage filters.



lation for monthly, bimonthly, and seasonal combinations ofthe data.

Daily maximum air temperature for selected stations wasanalyzed by temperature ranges indicative of the potentialfor thaw of nursery stream ice. Monthly counts of the num-ber of days that air temperature exceeded 1°C and 4°C and,by difference, the number of days between 1 and 4°C werecorrelated with the North American salmon landings indexusing Pearson’s product–moment correlation and also apply-ing the same lagging scheme used for the mean air tempera-ture, precipitation data, and rainfall.

Synchronization of smolt migration and ocean conditionsThe synchronization of smolt migration and ocean condi-

tions in post-smolt nursery habitats was evaluated by compar-ing the year-day of putative thermal conditions in freshwaternurseries that would trigger smolt migration versus the year-day of target SST conditions at first entry into the marine en-vironment. This measure is intended to provide an index ofthe fitness of smolt migration mechanisms in respect to themarine environment affecting post-smolt survival. The anal-ysis was based on the SST index areas used in the SST cor-relations (Fig. 1). Land-based air temperature observingstations were partitioned based on their representation of thedrainages emptying into the respective SST areas (Table 2).Within each area, a regression was fit to predict year-day ofoccurrence of spring mean monthly air temperature based onthe data from the member stations. Each annual regressionwas used to estimate the year-day for target air temperaturestested over a range of 7 to 13°C. The same procedure was

used with the SST data. SST data for each area was used todevelop regressions predicting year-day of spring SST; therange of target SST was 4–10°C, which was 3°C lower thanthe target air temperatures. The difference between the airtemperature target year-day and the SST target year-day,i.e., the number of days, was correlated over the full rangeof permutations between target air temperature and targetSST using Pearson’s product–moment correlation and alsoapplying the same lagging scheme used for the other datatypes.

Climate indicesTwo climate indices were compared with the North Amer-

ican Atlantic salmon landings index. First, the Gulf StreamNorth Wall (GSNW) indices, which are indicative of the dis-tribution of SSTs in the Northwest Atlantic, were correlatedwith the landings index. The GSNW indices are based on theanalysis of Taylor and Stephens (1980) and are characterizedas monthly and annual means of the first principal compo-nent of the position of the north wall of the Gulf Stream,1966–2000. Second, two versions of the widely applied win-ter North Atlantic Oscillation (NAO) index, which is an indi-cation of atmospheric effects across the North Atlantic (Hurrell1995), were correlated with the landings index. Both indiceswere correlated using Pearson’s product-moment correlationand at the same two time lags used with the other data sets.

Autocorrelation and degrees of freedomMost of the time series dealt with in this correlation study

are highly autocorrelated or persistent meaning that the

© 2003 NRC Canada


Fig. 6. Correlation between mean monthly air temperature and North American Atlantic salmon landings index at two time lags bylatitude.



catches from one year to the next are strongly related.Autocorrelation can lead to an overestimation of the signif-icance of the correlative relationship between variables. Thisis because standard significance tables for correlation coef-ficients assume that each point in the series is independent,i.e., the autocorrelation at lag of 1 or more is zero. For fish-eries and environmental data, this assumption is not valid.A number of approaches have been applied to correct forthe autocorrelation problem. One technique is to removethe low-frequency trends through differencing the data be-fore running the correlations. Another approach is to esti-mate the number of independent points (Na) in the timeseries by correcting for autocorrelation. We have used thelatter following Garrett and Toulany (1981), which simplycomputes an estimate of independent points for an auto-correlated time series:

1 1 22N N N

N j Aa j

n

j= + −=∑ ( )

1

Where N is the number of points in the original time series, jis the lag for the autocorrelation (j = 1,…,N), and Aj is theautocorrelation at lag j. For correlations between time seriesXj and Yj, Aj is the autocorrelation of the product time seriesXj times Yj. The number of lags used in the summation

should be to the first zero-crossing. Thus, the significance ofall correlations reported is evaluated using this criterion.

Results

Each of the data sets with which we worked had uniquecharacteristics with respect to the statistical properties oftheir time series and the relationship between observing sta-tions within a data type. Because we were interested in char-acterizing the impact of a season on the entire salmoncohort, it was critical to evaluate combinations of the envi-ronmental data that would hopefully capture the scale of thephenomenon. We tried to be conservative in the applicationof spatial and temporal smoothing, especially when therewas auxiliary data to suggest smoothing was unwarranted.

Stock size of the North American Atlantic salmon stockcomplex

The time series of North American Atlantic salmon land-ings has a distinct multidecadal pattern showing two peaksin abundance over the past century. The highest observed in-dex level was during the 1920s when landings reached 6000metric tons (t) (Fig. 4a). This peak catch was part of the firstperiod of high abundance that spanned approximately twodecades, the 1920s and 1930s. The stock is assumed to havedeclined during a period of low catch that occurred duringthe later 1950s to early 1960s, followed by a rapid increasein landings that also lasted approximately two decades. Thestock complex has decreased in size over the past two de-cades of the time series to the lowest levels observed. Thisdecline can be confirmed by comparing the landings indexwith the estimate of pre-fishery abundance for 1SW and2SW salmon from stock assessments (Anonymous 2002).Stock abundance has been as high as 1.7 million fish but hasdeclined over the past three decades to levels as low as400 000 fish (Fig. 4b). The landings are embedded in thestock abundance estimate, thus correlation would be inap-propriate. However, the main point is that the best estimatesof escapement and mortality outside the landing index donot alter the pattern of abundance suggested by the two datasets. The decline in stock abundance continues beyond theextent of the landings index, which ends in fishery year1997. This comparison supports our contention that changesin stock abundance are reflected in the landings index. Thereis a diversity of views on the use of catch data as an indica-tor of Atlantic salmon stock size (Bielak and Power 1986;Shearer 1986). We recognize these concerns but assert thatonly unrealistic changes in fishing mortality would reversethe interpretation of the landings data trends for NorthAmerican Atlantic salmon.

Thermal conditions in the marine nurseryA comparison of the landings index to monthly mean

SSTs 1 and 2 years previous from the each of the five ma-rine nursery areas was undertaken using correlation analysis.The initial analysis suggested that 37 of the 120 correlationswere significant at p < 0.5, more than by chance alone. Wethen repeated the significance tests adjusting for auto-correlation in the time series using the method of Garrettand Toulany (1981), which suggested that only three of the

© 2003 NRC Canada


Fig. 7. Mean air temperature for Chatham, New Brunswick, forthe winter months (a) December, (b) January, (c) February, and(d) March. Bold lines represent five-point moving average filters.



