UBC Faculty Research and Publications

Summer Sea-Surface Temperature Variability Off Vancouver Island from Satellite Data Fang, Wendong; Hsieh, William W. Aug 15, 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-Hsieh_AGU_1993_93JC01210.pdf [ 900.52kB ]
Metadata
JSON: 52383-1.0041794.json
JSON-LD: 52383-1.0041794-ld.json
RDF/XML (Pretty): 52383-1.0041794-rdf.xml
RDF/JSON: 52383-1.0041794-rdf.json
Turtle: 52383-1.0041794-turtle.txt
N-Triples: 52383-1.0041794-rdf-ntriples.txt
Original Record: 52383-1.0041794-source.json
Full Text
52383-1.0041794-fulltext.txt
Citation
52383-1.0041794.ris

Full Text

JOURNAL OF GEOPHYSICAL  RESEARCH, VOL. 98, NO. C8, PAGES 14,391-14,400, AUGUST 15, 1993  SummerSea SurfaceTemperatureVariability off VancouverIsland From Satellite  Data  WENDONG FArO • AND WILLIAM W. HSIEH Departmentof Oceanography,Universityof British Columbia,Vancouver,Canada  Satellite-sensedadvancedvery high resolutionradiometer(AVHRR) sea surfacetemperature (SST) data over eight summers(1984-1991) were used to analyzethe summerSST patternsof variabilityoff thewestcoastof VancouverIsland.Empiricalorthogonalfunction(EOF) analysisof the spatialvariancefor 133 nearlycloud-freesummerimageswasperformed.The first EOF mode, which resembledthe meanof all images,showeda strongcool waterband locatedat the northwest comerof VancouverIsland, a cool tongueextendingseawardfrom the Strait of Juande Fuca and a warm patchoff Barkley Sound.The secondmoderevealedtopographicallycontrolledupwelling: cool water over the shelf region with its seawardboundaryroughly following the 200-m depth contour,plusa coldeddylocatedjust northof the Juande FucaCanyon.The third modedisplayed cool water extendingsouthwestwardoff Brooks Peninsula,while the fourth mode showeda cool water plume extendingoff Cape Scott at the northerntip of VancouverIsland. These four modes accountedfor 33, 12, 10, and 5% of the SST variance,respectively.The temporalamplitudeof theseEOF modesrevealedhow the SST featureschangedas summerprogressed.From these images,we alsoconstructedan overall seasonalcoolnessindex, which revealedthe summersof 1986 and 1991 to havethe coolestcoastalwater,with bothsummersimmediatelyprecedingan E1 Nifio.  1. INTRODUCTION  With the increasing availability of satellite sea surface temperature (SST) data, it has become possible to use satellitedata to monitor SST variability in both spaceand time, improvingvastly on the very limited spatialcoverage offered by ship observations. Off the west coast of Vancouver Island, SST images of the summer coastal upwellingwere usedto studybaroclinicinstability [Emery and Mysak, 1980], eddies, upwelling events and coastal currents[lkeda and Emery, 1984; Thomsonand Gower, 1985; Burgert and Hsieh, 1989] and to compare with numericalmodelingresults[Ikedaet al., 1984]. Thomsonet al. [1989] gave a generaldescriptionof the circulationsoff Vancouver Island, showing the spatial and temporal variability of the coastal current and its effects on the coastalfishery.Formerstudiesrelatedto the SST variability off Vancouver Island used satellite data mainly for descriptivepurposesby examiningindividualimagesover a shortperiod. In this study,the nearly cloud-freeSST data extendingover eightsummerswereusedto studythe longer termSST variability. Empirical orthogonal function (EOF) analysis (also known as principal componentanalysis) is generally regardedas the most efficient way to extract information from large data sets.For the past few years,EOF analysis has been used to extractdominant satellite SST patterns  [Kelly, 1985; Lagerloef and Bernstein, 1988; Paden et al., 1991]. There are severalways to calculatethe EOFs, but no matter which methodis used,the solutionis uniquefor the nonmissing data [Kelly, 1988]. While the analysismethods are the same whether a temporally averaged mean or a spatiallyaveragedmean is first removedfrom the data set,  differentEOFs are obtained[Padenet al., 1991].Lagerloef and Bernstein [1988] suggestedthat for satellite images wherespatialfeaturestendedto be persistent or only weakly time varying, spatialEOFs (i.e., EOFs performedon data with their spatialmeansremoved)were moreappropriate thantemporalEOFs, whichmainlyexplainedstructures with strongtemporalvariability.This propertywasalsofoundby Paden et al. [1991] while using EOFs to analyze the patternsof SST variability in the Gulf of California in relationto tidal and wind forcing. In this paper, the purposeis to identify the dominant patternsof summerSST variability off Vancouver Island using EOFs, and to study the seasonaland interannual variabilityin the coastalupwelling.This paperis organized as follows:  Section  2  introduces  the  collection  and  processingof advancedvery high resolutionradiometer (AVHRR) SST data. After briefly describing the EOF method in section 3, we show the dominant EOF modes and  relatethem to oceanographic structuresin section4. Section 5 presentsthe coolnessindexderivedfrom the SST images, while section6 performscorrelationanalysisbetweenthe time seriesof EOF temporalamplitudesand variousindices 1Alsoat SouthChinaSeaInstitute of Oceanology, Academia whichaffectthe SST in the studyregion.  