Journal of the American Control Association,9(4):389-399, 1993 Copyright @ 1993 by the American Mosquito Control Association, Inc. SURVEILLANCE OF MOSQUITOES AND ARBOVIRUS INFECTION AT THE ROSS RIVER DAM (STAGE 1),

P. BARKER-HUDSON,T B. H. KAY', R. E. JONES,3I- D. FANNING2 AI'roL' D' SMYTHE4

ABSTRACT. This paper describes the temporal and spatial abundance of the mosquito fauna of the Ross River Dam (Stagel) in northern Queensland,Australia. Culex annulirostis, Anophelesannulipes s.1., uniformis, Mansonia septempunctata,and the nondam breeding vigilax were the major speciescollected by dry ice-supplementedlight traps set at various distancesfrom the edgeofthe reservoir. To estimatethe level ofarbovirus activity in thesedifferent zones,sentinel chicken flocks were bled 4 times a year and their antibody conversion rates determined by the hemaeglutination-inhibition test. Although mosquito abundanceat sites close to the reservoir were 1.5-6.1 times higher than at the more distani sites, arbovirus conversion rates, particularly to the alphaviruses Ross River and Sindbis, varied according to zone and year, suggestingthat risk ofinfection was no greater around the dam than elsewhere.

INTRODUCTION et al. ( I 990) were able to relate light trap returns to the activity of western equine encephalomy- In a previous article (Kay et al. 1990), we elitis and St. Louis encephalitis viruses. Trap pointed out that Australia is fortunate in lacking counts of Culex tarsalis Coq. were then used as widespread and serious vector-borne diseases, a basis for implementing control. although the alphavirusesRoss River (RR) and As the costs of surveillance systemsincrease Barmah Forest (BF), and flaviviruses Murray with their complexity, our basic approach to- Valley encephalitis(MVE), Kunjin (KUN), Ko- wards designingsuch a system was first through kobera (KOK), and dengue(DEN) causeclinical detailed understanding of spatial and temporal infection. In that paper we also described the trends ofvector populations and second,through Ross River Dam (19"26'5, 146o45'E)near association of those populations to arbovirus Townsville, north and indicated that Queensland, transmission.The potential elucidation ofa sim- mosquito and arbovirus activity was high in its plistic management tool based solely on abun- vicinity. Our present submission analyzesthose dance has advantagesover those which out of data further by examining the spatial and tem- necessityinclude other epidemiological consid- poral abundance of the mosquitoes, especially erations. those of public health importance, in relation to arbovirus activity at different distancesaway from the dam. MATERIALS AND METHODS In relation to surveillancemethodologies, our Study area: The Ross River Dam (Stage l) basic hypothesisrelates to mosquito abundance was constructedin 1973with an augmentedstor- as a reliable indicator of arbovirus transmission. age capacity of 109,000 megaliters and at full Bunnag et al. (l 979) found that anophelinesand supply level inundates approximately 3,000 ha. malaria prevalenceincreased with the construc- Details of the dam have been given by Kay et tion ofthe SrinagarindDam in westernThailand. al. (1990).In 1987-88,proposed works for en- In his concise review, Eldridge (1987) pointed largement to Stage 2a were completed, which out that measuresof mosquito abundancealone increased the storage capacity to 236,000 me- generally poor indicators have been ofarbovirus galiters. epidemics,but nevertheless,in California, Reeves Weekly recordingsof the level of water at the spillway, and monthly calculations of the amount of water stored in the dam, were maintained by tP. O. Box 42828, Casuarina,Darwin, Northern the Townsville City Council, and meteorological Territory 0810 Australia (formerly Vector Control Unit, data were recordedat JamesCook University, 2 QueenslandHealth). km away from the dam. 2 QueenslandInstitute of Medical Research,P. O. Mosquito sampling: Adult mosquitoes were Royal BrisbaneHospital, Brisbane4029 Australia (for- sampled with Encephalitis Virus Surveillance merly Director, Vector Control Unit, Queensland (EVS) light traps baited with dry ice (Rohe and Health). Fall 1979), modified by the addition of a pho- 3 Department of Zoology, JamesCook University of tosensitive switching device to turn the traps on North QueenslandP. O., Townsville 4811 Australia. a Laboratory of Microbiology and Pathology, and offat dusk and dawn. A bafle was installed QueenslandHealth, 63-76 GeorgeStreet, Brisbane 4000 between the fan unit and the collection bag to Australia. prevent the escapeof collectedmosquitoes when 389 390 Jounxer or tnE Arvrrrrcln Mosqurto Cowrnol Assocr,uroN Vor. 9, No.4 the light trap switched off. Traps were suspended crobiology and Pathology, QueenslandHealth, I m above the ground in shaded areas and op- Brisbane, for testing by the hemagglutination- erated for 2 consecutive nights per week from inhibition (HI) test, using a modified Clarke and January 1984to April 1985.From May 1985to Casals( I 958) procedure.Ether was deletedfrom September 1985, trapping frequency was re- the extraction procedure and acetone-extracted ducedto one night per weekfor 3 wk and 2 nights gooseerythrocyte adsorbedsera was used. per week for the other week. Ross River, Sindbis, Murray Valley enceph- Seven EVS trap sites for mosquito sampling alitis, and Getah hemagglutinationantigens were were selectedto represent3 distinct zones: preparedin sucklingmice usingsucrose acetone extraction procedures.Flavivirus l) Znne l, adjacent to the Stage I shoreline (3 antigenswere preparedinitially using procedure, trap sites). Two sites were chosen along the the above but were later prepared in C6/36 cell culture fluid northern shoreline(Big Bay, Ti-Tree Bay) and which was precipitated with polyethyleneglycol one on the easternshoreline (Round Island). and resuspendedin borate salinebuffer (pH The 3 sites were vegetatively similar, being 9.0). Virus was identified when dominated by open eucalypt woodlands with titers were at least 4-fold higher than other group reactors,or when ti-trees (Melaleuca species)also common in it was the sole reaction. the wetter soils, at the wetter end of slopes Analysis: All mosquito counts were trans- (Round Island), and in broad drainage de- pressions. formed to log(count + l) to stabilizethe variance in formal analyses,and all statistical analyses 2) Zolre 2, about 2 km from the shoreline(2trap werecarried out usingSAS Institute (1988)pro- sites,at Oak Valley and Toonpan).These were cedures.Three main setsof analyseswere carried sited close to the proposed shoreline of Stage out to examine the behavior of mosquito pop- 2 of the dam, thus providing baselinedata of ulations. future monitoring. Both comprised open woodland, interspersedwith the introduced l) To compare the speciescomposition of the shrub Zizyphus mauritania, and with open total catch at each trap location, cluster anal- areascleared ofvegetation for rural residen- yseswere used, with the data transformed to (Oak tial development Valley) or grazing standard scoreswith mean 0 and variance l. (Toonpan). The results presented use average linkage 3) Zone 3, about 4 km from the shoreline(2trap cluster analysis, but the same clustering se- sites, Kelso and Stanley). These 2 localities quenceswere obtained using alternative pro- will still be at some distance from the dam cedures (centroid, flexible beta, complete after the Stage2 expansion.As with tll'eTane linkage). 2 sites, Kelso and Stanley had been cleared 2) To examine the extent to which mosquito grazing, for residential development and for abundanceat different sitescovaried through respectively. Trapping did not begin until the sample period, Pearson'scorrelation co- April 1984at Stanley. efficientswere calculated for eachpair of sites, Sentinel chickenflocks and virus sumeillance: usingas data points the I I 5-l 49 separatecol- Sevenflocks eachof lG-30 chickenswere main- lections made at each site. These correlation tained within Zone | (Big Bay, Round Island), coefficients were calculated for the 4 most Zone2 (Oak Valley,Toonpan), and Zone 3 (Kel- common speciesbreeding at the reservoir,and so, Stanley) with a supplementary flock being also for Aedesrigllax (Skuse),which doesnot held at Pallarenda(an area I 5 km to the northeast breed at the reservoir. BecauseAe. vigilax is bordering coastalsaltmarsh, and here labelled as a coastal saltmarsh speciesthat hatchessyn- Znne 4). EachNovember, priorto the wet season, chronously in responseto tidal cycles,all in- which was recognizedas the probable period of dividuals of this speciesthat dispersedto the greatest arbovirus activity, antibody-positive dam environs were probably derived from a chickens were discarded and flocks replenished common source. We would therefore expect with nonimmunme birds. Chickens were bled that its abundancewould vary in similar ways from the wing vein 3 or 4 times per year: usually at all sites,and hencethat correlationsamong in mid-wet season(February-March), and at the all trapping sites would be uniformly high. end of the wet season(April-May) to define this Similarly, freshwater-breeding species de- activity and once thereafter until November. A rived from sites subject to the sameenviron- flock was maintained at each of the 6 sites from mental influencesshould similarly show high November 1983to June 1987.Sampling at the correlations.Conversely, a low correlation for 7th site, Pallarenda,was terminated in February a particular pair of sites may indicate that l 986. mosquito abundanceis determined by differ- Serawere dispatchedto the Laboratory of Mi- ent factors at each site. Drcrrrrssn1993 MosqurroEs eNp Apsovtnusrs rN NontHsRN AUSTRALIA 391