37 were reliably significant. Stock size was negatively corre-lated with spring SST conditions in area 1 (Table 3). Thenegative correlation weakens in the summer and is againprominent in the fall. The only significant correlation is be-tween May SST and the landings index at lag 1. The springcorrelation is temporally coincident to the time that salmonpost-smolts would be migrating to sea. The correlations inarea 2 occurred earlier in the year. Winter SST was nega-tively correlated with stock size, the only significant correla-tions being in February. In this case, lag 2 correlations wereslightly higher than lag 1. The correlations in areas 3–5 wereprogressively weaker. The correlation in area 3 mirroredarea 2 in those winter months that showed the highest corre-

lation, and the sign of the correlations were the same as forarea 2. The correlations were much weaker in area 4 and ab-sent in area 5. Therefore, the SST analysis reveals two areasof interest, a negative correlation between winter SST andstock size primarily focused over the Gulf of Maine and theScotian Shelf (Fig. 5b) and a negative correlation betweenstock size and SST conditions in the Gulf of St. Lawrenceduring spring (Fig. 5c). The landing index is repeated(Fig. 5a) for comparison. As can be seen in these time se-ries, the early decades show greater interannual variabilityand missing values, which reflects the nature of the COADSdata to be less reliable further back in time. Because manyof the features of the smoothed SST data qualitatively corre-

© 2003 NRC Canada


Fig. 8. (a) Mean February air temperature for New Brunswick observing stations and (b) the same data smoothed with five-pointmoving average filters.



late to the landings index time series, we feel that the ero-sion of correlative fit is in part due to observational noise inthe COADS data.

Thermal and rainfall conditions in the freshwater nurseryThere is no evidence of a relationship between winter air

temperature and stock size based on correlation betweenmonthly mean temperature and the landings index. Signifi-cance criterion (p = 0.05) for a time series of 88 observa-tions uncorrected for autocorrelation has an absolute valueof 0.21. A plot of correlation by latitude (axis rotated) foreach month, lag, and observing station yielded only a fewpotentially significant coefficients, none of which was foundto be significant after applying the autocorrelation correction(Fig. 6). The lower latitude station correlations were of neg-ative sign, switching to positive sign at the higher latitudes.The correlations at lags 1 and 2 track each other. Becausewe are concerned with the accumulated effects over a winterseason, it is not surprising that correlation to individual monthsyielded little information. As can be seen for the time seriesof mean temperature for Chatham (station 5, Fig. 2), locatedin the heart of the salmon production area, monthly data re-flected some of the features seen in the SST data and alsoshow important differences between months (Fig. 7). Janu-ary and February mean temperature at Chatham (Figs. 7b,7c) show an increase in temperature during the 1950s, whichis coincident with the increase in the SST data. March tem-peratures (Fig. 7d) suggest little contrast in conditions, andDecember temperatures (Fig. 7a) suggest a long-term in-crease in temperature during the past century. Though thetime course of temperature is difficult to evaluate over thewinter season, the spatial pattern within a region is coherent.Mean temperature for February in six New Brunswick sta-tions show very similar time trends (Fig. 8a). The smootheddepiction of these data reinforces the assertion that though

© 2003 NRC Canada


Meantemperature

Meantemperature

Area group Group members Period Lag 1 Lag 2 Lag 1 Lag 2

Newfoundland – Quebec – P.E.I. St. Anthony Dec. –0.07 –0.06 0.08 0.08St. John’s Jan. –0.07 –0.08 0.08 0.09Natashquan Feb. –0.07 –0.07 0.07 0.09Gaspé Mar. 0.08 0.03 –0.08 –0.03Sydney Dec.–Jan. –0.08 –0.09 0.10 0.11Charlottetown Jan.–Feb. –0.09 –0.10 0.09 0.11

Feb.–Mar. –0.01 –0.03 0.00 0.04Dec.–Mar. –0.05 –0.07 0.04 0.08

New Brunswick – Nova Scotia – U.S.A. Aroostook Nappan Dec. –0.06 –0.08 0.06 0.08Bagotville Portland Jan. –0.05 –0.09 0.04 0.09Chatham Presque Isle Feb. –0.13 –0.18 0.13 0.18Eastport Québec Mar. 0.01 –0.02 –0.02 0.01Fredericton Saint John Dec.–Jan. –0.07 –0.12 0.07 0.12La Tuque Sussex Jan.–Feb. –0.12 –0.19 0.12 0.19Moncton Yarmouth Feb.–Mar. –0.09 –0.13 0.07 0.12Mt. Joli Dec.–Mar. –0.11 –0.17 0.10 0.16

Note: Mean temperature refers to mean air temperature across the area group members and months designated; first principal components refers to thefactor scores for the data combined with principal components.

Table 4. Correlations between North American Atlantic salmon landings index and air temperature for two area groupings by monthand monthly combinations at 1 year before catch year (lag 1) and 2 years before catch year (lag 2).

Fig. 9. Number of days of daily air temperature (a) above 1°C,(b) above 4°C, and (c) between 1 and 4°C by year for Chatham,New Brunswick. Bold lines represent five-point moving averagefilters.



annual data are a poor match to the trend in the SST data,the filtered data are similar (Fig. 8b).

Two major groupings of air temperature observing sta-tions were identified in the principal component analysis(members of the groupings are listed in Table 4). One groupwas composed of stations from Newfoundland, Quebec, andPrince Edward Island while the other consisted of stationsfrom New Brunswick, Nova Scotia, and the United States.Sable Island, (station 19, Fig. 2), Deer Lake in Newfound-land (station 6), and the two Labrador stations (stations 3and 10) were excluded because of their location or shortenedtime series.

Correlations between stock size and mean air temperaturewere nonsignificant. Correlation coefficients for these sta-tion groups combined over various time periods and by twocombining methods were nonsignificant (Table 3).

