JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D7, PAGES 8781-8791, APRIL 20, 1997
Variations in tropospheric ozone related to transport at American
Samoa
JoyceM. Harris and SamuelJ. Oltmans NOAA Climate Monitoring and DiagnosticsLaboratory,Boulder,Colorado
Abstract. Ten yearsof isentropictrajectorieswere summarizedusingclusteranalysisto describeflow patternsfor AmericanSamoa. The trajectorieswerethenpairedwith surface ozonedatato determinethe dependence of surfaceozoneon transport.The two maintransport regimesaffectingsurfaceozonearetradewind transport,wheretrajectoriesshowflow bringing ozonefrom the eastin the tropicalmarineboundarylayer, and midlatitudetransport,where trajectoriesshowstrongwesterlyflow at higherelevationsof southernmidlatitudes,followedby descentwith anticycloniccurvature.Thesetwo transportregimesyield ozonefrom distinctly differentorigins,havingdifferentmixing ratios. The seasonallychangingfrequencyof transport typeis shownto be partlyresponsible for the seasonal cycleandchangesin variabilityof Samoa surfaceozone. On average,45% of winter ozonevariationcanbe explainedby differencesin transporttype. This strongrelationshipwasabsent,however,during1991,probablybecauseof UV blockingby aerosolsfrom the eruptionof Mount Pinatubo. Reducedtotal columnozone duringwinter 1992may havecontributed to this seasonhavingthe lowestsurfaceozonelevelsof the studyperiod.
1. Introduction
Observatoryand gave examplesshowingthe benefitsof using
The Climate Monitoring and DiagnosticsLaboratory(CMDL) of the National Oceanic and Atmospheric Administration (NOAA) operatesan observatoryon Cape Matatula, American Samoa(14.25øS,170.56øW) with the purposeof measuringtrace constituentsof the backgroundatmosphere,far removedfrom the pollutionof urbanenvironments.In an effort to identify origins of air arriving at Samoa Observatory (SMO), 10 years of isentropictrajectorieswere summarizedas averageflow patterns. These flow patternswere identified by "clustering"trajectories into groups using cluster analysis, a multivariate statistical technique. Trajectorieswere clusteredfor the entire 10 years,by individual month (i.e., 10 years of Januarytrajectories,10 years of February trajectories, etc.) and by season. The known meteorologicalfeaturesof the region were examinedto explain the clustersand their seasonalchanges. SMO surfaceozonedata were then pairedwith trajectoriesto identify ozone characteristics that canbe linked to transport,and possiblesourcesand sinksfor the region. Finally, analysisof variance (ANOVA) techniques were used to estimate the relative importance of transport, seasonality,and year-to-yeardifferenceson SMO surfaceozone levels. This methodologywas applied in a manner similar to Moody et al. [ 1995]. This studyis the fourth in a seriesof papersthat has detailed the flow regimes at each of CMDL's baseline observatories. Harris and Kahl [1990] presentedisobaricpatternsat both the 500- and 700-hPa levels for the Mauna Loa Observatory on Hawaii. That first paper in the seriesalso containsan appendix that documentsthe cluster analysis technique used in all four papers. Harris [ 1992] describedthe flow regimefor SouthPole Thispaperis notsubjectto U.S. copyright. Published in 1997 bytheAmericanGeophysical Union. Papernumber97JD00238. 0148-0227/97/97JD-00238509.00
isentropic instead of isobaric trajectories. Harris and Kahl [1994] showed flow patterns for the observatoryat Barrow,
Alaska, for the first time building the climatologyand cluster meanswith isentropictrajectories.That samepaperrecordsthe methods used by the CMDL transportmodel to produce isentropictrajectories. The presentstudybuilds on the work of Bortniak[1981], who detailedthe wind regimein the regionof SMO and listed sourcesof meteorologicaldata for the Pacific. Numerous studies have successfully linked trace gas characteristicsto transport;for example, Halter et al. [1988] showedeffects of transporton the SMO carbon dioxide record,
andHarris et al. [ 1992] examinedtrajectoriesin orderto explain seasonal characteristics
of methane
measured
at Mauna
Loa.
Savoie et al. [1992], Moody et al. [1995], and Oltmans et al. [1996] usedtrajectoriesto describeflow patternsthat affect the level of ozone observed at sites in the North Atlantic.
The trajectoriesanalyzedfor the currentstudy span the 10year periodfrom 1986 through1995. Ten-dayback isentropic trajectorieswere producedtwice per day at 0000 and 1200 UT arrival times. Trajectories were calculated to arrive at approximately 500 m abovesealevel, an elevationrepresentative of the boundarylayer abovethe observatory (77 m). Input to the trajectorymodel consistsof meteorologicaldata (U and V wind components,temperature,and height) from the EuropeanCentre for Medium RangeWeatherForecasts.Thesedata are produced on 2.5ø latitude,longitudegridson 10 isobaricsurfacesthrough the troposphere.
Sourcesof uncertaintyin trajectorieshave been discussed by Harris and Kahl [1994] and Merrill [ 1996] and include diabatic
effects, observationalerrors, interpolation,assumptionsabout vertical motion, and sub-grid-scalephenomena. The transport model used here assumesthat the air is dry and travels adiabatically; thatis, potentialtemperature is conserved alongthe trajectory. This condition determinesthe vertical motion of a modeled air parcel as it flows along an isentropic surface.