Sinica,Guangzhou,China. 2. DATA COLLECTIONAND PROCESSING  Copyright1993by theAmericanGeophysicalUnion. Papernumber93JC01210. 0148-0227/93/93JC-01210505.00  In this study, the primary data source was satellite imageryfrom polar-orbitingNOAA satellites.The Universi14,391  14,392  FANGAND HSIEH:SUMMERSEASURFACE TEMPERATURE VARIABILITY  I N ty of British Columbia (UBC) Satellite Oceanographyand Meteorology Laboratory has received and processed133 nearly cloud-free AVHRR images off Vancouver Island duringeightsummers(1984-1991) via theseNOAA satellite wheret represents a particularimagein the datasetandN series. All satellite imageswere navigated(i.e. corrected the totalnumberof images. for distortion and registeredto a map), and were nudged In contrast,for the studyarea where structuresvary very  T'(x,t) =T(x,t) - -• •'.T(x,t)  (i.e., entire image shiftedto fit map overlap)to correctfor receiving systemtiming errorsor satellitealtitude errors. The accuracyof the navigationwas about 1 pixel (1.1 km). The navigationand clouddetectiontechniquesusedin this studywere describedin Emery et al. [1986]. Only infrared images from band 4 (10.3-11.3 gm) were converted to brightness SST [Lauritson et al., 1979]. Since SST variability rather than absolute SST is important in this study, no correctionwas made for the atmosphericwater vaporeffect on the infraredtemperature. The irregular temporal coverage of the images might causeseriousaliasing as some short-termeventsmight be missingin the satellitedata. However, thereis little doubt that the main variability features can be covered by the satellitedata,asdiscussed in Kelly [1985]. In our study,we focuson the largerscaleand longerterm SST variability,so the effectsfrom the irregularityof time seriesmay not be too serious.A limited test on the seriousnessof missing temporalcoveragewill be shownin section5. The studyarea (Figure 1) is about350 km alongshoreby 150 km offshore,coveting the entire continentalshelf and sloperegion off VancouverIsland. For the convenienceof computationand analysis,we resampledthe spatialgrids (Figure 2), so that the SST spatialresolutionchangedfrom the original 1.1 km by 1.1 km pixel to 1.5 km by 1.5 km. A spatialweightedmoving filter matrix  stronglyin spacebut weaklyin time (e.g., fronts,upwelling or topographiceddies), spatially averaged mean in each imageis removedfrom the data,i.e., I  M  T'(x,t) =T(x,t)• ?_iT(x,t)  similar results were obtained. Hence, we will concentrate  mainlyon thegradientEOFs in the followingsection. The EOF decomposition of the data set T'(x,t) to form the  covariance  matrix  is  a  linear  combination  of  eigenfunctions, N  T'(x,t) = • an(t)F n(x)  was used to smooth the SST data and to reduce small scale  (4)  n=l  (1)  with thecoefficientan(t) obtainedby projectingthedataset ontoeachfunctionF n(x),  imagenoise[Wanget al., 1983].Eachselectedimagewasat least85% cloud-freeover the studyregion.There were 105 by 253 spatialdatapointsover the studyarea.The datawere arrangedinto a two-dimensionalarray T(x,t), wherex and t werethe spatialandtemporalindices,respectively.  M  a,,(t)= •T'(x,t)Fn(x )  (5)  3. Tim EOF METHOD  The applicationof EOF analysisto satelliteinfrareddata has previouslybeen presentedby Kelly [1985], Lagerloef and Bernstein [1988] and Paden et al. [1991]. EOFs are  the principal axes of a data covariance matrix, with the EOFs usuallyordereddecteasinglyby eigenvalue.The first few EOFs denotethe dominantpatternsof the variancewith the corresponding eigenvaluesrepresenting thepercentages of total variance accounted for.  In orderto explicitly decompose just the spatialor just the temporal variability of SST, one of two different means must be removed from T(x,t) before calculatingthe data covariancematrix. To studythe stronglytemporalpatterns associatedwith warming or cooling of some study area, temporallyaveragedmeanat eachpoint is removedfrom the data, i.e.,  (3)  whereM is the total numberof pixelsin an image. LagerloefandBernstein[ 1988] calledthe resultingEOFs from removingtemporalmeans,"temporalEOFs",andthose from subtracting spatial means, "spatial EOFs". For convenience of discussion and to avoid confusing terminology, Paden et al. [1991] called the former "covariance EOFs", and the latter "gradient EOFs". Lagerloef and Bernstein [1988] and Paden et al. [1991] used both approachesto analyze the SST patterns of variability. They suggestedthat the approachwith spatial meansremovedwas more suitablefor AVHRR SST patterns where oceanographicfeaturestendedto persistover time. This hypothesiswas testedwith our satelliteSST data and  121  F=1612 12421  (2)  130'W  128'W  126'W  124'W  Fig. 1. Studyareaoff the westcoastof VancouverIsland.  