Patterns in the daily maximum temperature data reflectthose in the mean monthly temperature. Cumulative wintercounts of the number of days above 1°C for the winter sea-son at Chatham station are shown (Fig. 9a). Though the time

series of counts show a great deal of annual variability, the5-year moving average data clearly show the same patternseen in the temperature data. The periods of high abundanceof salmon are associated with lower numbers of thaw days,and the periods of low abundances are associated with highercounts. A different trend can be seen in the time series ofcounts of days greater than 4°C; higher counts occurred withgreater frequency in recent years (Fig. 9b). The transitionaltemperatures, days between 1 and 4°C were highest at thebeginning of the time series (Fig. 9c). Despite the qualitativesimilarity between many of the features of the daily temper-ature counts time series and the landings index, the dailytemperature count data were for the most part uncorrelatedwith the landings index for the four stations considered (Ta-ble 5).

Two major groupings of rainfall observing stations wereidentified in the principal component analysis (members ofthe groupings are listed in Table 6). One group was com-posed of stations from Newfoundland, Nova Scotia, and PrinceEdward Island, and the other, New Brunswick and Quebec.

© 2003 NRC Canada


Days greater than 1°C Days greater than 4°C Days between 1 and 4°C

Station Period Lag 1 Lag 2 Lag 1 Lag 2 Lag 1 Lag 2

Chatham Dec. –0.14 –0.19 –0.17 –0.20 –0.03 –0.08Jan. –0.06 –0.05 0.06 0.06 –0.14 –0.12Feb. –0.13 –0.14 –0.18 –0.19 –0.02 –0.02Mar. 0.14 0.11 0.04 0.04 0.13 0.09Dec.–Jan. –0.15 –0.18 –0.11 –0.14 –0.11 –0.13Jan.–Feb. –0.14 –0.14 –0.11 –0.13 –0.11 –0.09Feb.–Mar. 0.03 0.00 –0.06 –0.07 0.10 0.06Dec.–Mar. –0.08 –0.13 –0.13 –0.16 0.00 –0.04

Fredericton Dec. –0.10 –0.13 –0.17 –0.19 0.05 0.02Jan. –0.04 –0.07 0.04 –0.01 –0.09 –0.10Feb. –0.09 –0.08 –0.11 –0.11 –0.03 –0.02Mar. 0.04 0.02 0.01 0.03 0.04 –0.01Dec.–Jan. –0.09 –0.13 –0.10 –0.14 –0.03 –0.06Jan.–Feb. –0.09 –0.11 –0.06 –0.09 –0.08 –0.08Feb.–Mar. –0.04 –0.04 –0.06 –0.04 0.01 –0.02Dec.–Mar. –0.09 –0.12 –0.12 –0.14 –0.01 –0.05

Moncton Dec. –0.09 –0.13 –0.12 –0.17 0.00 –0.03Jan. –0.05 –0.04 –0.01 –0.03 –0.07 –0.03Feb. –0.11 –0.15 –0.14 –0.19 –0.01 –0.03Mar. 0.10 0.15 0.03 0.03 0.09 0.16Dec.–Jan. –0.09 –0.11 –0.10 –0.14 –0.05 –0.04Jan.–Feb. –0.11 –0.13 –0.11 –0.15 –0.06 –0.04Feb.–Mar. 0.00 0.02 –0.05 –0.07 0.07 0.12Dec.–Mar. –0.06 –0.07 –0.10 –0.15 0.02 0.06

Eastport Dec. –0.11 –0.19 –0.11 –0.20 –0.01 –0.01Jan. 0.04 0.00 0.02 0.01 0.04 –0.02Feb. –0.20 –0.22 –0.24* –0.25* –0.07 –0.09Mar. –0.06 –0.04 –0.08 –0.09 0.03 0.07Dec.–Jan. 0.09 0.03 0.03 –0.03 0.13 0.10Jan.–Feb. –0.03 –0.05 –0.08 –0.08 0.05 0.01Feb.–Mar. –0.05 –0.03 –0.11 –0.10 0.05 0.08Dec.–Mar. 0.03 0.00 –0.05 –0.08 0.12 0.12

Note: Significance: *, p = 0.05.

Table 5. Correlations between North American Atlantic salmon landings index and counts of daily air temperaturegreater than 1°C, greater than 4°C, and between 1 and 4°C for stations in Canada and the United States by monthat 1 year before catch year (lag 1) and 2 years before catch year (lag 2).



Sable Island, Labrador stations, and Deer Lake were ex-cluded because of their location or the length of the time se-ries.

The correlations between the landings index and rainfallfor each station by months, arranged by latitude, were allnonsignificant (Fig. 10). Rainfall data combined over time

and space confirm the absence of any correlation betweenthe landings index and rainfall amounts (Table 6). However,the smoothed time series for average rainfall show many ofthe same features associated with the landing index time se-ries (Figs. 11a, 11b). Both time series show a peak in rain-fall associated with the low abundance years of the 1950s,

© 2003 NRC Canada


Mean rainfall First principal component

Area group Group members Period Lag 1 Lag 2 Lag 1 Lag 2

Newfoundland – Nova Scotia – P.E.I. St. Anthony Dec. –0.01 –0.06 0.01 0.06Sydney Jan. –0.11 –0.07 0.11 0.07Yarmouth Feb. –0.20 –0.15 0.18 0.17Charlottetown Mar. –0.10 –0.14 0.02 0.08Nappan Dec.–Jan. –0.08 –0.09 0.06 –0.01

Jan.–Feb. –0.20 –0.14 0.15 0.10Feb.–Mar. –0.21 –0.21 0.18 0.18Dec.–Mar. –0.17 –0.18 0.07 0.13

New Brunswick – Quebec Frederiction Dec. –0.05 –0.14 0.06 0.16Chatham Jan. –0.18 –0.13 0.20 0.15Moncton Feb. –0.14 –0.13 0.14 0.13Saint John Mar. 0.04 –0.01 –0.04 0.02Sussex Dec.–Jan. –0.16 –0.20 0.08 –0.02Gaspé Jan.–Feb. –0.23 –0.18 –0.06 –0.03Natashquan Feb.–Mar. –0.06 –0.10 –0.15 –0.13

Dec.–Mar. –0.15 –0.19 0.06 0.12

Note: Mean precipitation refers to rainfall across the area group members and months designated; first principal components refers to the factor scoresfor data combined with principal components. Significance: *, p = 0.05; **, p = 0.01.

Table 6. Correlations between North American Atlantic salmon landings index and rainfall for two area groupings by month andmonth combinations at 1 year before catch year (lag 1) and 2 years before catch year (lag 2).

Fig. 10. Correlation between mean monthly rainfall and North American Atlantic salmon landings index at two time lags by latitude.



and more so for the New Brunswick – Quebec group, rain-fall is low during the sustained period of high abundanceduring the first half of the century.