8781
8782
HARRIS
AND OLTMANS:
OZONE RELATED
TO TRANSPORT
AT SAMOA
Figure 1. The averagewinter surfacewind patternsin the Pacific,from Bortniak[ 1981].
Generallyspeaking,isentropicsurfacesare domedoverthe poles in the troposphere,and therefore air will descend if it is transportedfrom higher to lower latitudes. These simplifying assumptionsare necessarybecause of the complexity of the atmosphereand the inadequaciesof the griddedmoisturedata. We believethat despitethesesourcesof uncertainty,trajectories can be useful in determiningthe origins of air arriving at SMO. By analyzing such a large number of trajectories,we could smoothsomerandomerrors,and a reasonablerepresentation of the large-scale circulation was possible. Because of uncertainties,an individual trajectory at times may deviate somewhatfrom the actualpath taken by an air parcelmeasuredat SMO.
Surface ozone has been measured at SMO
Dasibi ozone monitor.
since 1976 with a
This instrument uses absorptionof
ultraviolet (UV) radiation at 254 nm to determine ozone amount
in ambient
air.
The measurement
is tied to a CMDL
network
standardand linked by intercomparison with the standardozone photometer maintained by the U.S. National Institute of Standardsand Technology. For the presentstudy, only ozone data since 1986 were consideredin order to overlap with trajectories.Most of 1993 and all of 1995 surfaceozonedataare missingbecauseof instrumentproblems.
2. MeteorologicalFeaturesAffecting Transport The broad meteorologicalfeaturesthat affect flow in the Pacific are shown in Figures 1 and 2 (winter and summer, respectively; from Bortniak[ 1981]). Thesefiguresmostclosely match conditionsfor August (austral winter) and February (australsummer). Note in Figure 1 the positionsof the two convergencezones, the intertropicalconvergencezone (ITCZ)
Figure 2. SameasFigure 1, but for summer.
HARRIS AND OLTMANS' OZONE RELATED TO TRANSPORT AT SAMOA
8783
and the intratropical or what we will refer to here as the South
Pacific convergencezone (SPCZ). The ITCZ has an average
20 o N
position during the year at about 5øN, but it marks the meteorological equator where air streaming off the two semistationarysubtropicalanticyclonesmeet. The SPCZ extends diagonally across the South Pacific and in winter is close to
SMO. Air flowing aroundthe Australiananticycloneconverges with southeasterlytrades in this region. The position of the SPCZ
to the south or north of SMO
determines
200 S
the two main
transportregimes in winter: trade wind flow or flow from the
40 o S
southern midlatitudes.
These main wintertime transporttypesmay be illustratedwith sampletrajectories. Figure 3 showstrajectoriesarriving at SMO in the trade wind regime. The arrival elevation is about 500 m 600 S above SMO. At most, 1 km of descentis seenin this type of 1200 E 140 '0E 160 '0E 180 '018 160 '018 140'ø18 120 '018 100ø 188' trajectory. Ten days back from SMO the air parcelsare in the ' SOLID - 00UT remotesoutheastern Pacific. The 10-day originsare never close •t DASH12 UT to the South American continent.An example of the second winter transporttype is depictedin Figure 4, showingflow from 1 2 3 4 5 6 7 9 the southernmidlatitudesto SMO. This trajectory type may •Z 0 DAYS FROM SMO descend3 km or more as it curves anticyclonicallyen route to SMO. Note that as the air parcelsdescendin elevation, wind Figure 4. SMO 10-dayback isentropictrajectoriesarrivingat speedsdecreasesignificantly. For this transporttype, approach 500 m on June30, 1990, showingan exampleof transportfrom to SMO is often from the southeast.
During summer the position of the SPCZ is usually to the south of SMO, and therefore flow from the south is blocked, as
shownin Figure 2. The tradesremain the dominant flow regime, although they are not as strong as in winter and can be interrupted by passing storms. To the west of SMO an area labeled "monsoon low" marks the average position of low pressure where airstreams from the north, south, and east converge. If the ITCZ retreats eastward to about 160øW, occasionallyair from the northern hemisphereis transportedto SMO [Bortniak, 1981]. Figure 5 showsexample trajectories/'or this third transporttype. Halter et al. [1988] /bund that carbon dioxide transportedto SMO in this mannerwas in phasewith that
200 N
•,
midlatitudes.The numbersalongthe trajectoriesgive daysback from SMO. The graph along the bottom shows trajectory elevation.
measuredat CapeKumukahi,Hawaii, althoughthe amplitudeof the signalwasreduced.In the next sectionthe frequencyof the threeSMO transporttypesis discussed.
3. Transport Clusters We employedclusteranalysisto partitionthe trajectoriesinto groupsor clusters. The procedureuses a mathematicalcriterion
%
200 N
0ø '-'•oL 200S
•
20 ø S
400 S i•} •• 600 S
1200 E
'•
140 %E
160 'øE
400 S
180 '018 160 %18
140 %18
' 2i
i '"'ø'4I........5i
3
i
6
i
7
i
8
I
9
60 ø S
100ø18
SOLID DASH -- 12 00UTUT
c•
•
120 %18
120 øE
140 %E
160 %E
160 %188' 160 '0188' 140 '0188' 120 '0188' 100 ø18
-•-
I
10
DAYS FROM SMO
SOLID - ooUT
•c• • •
0
1
2
3
4
5
6
7
DASH -12 UT 8
9
10
DAYS FROM SMO
Figure 3. SMO 10-day back isentropictrajectoriesarriving at 500 m on August 7, 1989, showing an example of trade wind typetransport.The numbersalongthe trajectoriesgive daysback from SMO. The graph along the bottom shows trajectory
Figure 5. SMO 10-day back isentropictrajectoriesarriving at 500 m on February1, 1987, showingan exampleof transport from the northern hemisphere. The numbers along the trajectoriesgive days back from SMO. The graph along the
elevation.
bottom showstrajectoryelevation.
