baseline climatology of viti levu (fiji) and current...

49

Baseline Climatology of Viti Levu (Fiji) and Current Climatic Trends1

Melchior Mataki, Kanayathu C. Koshy, and Murari Lal 2

Abstract: In this paper we characterize the climate at Nadi and Suva in Fijifrom 1961 to the present, providing a picture of ongoing climate trends. Thefocus is on surface observations of air temperature and rainfall, although someinformation on South Pacific Ocean climate is also discussed, given its relevancefor Fiji. Our findings suggest that surface air temperatures have registeredincreasing trends at both Suva and Nadi (the two observatory sites in Fiji iden-tified as Weather Observing Stations under the global network of the WorldMeteorological Organization) during the period 1961–2003. There has been asteady increase in the number of days per year with warmer nighttime temper-atures in recent decades. The rise in annual mean surface air temperature overSuva was @1.2 �C over the 43-yr period considered here (at a significantly in-creasing rate of 0.25 �C per decade, which is 1.5 times higher than the trendsin global average temperature increase during the past century). The rate of in-crease in annual mean surface air temperature at Nadi was 0.07 �C per decade.No trend in annual mean rainfall has been observed, however, at either of thelocations. Significant interannual variability in annual as well as summer rainfallobserved at both sites (including extreme rainfall events during the 43-yr period)can largely be attributed to El Nino–Southern Oscillation (ENSO) events andintraseasonal oscillations in the mean position of the South Pacific ConvergenceZone.

Baseline climatological data charac-terize climate during a recent period, servingas a reference against which to compare bothhistorical and likely future climate variations.By international meteorological convention,the baseline climate is usually of 30 yr dura-tion. This is long enough to encompasscharacteristic interannual and interdecadalvariations in climate and to capture a rangeof short-term extreme climate events but stillshort enough to allow comparison (for themajor climatic variables such as temperature

and precipitation) with earlier periods andhence detection of possible longer-termchanges in climate.

Climatological Normals

In the last two scientific assessment reportsof the Intergovernmental Panel on ClimateChange (IPCC) (IPCC 2001a,b), the period1961–1990 was adopted as the climatologicalbaseline, an official World MeteorologicalOrganization ‘‘normal’’ period. Convention-ally, these periods repeat at 30-yr intervals,the next being 1991–2020. Notwithstandingthis convention, with most climatologicaldata now available in digital form, it is a fairlystraightforward exercise to update 30-yr aver-ages on a decadal basis, and some countrieshave published climatological summaries forthe period 1971–2000.

In selecting the baseline period, one issueto bear in mind is the closure of manymanned weather stations worldwide duringthe 1980s and 1990s and their partial replace-ment with automatic weather stations. Thishas posed some problems for maintaining

Pacific Science (2006), vol. 60, no. 1:49–68: 2006 by University of Hawai‘i PressAll rights reserved

1 This study was undertaken as part of the activityunder the research project ‘‘Integrated Methods andModels for Assessing Coastal Vulnerability and Adapta-tion to Climate Change in Pacific Island Countries’’funded by International START Secretariat, Washing-ton, D.C. Manuscript accepted 29 April 2005.

2 Pacific Centre for Environment and SustainableDevelopment, The University of the South Pacific, P.O.Box 1168, Laucala Bay Campus, Fiji Islands (fax: þ679-3232891; e-mail: [email protected]).

high-quality, homogeneous records of cli-mate in many regions. It has also reducedthe total number of sites in the observationalnetwork, meaning that the station coverageand density for the 1971–2000 data are oftenpoorer than in the earlier (1961–1990) pe-riod. This paper therefore is confined to thebaseline period of 1961–1990 as standardWorld Meteorological Organization clima-tology in the analysis of ‘‘climatological nor-mals’’ of surface air temperature and rainfallat Nadi and Suva stations.

Geographic Location and Key WeatherCharacteristics

Fiji consists of 18,376 km2 of land and in-cludes about 330 islands, of which about 100are inhabited. The largest island and popula-tion center is Viti Levu, which has an area of10,388 km2 (Figure 1). Viti Levu is character-ized by high mountains, dry on the leewardwest side (Nadi) and wet on the windwardeast side (Suva). Vanua Levu, located north-east of Viti Levu, is the second largest islandand is slightly more than half the size of VitiLevu. Taveuni lies to the east of Vanua Levu.Equal in size to Taveuni is Kadavu, which liesto the south of Viti Levu. Fiji’s remainingislands are small and are divided into threemain groups: Lomaiviti, Lau, and the Yasa-was.

Fiji lies in the South Pacific trade-windbelt of predominantly southeasterly winds.The southeasterlies are relatively strong andpersistent in winter (between May and Octo-ber). In summer (between November andApril) the winds are generally light and vari-able with a predominant sea breeze duringthe day. Winds from the northerly quarterare also quite common during this periodbecause the South Pacific Convergence Zonelies just to the north of Fiji where the south-east trades and the divergent easterlies con-verge. Fiji’s climate is tropical. Overall,temperatures in summer are slightly higherthan in the winter months as is the humidityand the resulting rainfall. The two seasons arelargely controlled by the north to southmovements of the South Pacific Convergence

Zone, and rainfall distribution is largely con-trolled by the large-scale cloud bands as-sociated with the convergence zone, themigrating low pressure systems, the move-ment of the tropical upper tropospherictroughs, and surface midlatitude systemssuch as cold fronts moving across from thesouth. From November to April, the FijiIslands are frequented by tropical cyclonesoriginating in the Pacific Ocean, which re-sults in prolonged heavy rainfall and floodingof low-lying coastal areas. However, theweather in Fiji varies considerably betweenisland regions. This is due to the dominantsoutheast trade winds, which helps cloudsform over the mountains on the southeasternside of the larger islands.