Synchronization of smolt migration and ocean conditionsThe analysis of year-day of target air temperature and tar-

get SST reflects the trends seen in the SST correlations andreveals a discontinuity between spring air and ocean warm-ing dynamics. The analysis yielded a total of 490 correla-tions. The correlations with a positive sign suggest that whenthere were large differences in the target temperature year-days, i.e., the number of days between the two dates, stockabundance was high. Alternatively, negative correlation sug-gests that large differences between the year-days were asso-ciated with low abundance. The results are visualized byplotting the response surface of the correlation coefficientsby area and time lag. The largest magnitude correlationswere found for area 1 data, approximately centered on a tar-get SST of 6°C and a target air temperature of 11°C(Figs. 12a, 12b). The surfaces for the other areas are all oflower magnitude correlations and do not possess any distinctpeak or valley structures (Figs. 12c–12j). The only signifi-cant correlations are those associated with the large magni-tude negative correlations in area 1 at lag 2 (Fig. 12b).

Spring SST conditions have shifted dramatically duringthe study period, whereas spring air temperature trends show

very little variation. The strong correlation between differ-ences in target temperature dates in area 1 and the landingsindex can be traced to the shift in date of occurrence of 6°CSST (Fig. 13b). The progression of dates over the past cen-tury show that area 1 has warmed earlier in the year duringpoor production periods, characterized by a shift in datefrom the first week of June during better production periodsto the third week in May during poor production periods.There has been little change in the date of occurrence of11°C air temperature in the drainage area associated witharea 1; for the most part, the date of occurrence has beenduring the last week of May (Fig. 13a). Air temperatures forthe months April to June have remained relatively constantin area 1 during the past century (Figs. 14a–14c). However,it is worth noting that spring air temperatures in area 2 haveincreased over the past century as compared with area 1temperatures; the slopes of linear regressions were all signif-icant for April, May, and June air temperature for area 2 andnonsignificant for area 1 at p = 0.05 (Figs. 14a–14c). It isimportant to remember that this analysis is conditioned onthe landing index, which is greatly influenced by the produc-tion of stocks from the Gulf of St. Lawrence (area 1); hence,the response of stocks at the end of the range may be differ-ent than suggested by the trend of the stock complex.

Climate indicesBoth climate indices, the GSNW and NAO, were corre-

lated to the landings index. The strongest correlations be-tween the landings index and GSNW indices were for theJanuary and February principal components (Table 7). Theweaker correlations were associated with the data for June,November, and December. The annual GSNW index is sig-nificantly correlated and plotted in Fig. 15b (the landings in-dex is repeated in Fig. 15a) and shows an increasing trendsince 1970, exactly opposite to landing index trend. TheNAO index has also trended strongly positive over the pastthree decades (Fig. 15c), which probably accounts for thecorrelation observed between the NAO and landings indices;however, it is also clear that the NAO index during the firsthalf of the century is poorly correlated to the landings index.

Discussion

Broad-scale climate forcing appears to control the multi-decadal abundance of Atlantic salmon in North America.The strongest manifestation of this climate effect is a corre-lation to mid-winter ocean conditions focused off the coastsof northeastern United States and Nova Scotia; however, thecovariation of the winter signal with climate in the marinenursery areas during spring and in the freshwater nursery ar-eas during winter suggest that mechanisms other than factorsaffecting oceanic adults during winter are at work on thestock complex. The explanatory power of any single correla-tion is limited, more so in the case of this study where theapplication of statistical rigor has reduced the pool of poten-tially correlated variables to a smaller group. Nonetheless,taken in the context of other ancillary information, we feelthat certain hypotheses relating climate and salmon recruit-ment in North America are favored, though still unproven.

The fact remains that index stocks clearly show that ma-rine mortality for post-smolts has increased over the past

© 2003 NRC Canada


Fig. 11. Mean monthly rainfall in (a) Newfoundland – NovaScotia grouping and (b) New Brunswick – Quebec grouping dur-ing December through March by year. Bold lines represent five-point moving average filters.



two decades, coincident with the dramatic decline in stockabundance (Friedland 1994). The correlation related to springSST conditions is in the right place and time to suggest animpact on migrating smolts. Furthermore, surveys in freshwater show that conservation measures designed to increase

the production of juvenile salmon have succeeded in at leastproducing increased numbers of juveniles in many of themost important salmon-producing areas during the sametime period of the overall stock decline (Anonymous 2002).These factors would argue that the spring climate signal re-

© 2003 NRC Canada


Fig. 12. Response surfaces of correlation between differences in target dates (year-day for target air temperature – year-day for targetsea surface temperature (SST)) and North American Atlantic salmon landings index in SST index (a–b) area 1, (c–d) area 2, (e–f) area3, (g–h) area 4, and (i–j) area 5. (a, c, e, g, and i) Lag 1; (b, d, f, h, and j) lag 2.



lated to post-smolts is of greater ecological concern if a casefor overwintering mortality cannot be supported.

The correlations related to winter climate impacting onfreshwater nursery areas and overwintering mortality are dif-ficult to interpret. Temperature trends, considered on bothmonthly and daily time scales, are uncorrelated with stocksize. This would argue against temperature trends producingice destabilizing thaws in the nurseries and killing pre-migrant juveniles. Yet we must be mindful of the fact thatthe smoothed patterns in the temperature data do match stocksize and, in a way, appropriate to affect nursery stability. Thecoherent nature of the temperature fields and the limitednumber of recruitment years impacting the landings indexargue against filtering the data before the statistical analysis.Undoubtedly, in some years, unfavorable conditions for pre-migrant smolts would be more likely followed by poor post-smolt survival years, intensifying the climate effect on thestock complex. The filtered rainfall data for the NewBrunswick – Quebec group show many of the features asso-ciated with the landings index time series. In the absence ofdata to show otherwise, we must conclude that climate im-pacts on freshwater habitats must be a secondary factor. Di-rect observations show that ice instability is occurring infreshwater salmon habitats and that climate change appearsto be intensifying (Beltaos and Prowse 2001), but it is not

clear whether this impact is operating at a level commensu-rate with the stock complex.