FANG AND HSIEH: SUMMER SEA SURFACETEMPERATUREVARIABILITY  14,393  coast of Vancouver Island appearsto originate from the northerntip of Vancouver Island and then flow southward past Brooks Peninsula.This cool water band can often be observedfrom individual images during summer and its somewhatuncertainsource(from northerntip of Vancouver Islandor from BrooksPeninsula)hasbeenarguedby lkeda and Emery [1984] andEmery et al. [1986]. Another cool watercoreoriginatesfrom the mouthof Juande Fuca Strait.  The flow of thiscoolwatermustbe affectedstronglyby the dischargefrom the strait,and by upwellinginducedby the complex local topography,e.g., the Juan de Fuca canyon [Freeland and Denman, 1982]. From numerical simulations  [Weaver and Hsieh, 1987], the dischargeshould flow northwardalong the coast. However, with shallowbanksto Fig. 2. The solid grid indicatesthe center of re-sampledpixels, its northforminga northern"barrier",the cool surfacewater whereasthedashedgrid showsthecenterof theoriginalpixels. appears to deflect seawardsduring its northward flow [Thomsonet al., 1989]. The persistenceof this feature With,thispresentation, F• (x) is calledthe spatialamplitude, implies that despite upwelling favorable wind, which which presentsthe spatial covariancepattern of an EOF opposesthe flow, this cool water still flows northward mode, while the coefficienta• (t) is called the temporal becauseof peak dischargefrom the strait during summer. amplitude,which describesthe time variationsof an EOF The smallpatchof warm surfacewater off Barkley Sound separatesthe Juande Fuca cold tongueand the cool water mode. The EOFs are calculated by solving the eigenvalue problem of the data covariance matrix (formed by multiplying the data matrix T' by its transpose).Insteadof solvingfor the eigenvectorsof the covariancematrix (the "covariance method"), one can also calculate singular vectors by using singular-value decomposition(SVD). Hence, there are at least two different ways to calculate  off the northwest coast of Vancouver Island. This warm feature has also been observed from the summertime  individualimagesby Emery et al. [1986], who suggested that this warm water over the shallowbanksoff Barkley Soundwas relatedto solarheatingin thisregion,especially in Barkley Sound.Thomsonet al. [1989] systematically described thecirculationmechanism overtheshallowtopographically complex banks, pointing out that the summer EOFs, butthesolufion•is uniquewhenthereareno data circulation over the region was weak, and tidal rectification missing[Kelly, 1988]. With SVD, missing data must be estimated [Kelly, 1985; Paden et al., 1991], but fewer and outflow of the water in Barkley Sound played an computationsare required. In contrast, the use of the importantrole in the region. The velocity EOF analysisin covariancemethodallows one to simply skip over missing thisregionby Hickey et al. [1991], showedthe first EOF of data.The latter methodis usedin this study.For the details velocity at 30 rn depth off Barkley Sound to be directed of the covariancemethod and the processingtechniquefor weaklyoffshore,confirmingthe seaward-flowingfeatureof calculating the data covariancematrix from the data set this warm water band. GradientEOF mode 2, accountingfor 12% of the total T'(x,t) whererowsoumumbercolumns,seeLagerloefand variance,showsa zero-crossing alongthe continentalslope Bernstein [1988]. (Figure 3b). Except for being further offshore, the zero crossingis closely aligned with the 200-m depth contour 4. EOF ANALYSIS OF THE SST PATYERNSOFFVANCOUVER which is about65 km offshorein the southernportionand ISLAND  less than 5 km offshore off Brooks Peninsula. This 200m-  depth contour indicates the shelf-break region where The spatialpatternsof the first four gradientEOF modes upwelling is a commonfeature in summer[Freeland and arepresentedin Figure3, andtheircorresponding temporal Denman, 1982; lkeda and Emery, 1984; Denman and amplitudesin Figure 4. The first gradientEOF (Figure 3a), Freeland, 1985]. The seaward extension of the cool water accountingfor 33% of the total variance, showspositive beyond the shelf break may be causedby the offshore spatialamplitudesin the southwestcomer of the studyarea Ekman transportin wind-induced upwelling. lkeda and and in a small region off Barkley Sound. The negative Emery [1984] noticed that the cool water boundary spatialamplitudescover the northeastarea offshorefrom propagatedoffshore at 10 km/d, extending eventually the northwest coast of