8784
HARRIS AND OLTMANS:
OZONE RELATED TO TRANSPORT AT SAMOA
50ø N -
20 øN
q
•
•
•.5%
ø
6%
6
40ø • • 1 3• 70 øS•-
200 E
,•"-•
500 E
800 E
i
110ø E
1400E
1700E
1600W
i
i
1300W
100ø W
Figure 6. Atmosphericflow patternsfor SMO depictedby cluster-mean backtrajectoriesarrivingat 500 m for the period 1986-1995. Plus signsindicate1-dayupwindintervals. At 10 daysupwindthe clustermeansare labeled with two numbers: the top numbergives the percentageof trajectoriesthat occurin the cluster,and the bottom number identifies
the cluster.
(K-meansclustering)to assurethat within the clusterstrajectories havesimilarshapeand length[Harris and Kahl, 1990;Norusis, 1993]. This technique was first applied to atmospheric trajectories by Moody [1986]. Eachclusteris represented by an averagetrajectorycalled the clustermean. Figure 6 showssix clustermeansderived for the entire 1O-yearperiod, summarizing over 7000 trajectories. Each cluster mean is marked with plus signsindicatingthe 1-day upwind intervals. At 10 daysupwind, the clustermeansare labeled with two numbers;the top number gives the percentageof trajectoriesthat occur in the clusterand the bottom number identifies
the cluster.
The cluster identifiers
(1-6) are used to distinguishone transporttype from the other; the numbersthemselveshave no inherentsignificanceand are not related from one clusteringto the next. The number of clusters chosen(six) was based on the goal of describingall of SMO's importantmeteorologicalfeaturesin a succinctway. Figure 6 givesa quantitativeestimateof the frequencyof each transporttype. For example, the rapid transportfrom southern
midlatitudesis represented by cluster2 and cluster3 with 2 and 7% frequencyof occurrence,respectively. These two clusters differ in the averagelengthof trajectorycontainedin the cluster, but both bring air parcelsto SMO from 3-km elevationor higher. As a result of subsidenceon approachto the observatory,wind speedsdecrease significantlycomparedwith initial speedsseveral days earlier. Cluster 4, with transportfrom the south,exhibits muchlower wind speeds,but sharesthe characteristics of 10-day average origin south of 30øS and anticycloniccurvature on approachto SMO from the southeast.This cluster,constituting 18% of trajectories,has a mixed characterbecauseit contains sometrajectories with tropicalorigins. Clusters 1, 5, and 6 in Figure 6 contain trajectoriesthat correspondto the easterlytrade wind transporttype. Cluster5 trajectories havea northerlycomponenton approachto SMO and occur 15% of the time, Cluster 6 trajectoriesapproachfrom due eastand occur32% of the time. Cluster 1 trajectorieshave lower wind speedsand a southerlycomponent,and occur26% of the time.
Transportfrom the northernhemisphereoccursin cluster5. This cluster also contains trajectories that do not cross the
meteorologicalequator. To estimatethe frequencyof transportof true northern hemisphere air, we counted the number of
trajectories thathad10-dayoriginsnorthof 5øN. Trajectories of this transporttype have a frequencyof 2-3% per year. They occurmostoften in Februaryand March at about 10% frequency, but can also occur in December, January, and April, though usuallyat lower frequency. The occurrenceof transportfrom the
northernhemisphereis quite variablefrom year to year, with frequencyin Februaryand March (the two most likely months) rangingfrom 0 to 25%. Trajectorieswere also clusteredby month and quarterto learn more about the seasonalityof flow patterns at SMO. The transportextremesoccur in Januaryand July (not shown) and demonstratea strongseasonalityin SMO transport. Summer months are characterized by minimal flow from southern midlatitudes,a northerly componentin some of the trade wind trajectories,and a cluster with trajectoriesapproachingSMO from the northwest. Winter months are characterizedby both easterlytradesand flow from the south. In fact this "winter" patternis presentfrom May throughNovember,July havingthe highestfrequencyof strong southerlyflow. Trade winds are evidentduringevery month. Some seasonalclustersare shown in the next section.
The abovedescriptiongivesthe horizontalextentof transport to SMO. Isentropictrajectoriesalso provideinformationabout the verticaltransportof air parcels,assumingdry adiabaticflow. Becauseof this simplificationof actual conditions,a trajectory that travelsthroughan area of heavy precipitationor convection will be subjectto addeduncertainty.Howeverbecauseisentropic trajectoriesdo take into accountadiabaticmotion,they are more realistic than their isobaric counterpartsand have been used successfullyin numerouschemicaland aerosoltransportstudies [Harris and Kahl, 1994; Merrill et al., 1985; Merrill, 1996; Oltmans et al., 1996].