The correlation between thermal conditions during springin the Gulf of St. Lawrence and stock size is an indicationthat varying ocean conditions during first entry into the ma-rine environment is critical to Atlantic salmon. The spatialand temporal focus of this correlation is particularly usefulas it is the first indication from long-term data that springclimate effects could be the controlling mechanism for At-lantic salmon populations in North America. The aspect ofthis relationship that is particularly intriguing is the fact thatthe correlation is negative with temperature. The emergingparadigm from work with European Atlantic salmon stocksis that warm conditions are of benefit to the growth and sur-vival of post-smolts (Friedland et at. 2000). The reverse signof the correlation for North American stocks suggests that adifferent mechanism is at work. Because salmonid growthincreases linearly with water temperature given an adequatefood supply (Brett 1979), we are immediately faced with thedilemma of making sense of the correlations that showswarmer conditions associated with poorer survival. Growthmay be greater at lower temperature if ration is limited be-cause higher temperature is often associated with increasedmetabolic demands (Despatie et al. 2001); thus, a growth-based mechanism is still a possibility even if thermal condi-tions are not favorable.

Temperature, among a wide array of factors, can impactthe distribution and abundance of food items. Post-smoltsundergo an important ontogenetic change in feeding thatmay be key to their survival during the first year at sea.When post-smolts first enter the marine environment, theyfeed mainly on terrestrial insects and marine invertebratesand then make a transition to piscivory, usually concentrat-ing on the age-0 groups of a winter-spawned species (Hislopand Shelton 1993). For example, Salminen et al. (2001) re-ported that salmon posts-smolts first feed on young-of-the-year herring in the Baltic when they make the transition topiscivory. This is supported in the feeding data available forthe North American nurseries as well; however, the transi-tion fish prey is not always herring. Dutil and Coutu (1988)found that the dominant food item in the diet of Gulf ofSt. Lawrence post-smolts was sand lance larvae, Ammodytesamericanus. Sand lance larvae are one of the dominant spe-cies in the ichthyoplankton of most areas in the post-smoltnursery (Locke and Courtenay 1995; Lazzari 2001). Wrightand Bailey (1996) suggested that coupling between hatchingand the onset of spring secondary production is important tothe growth and survivorship of sand lance, Ammodytesmarinus, from the Shetland Islands area. Changes in avail-ability of the same species of sand lance have been impli-cated as critical to the success of seabirds (Rindorf et al.2000). Similarly, capelin larvae are also abundant in most ofthe nursery areas and may provide another source of foodfor salmon post-smolts as they switch to a piscivorous lifestyle. Spring climate variation may influence the availabilityof a preferred ichthyoplankton prey; however, most of thesespecies are only intermittently assessed in North America,precluding any firm conclusions at this time.

The contrast in climate conditions in the Gulf of St. Law-rence may be indicative of variation in the migration of Atlan-tic salmon post-smolts, especially in the context of movement

© 2003 NRC Canada


Fig. 13. (a) Estimated dates of air temperatures of 11°C in thegrouping of land observing stations associated with sea surfacetemperature (SST) index area 1 and (b) estimated dates of SSTsof 6°C in SST index area 1. Bold lines represent five-point mov-ing average filters.



of its predators and prey. Hinch et al. (1995) observed a neg-ative correlation between temperature conditions and theabundance of sockeye, Oncorhynchus nerka, and hypothe-sized that the higher temperatures were associated with greaterpredation pressure. Migratory pelagic predators like mack-erel, which could overlap spatially and temporally with post-smolts, are often associated with thermal transition zones(Garrison et al. 2000). Alternatively, the warm conditions in

spring post-smolt nurseries associated with poor survival mayreflect increased swimming demanded of these cohorts asthey seek optimal temperatures and thus not be directly re-lated to predation. Patterns of survival for salmon stocksthroughout the North Atlantic illustrates that stocks withshorter migrations to the feeding grounds have higher sur-vival rates (Bley and Moring 1988; Friedland 1994). Thereis reason to believe that the relationship between migration

© 2003 NRC Canada


Fig. 14. Mean air temperature in the grouping of land observing stations associated with sea surface temperature (SST) index area 1(solid lines) and area 2 (broken lines) for (a) April, (b) May, and (c) June. Bold solid and broken lines represent five-point movingaverage filters.



and survival is an important scalar for interannual variabilityas well.

The timing of ocean entry has been recognized as a criti-cal trait for locally adapted salmon populations. Our resultssuggest that emigrational cues are not in synchronizationwith target ocean conditions favorable to post-smolts in theGulf of St. Lawrence, which is the core area of the distribu-tion of salmon in North America. However, we might expecta different response for populations from other parts of therange. For Norwegian rivers, it has been shown that the tim-ing of smolt migration is a population-specific characteristicthat changes with latitude (Hvidsten et al. 1998). It is be-lieved that the specific timing of the migration is to provideoptimal feeding conditions outside the river of origin duringthe early marine phase. Genetic evidence suggests that thelocal adaptations go beyond simple migrational cues and in-clude adaptations related to growth and development as well(Nielsen et al. 2001). Smolt migration timing has been char-acterized for a group of Icelandic rivers, where despite widevariation in the timing of smolt runs, post-smolts enter theocean within a relatively narrow range of ocean temperatures(Antonsson and Gudjonsson 2002). However, the NorthwestAtlantic is far more dynamic than either the central or north-eastern regions of the Atlantic Ocean. Responding to primar-ily a temperature-related cue, migration timing of stocks intothe Gulf of St. Lawrence would appear to have changed littleover the past century. However, there have been dramaticchanges in the thermal conditions in the ocean and, by sup-position, ecosystem conditions impacting post-smolts. A re-lated effect may be at work in the Gulf of Maine and NovaScotia areas where air temperatures in spring have increasedsuggesting that migrations may have shifted to earlier dates.

Consideration of the impact of winter climate should notbe limited to mortality effects on the stock complex. With astock complex mainly composed of 1SW and 2SW salmon,variation in maturity schedules could have a profound effecton landings and the availability of fish to certain fisheries.Numerous authors have hypothesized that Atlantic salmonmaturity could be affected by winter migration routes (Mar-tin and Mitchell 1985; Reddin 1988; Friedland et al. 1998b);the variation in SST signal that we documented could repre-sent changes in the orientation cues used for migrations byadult fish. Noting the same concern over the role of winterclimate in freshwater nursery areas, variation in maturitycould also be represented as an additive affect that covarieswith other mortality effects and thus intensifies the impact ofclimate on the stock complex.