3) To examine the seasonalpattern of abun- recorded as larvae at the dam (Rae 19835).Be- dance. and the effects of zone and location cause some speciesknown to breed in the res- within zone, on the abundanceofparticular ervoir also breedin other freshwaterbodieswith- taxa and on the total catch, mixed model in all 3 zones,the extent to which the community analysesof variance were performed (using structure of the dam-breeding speciesfollows a the general linear model procedure to allow zonal pattern gives someevaluation ofthe extent for the unbalanceddesign). To accommodate to which mosquitoesderived from the dam dom- the data set to the computing capacity avail- inate the region's mosquito fauna, and also gives able, the sampling period was divided into 2 an estimate of which zones are subject to the monthly intervals, and trap-nights within an reservoir's influence. The non-dam-breeding interval were treated as replicates. The ana- speciesprovide a form of"control," becausethere lytical design therefore involved 2 crossed is no reason to expect them to follow a zonal factors, Zone and Month (both fixed factors), pattern. with Locations (random factor) nestedwithin Clustering only on the dam-breeding species Zones,and a variable numberofobservations (Fig. l) separatesthe sitesinto 2 majorgroupings: for each factor combination. This analysis was the 5 locations in Zones I and 2, vs. the 2 lo- applied to the total mosquito counts, and also cations in Zone 3. Within the first group, loca- to the 5 most abundant taxa. Four of these tions do not completely separateby zone: the (Culex annulirostris Skuse,Anopheles annu- speciescomposition of Round Island (Zone l) lipes s.l. Walker, Mansonia uniformis Theo- was more similar to the Zr:,ne2 sites than it was bald, and Mansonia septempunctata Theo- to the other Zane I sites. The main reason for bald) were known to breed in the reservoir. this was its relatively low numbers of the 2 Man- The 5th, Ae. vigilax, does not breed in the sonia speciescompared to the other 2 Zone I reservoir, as noted earlier. sites(see Table l). This analysissuggests that the effectsofthe dam on mosquito abundanceand The sentinel chicken data were analyzed by speciescomposition extendfor at least2 km from logistic regressionusing GLIM 3.77 (Baker and the shoreline,and that mosquitoesfrom the dam Nelder 1978) to compare the proportion of the probably do dominate the mosquito faunaudthin chicken flocks that developed antibodies to al- this area. phaviruses and flaviviruses among years and Clustering on speciesthat do not breed in the among sites. reservoir gave a very different pattern (Fig. l). Toonpan was clearly the most dissimilar to the other 6 locations, mainly ofits very high RESULTS because abundanceof a number of the Aedesspcies (Ta- Weatherand water levelsthroughout the study: ble l). Although Big Bay and Ti-Tree Bay proved In I 984, wet seasonrains were restricted to Jan- to be the most similar locations (note that they uary and February @a. I,OOOmm) but in 1985, are also the 2 closesttogether), site groupingsdid the wet seasonfailed. After February 1984 when not correspond in any way to the zones. the capacity was 101,000megaliters, the volume Temporal cotelations among sites in species steadily decreasedto 30,500 megaliters at the abundances: Examination of the patternsof cor- termination of the study in September 1985. relation for particular speciesin casesamong sites Minimum and maximum monthly temperatures (Table 2) strongly reinforces the conclusions varied from JuIy (ca. l3-24'C) to January (ca. above. We would, in general,expect to seepos- 24-32C)- itive correlations among sites,since they sharea Species composition and total catch of mos- common seasonalpattern. But we would expect quitoes: The number of speciescaptured (Table to see stronger correlations where the sites re- l) did not ditrer greatly between zones; if any- ceive mosquitoes from a common source: that thing, locations in Zone 2 were slightly more is, for dam-breeding species,among sites close species-richthan locations in the other 2 zones, to the dam, and for the highly mobile saltmarsh Traps in Znnes I and 2 caught more (average 80 mosquito Ae. vigilax, among all sites. For Ae. vigilax, and 55, respectively) individuals than traps in there are indeed strong cor- relations between all pairs of sites, supporting Zone 3. There are also smaller differencesbe- tween locations within each zone: within Zone 3, fewer mosquitoeswere caught at Stanleythan 5 at Kelso; and in Zr,ne 2, fewer were caught at Rae, D. 1983. The mosquito larvae of Ross Dam with particular reference Oak Valley to the ecology of Culex an- than at Toonpan. nulirostris. B.Sc. (Honors) thesis. Department of Bio- Mosquitoes were categorized into 2 groups: logical Sciences, James Cook University of North those known to breed at the dam. and those not Queensland,Townsville. 392 JounNer, or rne ArvGRrclN Moseurro Coxrnor Assocr.lrroll Vor.9, No.4

Table l. Mosquitoes captured by COr-baited EVS light traps at 7 sitesr near the Ross River Dam betweenJanuary 1984and September1985. Taxon BB TT RI OV Tp Ke St Aedes alboscutellatus 0 I 02t2 190 Aedesalternans I 0 l4 5 30 Aedeselchoensis l0 3 9499 20 Aedeskochi I I o234 93 Aedeslineatopennis 2 O 3 15 1,045 80 l Aedesmallochi 0 0 00 J 00 Aedesnormanensis 27 15 r92 36 I,139 163 s2 Aedes notoscriptus 0 0 049 28 15 22 Aedespurpureus I I 3377 00 Aedesquasirubithorax 0 0 00 5 0l Aedes vigilax 379 23O 94 313 373 323 444 Aedesvittiger 12 4 139 19 385 194 Aedeomyiacatasticta ll 99 61 365 l0l I 15 73 Anophelesamictus 97 37 548 42 65 44 133 Anophelesannulipes 8,850 2,919 6,381 1,273 I,175 3r7 293 Anophelesbancroftii 146 70 73 34 18 46 Anophelesmeraukensis 107 68 366 20 227 23 33 Coquillettidia rassipes 78 324 47 78 83 58 ll Culex annulirostris 2,478 2,138 l,806 3,533 4,885 928 351 Culex bitaeniorhynchus 29 44 29 13 20 60 Culex pullus 0 0 00 2 20 Culex quinquefasciatus 36 36 55 r73 122 8l I Mansonia uniformis 2,188 2,316 354 39 72 21 5 Mansonia septempuncteta 330 1,518 109 60 74 276 Uranotaenia albescens 2 O 25 .J 00 Uranotaenia nivipes 2 4 0l 0 00 Number collected 14.787 9,828 ro,272 6,083 10,052 2,259 r,439 Number of speciescollected 2l 19 19 23 25 2l 17 Number of collections 146 143 t45 149 146 t49 r20 r Zone I locations: BB-Big Bay; TT-Ti-Tree Bay; RI-Round Island. Zone 2 locations: OV-Oak Valley; Tp-Toonpan. Zone 3 locations: Ke-Kelso; St-Stanley.