Vancouver Island and the region beyond the shelf break. The largest negative spatial offshorefrom the Strait of Juande Fuca. As the temporal amplitudeoccurredjust northof the Juande Fuca canyon, amplitudeof this modewas almostalwayspositive(Figure wherea topographicallycontrolledcycloniceddy is often 4a), the negative spatial amplitudesrepresentthe cooler found [Freeland and Denman, 1982; Denman and Freeland, water. For the area more than 75 km offshore,the spatial 1985; Weaver and Hsieh, 1987]. When the temporal amplitudes decreasefrom south to north, revealing a amplitudeof this mode (Figure4b) is positive,this pattem northwardcooling trend. The spatial pattern of the first describes the topographically controlled, wind-induced mode is very similar to the mean temperaturefield from upwellingoverthe entireshelf-sloperegionwith the coolest averagingover all 133 images (not shown). But mode 1 surfacewaterlocatedin the southeast region.Freeland and provides more information than the mean pattern by Denman [1982] also found summertimeupwelling to be revealingthe temporalbehavior(Figure4a) of thispattern. especiallypronouncedat the edgesof the southernbanks In Figure 3a, the cool water band along the northwest owingto the presenceof majorcanyons.During upwelling,  14,394  FANG AND HSIEH:SUMMERSEA SURFACETEMPERATURE VARIABILITY  (A) SPATIALA•PLITUDE FOR EOF MODEONE  0  50  100  150  200  250  300  350  ALONGSHORE DISTANCE(I•)  (B) SPATIALAN[P••E  0  50  100  150  FOREOF MODETWO  200  250  300  350  ALONGSH01•DISTANCE (K•)  Fig. 3. Spatialamplitudepatternsfor theSST gradientEOF modes1 to 4. The spatialdomainextends150km offshoreand350 km alongshore. In thealongshore direction,themouthof theJuande FucaStraitstretches from 0 to 30 km, Barkley soundfrom 70 to 90 km, and Brookspeninsulaprotrudesfrom the coastat 250 km alongshore.  the temporal variabilityof this modemay alsobe relatedto theremotelywind-drivencoastallytrappedwaves.This type  of upwelling structuremay be present during summer despite weak or downwelling-favorable local winds [Thomson et al., 1989].  Explaining 10% of the total variance,the third gradient EOF mode shows a large tongue of negative amplitude extendingsouthwestward from Brookspeninsula.A similar cold water band has been observed  in 1980 summertime  amplitudeis positive,thismoderepresents thepresenceof a "squirt"or seawardjet of cool water.In the southernregion, the amplitudevaluesare positive,implyingan out-of-phase relationwith the squirtregion. The fourthmodeexplains5% of the total variance. When the temporalamplitudeis positive,this moderepresentsa cold plumeoriginatingoff CapeScottat the northerntip of VancouverIslandand extendingseaward.For thesehigher modes, we need to determine if they are statistically significant.A significance testintroduced by Overlandand Preisendorfer[1982] usesthe Monte Carlo techniquefor selectingEOFs. However,whenthe data set is very large,  images[lkedaandEmery, 1984].They arguedthatthisband appearedto be advectedby the southwardflow from the coolernortherndistrict,but could also be the expressionof upwelling at the local shelf break. When the temporal the cost of Monte Carlo simulations becomes excessive.  FANG AND HSIEH: SUMMER SEA SURFACE TEMPERATURE VARIABILITY  14,395  (C) SPATIALAMPLITUDEFOR EOF MODETHREE  0  50  100  150  200  250  300  350  ALONGSHORE DISTANCE(KM)  (D) SPATIALAMPLITUDEFOR EOF MODEFOUR  ::iili ...... i!i!i::::ii::iiii!iiiiiiiiii!iii•.i::i !ii !i::!• I iii::i !ii!i •i'1,i• k• .  o  0  50  100  150  •NCSHO•  200  250  300  350  DIST•CE (•)  Fig. 3. (continued)  Instead, theasymptotic theory forlargedatasets canbeused increasesmonotonicallyfrom Juneto September,revealing [Preisendorfer,1988, p. 204-205]. As the data set in this studyis large, we usedthe the asymptoticmethodto test the significanceof the EOFs. From Preisendorfer [1988, equation5.18] (but with the incorrectminussignpreceding  (ab)'/2 replaced by a plussign),we foundthatthefirst13  the intensificationof the mode 1 pattern(Figure 3a) as the summerprogresses.Mode 2 also intensifiesfrom June, peakingin August.Mode 3 showsa declinein thetemporal amplitudein Augustandespeciallyin September, implying a decreasein the likelihood of a strongsquirt off Brooks Peninsulaas summerprogresses. The temporalamplitudesfor the EOFs were alsoaveraged over eachsummer(Table 1), revealingthat the mode 1 type of SST patternwas strongestin 1990. The mode 2 type of upwellingwas strongestin 1991 and 1986. In order to comparewith the gradientEOFs, covariance EOFs (i.e., with the temporalmeanremovedfrom the data as in (2)) were also calculated.The first three covariance  EOF modeswereabovethe 99% significancelevel. In summary,the first four gradientEOF modes,altogether accountingfor 60% of the SST variance,representedsome