The specific climate impacts on Atlantic salmon appear tobe driven by broad-scale forcing related to the NAO andmanifested in the distribution of SST caused by the shiftingposition of the Gulf Stream. Friedland et al. (1993) noticedthe potential relationship between NAO and salmon stocksin North America; however, the refinement of an atmosphericto sea surface linkage described by the GSNW index may beuseful in considering a range of effects on Atlantic salmonstocks. It has been reported that much of the variation inGSNW can be attributed to NAO variation, at least for thelast half of the twentieth century (Taylor and Gangopadhyay2001). Therefore, the same climate forcing influencing con-ditions in freshwater salmon habitats, including those con-sidered in this paper, would be related to the same forcing

© 2003 NRC Canada


Index Method Period Lag 1 Lag 2

GSNW PC of position Jan. –0.48** –0.41*Feb. –0.60** –0.57**Mar. –0.39* –0.24Apr. –0.47 –0.41May. –0.40 –0.40June –0.35 –0.48*July –0.15 –0.27Aug. –0.48 –0.52Sept. –0.40 –0.41Oct. –0.42 –0.47Nov. –0.46* –0.43Dec. –0.40 –0.41*Annual –0.68* –0.67*

NAO Station-based index Winter –0.18 –0.23*PC of leading EOF Winter –0.19 –0.21*

Note: PC, principal component; EOF, Empirical Orthogonal Functionanalysis. Significance: *, p = 0.05; **, p = 0.01.

Table 7. Correlations between North American Atlantic salmonlandings index and Gulf Stream Northern Wall (GSNW) indexand North Atlantic Oscillation (NAO) index by month and sea-son at 1 year before catch year (lag 1) and 2 years before catchyear (lag 2).

Fig. 15. (a) North American Atlantic salmon landings index,(b) Gulf Stream Northern Wall (GSNW) index, and (c) NorthAtlantic Oscillation (NAO) index by year. Bold lines representfive-point moving average filters.



producing SST differences and the associated marine habitatvariations potentially impacting post-smolts.

We are intrigued by how the cyclic nature of the SST andputative response by the salmon stock complex relates to thefirst half of the NAO index versus the second half. Clearly,the NAO and GSNW indices in the second half of the twen-tieth century follow similar patterns to the SST and stocksize data, but that relationship falters for NAO during thefirst half of the century. NAO is not showing the same pat-tern as seen in the SST or stock size data. NAO and SST arestrongly correlated in the temperate portion of the AtlanticOcean (Robertson et al. 2000), which raises the question ofwhether SST is responsive to NAO in northern latitudes, orif the full NAO time series is equivalent?

The persistent state of the North Atlantic Oscillation in itspositive anomaly has been interpreted as an indication of aglobal-warming impact on the dominant atmospheric circu-lation pattern in the North Atlantic area (Visbeck et al.2001). The consequences of such a shift in climate regimeswill be manifest at all levels of the marine environment(Ottersen et al. 2001). For migratory species that depend onthe timing of seasonal events and that use environmentalvariables as migratory parameters, the consequences may bemore profound then for nonmigratory species. In the case ofsalmon, successful completion of the life cycle is dependanton locally adapted behaviors (Nicieza 1995; Verspoor 1997).In a practical sense, the importance of local traits can beseen in the husbandry literature, which is rife with examplesillustrating the erosion of survival rates for transplants re-leased at increasing distances from their natal rivers(Friedland 1994). Salmon stray between river systems, andthe evolution of new characters is believed to operate on a timescale of tens of generations (Hendry et al. 2000), whereas theshift in climate indicated by the phase change in the NAO ison a time scale of only a couple of decades. Atlantic salmonmay not be locally adapting quickly enough to this rate ofchange. Does that mean that the eroding stability of popula-tions in the southern part of the range in North America isan indication of a climate-induced range contraction? Appli-cation of the precautionary approach would dictate that ithas to be treated as such, with the burden of proof on thoseadvocating otherwise.

Acknowledgements

We thank M. Desjardins and L. Vincent for help with Ca-nadian environmental data and B. Dattore for help with ex-traction of the COADS data used in this study.

References

Allen, K.R., Saunders, R.L., and Elson, P.F. 1972. Marine growthof Atlantic salmon (Salmo salar) in the northwest Atlantic. J.Fish. Res. Board Can. 29: 1373–1380.

Anderson, J.T. 1988. A review of size dependent survival duringpre-recruit stages of fishes in relation to recruitment. J. Northw.Atl. Fish. Sci. 8: 55–66.

Anderson, J.M., Whoriskey, F.G., and Goode, A. 2000. Atlanticsalmon on the brink. Endangered Species Update, 17: 15–21.

Anonymous. 2002. Report of the Working Group on North AtlanticSalmon. ICES CM 2002/ACFM:14.

Antonsson, T., and Gudjonsson, S. 2002. Variability in timing andcharacteristics of Atlantic salmon smolt in Icelandic Rivers. Trans.Am. Fish. Soc. 131: 643–655.

Arndt, S.K.A., Cunjak, R.A., and Benfey, T.J. 2002. Effect of sum-mer floods and spatial–temporal scale on growth and feeding ofjuvenile Atlantic salmon in two New Brunswick streams. Trans.Am. Fish. Soc. 131: 607–622.

Beltaos, S., and Prowse, T.D. 2001. Climate impacts on extremeice-jam events in Canadian rivers. Hydrol. Sci. J. 46: 157–182.

Bielak, A.T., and Power, G. 1986. Catch records — facts or myths?In Atlantic salmon: planning for the future. Edited by D. Millsand D. Piggins. Timber Press, Portland, Ore. pp. 235–255.

Bley, P.W., and Moring, J.R. 1988. Freshwater and ocean survivalof Atlantic salmon and steelhead: a synopsis. Biological Rep.No. 88, Maine Cooperative Fish and Wildlife Research Unit,Orono, Maine.

Brett, J.R. 1979. Environmental factors and growth. In Fish physi-ology. Vol. VIII. Edited by W.S. Hoar, D.J. Randall, and J.R.Brett. Academic Press, London. pp. 599–675.

Chadwick, E.M.P. 1987. Causes of variable recruitment in a smallAtlantic salmon stock. Am. Fish. Soc. Symp. Ser. 1: 390–401.

Chaput, G., Allard, J., Caron, F., Dempson, J.B., Mullins, C.C.,and O.Connell, M.F. 1998. River-specific target spawning re-quirements for Atlantic salmon (Salmo salar) based on a gener-alized smolt production model. Can. J. Fish. Aquat. Sci. 55:246–261.