the expectation that abundancesofthis species bers of individuals collected at sites in Zones 2 would vary in a similar way at all sites (seealso and 3 in itselfsuggest that the reservoir produces Fig- 2). There was no particular tendency for all or most of these mosquitoes. Sites in Zones correlations among sites within a zone to be I and 2 tended to be well correlated with each stronger than correlations among sites in differ- other (except for Round Island, where few Man- ent zones. sonia were caught). Correlations involving Zone For Cx. annulirostris, Zone I and 2 sites were 3 sites are much less consistent. strongly correlated within their groupings, but Seasonal and spatial variation in abundance: Zone 3 sites, on opposite sides ofthe reservoir, Table 3 lists the significant factors influencing were only weakly correlatedwith each other. All the major species (Figs. 2-5) and that of total pairs ofsites gave significant correlations. mosquito abundance, as identified in analyses of Anophelesannulipes shows a clear zonal pat- variance. For total mosquito abundance there is, tern in its correlations. Zone I sites, closest to predictably, a range of effects. There are marked the dam, are all strongly correlated with each temporal changes in average abundance (as in- other. Correlations between Zone I and Zane 2 dicated by the significant effect of Month in the sites are less predictable: Zone I sites are all analysis of variance). There is also an overall strongly correlated with one of the Zone 2 sites difference between zones in average abundance, (Oak Valley) but not the other (Toonpan). Zone with Zones I and 2 having much higher total 3 sitesare only weakly correlatedor are not cor- abundances than Zone 3 (Table l), and there is related with sites from other zones, or indeed a smaller, but still significant, difference between with each other. locations within zones (Table l). There are also For the Mansonia species,the very low num- differences in the seasonal pattern ofabundance DscErvrasn Moseurrors lrvo Annovlnusrs rN NonrnsRN Ausrnlr:e 393

CLUSTERI}IGON DAN BREEDII{G 1.6 : + o u 1.4 o =F (to 1.2 =o a t.o 9 o.E +BIG BAY +TI-TREE +ROUND IS 0.6

0.4 + at ou o.2 E ot o ZONE2 o- AB TT RI ov Tp sr Ko =

+OAI(VALEY +TOONPAN CTUSTERIIIGON 1.0 ; o, l.a U P f ZONE3 1.2 o o- l.o 9 o.t

0.6 OKEISO +STANLEY Fig.2. Average nightly catchesof Aedesvigilaxby 0.4 month in relation to sites and zones. o.2 affecting each individual species reveals a sim- pler pattern in every case than that oftotal abun- BB TT OY K. 3r Rt Tp dance. Fig. I . Cluster analysesof mosquito specieswhich l) Ae. vigilax.' The pattern ofabundance of this do, and do not breedat the RossRiver Dam, in relation species clearly reflects that it does not breed to trapping sites. Abbreviations: BB-Big Bay, TT- in the dam (Table 2 and Fig. 2). There are Ti-Tree Bay, RI-Round Island, OV-Oak Valley, Tp- very marked seasonal changes in abundance, Toonpan, St-Stanley, and Ke-Kelso. with peaks occurring at similar times in all locations. There is a weak effect oflocation on overall abundance, arising because catches between zones, as indicated by the Zone x Month at Round Island were lower than elsewhere, interaction, and also between locations within but there are no differences between zones. zones. Thus the timing of peaks and troughs in 2) Culex annulirositris: Twoseasonal peakswere abundance is spatially variable. Total mosquito evident (Fig. 3): postwet season (March-May) abundance represents the aggregate ofall the spe- in both years, and a spring peak in Septem- cies sampled, and hence sums the behavior of ber-October, giving a strong effect of Month different species whose seasonal and spatial pat- (Table 3). There are marked differences be- terns of abundance may be quite different from tween zones, with substantially fewer indi- each other. Consequently, examination of factors viduals caught in Znne 3 than in the other 2 394 Jounrtel or rne AMBRTcIN Moseurro CoNrnor Assocr^lrroN Vor.9, No.4

Table 2. Pearson'scorrelation coemcients,expressed as levels of sigrrificance,for logarithmically transformedcatchest:olt!?ilti,l,i1lltT;.ff1-'j':t:i':",r1ffiffi0'000r'b:P<0'00r'

Znne I Znne 2 Zan'e3 Species/site BB TT RI ov TP ST K Species Anophelesannulipes Culex annulirostris Big Bay (BB) a a c b dd Ti-Tree Bay (TT) a a a a da Round Is. (RI) a a a aa Oak Valley (OD aaa AQ Toonpan (TP) NSNSb ab Stanley (ST) NSbb NS NS Kelso (K) bdNS NS NS Mansonia unifurmis M ansonia septempunctata Big Bay a a aa bNS Ti-tree Bay a b aaNS NS Round Is. a a ddNS NS Oak Valley a aa cNS Toonpan a aa NS NS Stanley b ac aa Kelso NS bd dd Aedes vigilax all a

Table 3. Significant effectson the abundanceof mosquito speciesat 7 sites near the Ross River Dam, as detectedby unbalancedmixed-model analysis of variance applied to log-transformed abundance data.