of the well-known featuresin the region. The first mode closelyresembledthe meansummerupwellingpattern,the second,the topographically controlledupwelling,the third, the squirtoff BrooksPeninsula,andthe fourth,the northern plume off Cape Scott. Figure 5 shows the temporal amplitude averaged over the particular summer months EOF modes(not shown) accountedfor 55.8, 6.9, and 6.6 % during 1984-1991. The temporal amplitude for mode 1 of the total variance, respectively.The covarianceEOF  14,396  FANGANDHSIEH:SUMlVlER SEASURFACE TEMPERATURE VARIABILITY  (a) Mode 1 1984  1985  1986  1987  1988  1989  1990  1991  '  J  A  S  J  A  S  J  A  S  J  J  A  S  J  J  A  S  J  J  A  S  J  A  S  J  A  S  (b) Mode2  984 1985 1986 . 1987 I ' 1988 1989 1990 1991 ß  I A  I  I  S  I J  I A  I  t ' t['L*J t' I  S  I  J  I  A  I  I  S  I  J  I  J  I  A  I  I  S  I  J  I  J  A  ,  '  I  I  S  I J  J  A  S  J  A  ! I S  I J  !  A  I  S  (c) Mode 3  984  19  1986  1987  !  1988  T  1989  1990  .  •  .  I J  A  S  I  I J  I A  I S  I  I J  I A  I S  I  , J  I  J  A  1991  S  , J  J  ! A  .  I S  I J  I J  I A  I S  •  '  .  I  I J  I A  I S  I  I J  I A  I S  (d) Mode 4  984 1985 1986 1987 1988 1989 • 1990 1991 J  A  S  J  A  S  J  A  S  J  J  A  S  J  J  A  S  J  J  A  S  Fig. 4. 'remporalamplitudefor the SST gradientEOF modes1 to 4.  J  A  S  J  A  S  FANG AND HsmH: SUMMER SEA SURFACETEMPERATUREVARIABILITY AVERAGED  TEMPORAL  AMPLITUDE  TABLE  OF EOF MODES  0.1  14,397  1. Summer SST Indices off Vancouver Island from 1984 to 1991  Mode 1  Summer a• 1984 1985 1986 1987 1988 1989 1990 1991  0.05  4  .......  .........  0-  •"----.--...,.•.' _•,..•.  '"'1-  • .,•..•"•2 •' '•" ''w'x ß  \  -0.05  2  7.5 6.3 8.1 5.9 9.0 7.8 9.4 8.0  a2  a•  an  d  C  1.0 1.9 4.7 1.2 -2.2 -5.5 0.5 4.9  1.0 5.2 -1.4 -1.3 -3.0 -7.7 0.4 0.4  -4.1 4.5 3.1 -0.3 -0.3 7.8 -1.6 -4.0  70 73 71 65 65 72 65 80  0.59 0.64 0.92 0.54 0.51 0.42 0.81 0.86  Monthly averagesof the indiceswere first calculatedfrom the imagesfor the monthsJuneto September,thenaveragedto form thesummerindex.,Thea. arethesummervaluesof thetemporal amplitudes(in %) of the gradientEOF modes1 to 4. The d are the offshore locations of the from in kilometers, and C are the  \.3  coolnessindicesin degreesCelsius.  each summer, indicated that coolest coastal water occurred  in 1986 and 1991 (Figure 7). As the C index measuresthe average temperature anomaly in the coastal upwelling JUN JUL AUG SEP regionwith respectto the temperature in theoffshoreregion, thesestronganomaliesin theC indexindicatedthattheshelf Fig. 5. The monthlyaveragedtemporalamplitudesof the gradient waterwasrelativelycool in the summersof 1986 and 1991. EOF modes 1 to 4 over all years, revealing the monthly The mean offshore location of the front (Table 1) was also progression of the SST patternsas summeradvances. greatestin 1991 (80 km offshore).The years1987, 1988 and 1990 had the front closestto the shore(at 65 km offshore). spatialpatternsfor modes 1, 2, and 3 resembledgradient As the satellitedata are availableonly during cloud-free EOF modes1, 3 and 2, respectively,thoughthe covariance periods,we have to questionthe reliability of our summer EOF patternswere much noisier than the corresponding upwellingindex. To test the interferenceof clouds,we gradientEOF patterns. deliberately computed C with data missingat a 2-week period during different times in each summer.Figure 8 showsthatthe fluctuationsof theseC valuescomputedwith 5. A CooLmss INDEX missingdata were in generalsmaller than the interannual variability in C, therefore giving us confidencethat the It would be useful to define a simple index that interannualvariability in C observedwas above the noise characterizesthe relative strengthof the coastalcool water. generatedby missingdata from cloudy periods.Figure 8 By calculatingthe differencebetweenadjacentpixel values, also showsthat missing data in June(when upwellingwas we located the maximum temperature gradient in the still immature)tendedto result in C being shifted higher, offshoredirection,therebydefiningan offshoreSST front a whereasmissingdata in September(when upwelling was distanced from the coast(Figure 6). We then chosea line, mature)tendedto resultin C being shiftedlower. 7.5 km seawardfrom the SST front, as a boundaryto divide The coolnessindex C covers two E1 Nifio events (1986the studyarea into two: the shorewardpart representingthe 1987 and 1991-1992). As the strongestC occuredin the cool, upwelling region, and the seawardpart, the warmer summersof 1986 and 1991, this suggeststhat the C index openocean.The overall coolnessindex C, which represents leads the E1 Nifio events. This result was tested by a regional average of the relative massof cold water, is calculatingthe laggedcorrelationbetweenthe C index and defined as the seasonalSouthern Oscillation Indices (SOI) (i.e., the -0.1  -  I  ......  