Figure 7 addressesthe vertical motion of air parcels transported to SMO. The maximumheightalongeachtrajectory is plotted againstthe latitude at which this height was attained. One point is plotted for each trajectory in the 10-year set. Althoughthis plot saysnothingabouthow many daysfrom SMO
HARRIS ANDOLTMANS' OZONERELATED TOTRANSPORT AT SAMOA
8785
10000
8000,
6OOO
4000,
ß
2001 -70
-60
-50
-40
-30
-20
-10
0
10
20
Latitude
Figure 7. Maximum height attained along each trajectory shown versus latitude atwhich thatheight wasattained.
Onepointisplottedforeachtrajectory calculated for 1986-1995.
themaximum heightis attained, it doesshowhowair parcels transport from midlatitudes and that the low ozone values from southernmidlatitudesmust descend3-8 km en route to
correspondto trade wind flow.
SMO. This dramaticchangein air parcelelevationoccurs Next we combined the wintermonths(July,August,and primarilywhere the maximumtemperature gradientexists September) to getanother viewof transport duringthepeakof betweenthe tropicsand midlatitudes.Thusthe shapeof thesurface ozoneseasonal cycle.Wechose tolookatonlythree isentropic surfaces precludes transport frommidlatitudes in the monthsto avoid a possibleseasonalbias in ozonelevel that is boundarylayer. independent of transport. Thisseason is of greatinterest because In termsof 10-day origins,air transported to SMO is it hasthehighest variability in transport andozonemixingratio. predominantly from remote,tropical,marineregions. The Figure1l a shows thetransport summary obtained by clustering frequencyof suchorigins is about 50% in winter to 80% in thesemonths.Note that transport typesare fairly balanced summer.The remaining trajectories are frommoresoutherly betweentradewindandsoutherly flow. The clusters arefurther latitudes, usuallyfromelevations of 3 km or higher. Northern grouped by windspeedor trajectory length.Figure1lb relates hemispheric airistransported to SMOata rateof 2-3%peryear, surface ozonemixingratiosto eachof theseclusters.Only exclusivelyduringsummerandfall months. surface ozonehourlyaverages corresponding to trajectory arrival timeswereused.Notethattrade-wind-type transport (cluster1) is mostfrequent,and corresponds to the lowestsurfaceozone 4. SeasonalCharacteristics of SMO Surface Ozone
mean. Air parcelstransportedto SMO in this manner have
Theannual cycleof SMOsurface ozone isshown in Figure8 for the periodfor whichboth trajectories and ozonedataare
available.Ozonemixingratiosarereported in partsperbillion
(ppb)byvolume. Thesummer minimum occurs during January through Marchatthesame timethatvariability in ozone mixing ratiosis at its lowest. The long-termmeanfor this seasonis
about 9 ppb.Themaximum in theannual cycleoccurs during JulyandAugust, withlong-term monthly means close to 19ppb. Australwinteris characterized by increased variationin surface ozonemixingratios,whichperiodically surpass 30 ppb. In an effortto determinehow surfaceozonecharacteristics are
related totransport, wefirstfocused onshorter timeperiods.For example, Figure9ashows SMOsurface ozonemeasured during February1989. This "quiet"monthwaschosenbecauseof its
verylow mixingratiosas well as low variability.In contrast, July1989(Figure9b) waschosen asa "noisy"monthin which threeevents of highozone mixingratioswereinterrupted bydays J F M n j j n s o N B whenmixingratiosdippedto 12 ppb. Figures10aand10bshow P10NTH SMO trajectories arrivingduringFebruaryand July 1989, respectively. Theuniformity of tradewindflowduringFebruary Figure8. SMOsurface ozone mixing ratios (parts perbillion) by is in contrast to thegreatlyvaryingtransport depicted for July. monthfor 1986-1994. The soliddot is themean,the bar is the We foundby comparing individualtrajectories to ozoneamounts median,theboxis theinner50thpercentile, andthewhiskers are thatthe threeozonepeaksshownin Figure9b correspond to theinner90th percentile.
8786
HARRIS AND OLTMANS' OZONE RELATED TO TRANSPORT AT SAMOA
(b)
(a)
r,1
I',1 "•
•
Z ¸ N 0
Z ¸ N ¸
$
10
15
20
25
1
30
5
10
15
20
25
Figure9. SMOsurface ozonemixingratios(partsperbillion)for(a)February 1989and(b)July1989.
a0øN-• ••7 •, 400 S-
•7 •/•.
ß
700E
(a)
i
i
100øE
1300E
i
1600E
1700W
i
1400W
110øW
(b)
20øN• 400 S -
70øE
i
t
100øE
130øE
i
1600E
170øW
i
1400W
i
110øW
Figure10. SMO 10-daybackisentropic trajectories arrivingat500m during(a)February 1989and(b)July1989.
30
HARRIS AND OLTMANS:
70ø S-½--'•/•-' ,• 0ø
300 E
600E
OZONE RELATED
900E
TO TRANSPORT
x"x180'øW
1200E
1500E
AT SAMOA
150 '0W
8787
120 'øW
I
(b)[
I
30-
25-
•
20-
¸
O15-
2OH.
would
result
in
an increase
in
the rate
of
ozone
destructionby roughly 15%, other factors being constant. However, increased water vapor is often concomitant with increasedcloud cover. Cloud cover would attenuate UV flux, reducingsink strength. To further investigatethe influence of
watervaporon surfaceozonewe correlatedthe noontimevalues A radiative transfer model was used to estimate the decrease in
for two summer
actinicflux at 310 nm from February15 to August15 to be 2.7 x
virtually no relationshipbetweenwater vapor and surfaceozone duringthe summermonths,and moderateanticorrelationduring winter months. The anticorrelationlikely resultsfrom transport effectswhich are much more pronouncedin winter, rather than
10-3 W cm-2 gm-1 or 32% with respectto the latter date (E. Dutton,personalcommunication, 1996). This estimateis for 14øS at noon and considersonly single scattering. We also
months
and two winter
months.