Taxon Effect dfFP Culex annulirostris Zone 2x4 68.92 0.007 Month 10x39 9.26 0.0001 Month x Location 39 x 921 3.90 0.0001 Anophelesannulipes Zone 2x4 15.85 0.0r Month 10x39 ).o) 0.0001 Location 4x42 4.76 0.003 Month x l.ocation 39 x 921 4.86 0.0001 Mansonia uniformis Month 10x39 1o.79 0.0001 Location 4x4l 7.19 0.0002 Month x Z.o.ne 20x39 4_60 0.0001 Month x Location 39 x 919 6.00 0.000r M a nso nia sept empunct at a Month 10x39 4.98 0.0001 Location 4x4l I l.l9 0.0001 Month x L.ocation 39 x 921 7.96 0.0001 Aedes vigilax Month 10x40 23.9r 0.0001 Location 4x6l 2.59 0.05 Total mosquitoes Znne 2x4 r7.56 0.01 Month 10x39 11.79 0.0001 Location 4x42 2.94 0.03 Month x z,one 2Ox39 1.86 0.05 Month x Location 39 x 921 4.27 0.0001 DeceMBen1993 MoseurroEseNn AnrovrnusnsrN NonrnBu Ausrnarn 395

+ Er, o ZONE 1 -F I =o (,o. =

JFMAMJJASONDJFMAMJJA JFMAMJJASONDJFMAMJJA +BIO BAY+TI.TREE +ROUNO IS +B|q BAY+TFTREE +ROUND IS

; qt. ; ZONE 2 o ZONE2 F U 5 o o te o = o8 go - I o 9 JASONDJFMAMJJA J FMAMJ J ASOND J FMAI'J J A +OAI(VALLEY +TOONPAN +OAKVALLFT/ -TOONRAN

+ at : U + o o E ZONE3 u l ZONE 3 I -P o o - o O, o 9 = = J FIIAf J J A gON D J FIIAI' J J A OXETT}O +STANLEY JJASON AIIJ J A Fig. 3. Average nightly catches of Culex annuli- +KELSO+STANLry rostris by month in relation to sites and zones. Fig. 4. Average nightly catches of Anophelesan- nulipes by month in relation to sites and zones.

zones. The average abundance was similar at locations within each zone, but the timing of nulirostris and were generally characterized abundance peaks varied between locations by a buildup during the winter-spring dry sea- within Zone l, giving a significant Month x son. However, in 1985 patterns ofabundance Location interaction: patterns of abundance were similar to that for Cx. annuliroslrus with at Big Bay were clearly different from those an autumn peak. at the other 2 Zone I locations, showing a 4) Mansonia uniformis: This species achieved progressive winter-spring buildup similar to high abundance only at locations in Tnne I An. annulipes at Z,one I sites (Fig. 4). (Fig. 5), but effects ofZane on average abun- 3) Anopheles annulipes: Here too, there were dance throughout the period were swamped marked differences in abundance between in the analysis ofvariance by the variability zones. In this case Zone l, on the shoreline. between locations within Zane I and by the had by far the highest numbers of this taxon, strong Month x Zone interaction. Abun- and abundance decreased progressively with dances ofthis species in Zone I appeared re- distance from the dam (Table I and Fig. 4). lated to water level of the dam, with a peak Especially in Zane l, there were also differ- early in the study followed by a steady decline ences between locations: peak abundances at to 0 at all locations by the end ofthe study. Ti-Tree Bay were consistently lower than at The species was rare at any time inZone 3. the other 2 shoreline sites. In 1984 seasonal 5) Mansonia septempunctatq: Patterns of patterns ofabundance differed from Cx. an- abundance for this species are very similar to 396 JounNlr or rHp AvsRrclN Moseulro Colvrror, Assocr,crIoN Vou 9, No. 4

not consistentwith zone (y2 : 27.46,8 df, P < 0.001). Proportions remained steadyor dropped over time for Zones | , 2, and 3 whereasthey rose for Znne 4 (although the precision ofthese latter ZONE 1 data is inferior to those for Zones l-3). For the flaviviruses, zone (y2 : 30.93, 8 dl P < 0.001)and year(x2 : 17.17,3df, P < 0-001) were both significant as was their interaction (12 : 30.93,8 dt P < 0.001),once again indicating that differencesdepended on the site under con- UAMJJASONDJ sideration. Proportions remained consistently +BIO 6AY +TI.TREE +FIOUND IS high for Zone 2 and proportions increasedsteadi- ly atZnne 3 sites.Fewconversionsto flaviviruses occurred it Zone 4 (Pallarenda). There was evidence of the activity of the fol- lowing alphaviruses,Ross River (all years),Sind- bis (all years),and Getah (1985-86, 1986-87). ZONE2 Barmah Forest virus was not included in the testing panel. Flavivirus activity was consider- ably less than that of the alphaviruses but the following were consideredto have been present: Valley encephalitis (1982-83, 1986)' JFMAMJ At,JJA Murray Kunjin (1984), Kokobera (1985-86),Alfuv AOAKVALLEY +TOONFAN (1984-86),and EdgeHill (1984).