Nadi is situated on the western side of VitiLevu (Fiji’s western division). The westerndivision (Sigatoka to Rakiraki) is a relativelydry zone and is subject to large seasonaland interannual climatic variation. The an-nual rainfall in the dry zones averages around200 cm. Typically, winter weather in theNadi region has clear or partly cloudy skies,much sunshine, a well-developed west tonorthwesterly sea breeze by day (south-easterlies are predominant at night and in allmonths except from August to March whenwind is generally from the northwesterlyquarter), and relatively cool nights. In sum-mer, thunderstorm activity is common andresponsible for brief spells of high-intensityrainfall. Suva, the capital of Fiji, is situatedon the wet eastern side of the island of VitiLevu (in Fiji’s central division where the an-nual rainfall ranges from 300 cm around thecoast to 600 cm on the higher mountains).

The Fiji group of islands is surrounded by ahuge body of water and is affected by global-and regional-scale climatic events such as ElNino and La Nina resulting from interannualvariations of the strength of the ‘‘Walker Cir-culation’’ over the equatorial Pacific Oceanand associated droughts and/or floods, tropi-cal cyclones, and high waves. The islandshave experienced extreme weather events re-cently. In 2003, tropical cyclone Ami devas-tated parts of Vanua Levu, Tavueni, and theislands in the Lau Group. Heavy rain associ-

50 PACIFIC SCIENCE . January 2006

ated with the cyclone brought widespreadfloods in Vanua Levu and other parts of thecountry. Floods of rare magnitude such asone in the summer of 2000 occurred in west-ern Viti Levu leaving many homes under

water and millions of dollars worth of dam-age. In 1998 western Viti Levu, northern Va-nua Levu, and islands in eastern Fiji suffered asevere drought that almost crippled the sugarindustry, stressing moisture-reliant and non-

Figure 1. The Fiji Islands showing location of Nadi and Suva sites on Viti Levu.

Baseline Climatology of Viti Levu . Mataki et al. 51

heat-tolerant flora and fauna, resulting in lossof biodiversity. The cost to the central gov-ernment for relief and recovery, such as foodand water, fertilizers, cane seeds, etc., washundreds of millions of dollars.

However, until now only fragmentedstudies on a few meteorological variableshave been done on the climatological detailsof Nadi and Suva, and no systematic studyon the subject exists as a published document.A general account of the climate and weatherof Fiji was provided by Derrick (1951) identi-fying three classes of regional climate for theFiji Islands. A similar treatment of the subjectwas given by Twyford and Wright (1965) butwith lesser details. An updated and compre-hensive review of Fiji’s weather and climateduring the period from 1950 to 1975 waspublished by Prasad (1979). Some aspects ofthe climate of the Nadi region for planningpurposes were also reported in Sharma(1982) and Krishna (1989). Databases offeringmean baseline climatology and current trendsin surface air temperature and rainfall atselected meteorological stations in Fiji are,however, crucial for assessing the impact ofany changes in climate and its variability onvarious sectors of the economy in the region.Keeping this in view, the findings on theanalysis carried out and reported here mayprovide a helpful reference.

materials and methods

The study reported here is based mainly onclimatological observations made at Nadi(civil aviation compound, Korowai Road inNamaka) and Suva (Laucala Bay) during theperiod from 1961 to 2003. Although meteo-rological data records exist from 1942 forboth stations at Fiji Meteorological Service,the data for the period 1942 to 1960 are con-sidered not reliable because they have notundergone quality control checks and containlarge gaps; hence in this study we did not usethat data. The Fiji Meteorological Serviceprovided the daily data for maximum andminimum temperatures and rainfall for Suvaand Nadi for the period from 1961 to 2003from its climate database. The meteorologicalobservations at the civil aviation compound

in Namaka in Nadi (17.75� S, 177.45� E) andat Laucala Bay in Suva (18.09� S, 178.27� E)can be considered generally representative towithin 10 km of the two sites considered inthis study.

The daily temperature and rainfall datafor the baseline period from 1961 to 1990were analyzed to obtain monthly and annualclimatological means (also referred to as‘‘normals’’). Subsequently, a time series anal-ysis was carried out using the daily data forthe period 1961 to 2003 to find monthly, sea-sonal, and annual anomalies and to investigateany significant changes in the trends and pe-riodicity in the surface air temperature andrainfall of the two locations over the periodon a seasonal, annual, and decadal basis.Further, the observed instances of and trendsin extreme weather events such as number ofhot days, number of wet and dry spells, andrecurrence periodicity of droughts and floodswere examined as part of the climatology ofthe selected sites.

An attempt has also been made to interpretand explain the observed trends, periodicity,and extreme events on a seasonal, annual,and decadal basis. For this purpose, we per-formed stringent statistical tests using analysisof variance (ANOVA) (MINITAB and SPSSsoftware). Possible links between changes intotal precipitation and frequency of extremeswere explored including the influences, ifany, due to El Nino and La Nina telecon-nections and the intraseasonal oscillations onthe position and strength of the South PacificConvergence Zone. To critically examine therelationship between El Nino–Southern Os-cillation (ENSO) events and the rainfall inNadi and Suva in this study, the years withcold and warm events during the 1961–2003period were identified and seasonal mean sur-face air temperatures and seasonal total rain-fall amounts for those years were computed atNadi and Suva. A two-tailed significance testwas performed on the differences between thecomposite rainfalls for the samples represent-ing the warm phases of El Nino versus that ofthe remaining (neutral plus La Nina) yearsand those representing the cold phases of LaNina versus remaining (neutral plus El Nino)years.


results

Climatological Surface Air Temperature andRainfall Normals

The mean monthly surface air temperaturesand their extremes as inferred from averagingof daily data records collected for Nadi andSuva during the period from 1961 to 1990are depicted in Figure 2. The annual averagesurface temperature was 25.4 �C at both Nadiand Suva, with January and February beingthe warmest months with average dailytemperatures exceeding 27.0 �C. The highestmean temperatures and their extremes atNadi (Table 1) occurred in January andthose at Suva occurred in February (Table2). The daily maximum temperatures peakedat higher than 32 �C in February at boththese locations. The nighttime temperaturesshowed a high of 24 �C or more in Februaryat both Nadi and Suva. Both Nadi and Suvaexhibited the lowest daily mean temperaturesin July. However, the mean nighttime tem-peratures for July in Suva (20.5 �C) werehigher than in Nadi (17.8 �C) by about 2.5�C. The range of extremes in daily mean andmaximum surface air temperatures was leastin March and greatest in July at both loca-tions. The range of extremes in daily mini-mum surface air temperature was also theleast in March, but the range was greatest inJuly at Nadi with a secondary maximum inSeptember; at Suva, the range of these ex-tremes was the least in March and greatestin May with a secondary maximum in Sep-tember.