253  I  105  I  I  --  Y• Z(Toc.a.(J)- T(I,J))S(I,J)  C--  •  •o•  (6)  J=[t=[(•) Pacific  Tocean(J)  (I,J)  where I, J denote the onshoreand alongshoreindices of eachpixel, S(I,J) the pixel area and T(I,J) the temperatureat  thatpixel. To•ea , (J) denotesthetemperature averagedalong row J over the region seaward of the partitioning line I = f(J) (shown as the dashedcontour in Figure 6). The summationsin (6) are over the shorewardregion from the partitioningline I = f(J). The mean coolness index C for each summer, estimated  by averagingthe individualC valuesfrom eachimageover  Vancouver  Island  Warmer :ø1 /  Cooler  water  oo?water  Fig. 6. Cool and warm regionsusedfor calculatingthe coolness index C as discussed in the text.  14,398  FANGANDHSIEH:SUMMERSEASURFACE TEMPERATURE VARIABILITY  Data omitted I --- 16-30/Jun 2 -- 01-15/Jul 3-- 16-31/Jul  --  4 -- 01-15/Aug 5--- 16-31/Aug  1984  1985  1986  1987  1988  1989  1990  6 -- 01-15/Sept 7 -- 16-30/Sept  1991 .  Fig. 7. Summercoolnessindex C off VancouverIsland derived from satellite-sensed SST data (using images from June to September).The index C definedby (6) indicateson averagehow much the SST in the coastalregion is coolerrelativeto the offshore SST.  o  •  I  , I  1984  1985  ,  I  ,  1986  I  I  1987  I 1988  ,  I  ,  1989  I 1990  ,  I 1991  Fig. 8. Fluctuations in the summercoolnessindexC asdatafrom various 2-week intervals are omitted in the computationof C to simulatemissingdata from clouds.The C valuescomputedwith dataomittedare markedby the symbols1 to 7 corresponding to the  spring, summer,autumn and winter meansof the monthly SOI as definedby Chelliah [1990]). The highestcorrelation sevenpossible2-weekintervalsomitted,andthe C valueswith no of-0.92  was attained when the summer C was correlated  with the SOI from the following spring,which meansthat cool summer coastalSST (high C index) off Vancouver Islandprecededthe E1 Nifio warm events(low SOI), which peakedat aroundthe following spring.From assumingone degreeof freedom for each year, a correlationof-0.92 is significant at the 99% level, though for such short interannual time series, the significancelevel cannot be determinedreliably. Clearly a muchlongertime seriesfor C is neededbefore one can confirm this relationshipbetween the two indices.  6. CORRELATIONSBETWEENEOFs AND VARIOUS INDICES  missingdata are markedby the + symbol(linked by the dashed curve).  temporalamplitudesshowedthe highestcorrelationof 0.70 with mode2, the next highestcorrelationof 0.59 with mode 1, and insignificantcorrelationswith modes3 and 4. This suggeststhat the first two EOF modesstronglyinfluenced the overall coolness index C.  The satellite data tend to suffer from fair-weather bias,  i.e., satellite images are available only during cloud-free days with high pressureover the northeastPacific, which correspondto days with upwelling favorablewinds. Thus, despitethe bias in the satellitecollection,the occurrenceof strongupwellingshouldcoincidewith daysof highpressure and clear weather, and hence would have been included in  To examinetherelationshipof the dominantgradientEOF modes with various indices, we calculated the correlations  between the weekly averagedtime seriesof the first four gradientEOF temporalamplitudeswith variousrelatedtime series. From the daily Bakun upwelling indices (kindly providedby NOAA/NMFS PacificFisheriesEnvironmental Group, Monterey,California, at two offshoresites,(48øN, 125øWand 51øN, 131øW), we computedthe weekly Bakun indicesby averagingthe daily data over the days when satellite images were available. The Bakun coastal upwelling indices were based on calculationsof offshore Ekman surface wind transport from surface atmospheric pressuredata [Bakun, 1973]. Weekly time seriesfor the coolness index C and for the Fraser River discharge measuredat Hope, British Columbia, were preparedin a similarway. The correlation coefficients between the time series of the EOF modes and various indices are listed in Table 2. Correlations of the coolness index C with the first four EOF  TABLE  2. The Cross Correlations Between Time Series of the Gradient EOF Modes and Various Indices  C  C  Mode  1  0.59/99%  Mode2  0.70/99%  Mode3  0.07/30%  Mode4  0.01/6%  Bakun. S -0.13/50% -0.10/48% -0.05/21% 0.24/88% -0.10/54% Bakun. N -0.06/39%-0.30/96%  0.20/80%  0.30/96%  0.34/98%  Discharge-0.43/98% -0.23/84% -030/92% 0.24/91% -0.26/91% The levels of significance given after the correlation coefficients, were calculated with the number of degrees of freedomestimatedby the methodof Davis [1976]. Correlations above the 90% significancelevel were printed in bold, and they  all have at least30 degreesof freedom.C denotesthe coolness index, while Bakun.S and Bakun.N represent the Bakun upwelling index at 48øN, 125øW and at 51øN, 131øW, respectively.  