DISCUSSION From January I 984 to September1985' 54,720 adult mosquitoeswere collected by the EVS traps in Zone I on the Stage I water margin (average collection 80 mosquitoes/trap), inZone 2 on the planned Stage 2 water maryin (55 mosquitoes/ trap), and in Zone 3, away from the dam (13 mosquitoes/trap). Mosquito prevalenceat each +KELSO +STANLEY of the 3 zones was to a large extent influenced Fig. 5. Averagenightly catches of Mansoniaunt' by the proximity of suitablebreeding habitats for formis by month in relation to sitesand zones. 3 taxa, An. annulipes, Cx- annuliroslrrs' and Mansonia spp., which comprised81.2olo of the total catch. Both Cx. annulirostris and An. an- those for Ma. uniformls (Table 2 and Fig. 5)' nulipesbreedextensivelyin the reservoir but also (Ti-Tree except that at one Zone I location in permanent water holes, semipermanent Bay), abundancesremained relatively high for swamps, and temporary rain-filled pools away a longerperiod than elsewhere,giving a strong from the dam. Both Mansonia spp. breed in as- Month x Location interaction. As fot Ma. sociation with aquatic plants, most common in uniformis, high abundancesonly occurred at the Big Bay and Ti-Tree Bay areas of the lake shoreline (Zone l) locations, and even there (Rae 1983s). declined to 0 as the study progressed. Examination of numbers of Cx. annulirostris, (dam-breeo- Arbovirus conversionrates: Hl-antibody con- An. annulipes, and Mansonia spp- (Fig. 3) demon- version rates for the 4 years from 1983-84 to ing species)by cluster analyses Zones I and 2, 1986-87 were analyzedfor Zones 1,2, and 3 strated the similarities between spp. from with that for Pallarenda(Zone 4) from 1983-84 although reduced numbers of Mawonia in this site being more to 1985-86(Table 4). Round Island resulted Island lacked For the alphaviruses,there weresignificant dif- closely linked to Zane 2. Round Mansonia breed' ferencesin the proportion ofthe chickens con- floating vegetation in which separatecol- vertingby both zone(12 : 15.O2,3df' P < 0.005) Pearson'scorrelations of I l5-149 speciesfrom the 7 and year(x2 : 69.19,3 df, P < 0.001). lectionsofeach ofthe above also indicated There wasa significantinteraction effectin that sites further illustrated this, but from Toonpan were the differencesseen over the 4-year period were that An. annulipes catches DscEMsnR1993 Moseurronslxo Annovrnusnsw NonrNsnn Ausrnern 397