Average annual mean daily surface airtemperatures were 25.4 �C at both Suva andNadi, and the range of extremes in the annualmean daily surface air temperature at theselocations was also identical (high at 26.5 �Cand low at 24.4 �C). However, as is evidentfrom Tables 1 and 2, the average annual day-time maximum surface temperatures at Nadi(on the lee side of mountains) were @1.6 �Chigher than at Suva (on the windward side).This anomaly was least during summer (0.3�C in March) and more pronounced in winter(2.2 �C in July). It is interesting that the aver-age annual nighttime minimum surface tem-peratures at Nadi were lower than at Suva

by about 1.6 �C. This implies that the annualmean diurnal temperature range for Nadi (9.6�C) was larger than that for Suva (6.5 �C) onan average. Moreover, the diurnal tempera-ture range in July was about 1 �C higherthan in January at Nadi, but the reverse istrue at Suva, suggesting a significant differ-ence in the microclimate of the two locationsexamined in this study.

Due to the influence of the surroundingocean, changes in the surface air temperaturefrom day to day and from season to season atboth these locations are, in general, relativelysmall. However, it is noteworthy that duringthe 1961–1990 period the lowest nighttimesurface air temperature at Nadi was recordedat 11.3 �C on 7 August 1965, and the highestnighttime surface air temperature here wasrecorded at 26.9 �C on 3 February 1962(Table 1). At Suva, the lowest nighttime sur-face air temperature was recorded at 13.3 �Con 18 October 1965, and the highest night-time surface air temperature was recorded at27.6 �C on 24 March 1979 (Table 2). Thehighest daytime surface air temperature atNadi was recorded at 36.7 �C on 11 January1987, and the lowest daytime surface air tem-perature was recorded at 20.8 �C on 2 August1965. At Suva, the highest daytime surfaceair temperature was recorded at 34.5 �C on10 February 1987, and the lowest daytimesurface air temperature was recorded at 21.0�C on 18 June 1985.

Rainfall in Fiji is highly variable andmainly orographic (influenced by island to-pography and the prevailing southeast tradewinds). The southeast trade winds are oftensaturated with moisture, causing any highlandmass lying in their path to receive largeamounts of precipitation. The mountains ofViti Levu create a wet climatic zone on theirwindward side and a dry climatic zone ontheir leeward side. The climatological normalfor the period 1961–1990 shows that Suva(windward side) received an annual total rain-fall of 304 cm, but Nadi (leeward side) re-ceived a total of only about 181 cm rainfallannually on an average basis (Tables 1 and2). The fact that the total annual rainfall atNadi is only about 60% of that recorded atSuva highlights the dominating role of orog-


raphy in the spatial distribution of rainfallover Viti Levu Island. The highest annualtotal rainfall for the period 1961–1990 wasrecorded as 438 cm (144% of the normal) in1975 at Suva and 298 cm (165% of the nor-

mal) in 1974 at Nadi. Similarly, the lowestannual total rainfall for the period 1961–1990 was recorded as about 181 cm (60% ofthe normal) in 1987 at Suva and about 86 cm(48% of the normal) in 1969 at Nadi.

Figure 2. Climatological monthly normal (1961–1990, mean, high, and low) average surface air temperatures at (a)Nadi and (b) Suva.


TA

BL

E1

Cli

mat

olo

gica

lN

orm

als

of

Su

rfac

eT

emp

erat

ure

and

Rai

nfa

llfo

rN

adi

(17.

75�

S,

177.

45�

E)

Bas

edo

n19

61–

1990

Per

iod

Par

amet

ers

Jan

.F

eb.

Mar

.A

pr.

May

Jun

eJu

lyA

ug.

Sep

t.O

ct.

No

v.D

ec.

An

nu

al

Dai

lyT

mea

n,� C

27.2

627

.21

26.8

526

.02

24.9

924

.14

22.8

223

.58

24.2

925

.13

26.1

426

.83

25.4

4H

igh

Tm

ean,� C

28.1

428

.12

27.5

326

.97

26.0

825

.34

24.4

524

.93

25.7

726

.35

27.0

027

.70

26.5

3L

ow

Tm

ean,� C

26.3

826

.30

26.1

625

.07

23.9

022

.95

21.2

022

.24

22.8

023

.91

25.2

825

.95

24.3

4A

vera

geT

max

,� C

31.7

531

.46

30.9

130

.75

29.9

529

.33

28.2

128

.72

29.5

130

.05

30.9

131

.48

30.2

5H

igh

Tm

ax,� C

33.0

832

.59

31.8

631

.88

30.9

630

.50

29.5

430

.09

31.0

431

.47

31.9

832

.68

31.4

7L

ow

Tm

ax,� C

30.4

230

.33

29.9

529

.61

28.9

428

.15

26.8

927

.35

27.9

928

.63

29.8

530

.29

29.0

3A

vera

geT

min

,� C

22.6

623

.05

22.5

421

.51

20.0

318

.90

17.8

518

.26

19.3

520

.15

21.2

822

.30

20.6

6H

igh

Tm

in,� C

23.9

024

.40

23.5

123

.19

21.7

020

.79

20.2

119

.83

21.5

921

.46

22.6

523

.78

22.2

5L

ow

Tm

in,� C

21.4

321

.70

21.5

619

.82

18.3

517

.01

15.5

016

.69

17.1

018

.85

19.9

120

.81

19.0

6A

vera

gera

in(m

m)

299.

230

2.3

323.

916

3.0

78.3

61.9

46.3

58.4

76.6

102.

713

8.2

159.

01,

809.