Cunjak, R.A., Prowse, T.D., and Parrish, D.L. 1998. Atlantic salmon(Salmo salar) in winter: “the season of parr discontent”? Can. J.Fish. Aquat. Sci. 55(Suppl): 161–180.

Despatie, S.-P., Castonguay, M., and Chabot, D. 2001. Final ther-mal preferendum of Atlantic cod: effect of food ration. Trans.Am. Fish. Soc. 130: 263–275.

Dutil, J.D., and Coutu, J.M. 1988. Early marine life of Atlanticsalmon, Salmo salar, postsmolts in the Northern Gulf of St.Lawrence. Fish. Bull. 86: 197–211.

Eriksson, T. 1994. Mortality risks of Baltic salmon during down-stream migration and early sea-phase: effects of body size andseason. Nord. J. Freshw. Res. 69: 100.

Fisher, J.P., and Pearcy, W.G. 1988. Growth of juvenile coho salmon(Oncorhynchus kisutch) off Oregon and Washington, U.S.A., inyears of differing coastal upwelling. Can. J. Fish. Aquat. Sci. 45:1036–1044.

Friedland, K.D. 1994. Marine survival of restoration stocks. In Ahard look at some tough issues. Edited by S. Calabi and A.Stout. Silver Quill Books, Camden, Maine. pp. 223–239.

Friedland, K.D. 1998. Ocean climate influences on critical Atlanticsalmon (Salmo salar) life history events. Can. J. Fish. Aquat.Sci. 55(Suppl. 1): 119–130.

Friedland, K.D., Reddin, D.G., and Kocik, J.F. 1993. Marine sur-vival of North American and European Atlantic salmon: effectsof growth and environment. ICES J. Mar. Sci. 50: 481–492.

Friedland, K.D., Haas, R.E., and Sheehan, T.F. 1996. Post-smoltgrowth, maturation, and survival of two stocks of Atlantic salmon.Fish. Bull. 94: 654–663.

Friedland, K.D., Hansen, L.P., and Dunkley, D.A. 1998a. Marinetemperatures experienced by post-smolts and the survival of At-lantic salmon (Salmo salar L.) in the North Sea area. Fish.Oceanogr. 7: 22–34.

Friedland, K.D., Reddin, D.G., Shimizu, N., Haas, R.E., andYoungson, A.F. 1998b. Strontium:calcium ratios in Atlantic salmonotoliths and observations on growth and maturation. Can. J. Fish.Aquat. Sci. 55: 1158–1168.

Friedland, K.D., Hansen, L.P., Dunkley, D.A., and MacLean, J.C.2000. Linkage between ocean climate, post-smolt growth, and

© 2003 NRC Canada




survival of Atlantic salmon (Salmo salar L.) in the North Seaarea. ICES J. Mar. Sci. 57: 419–429.

Gargett, A.E., Li, M., and Brown, R. 2001. Testing mechanistic ex-planations of observed correlations between environmental fac-tors and marine fisheries. Can. J. Fish. Aquat. Sci. 58: 208–219.

Garrett, C., and Toulany, B. 1981. Variability of the flow throughthe Strait of Belle Isle. J. Mar. Res. 39: 163–189.

Garrison, L.P., Michaels, W., Link, J.S., and Fogarty, M.J. 2000.Predation risk on larval gadids by pelagic fish in the GeorgesBank ecosystem. I. Spatial overlap associated with hydrographicfeatures. Can. J. Fish. Aquat. Sci. 57: 2455–2469.

Ghent, A.W., and Hanna, B.P. 1999. Statistical assessment of Hunts-man’s 3-y salmon–rainfall correlation, and other potential correla-tions, in the Miramichi fishery, New Brunswick. Am. Midl. Nat.142: 110–128.

Hendry, A.P., Wenburg, J.K., Bentzen, P., Volk, E.C., and Quinn,T.P. 2000. Rapid evolution of reproductive isolation in the wild:evidence from introduced salmon. Science (Wash., D.C.), 290:516–518.

Hesthagen, T., and Garnås, E. 1986. Migration of Atlantic salmonsmolts in the River Orkla, central Norway, in relation to man-agement of a hydroelectric station. N. Am. J. Fish. Manag. 6:376–382.

Hinch, S.G., Healey, M.C., Diewert, R.E., Thomson, K.A., Hourston,R., Henderson, M A., and Juanes, F. 1995. Potential effects of cli-mate change on marine growth and survival of Fraser Riversockeye salmon. Can. J. Fish. Aquat. Sci. 52: 2651–2659.

Hislop, J.R.G., and Shelton, R.G.J. 1993. Marine predators andprey of Atlantic salmon (Salmo salar L.). In Salmon in the sea.Edited by D. Mills. Fishing News Books, London. pp. 699–708.

Holtby, L.B., Andersen, B.C., and Kadowaki, R.K. 1990. Importanceof smolt size and early ocean growth to interannual variability inmarine survival of coho salmon (Oncorhynchus kisutch). Can. J.Fish. Aquat. Sci. 47: 2181–2194.

Hurrell, J.W. 1995. Decadal trends in the North Atlantic Oscilla-tion: regional temperatures and precipitation. Science (Wash.,D.C.), 269: 676–679.

Hvidsten, N.A., Heggerberget, T.G., and Jensen, A.J. 1998. Seawater temperatures at Atlantic salmon smolt entrance. Nord. J.Freshw. Res. 74: 79–86.

Karl, T.R., Williams, C.N., Jr., Quinlan, F.T., and Boden, T.A. 1990.United States Historical Climatology Network (HCN) serial tem-perature and precipitation data. Environmental Science Division,Publication No. 3404, Carbon Dioxide Information and AnalysisCenter, Oak Ridge National Laboratory, Oak Ridge, Tenn.

Lazzari, M.A. 2001. Dynamics of larval fish abundance in PenobscotBay, Maine. Fish. Bull. 99: 81–93.

Locke, A., and Courtenay, S.C. 1995. Effects of environmental fac-tors on ichthyoplankton communities in the Miramichi Estuary,Gulf of St. Lawrence. J. Plankton Res. 17: 333–349.

Martin, J.H.A., and Mitchell, K.A. 1985. Influence of sea tempera-ture upon the numbers of grilse and multi-sea winter Atlanticsalmon (Salmo salar) caught in the vicinity of the River Dee(Aberdeenshire). Can. J. Fish. Aquat. Sci. 42: 1513–1521.