Table 4. Percentagecumulative seroconversionsin domestic chicken flocks to alphaviruses (Alpha) and flaviviruses (Flavi) tested by haemagglutination-inhibition test, Ross River Dam Zones I and 2 (on or nearby the reservoir) compared to Zones 3 and 4 (removed from the dam). Month of Zane I Zone 2 Zone3 Zone4 serum Year collection Alpha Flavi Alpha Flavi Alpha Flavi Alpha Flavi 1983-84 January 25.7 l.l 24.4 0.5 t7.2 0 00 March 45.7 l.l 31.3 t.2 48.2 1.3 200 November 68.3 4.3 74.8 1.7 86.7 1.8 32.5 0 1984-85 January 0 0 34.9 1.0 t7.2 1.5 44.4 0 February 0 0 35.8 1.0 l9.l 1.5 44.4 0 August 25.8 l.l 59.2 t.5 43.1 3.1 88.8 I l. l November NA NA NA NA NA NA NA NA 1985-86 February 18.4 o.4 34.2 4.5 t5.7 o.4 75.0 0 March 35.9 2.6 4t.l 7.4 35.7 o.4 May 66.5 5.8 66.r 9.5 63.0 6.5 November NA NA 84.4 9.5 9l.l 10.0 1986-87 December 0 0 0 0 NA NA March 1.4 o.7 14.3 4.8 20.8 6.9 April 39.9 0.7 19.6 5.7 42.4 12.8 June 51.0 0.7 22.6 5.7 49.5 12.8 Note: New sentinel flocks were established each year after the November bleeding. dissimilar (Fig. 4) to those from Big Bay and Ti- Cx. annulirostris and,An. annulipes was restrict- Tree Bay. ed to 150 m from water. More recently, RusselL In generalwithin Zones I and 2, correlations (1986) demonstrated that at Echuca, Victoria, between sites in speciesabundance were highly some Cx. annulirostris dispersedat least 7 km, significant but this was not the case for Zone 3 subsequently confirmed in southwestern New sites,Stanley and Kelso, where within-group sig- SouthWales (O'Donnell et al. 1992).The relative nificancevalues were P < 0.05, < 0.01,NS, and contributions of the different breeding sites of NS for Cx. annulirostris, An. annulipes,and the Cx. annulirostris and An. annulipes in relation 2 Mansonia species,respectively (Table 2). Al- to the numbers collectedin the 3 zonesis difficull though both ofthese trapping siteswere 3-4 km to interpret because of the diversity of their away from the dam and in similar topography, breeding habitats. However, the strong correla- they also were the most distant from each other, tions for the major numbers of adult An. an- being on opposite sides of the dam. Whereas nulipes and, Cx. annulirostris at Zrlne I and 2 numbers of Cx. annulirostris and An. annulipes sites, respectively, suggestthat extensive breed- would be affected by common factors such as ing habitat in the reservoir extendsan influence generalized rainfall and flooding, we were not for at least 2 km. The more rapid falloffin num- able to record localized falls or the various water bers of ln. annulipes compared to Cx. annuli- managementpractices of residentson what pre- rostris is probably related to their relative dis- dominantly are rural blocks. There was, how- persal powers. ever, limited suitable breeding habitat. The temporal patternsof Cr. annulirostris,An. That all Ae. vigilax correlations were highly annulipes, and,Ae. vigilax collected at Big Bay, significant (P < 0.0001) for all 7 sites suggests: Oak Valley, and Stanley, and for Ma. uniformis l) a common influence(i.e., saltmarsh inundated and Ma. septempunctataat Big Bay, have been by tides), and 2) wide dispersal.Aedes vigilax is figuredfrom April 1984 to September1985 in well known for movement of up to 64 km from Joneset al. (1991). A more extensiveanalysis its coastalsaltmarsh breeding sites (Marks 1969) with additional sitesindicates that temporal pat- and therefore, an equitable distribution at all terns can difer within a particular zone (e.g.,for trapping siteswould be expected.In contrast,the Cx. annuliroslrusat Big Bay comparedto Ti-Tree restricteddistibution of Mansonia spp.to main- Bay and Round Island in 1984), from zone to ly the Zone I sites(Fig. 5) suggeststhatboth Ma. zone (e.g.,Cx. annulirostris atTi-Tree Bay com- undormis and Ma. septempunctatado not far. pared to Oak Valley and Toonpan), and from Along the low-lying flats of the Murray River year to year (e.g., An. annulipes in 1984 com- in Victoria, Myers (1954) noted that biting of pared to 1985).Although the averageminimum 398 Jounrqnr or rnr Avep.rceN Moseulro Cotvrnol AssocnnoN Vor. 9, No.4 temperaturefor June I 984 was 2.8'C higher than t) Trap locations to monitor Cx. annulirostris, for June 1985, July-August 1984 temperatures An. annulipes, and the Mansonia species were0.8-1.5oC colder. Hence, there does not seem should be stratified accordingto their distance to be an obvious explanation for this temporal from major breedingsites. In this case,where difference. the Ross River Dam is clearly of major im- Peakabundance of Cx. annulirostris