7

Max

.m

on

thly

tota

lra

in(m

m)

(yea

r)

597.

5(1

977)

643.

6(1

989)

918.

2(1

964)

580.

1(1

986)

332.

1(1

989)

260.

3(1

973)

189.

8(1

963)

192.

7(1

974)

278.

8(1

976)

341.

8(1

972)

460.

6(1

975)

562.

4(1

988)

2,98

3.5

(197

4)

Min

.m

on

thly

tota

lra

in(m

m)

(yea

r)

12.7

(198

7)90

.1(1

990)

75.6

(196

1)21

.6(1

983)

0.8

(198

7)0.

0(1

987)

0.0

(198

6)0.

0(1

970)

4.2

(198

7)1.

8(1

987)

14.2

(197

7)29

.5(1

965)

863.

4(1

969)

Nu

mb

ero

fw

etd

ays

14.5

15.3

15.6

9.7

5.2

4.2

3.2

3.7

4.6

6.1

8.3

10.3

100.

6a

a73

.6(s

um

mer

),27

(win

ter)

.

TA

BL

E2

Cli

mat

olo

gica

lN

orm

als

of

Su

rfac

eT

emp

erat

ure

and

Rai

nfa

llfo

rS

uva

(18.

09�

S,

178.

27�

E)

Bas

edo

n19

61–

1990

Per

iod

Par

amet

ers

Jan

.F

eb.

Mar

.A

pr.

May

Jun

eJu

lyA

ug.

Sep

t.O

ct.

No

v.D

ec.

An

nu

al

Dai

lyT

mea

n,� C

27.1

527

.37

27.0

926

.19

25.2

824

.46

23.2

823

.54

23.8

024

.81

25.7

326

.74

25.4

5H

igh

Tm

ean,� C

27.9

828

.41

27.7

327

.14

26.3

725

.48

24.6

124

.50

25.1

126

.18

26.8

427

.68

26.5

0L

ow

Tm

ean,� C

26.3

226

.34

26.4

525

.24

24.1

923

.45

21.9

422

.58

22.5

023

.43

24.6

325

.80

24.4

1A

vera

geT

max

,� C

30.7

331

.06

30.5

829

.62

28.1

127

.48

25.9

726

.68

27.1

528

.13

28.7

429

.73

28.6

7H

igh

Tm

ax,� C

31.8

132

.14

31.4

630

.75

29.3

328

.61

27.8

028

.06

28.7

329

.89

30.0

930

.81

29.9

6L

ow

Tm

ax,� C

29.6

629

.98

29.7

028

.49

26.8

926

.36

24.1

525

.30

25.5

726

.37

27.3

828

.65

27.3

7A

vera

geT

min

,� C

23.7

923

.76

23.5

522

.88

21.8

321

.42

20.4

920

.51

20.6

721

.54

22.7

723

.37

22.2

1H

igh

Tm

in,� C

24.7

124

.81

24.4

724

.00

23.5

022

.42

21.5

221

.40

22.2

422

.60

23.9

924

.55

23.3

5L

ow

Tm

in,� C

22.8

822

.70

22.6

321

.76

20.1

620

.41

19.4

519

.62

19.1

120

.48

21.5

422

.20

21.0

8T

ota

lra

in(m

m)

315.

228

6.3

370.

939

0.7

267.

316

3.8

141.

715

8.8

183.

923

4.3

264.

426

2.8

3,04

0.14

Max

mo

nth

lyto

tal

rain

(mm

)(y

ear)

718.

4(1

982)

530.

7(1

964)

799.

2(1

973)

1116

.1(1

986)

584.

3(1

989)

419.

3(1

979)

408.

8(1

969)

522.

8(1

963)

480.

5(1

980)

900.

3(1

967)

790.

2(1

981)

571.

2(1

975)

4,38

4.7

(197

5)

Min

mo

nth

lyto

tal

rain

(mm

)(y

ear)

126.

5(1

984)

94.6

(197

4)82

.8(1

967)

97.8

(198

3)49

.9(1

980)

24.0

(196

9)24

.1(1

986)

33.1

(198

7)16

.5(1

987)

38.6

(198

4)45

.6(1

967)

51.9

(196

5)18

15.0

(198

7)

Nu

mb

ero

fw

etd

ays

17.8

18.2

19.4

17.5

15.0

12.8

11.6

11.4

12.7

12.9

14.2

15.7

178.

6a

a10

3.0

(su

mm

er),

75.6

(win

ter)

.

In general, Fiji experiences a distinct wetseason (November to April) and a relativelydry season (May to October), largely con-trolled by the north and south movements ofthe South Pacific Convergence Zone, themain rainfall-producing system for the re-gion. Nadi gets as much as 76.6% and Suvagets 62.2% of the annual total rainfall duringthe summer. Based on averages for the periodfrom 1961 to 1990, the monthly total normalrainfall at Nadi peaked at 32.4 cm in Marchwith a low of 4.6 cm in July (Tables 1 and2). In Suva, the monthly total normal rainfallpeaked at 39.1 cm in April with a low of 14.2cm in July. The highest total rainfall was re-corded in Nadi in March 1964 (91.8 cm) andin Suva in April 1986 (111.6 cm). No rainfallwas reported at Nadi in July 1986 while Suvarecorded the lowest monthly rainfall of 2.4cm. The climatological normal for numberof wet days in a year was 101 days for Nadiand 179 days for Suva. Of these, Nadi had74 wet days during summer and 27 wet daysduring winter on average, and Suva had 103wet days during summer and as many as 76days during winter. It is thus evident thatthe wet and dry seasons are more distinct atNadi than at Suva and that Suva is wetterthan Nadi both on a seasonal and an annualbasis.