May, A.W., and Lear, W.H. 1971. Digest of Canadian Atlantic salmoncatch statistics. Fish. Res. Board. Can. Tech. Rep. No. 270.

Mekis, É., and Hogg, W.D. 1999. Rehabilitation and analysis of Ca-nadian daily precipitation time series. Atmos.-Ocean, 37: 53–85.

Minns, C.K., Randall, R.G., Chadwick, E.M.P., Moore, J.E., andGreen, R. 1995. Potential impact of climate change on the habi-tat and population dynamics of juvenile Atlantic salmon (Salmosalar) in Eastern Canada. In Climate change and northern fishpopulations. Edited by R.J. Beamish. Can. Spec. Publ. Fish.Aquat. Sci. No. 121. pp. 699–708.

Nicieza, A.G. 1995. Morphological variation between geographi-cally disjunct populations of Atlantic salmon: the effects of on-togeny and habitat shift. Funct. Ecol. 9: 448–456.

Nielsen, C., Holdensgaard, G., Petersen, H.C., Björnsson, B.Th., andMadsen, S.S. 2001. Genetic difference in the physiology, growthhormone levels and migratory behavior of Atlantic salmon smolts.J. Fish Biol. 59: 28–44.

Ottersen, G., Planque, B., Belgrano, A., Post, E., Reid, P.C., andStenseth, N.C. 2001. Ecological effects of the North AtlanticOscillation. Oecologia, 128: 1–14.

Pepin, P. 1991. The effect of temperature and size on development,mortality and survival rates of the pelagic early life stages ofmarine fishes. Can. J. Fish. Aquat. Sci. 48: 503–518.

Prowse, T.D., and Beltaos, S. 2002. Climatic control of river-icehydrology: a review. Hydrol. Process. 16: 805–822.

Reddin, D.G. 1988. Ocean life of Atlantic salmon (Salmo salar L.)in the Northwest Atlantic. In Atlantic salmon: planning for the fu-ture. Proceedings of the Third International Atlantic salmon Sym-posium, Biarritz, France, 21–23 October 1986. Edited by D.H.Mills and D.J. Piggins. Croom Helm, London. pp. 483–511.

Reddin, D.G., and Friedland, K.D. 1993. Marine environmental fac-tors influencing the movement and survival of Atlantic salmon. InSalmon in the Sea. Edited by D. Mills. Fishing News Books, Lon-don. pp. 79–103.

Reddin, D.G., and Friedland, K.D. 1999. A history of identificationto continent of origin of Atlantic salmon (Salmo salar L.) atWest Greenland, 1969–1997. Fish. Res. 43: 221–235.

Rindorf, A., Wanless, S., and Harris, M.P. 2000. Effects of changesin sandeel availability on the reproductive output of seabirds.Mar. Ecol. Prog. Ser. 202: 241–252.

Ritter, J.A. 1989. Marine migration and natural mortality of NorthAmerican Atlantic salmon (Salmo salar L.). Can. Manuscr. Rep.Fish. Aquat. Sci. No. 2041.

Robertson, A.W., Mechoso, C.R., and Kim, Y.-J. 2000. The influ-ence of Atlantic sea surface temperature anomalies on the NorthAtlantic Oscillation. J. Clim. 13: 122–138.

Salminen, M., Kuikka, S., and Erkamo, E. 1995. Annual variabilityin survival of sea ranched Baltic salmon, Salmo salar L.: signifi-cance of smolt size and marine conditions. Fish. Manag. Ecol. 2:171–184.

Salminen, M., Erkamo, E., and Salmi, J. 2001. Diet of post-smoltand one-sea-winter Atlantic salmon in the Bothnian Sea, North-ern Baltic. J. Fish Biol. 58: 15–35.

Shearer, W.M. 1986. Relating catch records to stocks. In Atlanticsalmon: planning for the future. Edited by D. Mills and D. Pig-gins. Timber Press, Portland, Ore. pp. 256–274.

Swansburg, E., Chaput, G., Moore, D., Cassie, D., and El-Jabi, N.2002. Size variability of juvenile Atlantic salmon: links to envi-ronmental conditions. J. Fish Biol. 61: 661–683.

Taylor, A.H., and Gangopadhyay, A. 2001. A simple model ofinterannual displacements of the Gulf Stream. J. Geophys. Res.C Oceans, 106: 13 849 – 13 860.

Taylor, A.H., and Stephens, J.A. 1980. Latitudinal displacementsof the Gulf Stream (1966 to 1977) and their relation to changesin temperature and zooplankton abundance in the NE Atlantic.Oceanol. Acta, 3: 145–149.

Verspoor, E. 1997. Genetic diversity among Atlantic salmon (Salmosalar L.) populations. ICES J. Mar. Sci. 54: 965–973.

Vincent, L.A. 1998. A technique for the identification of inhomo-geneities in Canadian temperature series. J. Clim. 11: 1094–1104.

Vincent, L.A., and Gullett, D.W. 1999. Canadian historical and ho-mogeneous temperature datasets for climate change analyses.Int. J. Climatol. 19: 1375–1388.

© 2003 NRC Canada




© 2003 NRC Canada


Visbeck, M.H., Hurrell, J.W., Polvani, L., and Cullen, H.M. 2001.The North Atlantic Oscillation: past, present, and future. Proc.Natl. Acad. Sci. U.S.A. 98: 12 876 – 12 877.

Whalen, K.G., Parrish, D.L., and Mather, M.E. 1999a. Effect of iceformation on selection of habitats and winter distribution of post-young-of-the-year Atlantic salmon parr. Can. J. Fish. Aquat. Sci.56: 87–96.

Whalen, K.G., Parrish, D.L., and McCormick, S.D. 1999b. Migra-tion timing of Atlantic salmon smolts relative to environmentaland physiological factors. Trans. Am. Fish. Soc. 128: 289–301.

Woodruff, S.D., Diaz, H.F., Elms, J.D., and Worley, S.J. 1998.COADS release 2: data and metadata enhancements for im-provements of marine surface flux fields. Phys. Chem. Earth,23: 517–527.

Wright, P.J., and Bailey, M.C. 1996. Timing of hatching in Ammo-dytes marinus from Shetland waters and its significance to earlygrowth and survivorship. Mar. Biol. 126: 143–152.



Multidecadal trends in North American Atlantic salmon ... · Multidecadal trends in North American...

Documents

Transcript of Multidecadal trends in North American Atlantic salmon ... · Multidecadal trends in North American...