We
found
HARRIS AND OLTMANS:
OZONE RELATED
from the in situ effects of changing water vapor amounts on ozonephotochemistry at SMO. Apart from sink strength,a possiblecauseof a regionalrise in
TO TRANSPORT
8789
Table 2. SMO SurfaceOzoneMeansfor July-September and PercentVariation Explainedby TransportType and Year Both
tropospheric ozone during winter may be widespreadbiomass burningin tropicaland subtropicalAfrica [Fishmanet al., 1991]. The greater"background"levels of ozoneduring winter may also resultfrom more frequentinfusionand mixing of ozone-richair from higherelevationsof the midlatitudesinto the tropics. Comparing ozone levels from winter to summer for the midlatitudetransporttype alsoyieldsa seasonaldifference. For example,cluster4 in Figure 12 correspondsto a summermean ozonelevel of 14.1 ppb, about 10 ppb below the corresponding winter clusters4 and 6 in Figure 11. The strengthof the ozone sink could have an effect on air parcels transported from midlatitudesas they descend,slow down, and enterthe boundary layer closeto SMO. In generalduring summerthe polar front will shift southward,causingthe elevationof a given isentropic surface to be lower. Hence air parcels transported from midlatitudes duringsummerwouldbe expectedto originatefrom lower elevations. In a similar study of transportof tropospheric ozoneat Bermuda,Moody et al. [1995] found that with respectto subsidingflow from the northwest,a seasonallyvaryingsourceor
AT SAMOA
% Variation
Type 1: Type 2: Transport Explainedby TradeWind, Midlatitude, Types, Transport ppb (n) ppb (n) ppb (n) Differences 1986
15.5 (92)
26.9 (34)
18.6 (126)
50%
1987
10.7 (11)
23.9 (62)
21.9 (73)
45%
1988
16.4 (113)
26.6 (20)
17.9 (133)
42%
1989
15.8 (41)
24.3 (41)
20.1 (82)
57%
1990
14.1 (73)
24.2 (34)
17.3 (107)
58%
1991
20.3 (35)
23.7 (27)
21.8 (62)
10%
1992
11.9 (91)
19.2 (19)
13.1 (110)
36%
1994
12.5 (44)
23.6 (24)
16.4 (68)
58%
All years
14.8 (500)
24.2 (261)
18.0 (761)
45%
% Variation
22%
13%
16%
explainedby differences
betweenyears
atmospheric processmustexistapartfrom transporttype, which resultsin higherozonein spring. variance,we then estimatedthe percentvariationattributableto For the next part of this study we clusteredtrajectoriesover differencesbetween transport and between season. Percent the entire time period for which we have both trajectoriesand variationin thiscontextrefersto the sumof squares of betweenozone data, a subset of trajectories summarized in Figure 6. groupdifferences dividedby the total sumof squares.The most Transportpatternsdeterminedwere quite similar, although1995, variation(45%) can be explainedby differencesin wintertime most of 1993, and other short breaks in the ozone record were not transport(July-September)betweenthe tropical (type 1) and included. We then consideredonly two main transport types: midlatitude(type 2) clusters. Transportdifferencesaccountfor Type 1 is trade wind transport,wheretrajectoriesshow airflow the leastpercentageof variationduring summer(7%), probably from the eastin the tropical marine boundarylayer, and type 2 is becausetransporttype 2 occursso rarely,only 17 timesout of a midlatitudetransport,where trajectoriesshow strong westerly total of 1003 occurrencesover the entire period. This is in flow at higher elevationsof southernmidlatitudes,followed by contrastto the winter months,when transporttype 2 is most descent with anticyclonic curvature. This grouping into two frequent(261 out of 761 occurrences). transporttypes consolidatessome clusters and ignores two It is interestingto notethat the percentvariationexplainedby clustersof mixed character.We employedthis methodto clarify seasonal differences is very nearlyidenticalfor the two transport transporteffects. Table 1 presentsozone means and percent types(22-24%). This overridingseasonaleffect was notedin our variation explained by transport and season. Using earlierdiscussionof Figures 11 and 12. nonparametric statistics,we determinedthat within each season It is evident from the spread of ozone values within each the mean ozone amountsfor each transporttype are statistically cluster(Figures1lb and 12b) that transporttype is not the sole different at a significancelevel less than 1%. Likewise a factordeterminingozonelevels at SMO. In fact our analysisof statistical difference
in the mean was found between seasons for
each transport type.
Using standard parametric analysis of
year-to-year variations in the next section emphasizesthe complexityof factorsinfluencingSMO surfaceozone mixing ratios.
Table
1. SMO
Surface Ozone Means and Percent Variation
Explainedby TransportType andSeason
5. Year-to-Year Variability in SMO Surface % Variation
Type 1: Trade Wind,
Type 2: Midlatitude,
ppb (n)
ppb (n)
Explainedby Transport Differences
Season:
Jan. - March
8.5 (986)
14.9 (17)
7%
April- June July - Sept.