Past and Current Climatic Trends

Using the daily data records, trends in dailymaximum, minimum, and mean temperaturesand rainfall were analyzed from 1961 to 2003for Nadi and Suva. Changes in temperatureextremes were also studied. The analysis sug-gests that the average annual day mean tem-peratures registered increasing trends at bothlocations during the period 1961–2003. Thistrend was more prominent and statisticallysignificant at Suva (at a rate of 0.25 �C per de-cade). The rate of rise in average annual meansurface temperature at Nadi was only 0.07 �Cper decade. In contrast, the rate of increase inglobal mean surface air temperature in thetwentieth century has been reported to beabout 0.17 �C per decade (IPCC 2001a).From the average annual day mean tempera-ture trends (Figure 3a) it is evident that the

year 1998 was the warmest year at Nadi,when the annual mean surface air tempera-ture peaked at 0.66 �C higher than the clima-tological normal of 25.4 �C, with 1988 and2002 being the second and third warmestyears (anomalies were 0.61 �C and 0.58 �C,respectively). In Suva, the year 2001 was thewarmest, when the annual mean surface airtemperature peaked at 0.95 �C higher thanthe climatological normal of 25.4 �C, with1998 and 2002 being the next warmest years(anomalies were 0.90 �C and 0.87 �C, respec-tively [Figure 4a]). Both Nadi and Suva expe-rienced the lowest average annual daily meansurface air temperatures in the year 1965,with anomalies being 0.5 �C and 0.7 �C, re-spectively, below the normal. At Nadi, thetrend in annual mean maximum temperatureis not significant, but the increasing trend dis-played in annual mean minimum temperatureis significant and is most prominent since theearly 1990s when anomalies exceeded theG1s level (Figure 3b, c). At Suva, both maxi-mum and minimum temperatures exhibit sig-nificant increasing trends (Figure 4b, c). Theobserved increases in nighttime minimumtemperature at both Nadi and Suva (perhapslinked to a decline in cloud cover during the1980s and 1990s relative to the 1960s and1970s) are consistent with global mean night-time minimum temperature, which has in-creased at twice the rate of daytimemaximum temperatures during the period of1950 to 2000 (IPCC 2001b).

Long-term precipitation changes in Nadiand Suva were also examined by studying thedaily rainfall data over a period of 43 yr. Nodiscernable increasing or decreasing trendswere observed in the annual total rainfall ateither of the two locations during the 1961–2003 period (Figure 5). The data show thatboth Suva and Nadi experience a maximumnumber of days with b30 mm of rainfall inMarch and a minimum number of daysin July. There were no significant changes inheavy rainfall events (daily rainfall higherthan 100 mm) in recent years. For furtherinvestigation, we divided the whole periodinto two 20-yr intervals: 1961–1980 and1981–2003. Two well-defined rainfall peaksoccurred during summertime in both inter-


Figure 3. Trends in average day mean, maximum, and minimum surface temperature anomalies (departures withrespect to 1961–1990 baseline period) at Nadi during the period from 1961 to 2003.

Figure 4. Trends in average day mean, maximum, and minimum surface temperature anomalies (departures with re-spect to 1961–1990 baseline period) at Suva during the period from 1961 to 2003.

vals in Suva. During the earlier interval,primary and secondary rainfall peaks werefound in late March and early November, re-spectively. In the later interval, on the otherhand, the primary peak was found in mid-April, mainly attributed to enhanced heavyrainfall (b30 mm/day) events. Although a

similar shift occurred in the secondary peak,it was much smaller. Thus, the relatively dryspell between the two peaks became longerin the later interval compared with the earlierone at Suva. At Nadi, however, only one well-defined rainfall peak occurred in March dur-ing summertime in both the time intervals.

Figure 5. Trends in observed annual mean rainfall anomalies (%) at Nadi and Suva during the 1961–2003 period.


discussion

Trends in Surface Temperature and Rainfall

The findings presented here suggest that theannual mean minimum nighttime tempera-tures have increased significantly (especiallyin Suva) during the past 43 yr and at a muchfaster rate than the annual mean maximumdaytime temperatures. As a consequence thediurnal temperature range has registered adeclining trend at both locations. The ob-served decrease in diurnal temperature rangeis consistent with global observations of de-crease in diurnal temperature range overland during the 1950 to 2000 period (IPCC2001b). Meanwhile, annual mean maximumdaytime temperature displayed no significanttrend for Nadi (Table 3). However, increas-ing summer mean maximum temperaturesare obvious at both locations, and the numberof hot days during the season went up in themost recent decade. An examination of thechanges in the frequency of extreme temper-ature events suggests that significant increaseshave taken place in the annual number ofhot days and warm nights for both Suvaand Nadi, with decreases in the annual num-ber of cool days and cold nights at both loca-tions. The number of hot days (Tmax b 32�C) showed a significant increasing trend

(Table 3), and the number of colder nights(Tmin < 18 �C) showed a decreasing trend atSuva (Figure 6). A systematic increase in the90th percentile of daily minimum tempera-tures was also observed. This increase wasaccompanied by a similar reduction in thenumber of colder days. There was a declinein the frequencies of cool days and an evenstronger decreasing trend in the frequenciesof cool nights.

The results of analysis on trends in ex-treme daily precipitation (defined as thoselarger than its 95th percentile for the year)during summer and winter seasons for theperiod 1961–2003 indicate that there is littletrend in total precipitation, but there aredistinctive trends in seasonal patterns at bothlocations. The number of wet days per year(with at least 2.5 mm of rain per day) in-creased marginally at Nadi. No significantincrease or decrease in normal rain days inwinter or summer season or per year wasdetected at Nadi or Suva (Figure 7). Thefraction of annual total rainfall from extremeevents (events exceeding the 1961–199095th percentile), however, increased at bothlocations. The frequency of extreme rainfallevents seems to have increased marginallybut is not significant. Increasing trends inthe average intensity of the wettest rainfall

TABLE 3

Statistical Significance of Past and Current Trends in Surface Temperatures and Rainfall at Nadi and Suva

Suva Nadi

Statistics Mea

nT

emp

erat

ure

Max

imu

mT

emp

erat

ure

Min

imu

mT

emp

erat

ure

No:

of

Day

sb

32� C

Rai

nfa

llA

no

mal

y

Mea

nT

emp

erat

ure

Max

imu

mT

emp

erat

ure

Min

imu

mT

emp

erat

ure

No:

of

Day

sb

32� C

Rai

nfa

llA

no

mal

y

P valuea 0.000 0.000 0.000 0.000 0.312 0.061 0.841 0.015 0.757R2 (Ad)b 54.0% 42.7% 53.2% 63.5% 0.1% 6.1% 0.0% 11.6% 0.0%

a P value: the smallest level of significance (a) for which the observed data indicate that the null hypothesis should be rejected.b R2 (Ad): the fraction of variance in Y (variable) accounted for by the variance in X (year) adjusted for degrees of freedom (linear

regression equation).


events each year were generally weaker atNadi than at Suva but were not significant ateither location.

