490 JounNe,r,oF THE At'lnRIcaNMoseurro CoNtnor, AssoctlrroN Vor,. 6, No.

ESTIMATION OF ABSOLUTE NUMBERS OF NYMPHS (: )BY DIPPER SAMPLING IN CALIFORNIA RICE FIELDS WITH SEASONAL,SPATIAL DISTRIBUTIONS AND VEGETATION ASSOCIATION

T. MIURA, R. M. TAKAHASHI rNn R. J. STEWART

Uniuersity of California Mosquito Control ResearchLaboratory, 9240 S. RiuerbendAuenue, Parlier, CA 93648

ABSTRACT. Estimates of relative and absolutedensity in rice field populations of damselfly nymphs (predominantly Enallagma ciuib with few d.enticollis)were compared using the regression method. An equation, X : Y + 0.0016,allows estimation of absolutedensity (X) from a relative density index (Y, dipper count). In the rice growing area of Fresno, California, nymphal population peaks appearedduring June and August approaching3-5 million per 0.405ha (1 acre). Spatial distribution was theoretically representedfairly well by a negative-binomialdistribution. The degreeof clumping is one of overdispersedtypes; it is especiallyclassified as a model of randomly distributed colonieswith mean colony size fixed. Presenceor absenceof submergedvegetation narkedly affected damselfly nymphal density but the biomassof submergedor emergedvegetation was not a significant factor.

INTRODUCTION to define the relationship between damselfly nymphs and vegetational characteristics in the Damselfly nymphs are the most abundant rice field. aquatic inhabiting San Joaquin Valley rice fields in California (Miura et al. 1981). MATERIALS AND METHODS Odonata nymphs are generalizedpredators and usually prey on small crustacea, annelids and Study area: A series of 8.0 to 11.0 ha fields immature , including mosquito and chi- of commercialrice at the western edgeof Fresno ronomid larvae (Sailer and Lienk 1954,Merrill County, California, were used for this study. and Johnson1984, Miura and Takahashi1988). Rice has been grown in this area for many years Although there are many other predators in the on a continual basis becausethe land is poorly Central California rice fields, Miura et al. (1981) drained (Lost Hills clay loam) and easily acces- predicted that are a major contrib- sible to canal water. Normally, rice culture starts utor to mosquito predation in rice fields because during mid-April to mid-June. Lawae of Cx. damselflies dominate overwhelmingly in num- tarsalis appear in fields 3-5 weeks after initial bers and also sharethe samehabitat as mosquito flooding. Predominant weeds within the rice Iarvae. fields are barnyard gass (Echinochlaacrusgalli Determinations of the total numbers of mos- (Linn) Beauv.),sedge family (Cyperusspp., Scir- quito larvae and their predators in a given area pus spp.), water-nymph (Nojos sp.) and stone are essential for the development of an inte- woft (Chara sp.). The water-nymphsare espe- grated mosquito management program. Abso- cially abundant and form mats covering entire Iute densities of Culex tarsalis Coquillett larvae basins of rice fields which provide an ideal hab- in Fresno rice fields have been estimated using itat for damselfly nymphs. dipper counts (Stewart and Schaefer1983), and Dipper efficiency for collecting damselfly population densities and growth rates of the nyrnphs: Circular sampling enclosures were important mosquitofish predator, Gambusiaaf- made from 30-cm high aluminum sheets (Stew- finis, in the same area were determined using art and Schaefer 1983) to.give a sampling area capture-recapturemethods (Miura et al. 1982, of 1 m2. At the start of each sampling day, 5 Stewart and Miura 1985). enclosureswere placed at random within a rice Cochran (1953), Southwood (1978) and paddy. A white enameledmosquito larval dipper Downing (1986)suggested that regressionanaly- (1-pint capacity)with a long handle (1.5m) was sis might be used to predict absolutepopulation used as the sampling device for relative abun- densities from relative indices. where a variable danceestimates, such as relative index is correlatedwith another Tests were performed 4 times between June variable such as absolute density of the popula- and August when rice plant heights ranged from tion from which the relative index was deter- 0.4 to 1 m. Dip samples inside the enclosures mined. were taken between 1000 and 1800 h with the The objectives of this study were: 1) to esti- dipper. Each enclosure was divided into quad- mate the absolute population of damselfly rants by marks on the outside. The quadrants nymphs from dipper counts, 2) to determine aided in sampling over the entire enclosedarea their seasonaland spatial distributions and 3) and reducedbias causedbv the individual sam- SEPTEMBER1990 Devssr,rr,v NyMpH Slrrlpr,ruc rt.t Rrcp