toward or portance, the minimum requirement would at the end of the wet season(March-May) is well be for a trap closeto the water's edge,chosen known (e.9., Charleville and Kowanyama [Kay for proximity to aquatic vegetation, and a trap 1979], Darwin and Casuarina[Russell and Whe- at least 3 km away. The existenceof substan- lan 19861).Increasing temperatures after winter tial variation betweenlocations within a zone generallyproduced a spring peak in September- of some speciessuggests that several traps October and numbers generallyrose until heavy within each stratum would be a significantly wet seasonrains arrived. At this time, extensive better monitoring design if resources were gowth of Hydrilla verticillata formed extensive available to support it. Any ofthe above com- mats over up to 30o/oof the reservoir's surface binations would be suitable for Ae, vigilax. due to decreasedwater depth, thereby providing 2) Eldridge ( I 987) has reviewedstudies in which extensive additional breeding habitat for both there was either a positive or no correlation Cx. annulirostris and An. annulipes (Barker- between mosquito abundance and arbovirus Hudson et al. 1986).Although Mansonia spp. transmission. Clearly in this case,the situa- numbers decreasedalmost from the commence- tion at the RossRiver Dam falls into the latter ment of the trapping program, those for Cx. an' category where other entomological and ep- nulirostris and An. annulipes in Zones I and 2 idemiological factors require consideration. generallydid not. The cluster analysisof non-dam-breedingspe- cies clearly separatedToonpan from the other ACKNOWLEDGMENTS sites, probably due to the suitability of its clay- basedsoils for Aedesoviposition and for creation We thank George Hapgood, Ron Anderson, of many temporary pools after rainfall. Thus, and other state and local health department per- 76.60/o of the 3.352 Aedesnormanensis (Taylor), sonnel for assistance with mosquito trapping and Aedeslineatopennls (Ludlow), and Aedesvittiger bleeding of chickens; Ted Plum, Townsville City (Skuse)were collected in the COr-EVS trap set Council for his local expertise; Diana Battistutta at Toonpan. for doing the statistical analysis on the serological Alphavirus activity was generallyhigh and un- data: and Mark Hearnden and Don McManus predictableat the 4 zoneswhere sentinel chickens for reviewing the manuscript. The study was were maintained (Table 4). Flavivirus activity, funded by the National Diseases Control Pro- including Murray Valley encephalitis and Kun- gramme, and through a partnership between the jin, was detected but at low levels. However, Townsville-Thuringowa Water Supply Board and sentinel chickens develop an HI antibody re- the Land and Water Resources Research and De- sponsemore readily to flaviviruses than to Ross velopment Corporation. River virus (Kay et al. 1986),one ofthe common alphaviruses,so seroconversionrates to alphavi- ruses may be underestimated. REFERENCES CITED abundanceof both mosquitoes Although the Baker, R. J. and J. A. Nelder. 1978. The GLIM sys- I and and vertebrate hosts was greatestin Zones tem, releaseIII. Numerical Algorithms Group, Ox- 2 closest to the reservoir, arbovirus conversion ford. ratesdid not necessarilyfollow this pattern. Thus Barker-Hudson, P., R. Piper and B' H. Kay. 1986. in terms of the potential hazard posed by the Water impoundments and their implication for the Ross River Dam (Stagel), the risk of arbovirus development of northern Australia. Arbovirus Re- infection would seem to be no greater than in searchin Australia 4:179-184- Har- suburbs such as Kelso, Bunnag,T., S. Sornmani, S. Pinithpongseand C' Townsville-Thuringowa of water-borne para- In fact, high alphavirus inasuta. 1979. Surveillance Stanley,and Pallarenda. studieson the impact ofecolog- Pallarenda sitic infections and activity in 1984-85 and 1985-86 at ical changes on vector mosquitoes of malaria after may be due to its proximity to both coastal salt- dam construction. SoutheastAsian J. Trop. Med. marsh and an environmental park whereAe. vi- Public Health l0:656-660. gilax and Cx. annulirostris ate plentiful. Clarke, D. H. and J. Casals. 1958' Techniques for The results of this study have several impli- hsrneeglutination and hemagglutination-inhibition cations for the spatial design of a minimal sur- with -borne viruses. Am. J. Trop. Med. veillance program using EVS traps for mosquito Hyq--7:,561-573. Strategiesfor surveillance,pre- vectors of arboviruses: Eldridge, B. F. 1987. Drcurrn 1993 Moseurrors exo AxsounusEs rN Nonrrcnll AUSTRALIA 399