Climate Variability and Extreme Events

El Nino has been recognized as the majorcause of interannual variability in rainfall inFiji and associated droughts. Conversely,during La Nina episodes rainfall is enhanced

across the western equatorial Pacific. Thechanges in ocean temperatures during ElNino and La Nina events are also accompa-nied by large-scale fluctuations in air pressureknown as the Southern Oscillation. TheSouthern Oscillation Index is one measureof the large-scale fluctuations in air pressureoccurring between the western and easterntropical Pacific (i.e., the state of the SouthernOscillation) during El Nino and La Nina epi-

Figure 6. Trends in number of days in a year with (a) day maximum temperature exceeding 32 �C and (b) nightminimum temperature below 18 �C at Suva during the 1961–2003 period.


sodes. Traditionally, this index is calculatedbased on the differences in air pressureanomaly between Tahiti and Darwin, Austra-lia. Prolonged periods of negative SouthernOscillation Index values coincide with abnor-

mally warm ocean waters across the easterntropical Pacific typical of El Nino episodes.The presence of the El Nino events in theequatorial Pacific region shifts the weathersystems and the tropical cyclones to the north

Figure 7. Trends in number of wet days during summer and winter seasons at Nadi and Suva during the 1961–2003period.


and east of the dateline, decreasing the rain-fall to the west of the dateline. The La Ninahas the opposite effect and the weather sys-tems are displaced to the south and east ofthe dateline, increasing the storminess andthe rainfall episodes over the western Pacificand generally resulting in above-average rain-fall in the region.

The 1980s and 1990s featured a very activeENSO cycle, with five El Nino episodes(1982/1983, 1986/1987, 1991–1993, 1994/1995, and 1997/1998) and three La Ninaepisodes (1984/1985, 1988/1989, 1995/1996)occurring during the period. This periodalso featured two of the strongest El Nino ep-isodes of the century (1982/1983 and 1997/1998), as well as two consecutive periods ofEl Nino conditions during 1991–1995 with-out an intervening cold episode. El Ninoconditions also prevailed during the year2001–2002. During these El Nino episodesthe main rain-producing system, the SouthPacific Convergence Zone, was displacednorth and eastward away from Fiji, resultingin suppressed rainfall in the western SouthPacific region (Folland et al. 2002). However,geographically Fiji lies in the transition cli-matic zone from dry air (high-pressure re-gion) to moist air (low-pressure region) ofthe Southwest Pacific, and the effects of ElNino on the rainfall of the two locations inFiji considered here were not always distinct.

Seasonal mean surface air temperaturesand seasonal total rainfall amounts at Nadiand Suva for the years with cold and warmevents during the 1961–2003 period are listedin Table 4. Also, composite rainfall amountsfor each of the two stations are plotted byENSO status in Figure 8. The dotted line inthis figure denotes the climatological meanrainfall for all years in the 1961–2003 period,the solid line the mean for the 12 compositeEl Nino years, and the dashed line the meanfor the eight La Nina years. Vertical lines de-note calendar year changes. Each plot showscomposite rainfall for the period beginningabout five seasons before the boreal winterof a mature episode and ending about fiveseasons after that winter. Differences betweenthe composite rainfalls for the samples repre-senting the warm phases of El Nino versus

that of the remaining (neutral plus La Nina)years passing a two-tailed significance testat the 0.05 level are indicated with a hollowsquare along the solid line. Significant dif-ferences at the 0.05 level with respect to theLa Nina composite rainfalls versus remaining(neutral plus El Nino) years are indicated inthe figure with a solid square along thedashed line. It is clearly evident that a moder-ate to strong El Nino event has a significantimpact on the rainfall at Nadi (in the westerndivision of Fiji) leading to serious to severedrought conditions (Figure 8). Though a sim-ilar relationship holds true for the rainfall inSuva (central region) as well, the magnitudeof impact is rather mild at Suva relative tothat at Nadi.

Year-to-year (interannual) variability acrossFiji is also related to the South Pacific Con-vergence Zone, a dominant feature of theatmospheric circulation in the SouthwestPacific that plays, in general, a pivotal role inboth variability and trends in South Pacificclimate. The South Pacific ConvergenceZone occurs where the southeast tradesfrom transitory anticyclones to the southmeet with the semipermanent easterly flowfrom the eastern South Pacific anticyclone(Kiladis et al. 1989). The South Pacific Con-vergence Zone exists both in summer andwinter, with approximately the same orienta-tion and location. It appears that the SouthPacific Convergence Zone moved eastwardabout 150 km around the year 1977 (Salingeret al. 2001). This has been confirmed fromsatellite pictures, which portray a shift in thelongitude of the sea-level pressure trough andthat of the minimum in outgoing longwaveradiation. The latter is a measure of thunder-storm activity: the less longwave radiation isemitted into space, the colder the cloud topsare. This shift of the South Pacific Conver-gence Zone caused relatively drier conditionsin Fiji, particularly in Suva. For example, thedaily rainfall, sunshine, and cloud cover datasuggest that Suva received @7% less rain andhad @12% less clouds on an annual meanbasis during the 1980s and 1990s relative tothe 1960s and 1970s. At Nadi, however, thesedeviations are rather marginal and inconclu-sive.