plers. Ten dip samples were taken randomly vegetation(36 kg rice seed/ha;ca.70% of soil from eachenclosure. The contents were counted surface was coveredby mixed grou'th of Nojas and returned to the same place where taken sp. and Chara sp.). after each dip. To reduce disturbances to the The 3 vegetational conditions were simulated samples, dips from the same enclosure were in the rice plots which were arranged in a 3 x 3 taken at 2- to 3-min intervals. Latin squaredesign (Cochran and Cox 1957). After the dip sampleswere completed, abso- Damselfly nymph densities were determined by lute population densities of damselfly within taking 30 dip samplesfrom each plot. The data enclosures were determined by the removal obtained were examined by the Latin square method (Southwood1978); 5 sweepcollections method. This study was conducted during the made by dragnets (20 x 14 cm with 1.5-mm 1985rice growingseason. opening mesh net) were pooled into one sample Relationship betweendamselfly "super densities and unit (Wada's unit," Wada 1962a)and biomassof submergedvegetation at the western preservedin a jar with 75% ethanol. Four or 5 Fresno County rice fields was studied by taking sample units, depending on the density of the random area sampleswithin the rice field with damselfly, were collected from each enclosure a 0.045-m'(40-cm high) area sampler.The sam- and these samples were examined under the pler was thrust as quickly as possible into the stereoscopein the laboratory. Absolute popula- water and down into the substrate for about 5 tion density of each enclosure was then deter- cm, creating a watertight seal from outside the mined by the regressionmethod (Wada 1962a, sampler. Damselfly nymphs were collected by Miura 1980). scoopingout all water contained within the sam- The relationship between the relative esti- pler, screeningthrough a 0.5mm meshsieve and mate (dipper count) and the absolute estimate releasing the contents in a shallow, white pan (removal method) for the enclosureswas exam- containing clear filtered water from the field. ined by the correlation method (Ostle 1954)and The nymphs were counted and recorded. All a linear regressionequation was calculated. vegetationwithin the sampler from ground level Seasonaland spatial distributions: Concurrent up was cut and washed free of clinging insects with sampling Cx. tarsalis larvae with a diFper, and debris and then placed in plastic bags. we also conducted a field sampling program to The vegetation samples were taken to the study seasonalpopulation changes and spatial laboratory within a few hours of collection. Once distribution of damselfly nymphs in rice fields. there, they were removedfrom the bags,further Detailed sampling methods used were reported washedof debris and then blotted dry with paper by Stewart et al. (f9$). In brief, the sampling towels. They were then placed in a drying oven sites (100 sites per 8 ha) were selected evenly at 105'C for 24 h. The vegetation was then over the fields by using aerial photographs and cooledin desiccationjarsfor t h beforeweighing. the sites were marked with wooden stakes (2 x There were 132 samples collected periodically 2 x 100 cm) numbered over each field. One dip throughout the 1984 rice growing season.The sample was taken at each stake and examined data were analyzedusing correlation methods. by eye for nymphs and other nontarget orga- nisms. All captured taxa were recordedsepa- RESULTS AND DISCUSSION rately for each sample site. Sampleswere taken weekly throughout the rice growing seasondur- Dipper efficiencyand the deriuatinn of absolute ing 1980,1981 and 1982. estimate:Table I summarizesresults of the dip- Absolute abundance was estimated by using per efficiency study. Average number of dipper the linear regressionequation calculated in the counts from 20 enclosureswas 1.35 nymphs per dipper effrciency study. The dipper collection dip ranging from 0.2 to 2.13. Averagepredicted data were also examined for goodnessof fit to absolute population density per enclosure was the theoretical distributions of Poisson (An- 813.7nymphs (range;230-1,150). scombe 1948) and negative-binomial distribu- Removal method was usedto predict absolute (Bliss "Patchiness tion and Owen 1958).Iwao's densities within each enclosure (Table 1) be- regression indices" (Iwao 1968, Southwood cause it is relatively simple and able to obtain 1978)were also calculated. reasonablyprecise estimations. We have found Vegetational associntion:Three vegetational this method useful for estimating population conditions were investigated on newly con- densities of mosquito larvae in small confine- structed rice plots at the University of Califor- ments such as small artificial ponds, catch bas- nia, Kearney Agricultural Center, Parlier. They ins of storm drain systems,tree holesand cem- were: 1) normal rice stand with bare soil bottom etery flower vases(Miura 1980).Wada (1962b) (36 kg rice seed/haplanted), 2) denserice stand used this method to estimate population size of (45 kg rice seed/haplanted) with bare soil bot- immature stagesof mosquitoesbreeding in fer- tom and 3) normal size stand with submerged tilizer pits and tide pools.Mogi et al. (1984)also 492 JouRNu,oF THEArupnrclN Moseurro ConrRol AssocrlrroN VoL. 6. No. g

Table 1. Relationship betweendipper counts, absolutedensity and removal numbers by net in 1-m2 enclosures. Estimated" Actually total no. removed Proportion Enclosure Mean no. nymphs/ by drag net of population no. nymphs/dip m2 from 1 m2 removed t 2.r3 1,150 675 0.59 1.73 820 473 0.58 L.t7 1,090 672 0.62 4 1.73 1,100 682 0.62 1tn 940 594 0.63 o 1.13 800 473 0.59 7 1.53 1,130 504 0.45 8 1.50 900 365 0.41 q t.20 800 473 0.59 10 1.07 530 406 0.77 11 1.30 720 410 0.57 t2 r.alt 900 547 0.51 13 0.75 500 183 0.37 T4 1.90 1,000 330 0.33 15 1.60 800 474 0.59 16 1.55 790 318 0.40 t7 0.65 456 372 0.82 18 0.20 230 r54 0.67 19 1.15 740 JD' 0.48 20 2.00 878 463 0.53 Total 27.01 16,274 8,835 11.10 Mean I..JJ 813.7 44r.75 0.55 " By removal method.