10.9 (592)
20.1 (146)
39%
14.8 (500)
24.2 (261)
45%
Oct. - Dec.
11.3 (700)
18.1 (71)
15%
% Variationexplained by seasonal
22%
24%
differences Here the value n denotes the number of occurrences.
Ozone
To examineozoneyear-to-yearvariations,we againfocuson the winter months,July - September.As mentionedearlier,this is the time when ozone reachesits peak level and also when transport plays the greatest role in determining ozone characteristics.Table 2 presentsozone means by year and transporttype for winter months. We determinedthat the ozone meansfor eachyearwere at timesstatisticallydifferent. This was tested with a nonparametrictechniqueat the 1% significance level. Again usingstandardparametricANOVA techniques,we estimatedthat 22% and 13% of ozone variationis explainedby year-to-year differences with type 1 and type 2 transport, respectively.We alsodeterminedthe percentvariationexplained by differencein transportfor each year, shownin the far right
8790
HARRIS
AND OLTMANS:
OZONE RELATED
column. These data show a wide spreadin the strengthof the relationshipof ozoneto transport,from a low of 10% to a high of 58%. In this sectionwe attemptto explainsomeof theseyear-toyear differences. The causesof year-to-yearvariability in ozone are varied and complex. Wuebbleset al. [ 1991] modeledthe effect of the 11year solar cycle on ozone and found a changein tropospheric
TO TRANSPORT
AT SAMOA
duringthisseasonto explainin parta declinein 13Candmethane growthrate. The averagewatervapormixingratio duringwinter 1992 at SMO was slightlybelow the averagevalue basedon six
winters, so its effect on sink strengthwas not a factor. But it does appearthat in addition to transport,severalother factors may have beenpresentto resultin low ozonelevels in 1992. On the other hand, the year with the least trade wind flow and ozone (42øN) of 0.7% from solar maximum to solar minimum, mostmidlatitudetransportis 1987. This yearalsohasthe highest so the solarcycle may have a small effect on SMO ozonelevels. averageozone (21.9 ppb), as might be expected. However, the Oltrearis and Levy [1994] found a quasi-biennialoscillation averageozone for 1991 (21.8 ppb) came very closeto that level (QBO) in SMO surfaceozone,with an amplitudeof about2 ppb. althoughtransportwas morebalancedbetweenthe two transport This cycle seemsto be relatedto the stratospheric wind QBO, types. To try to explain the high ozone values of 1991, we with the maximum in ozone usually following the east wind examinedTable 2 and noted that during 1991 the relationship maximum by several months. The mechanismcausing this betweenozoneandtransportwasextremelyweak, with only 10% relationshipis not known. of the ozone variation explainedby differencesin transport. Other elementsinfluencing ozone mixing ratios from year to Theseaustralwintermonthsfollowedthe June1991 eruptionof year are mentionednext with examplesfrom the period of study. Mount Pinatubo volcano. Dlugokenckyet al. [1996] used a A closelook at theseyearswill, if not explain someunexpected radiativetransfermodel to estimatethe impact of the volcanic results,then at leastpoint out factorsotherthan transportthat can cloud,including SO2 and sulfateaerosols,on tropicalUV actinic also have a strong influence on ozone levels. Becauseseveral flux. They calculated a significant attenuation,which they factors may be present at the same time and they may work in proposed affected OH concentrationsand in turn led to oppositedirections,the effect of transportalone is difficult to anomalousincreasesof methaneand CO. Using the samemodel, comparefrom year to year. E. Dutton (personal communication, 1996) estimated the maximum attenuation of UV actinic flux at 310 nm to be about For example,two years that look very much alike in terms of transportare 1988 and 1992. Of all the years,thesetwo have the 10%. This level was calculatedfor the middle of August 1991 leasttransportfrom midlatitudesand correspondinglymore trade when sulfate aerosol was maximum. The value calculated was for the 10øN-10øS latitude band where the cloud was densest. wind flow (Table 2). One might expect suchyearsto result in low averageozone becauseof fewer episodeswhere ozone-rich Over SMO, UV attenuationwas probably somewhatless, but air is transportedfrom the southcombinedwith more exposureto nonetheless,this photochemicalperturbation appears to have the sink. The year with the lowest averageozone (considering affectedozonelevels as well by reducingthe strengthof the sink. just thesetwo well-defined transporttypes)is indeed 1992 (13.1 This could explain the very high ozone mixing ratios we saw ppb), but the averagefor 1988 (17.9 ppb) is closeto the long- under the trade wind regime during the winter of 1991. The term mean (18.0 ppb). During the austral winter of 1988 the averagefor this time and transporttype was 20.3 ppb, more than SouthernOscillationIndex was stronglypositive,signalingnon5 ppb abovethe long-termaverage. E1 Nino conditions in the South Pacific. These conditions Other possibleinfluenceson ozone levels from year to year prevailedfor about 1 year, affecting only this particularwinter include cloudiness, which scatters UV and thus diminishes sink season.They may be linked to the predominanceof easterlyflow strength[Fredericket al., 1989]. Amountof cloudiness depends shown in Table 2. A large fluctuation in methanegrowth rate on meteorologicalconditionssuch as proximity of the monsoon detectedby a global network may also be attributableto these low or the SPCZ to SMO. The frequencyof mixing with higher anomalousconditions. Steele et al. [1992] proposeda --10% layersin the atmosphere,due to convectiveepisodes,may alsobe increasein interhemispherictransportduring this time to explain a factor. Oltmans et al. [1989] presentedaverageprofiles for the methaneanomaly. Prinn et al. [1992] proposedincreased ozoneover SMO that showhigher ozone mixing ratios abovethe interhemispheric transportduring tropical Pacific cold eventsto boundarylayer. This is expectedbecausethe drier air in the free explaincharacteristics of SMO methyl chloroformdata. It is not troposphere will diminishthe sink strengthrelativeto the marine known whether such an enhancement in transport from the boundarylayer [Crutzen, 1988]. Another factor that could cause northernhemispherewould affect surfaceozone levels, but this year-to-yeardifferencesin the amount of ozone transportedto may partly explain the differencein ozone levels seenin 1988 SMO from midlatitudesis the timing and vigor of stratosphericand 1992. troposphericexchangenear air parcel origins,but we have not However,1992 wasanomalousin its own way. Gleasonet al. quantifiedthiseffect aspart of this study.