TA

BL

E4

Yea

rsw

ith

El

Nin

o(W

arm

)an

d/o

rL

aN

ina

(Co

ld)

Eve

nts

and

Cli

mat

ein

Nad

ian

dS

uva

El

Nin

o(W

)/L

aN

ina

(C)

Inte

nsi

tya

Nad

iT

emp

erat

ure

s(�

C)

Nad

iT

ota

lR

ain

fall

(mm

)S

uva

Tem

per

atu

res

(�C

)S

uva

To

tal

Rai

nfa

ll(m

m)

Yea

rsJF

MA

MJ

JAS

ON

DM

ay–

Oct

ob

erN

ove

mb

er–

Ap

rilb

May

–O

cto

ber

No

vem

ber

–A

pri

lbM

ay–

Oct

ob

erN

ove

mb

er–

Ap

rilb

May

–O

cto

ber

No

vem

ber

–A

pri

lb

1961

NN

NN

25.0

26.8

455

1,62

821

.723

.199

42,

788

1965

C�

NW

Wþ

23.4

26.5

410

990

20.7

23.0

1,00

51,

247

1966

WW�

W�

N23

.826

.213

11,

550

20.8

23.0

670

1,34

319

69W

W�

W�

W�

24.3

26.7

113

1,30

120

.923

.31,

076

1,59

419

71C

C�

C�

C�

24.5

26.3

539

1,43

321

.622

.81,

364

2,38

819

72N

W�

WWþ

24.3

27.1

518

1,18

821

.223

.71,

534

2,91

119

73W

NC�

Cþ

24.7

26.4

820

2,09

622

.123

.195

71,

544

1974

Cþ

CC�

C�

24.6

26.7

799

1,22

421

.523

.31,

240

2,14

919

75C�

C�

CCþ

25.1

26.5

514

1,64

921

.923

.01,

805

2,20

219

82N

W�

WWþ

24.3

26.9

484

825

21.6

23.5

1,10

01,

368

1983

Wþ

WN

C�

24.5

27.0

276

1,52

221

.323

.378

11,

653

1984

C�

C�

NC�

24.6

27.1

474

1,35

021

.823

.590

31,

948

1987

WW

Wþ

W23

.627

.415

1,20

621

.224

.032

71,

820

1991

W�

W�

WW

24.3

26.8

239

624

21.5

23.9

1,00

01,

697

1992

Wþ

Wþ

W�

W�

24.2

26.5

305

1,91

222

.323

.899

62,

507

1993

W�

WW

W�

23.6

27.2

312

1,37

921

.324

.01,

000

1,58

219

97N

WWþ

Wþ

23.9

27.4

475

518

21.8

24.2

729

1,02

019

98Wþ

WC�

C24

.827

.012

12,

228

21.7

24.1

528

1,72

519

99Cþ

CC�

C24

.926

.791

21,

787

22.4

24.0

1,42

21,

894

2000

CC�

NC�

24.9

27.0

620

1,59

922

.423

.91,

930

1,95

920

02N

W�

WW

24.9

27.3

440

882

22.4

24.2

1,09

01,

944

2003

W�

NN

N24

.3—

250

—21

.7—

750

—

aS

easo

nal

bre

akd

ow

no

fE

lN

ino

(W)

and

La

Nin

a(C

)ep

iso

des

inth

etr

op

ical

Pac

ific

(alo

ng

the

equ

ato

rfr

om

150�

Wto

the

dat

eli

ne)

usi

ng

the

rean

alyz

edse

asu

rfac

ete

mp

erat

ure

anal

yses

.W

eak

per

iod

sar

ed

esig

nat

edas

C�

or

W�

,m

od

erat

e-st

ren

gth

per

iod

sas

Co

rW

,str

on

gp

erio

ds

asWþ

or

Cþ

,an

dn

eutr

alp

erio

ds

asN

.Th

eye

ar19

61h

asb

een

incl

ud

edh

ere

asa

refe

ren

ceye

arw

ith

neu

tral

con

dit

ion

s.b

No

vem

ber

toA

pri

lte

mp

erat

ure

san

dra

infa

llco

rres

po

nd

toA

pri

lo

fth

ere

fere

nce

year

toA

pri

lo

fth

en

ext

year

.

Rainfall in Fiji also exhibits strong vari-ability on subseasonal time scales. These fluc-tuations in rainfall often go through an entirecycle in 30–60 days and are referred to as‘‘intra-seasonal oscillations’’ (also known as

‘‘Madden-Julian Oscillation’’). The Madden-Julian Oscillation is a naturally occurringcomponent of the coupled ocean-atmospheresystem and is regarded as the main fluctuationof atmospheric circulation that explains varia-

Figure 8. Relationship between ENSO events and rainfall at (a) Nadi and (b) Suva.


tions of weather in the tropics (Madden1986). Variation of the Madden-Julian Oscil-lation involves variations in wind, sea-surfacetemperature, cloudiness, and rainfall. TheMadden-Julian Oscillation is characterizedby a large-scale eastward movement of airin the upper troposphere with a period ofabout 20–70 days over the tropical easternIndian and western Pacific Oceans at approx-imately 200 hPa level (Rui and Wang 1990).During the summer the Madden-Julian Os-cillation has a modulating effect on cyclonicactivity in the western Pacific basin. TheMadden-Julian Oscillation organizes convec-tion into an eastward-propagating envelopeof smaller-scale disturbances. This envelopemoves eastward at between 5 and 10 m/secon average, generally forming in the tropicalIndian Ocean and dissipating over the centralPacific (Madden and Julian 1994). In thetropical Pacific, summers with ENSO-neutralconditions are often characterized by en-hanced 30- to 60-day Madden-Julian Os-cillation activity. Such summers are alsocharacterized by relatively small sea-surfacetemperature anomalies in the tropical Pacificcompared with stronger warm and cold epi-sodes (Hayashi and Golder 1993, Jones et al.1998). In these summers there is a strongerlinkage between the Madden-Julian Oscilla-tion events and extreme precipitation eventsin the western South Pacific region ( Jonesand Weare 1996).