vention, and control ofarbovirus diseasesin western ments, August-November 1950, and the first epi- North America. Am. J. Trop. Med. Hyg. 37(Suppl.): zootic of myxomatosis in the riverine plain of south- 77S-86S. easternAustralia. J. Hyg. 52:47-59. Jones.R. E., P. Barker-Hudsonand B. H. Kay' 1991' O'Donnell, M. S., G. Berry, T. Carvan and J. H. Bryan. Comparison of dry ice baited light traps with 1992. Dispersal of adult females of Culex annuli- bait collections for surveillance of mosquitoes in rostris it Griffith, New South Wales, Australia. J. northern Queensland,Australia. J. Am. Mosq. Con- Am. Mosq. Control Assoc.8:159-165. trol Assoc. 7:387-394. Reeves,W. C., M. M. Milby and W. K. Reisen. 1990. Kay, B. H. 1979. Seasonalabundance of Culex an- Development of a statewide arbovirus surveillance nulirostris and other mosquitoes at Kowanyama, program and models ofvector populations and virus north Queensland, and Charleville, south west transmission, pp. 431-459. In: W. C. Reeves(ed.). Queensland.Aust. J. Exp. Biol. Med. Sci. 5'l:497- Epidemiology and control of mosquito borne ar- 508. bovirusesin California. 1943-1987.California Mos- Kay, B. H., P. Barker-Hudson, R. G. Piper and N. D. quito and Vector Control Assoc., Sacramento,CA. Stallman. 1990. Arbovirus disease and surveil- Rohe, D. and R. P. Fall. 1979. A miniature battery lance methodology related to water resource devel- powered CO, baited light trap for mosquito borne opment in Australia. Water Resour. Devel. 6:95- encephalitis surveillance. Bull. Soc.Vector Enol. 4:24- 103. 27. Kay,8.H., R. A. Hdl, I. D. Fanning,P. Mottram, P. Russell,R. C. 1986. Dispersalofthe arbovirus vector L. Young and C. C. Pollitt. 1986. Experimental Culex annulirostns Skuse(Diptera: Culicidae) in the infection of vertebrates with Murray Valley enceph- Murray Valley of Victoria, Australia. Gen. Appl. En- alitis and Ross River viruses. Arbovirus Research tomol. l8:5-9. in Australia4:71-75. Russell,R. C. and P. I. Whelan. 1986. Seasonalprev- Marks, E. N. 1969. Mosquitoes (Culicidae) on alenceofadult mosquitoes at Casuarina and Lean- Queensland'scoral cays. Queensl.Nat. l9:9zl--98. yer, Darwin. Aust. J. Ecol. I l:99-105. Myers, K. 1954. Studies in the epidemiology of SAS Instirute. 1988. SAS/STAT user'sguide, release infectious myxomatosis of rabbits II. Field experi- 6.03 ed. SAS Institute, Cary, NC.