FANGANDHSmH:SUMMERSEASURFACE TEMPERATURE VARIABILITY  our data collectionand incorporatedinto our indices.Table 2 also showsthe correlationbetweenour indiceswith the Bakunupwellingindices.The Bakun index at 48øN, 125øW (Bakun.S), lying just south of our study region, showed generallylow correlationswith the EOF amplitudes.The Bakunindex at 51øN, 13IøW (Bakun.N), lying just to the northwestof our domain showedhigher correlationswith the EOF amplitudes.Except with the first EOF amplitude, the positive correlationsbetween Bakun.N and the other EOF amplitudesmeant that an upwellingfavorableBakun index  tended to concur  with  cool coastal  SST  in these  modes.Sincemodes3 and 4 involved,respectively,plumes off BrooksPeninsulaand Cape Scott, both in the northern part of our studyregion,it is not surprisingthat they were correlatedwith the Bakun index to the north (Bakun.N) than with Bakun.S. The highest correlation, which occurred betweenmode4 and Bakun.N, is expectedsinceof the four modes,mode4 with the Cape Scott plume must have its upwelling most strongly affected by winds around the northern part of our domain. The negative correlation betweenthe first EOF amplitudeand Bakun.N seemedto suggestthat an upwellingfavorableBakun index tendedto  concurwith theweakeningof themode1 coastalupwelling pattern,in contradiction to ourexpectations.To explainthis paradox,we note that the mode 1 amplitude time series (Figure 5) has a notable seasonaltrend, indicatingan intensificationof the pattern of Figure 3a from June to September. The correspondingseasonal trend for the Bakun.N index showeda decreasefrom Juneto September (not shown). Hence the negative correlation between the two couldsimplybe due to oppositeseasonaltrends.  14,399  fourthmodeshoweda cool waterplumeextendingoff Cape Scott at the northern tip of Vancouver Island. These 4 modesaccountedfor respectively,33, 12, 10 and 5% of the SST variance.  As 60%  of the total variance  can be  accountedfor with only four modes, we concludethat the  EOF methodis highly effective in condensingthe huge amount of satellite SST data off Vancouver Island.  CorrelationsbetweenBakun upwellingindicesand our EOF temporalamplitudesshowedthat SST in our study regionwas significantlyinfluencedby the Bakun index to the northwestof our domain but insignificantly by the Bakun index immediatelyto the southof our domain.That the wind (as representedby the Bakun index) to the north had a greater influence than the wind to the south of our regioncouldbe explainedby the fact thatmodes3 and4 had plumes extending southwestwardfrom Brooks Peninsula and Cape Scott, both in the northernpart of our domain. TheseSST EOF time seriescan potentiallybe usedas new upwellingindices,providinga measureof the fine features of coastalupwelling, which the Bakun index, basedon large-scalegeostrophicwind, cannot. From these images, we also constructedan overall coolnessindex C, measuringthe coolnessof the coastal upwelling region relative to the offshore region. The negative correlation between C and the Fraser River dischargetendsto suggestthat high dischargeconcurswith warmerSST in our studyregion. Interannualvariabilityin theaveragevalueof C over eachupwellingseasonrevealed the summers of 1986 and 1991 to have the coolest coastal  water. Thesetwo cool anomaliesslightlyprecededthe last two ENSO events(1986-1987 and 1991-1992).  While the coolness index C was not correlated with either  Bakun.Nor Bakun.S,C andtheFraserRiver dischargehad a significantnegativecorrelationof-0.43 (Table 2). In summer, Fraser River water and thus surface water in the  Acknowledgments.We are indebtedto Denis Laplanteand GordonStaplesfor the satelliteimages,and to W. J. Emeryfor startingthe satellite collection. D. Bancroft and W. R. Crawford  Straitof Georgiaarenormallywarmerthanthatin the Strait kindly drew our attentionto the fair-weatherbias in the data.This of Juande Fuca,wherethereis muchtidalmixing.Hickeyet researchwas supportedby the Canadian Natural Sciencesand al. [1991] reviewedthe transportprocessof the monthly EngineeringResearchCouncil and the Departmentof Fisheries pulseof warm, fresh water exiting the Strait of Georgia and Oceans. from satelliteSST images.They pointedout thatthe outflow of this anomalouslywarm freshwater was one of the causes REFERENCES for warmer water appearingat the mouth of the Juan de FucaStrait.A largeoutfloweventfromFraserRiverimplies Bakun, A., Daily and weekly upwelling indices,west coastof that more warm water flows throughthe Strait of Juande North America, 1946-71,NOAA Tech.Rep., NMFS-SSRF-671,  Fuca, which diminishes the overall coolness index of our  U.S. Dep. of Cornmet.,Washington,D.C., 1973.  