The Madden-Julian Oscillation seems tohave a substantial impact on islands locatedin the western South Pacific that experiencerainy seasons during both winter and summerseasons such as Fiji. These disturbances areresponsible for many of the extreme precipi-tation events that are linked to flooding, asobserved in preliminary scrutiny of recordsof synoptic scale disturbances over Suva andthe daily rainfall records. Under the influenceof Madden-Julian Oscillation, reversal of thetrade winds (which blow from the SouthAmerican coast toward Asia) also takes place,which generally triggers Kelvin waves andtwin cyclones (Maloney and Kiehl 2002).The equatorial westerly winds generate ananticlockwise vortex in the Northern Hemi-sphere and a clockwise vortex in the Southern

Hemisphere. Twin cyclones, one in theNorthern and one in the Southern Hemi-sphere, are common in the months of April,May, and June during ENSO-neutral years.The Southern Hemispheric cyclonic systemcontributes substantially to the intense rain-fall over Fiji during those months. They alsogenerate Kelvin waves that travel across theentire Pacific in about 1 to 2 months (Malo-ney and Hartmann 1998). The increase infrequency and strength of the Kelvin wavesmay even lead to warming in the east and ini-tiation of an El Nino event.

Due to its slowly evolving nature, accurateprediction of the Madden-Julian Oscillationis fundamentally related to our ability tomonitor the feature and to assess its relativeposition and strength. Dynamical modelsgenerally do not predict the Madden-JulianOscillation well, partly because of the inher-ent difficulties that still remain regarding thecorrect mathematical treatment of tropicalconvective (rainfall) processes. Therefore,meteorologists have to use a variety of con-ventional data and analysis techniques tomonitor, study, and predict tropical intra-seasonal oscillations and their evolution.Satellite-derived data could be useful to indi-cate regions of strong tropical convectiveactivity and regions in which the convectiveactivity departs substantially from the long-term mean. These departures from normalare a fundamental diagnostic tool that shouldbe used directly to monitor and predict theMadden-Julian Oscillation as it propagatesover the islands located in the western SouthPacific Ocean and hence enhance the fore-casting skill of extreme precipitation events.

It is evident from our study that themoderate to strong ENSO events had a sig-nificant impact on the rainfall at Nadi (in thewestern division of Fiji) leading to severedrought conditions. Though a similar rela-tionship exists for the rainfall in Suva (centralregion) as well, the magnitude of impact israther mild at Suva relative to that at Nadi.The study further suggests that the twincyclonic vortexes forming under the influenceof Madden-Julian Oscillation and reversal oftrade winds could contribute substantiallyto the intense rainfall events in Fiji in


April, May, and June during ENSO-neutralyears.

acknowledgments

We gratefully acknowledge the support of theDirector of Meteorology, Fiji MeteorologicalServices, and the Head of Climate ServicesDivision in providing relevant archived clima-tological data for this study.

Literature Cited

Derrick, R. A. 1951. Weather and climate.Pages 102–119 in The Fiji Islands. FijiGovernment Press, Suva.

Folland, C. K., J. A. Renwick, M. J. Salinger,and A. B. Mullan. 2002. Relative influ-ences of the Interdecadal Pacific Oscilla-tion and ENSO on the South PacificConvergence Zone. Geophys. Res. Lett.29 (13): 1–4.

Hayashi, Y., and D. G. Golder. 1993. Tropi-cal 40–50 and 25–30 day oscillations. J.Atmos. Sci. 50:464–494.

IPCC. 2001a. Climate change 2001—Thescientific basis. In J. T. Houghton et al.,eds. Third Assessment Report of theWorking Group I of IntergovernmentalPanel on Climate Change. CambridgeUniversity Press, United Kingdom.

IPCC. 2001b. Climate change 2001—Thesynthesis report. In R. T. Watson and theCore Writing Team, eds. A contributionof the Working Groups I, II and III tothe Third Assessment Report of the Inter-governmental Panel on Climate Change.Cambridge University Press, United King-dom.

Jones, C., and B. C. Weare. 1996. The role oflow-level moisture convergence and oceanlatent heat fluxes in the Madden and JulianOscillation. J. Clim. 9:3086–3104.

Jones, C., D. E. Waliser, and C. Gautier.

1998. The influence of the Madden-JulianOscillation on ocean surface heat fluxesand sea-surface temperatures. J. Clim.11:1057–1072.

Kiladis, G. N., H. von Storch, and H. vanLoon. 1989. Origin of the South Pacificconvergence zone. J. Clim. 2 (10): 1185–1195.

Krishna, R. 1989. Climatology of Nadi Re-gion for planning purposes. Fiji Meteoro-logical Service Technical InformationCircular 89:3.

Madden, R. A. 1986. Seasonal variations ofthe 40–50 day oscillation in the tropics. J.Atmos. Sci. 43:3138–3158.

Madden, R. A., and P. R. Julian. 1994. Ob-servations of the 40–50 day tropical oscil-lation: A review. Mon. Weather Rev.122:814–837.

Maloney, E. D., and D. L. Hartmann. 1998.Frictional moisture convergence in a com-posite lifecycle of the Madden-Julian Os-cillation. J. Clim. 11:2387–2403.

Maloney, E. D., and J. T. Kiehl. 2002. MJO-related SST variations over the tropicaleastern Pacific during Northern Hemi-sphere summer. J. Clim. 15:675–689.

Prasad, R. 1979. Climate of Fiji. Fiji Meteo-rological Service, Nadi.

Rui, H., and B. Wang. 1990. Developmentcharacteristics and dynamic structure oftropical intraseasonal convection anoma-lies. J. Atmos. Sci. 47:357–379.

Salinger, M. J., J. A. Renwick, and A. B. Mul-lan. 2001. Inter decadal Pacific Oscillationand South Pacific climate. Int. J. Climatol.21:1705–1722.

Sharma, S. K. 1982. Climatology of NadiRegion—Fiji. Fiji Meteorological Service,Technical Note 16:1–23. Nadi.

Twyford, I. T., and A. C. S. Wright. 1965.Climate. Pages 72–79 in The soil re-sources of the Fiji Islands. Eyre and Spot-tiswoode, London.


baseline climatology of viti levu (fiji) and current...

Documents