[1993] reporteda significantdecreasein southernhemisphere stratosphericozone, which may have been relatedto destruction of ozoneon aerosolsfrom the eruptionof Mount Pinatubo. The averagevalue of total column ozone as measuredby the SMO Dobson spectrophotometer during July-September 1992, represented a decreaseto 5.2% below the long-termaverage(W. Komhyr, personal communication, 1996). Liu and Trainer [1988] modeledthe responseof surfaceozone to a reductionin the total column. In the remoteboundarylayer of oceanicareas they calculatedthat the percentdecreasein surfaceozone would be somewhatgreater than the percent decreasein the total column.
The fact that the mean surface ozone value is lowest in
1992 comparedwith otheryearsis consistentwith thesefindings. In addition,Lowe et al. [ 1994] proposereducedbiomassburning
6. Conclusions
Our analysisof trajectoriesand SMO surfaceozone data has
shownthat ozone variability is lowest in summer,owing to homogeneityof transportfrom tropical marine origins. The minimumin the seasonal cycleduringJanuary-March is probably due to enhancedphotochemicalsink, which is most effective during this season, as well as relative isolation from ozone
sources. Variability in winter is highestbecauseof transport shifts,whichresultin air parcelswith distinctlydifferentorigins and ozone mixing ratios. The July-Augustmaximumin SMO surfaceozoneprobablyreflectsthe diminishedsink strength,the frequent ventilating with ozone-rich parcels from higher
HARRIS AND OLTMANS:
OZONE RELATED
TO TRANSPORT
AT SAMOA
8791
elevations (3-8 km) of the midlatitudes, and possibly an climatology for the Mauna Loa Observatory,using clustered trajectories,J. Geophys.Res.,95, 13,651-13,667, 1990. enrichment of "background" ozone levels due to biomass burning. It is importantto note that in the caseof SMO, direct Hams,J. M., andJ. D. Kahl, Analysisof 10-dayisentropicflow patterns for Barrow,Alaska:1985-1992,J. Geophys.Res.,99, 25,845-25,855, transportfrom anthropogenic sourcesof ozone is not indicated. 1994. Aside from a possible hemisphericozone enhancementfrom Harris,J. M., P. P. Tans,E. J. Dlugokencky,K. A. Masarie,P.M. Lang, S. Whittlestone, andL. P. Steele,Variationsin atmospheric methane biomassburning, the annual cycle in SMO ozone is completely at Mauna Loa Observatoryrelated to long-rangetransport,J. natural and reflectsthe timing and action of the pathwaysof Geophys.Res., 97, 6003-6010, 1992. ozone from a source in the midlatitude upper troposphereor Liu, S.C., andM. Trainer,Responses of the tropospheric ozoneandodd stratosphere to the sinkin the tropicalboundarylayer. hydrogenradicalsto columnozonechange,J. Atmos.Chem.,6, 212233, 1988. Our analysis showed that in most winter seasons,transport Lowe, D.C., A.M. Brenninkmeijer,G. W. Brailsford,K. R. Lassey,A. J. playsan importantrole in determining ozonelevels. On average, Gomez, and E.G.
Nisbet, Concentration and 13C records of
45% of winter ozone variationcan be explainedby differences atmosphericmethanein New Zealand and Antarctica:Evidence for betweenthe trade wind transportand the southernmidlatitude changesin methanesources,J. Geophys.Res., 99, 16,913-16,925, 1994. transport.During 1990 and 1994 this number was as high as andR. 58%. On the other hand, during 1991 the transportrelationship MacFarland,M., D. Kley, J. W. Drummond,A. L. Schmeltekopf, H. Winkler, Nitric oxide measurements in the equatorialPacific was quite weak, possiblydue to effects of aerosolsfrom the region,Geophys.Res.Lett., 6(7), 605-608, 1979. eruptionof Mount Pinatubo. A seasonalvariationof about23% Memll, J. T., Trajectoryresultsand interpretationfor PEM-West A, J. that is independentof transport is probably linked to the Geophys.Res., 101, 1679-1690, 1996. transportto seasonallyvarying photochemicalsink. About 16% of ozone Memll, J. T., R. Bleck, andL. Avila, Modelingatmospheric variation is attributed to differencesfrom year to year. Possible
causesof year-to-yeardifferencesinclude blocking of UV by volcanic aerosolsas well as changesin biomassburning, total column ozone, interhemispherictransport,and timing or degree of stratospheric-tropospheric exchange.