study region, hence the observed negative correlation Burgert,R., and W. W. Hsieh, Spectralanalysisof the AVHRR betweendischargeand C. Dischargewas alsonegatively sea surface temperaturevariability off the west coast of Vancouver Island, Atmos.-Ocean, 27, 577-587, 1989. correlated with theEOF amplitudes in Table2 exceptwith mode 3, where the marginally significant positive Chelliah,M., The globalclimatefor June-August,1989:A season of near normal conditionsin the tropicalPacific. J. Clim., 3, correlationsuggests that strongdischarges tendedto concur 138-162, 1990. with theappearance of thecoldsquirtoff BrooksPeninsula. 7. SUMMARY AND DISCUSSION  Davis,R. E., Predictabilityof seasurfacetemperature andsealevel pressureanomalies over the north Pacific Ocean. J. Phys. Oceanogr.,6, 249-266, 1976. Denman,K. L., andH. J. Freeland,Correlationscales,objective mapping,andstatisticaltestof geostrophyover the continental  EOF analysisof the spatialvariancesfor 133 nearly cloud-freeSST imagesprovidedus the first systematic shelf, J. Mar. Res., 43, 517-539,1985. classificationof the summerSST patternsoff Vancouver Emery, W. J., and L. A. Mysak, Dynamicalinterpretationof thermalfeaturesoff VancouverIsland, J. Phys. Island,andto follow theirevolutionas summerprogresses. satellite-sensed Oceanogr.,10, 961-970, 1980. The first EOF modeof spatialvarianceresembledthe mean Emery, W. J., A. C. Thomas, M. J. Collins, W. R. Crawford, and SST patternobtainedfrom averagingall images,while the D. L. Mackas,An objectivemethodfor computingadvective secondmoderevealedupwellingcontrolledby the bottom surfacevelocitiesfrom sequentialInfraredsatelliteimages,J. topography.The third mode correspondedto cool water Geophys.Res., 91(Cll), 12,865-12,878, 1986. extendingsouthwestwardoff Brooks Peninsula, while the  Freeland,H. J., and K. L. Denman,A topographically controlled  14,400  FANGANDHSmH:SUMMERSEASURFACE TEMPERATURE VARIABILITY  upwelling center off southernVancouverIsland. J. Mar. Res.,  Part I, Sea surfacetemperaturevariability,J. Geophys.Res.,  40, 1068-1093, 1982.  96(C10), 18,337-18,359, 1991.  Hickey, B. M., R. E. Thomson, H. Yih, and P. H. LeBlond, Preisendorfer, R. W., Principal component analysis in Velocity and temperaturefluctuationsin a buoyancy-driven meteorologyand oceanography,Elsevier,New York, 425 pp., 1988. currentoff VancouverIsland,J. Geophys.Res., 96(C6), 10,50710,538, 1991.  Thomson, R. E., and J. F. R. Gower, A wind-induced mesoscale  Ikeda, M., and W. J. Emery, A continentalshelf upwelling event off VancouverIsland as revealedby satelliteinfraredimagery, J. Mar. Res., 42,303-317,  1984.  Ikeda, M., L. A. Mysak, and W. J. Emery, Observation and  eddyover the VancouverIsland continentalslope,J. Geophys. Res.,90(C5), 8981-8993, 1985. Thomson,R. E., B. M. Hickey, andP. H. LeBlond,The Vancouver Island coastal current: fisheries barrier and conduit, Effects of  modelingof satellite-sensed meaders andeddiesoff Vancouver  Ocean Variability on Recruitment and an Evaluation of Parmeters Used in StockAssessment Models, editedby R. J. Island,J. Phys.Oceanogr.,14, 3-21, 1984. Beamishand G.A. McFarlane,Can. Spec.Publ. Fish. Aquat. Kelly, K. A., The influenceof winds and topographyon the sea Sci., 108, 265-296, 1989. surfacetemperaturepatternsover the northernCaliforniaslope, J. Geophys.Res.,90(C6), 11,783-11,798,1985. Wang, D.C. C., A.H. Vagnucci, and C. C. Li, Digital image Kelly, K. A., Commenton "Empiricalorthogonalfunctionanalysis enhancement:A survey, Comput. Vision Graphics Image Processing,24, 363-381, 1983. of advancedvery high resolutionradiometersurfacetemperature patternsin SantaBarbaraChannel"by G. S. E. LagerloefandR. Weaver, A. J., andW. W. Hsieh, The influenceof buoyancyflux from estuaries on continental shelf circulation, J. Phys. L. Bernstein,J. Geophys.Res.,93(C12), 15,753-15,754,1988. Oceanogr.,17, 2127-2140, 1987. Lagerloef, G. S. E., and R. L. Bernstein, Empirical orthogonal function analysisof advancedvery high resolutionradiometer surfaceseatemperature patternsin Samta BarbaraChannel, J. Geophys.Res.,93(C6), 6863-6873, 1988. Lauritson, L., G. Nelson, and R. W. Porto, Data extraction and W. Fang and W. W. Hsieh, Departmentof Oceanography, calibration of TIROS-N/NOAA A-G radiometers, NOAA Tech. Universityof British Columbia,Vancouver,B.C., Canada V6T Memo., NESS 107, 44-46, 1979.  1ZA.  Overland,J. E., and R. W. Preisendorfer,A significancetest for principal componentsapplied to a cycloneclimatology,Mon. Weather Revi., II0, 1-4, 1982. Paden, C. A., M. R. Abbott, and C. D. Winant, Tidal and  atmospheric forcingof theupperoceanin theGulf of California,  (ReceivedJuly29, 1992; revisedApril 8, 1993; acceptedMay 4, 1993.)  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0041794/manifest

Comment

Related Items