Acknowledgments. Many thanks to E. Hackathorn /'or computinghelp. The authorsare alsoindebtedto C. M. Wang, J. Kahl, E. Dlugokencky, and E. Dutton for valuable advice and in/'ormation.
References Bortniak, J. C., The wind climatologyof American Samoa,NOAA Tech. Memo. ERL ARL-98, 67 pp., NOAA Environ. Res. Lab., Silver Spring,Md., 1981. Crutzen,P. J., Troposphericozone:An overview,in TroposphericOzone, editedby I.S.A. Isaksen,pp. 3-32, D. Reidel,Norwell, Mass., 1988. Dlugokencky, E. J., E. G. Dutton, P. C. Novelli, P. P. Tans, K. A. Masarie,K. O. Lantz, and S. Madronich,Changesin CH4 and CO growthratesafter the eruptionof Mount Pinatuboand their link with changes in tropical troposphericUV flux, Geophys. Res. Lett., 23(20),2761-2764,1996. Fishman,J., K. Fakhruzzaman,B. Cros, and D. Nganga,Identificationof widespread pollution in the southern hmnispherededuced from satelliteanalysis,Science,252, 1693-1696, 1991. Frederick, J. E., H. E. Snell, and E. K. Haywood, Solar ultraviolet radiation at the Earth's surface, Photochem. Photobiol., 50(8), 443450, 1989. Gleason, J. F., P. K. Bhartia, J. R. Herman, R. McPeters, P. Newman, R.
S. Stolarski,L. Flynn, G. Labow, D. Larko, C. Seftor,C. Wellemeyer, W. D. Komhyr,A. J. Miller, and W. Planet,Recordlow globalozone
the MarshallIslands,J. Geophys.Res.,90, 12,927-12,936, 1985. Moody,J. L., The influenceof meteorology on precipitationchemistryat selectedsitesin the EasternUnited States,Ph.D. thesis,176 pp., Univ. of Mich., Ann Arbor, 1986.
Moody, J. L., S. J. Oltmans,H. Levy II, and J. T. Merrill, Transport climatologyof tropospheric ozone:Bermuda,1988-1991,J. Geophys. Res., 100, 7179-7194, 1995.
Norusis,M. J., SPSS ProfessionalStatistics6.1, 385 pp., SPSS, Inc., Chicago,Ill., 1993. Oltmans, S. J., and H. Levy II, Surface ozone measurementsfrom a global network,Atmos.Environ., 28(1), 9-24, 1994. Oltmans, S. J., W. D. Komhyr, P. R. Franchois,and W. A. Matthews, Troposphericozone: Variations from surface and ECC ozonesonde observations, in Ozonein theAtmosphere, editedby R.D. Bojkovand P. Fabian,pp. 539-543, A. Deepak, Hampton,Va., 1989. Oltmans,S. J., H. Levy II, J. M. Hams, J. T. Merrill, J. L. Moody, J.A. Lathrop,E. Cuevas,M. Trainer, M. S. O'Neill,, J. M. Prospero,H. V6mel, and B. J. Johnson,Summerand springozoneprofilesoverthe North Atlantic from ozonesondemeasurements, J. Geophys.Res., 101(D22), 29,179-29,200, 1996.
Prinn, R., D. Cunnold,P. Simmonds,F. Alyea, R. Boldi, A. Crawford,P. Fraser,D. Gutzler, D. Hartley, R. Rosen, and R. Rasmussen,Global averageconcentrationand trend for hydroxyl radicalsdeducedfrom ALE/GAGE trichloroethane(methylchloroform)data for 1978-1990, J. Geophys.Res., 97, 2445-2461, 1992. Savoie, D. L., J. M. Prospero,S. J. Oltmans, W. C. Graustein, K. K. Turekian,J. T. Merrill, and H. Levy II, Sourcesof nitrateand ozonein the marineboundarylayer of the tropicalNorth Atlantic,J. Geophys. Res., 97, 11,575-11,589, 1992.
Steele,L. P., E. J. Dlugokencky,P.M. Lang, P. P. Tans, R. C. Martin, and K. A. Masarie, Slowing down of the global accumulationof atmosphericmethaneduring the 1980s,Nature, 358, 313-316, 1992. Wuebbles, D. J., D. E. Kinnison, and J. L. Lean, Solar variations and
their influence on trends in upper stratospheric ozone and temperature,in Proceedingsof the SecondSymposiumon Global Change Studies,pp. 108-113, Am. Meteorol. Soc., Boston, Mass., 1991.
in 1992, Science. 260, 523-525,1993.
Halter, B. C., J. M. Hams, and T. J. Conway, Componentsignalsin the record of atmosphericcarbon dioxide concentrationat American Samoa,J. Geophys.Res.,93, 15,914-15,918, 1988. Hams, J. M., An analysisof 5-day midtropospheric flow patternsfor the South Pole: 1985-1989, Tellus, 44B, 409-421, 1992.
Harris, J. M., and J. D. Kahl, A descriptive atmospherictransport
J. M. Harris and S. J. Oltmans, NOAA Climate Monitoring and DiagnosticsLaboratory,325 Broadway, Boulder, CO 80303. (e-mail: j
[email protected])
(ReceivedSeptember6, 1996;revisedJanuary16, 1997; acceptedJanuary16, 1997.)