2.4 each sampling, whereby subsequent catches would be reduced, thus a large portion of the 2.2 + y=O.O48 O.OO16x parent population should be removedby collect- -o 2.o ing (Zippin 1956). In this study, we have re- ; 1.8 movedca.55.5% (33-82%\of nymphs from en- 6 1.6 closuresthus, we have obtained reasonably re- 1.4 liable estimations (Table 1). 5 Oncedipper counts (Yi) and absolutedensities .a 1.2 (Xr) from enclosureswere determined, a scatter H 1.o diagram (Fig. 1) was constructed to inspect the F o.8 approximate form of the relationship between q 0.6 these 2 measurements(Yi and Xr). The scatter diagram showeda strong tendency of a straight 2 o.4 line relationship. The linearity was tested by o.2 correlation method; the correlation coefficient o- (r) was 0.805,highly significant at the l% ctiti- 1200 o 200 4o(t 60(} 800 1000 cal point. A linear regressionequation was then l.lo. Damelfh, nynph6./ 1-m2 enclosure constructedas Y, : 0.048+ 0.0016X1.If meas- Fig. 1. Scatterdiagram showing the relationship urements of Yi and Xi are perfectly proportional, betweenabsolute and relative densities with a regres- the line must intercept at the origin (Y : 0, X sionline andcorrelation coefficient. : 0). Our calculateda was near zero and the hypothesis a = 0 was unable to be rejected (t : used this method to estimate absolute popula- 0.538) statistically. Equation Y' = a * 0Xi was tions of immature stagesof Aedesbaisasi Knight reformulated in terms of our model, Y' : and Hull, Culex tuberis Bohart and Uranotaenin 0.0016Xi, then absolute population densities ohamai Ohama inhabiting crab holes. The con- (Xi) were derived from dividing dipper counts ditions that must be satisfied to use this method (Yr) bY 0.0016. are summarizedby Southwood(1978) and Serv- Seasonal and spatinl distribution: Our field ice (1976).An important factor is having ade- observationsof adults indicated that Enallagma quate sampling sizebecause a sufficient number ciuile (Hagail was the most abundant (ca. 80%) of organismsshould be removedfrom the site at adult damselfly flying over the rice fields, fol- Snprnrraspn1990 Deusnr-rr-v NyMpH Slupr,rt.tc rN RrcE lowed by Ischnura denticollis (Burmeister). A Table 2 showstypical field collection data by few Ischnura perparua Selys and Ischnura cer- dipper, e.g., the July 2 collection contained 45 uulo Selys adults were also observed.However. dips with no nymphs; 31 dips of 1 nymph, etc. only E. ciuilc and I. denticollis nymphs were Overall, nymphs were collectedfrom throughout collectedfrom the study area. the rice fields indicating near random to Figure 2 shows the seasonal abundance of clumped distribution patterns. Comparison of damselfly nymphs collected from the study goodnessof fit tests of nymphal distribution to fields over the 3-year sampling period. In the Poissonand negative-binomialdistributions are Fresno area, nymphal population peaks ap- summarized in Table 3. Near the start of the peared during July and August each year ap- season(first 2 collections) the population den- proaching 3-5 million per 0.405 ha (1 acre). sities and varianceswere near equal, indicating Their population reached to 2-4 million per a near random distribution; however, as the 0.405ha even in the fields planted with mosqui- seasonprogressed, the variancesto mean ratios tofish (predators of many invertebrates). increased, indicating increases of clumpings. Later in the season the ratios were reduced ; sxlo6 again.This clumping pattern was also examined by Iwao's patchiness reglession method (Iwao EElllS: 1968,Southwood 1978) and the results,i.e., the 99 zxro" relation of mean crowding (i) to mean density (i) are illustrated in Fig. 3. The index of basic 8fr,*,, contagion(a) was greaterthan zero (0.716)and I axro" the density contagiousnesscoefficient (0) was near unity (0.945).Thus the distribution pattern 18253 10172431 7 14212A of immature stagesof damselfly in Fresno West- June July August side commercial rice fields is one of overdis- - Fig. 2. Seasonalabundance ofdamselfly nymphs in persed distributions. Iwao classified this type Fresno rice fields during 1980,1gg1 ana f"g8dseasons. patchiness distribution, a ) 0 and 0 = 1, as a

Table 2. Spatial distribution patterns of damselfly nymph in a Fresno County rice field in 1980. Observedfrequency (expectedfrequency from best fit distribution) Frequencyclass July 2 JuIy 9 July 16 August 27 (no. nymphs/ diP) NB' NB Poisson NB 0 45 (48.2r) 41 (44.93) 30 (27.30) 57 (58.60) I 31(27.60) 34 (28.24) 33 (34.61) 23 (23.9e) 2 16(13.38) 13 (14.46) 18 (21.94) 13 (9.76) Jf 8 (10,81) 12 (12.37) 16(13.15) 6 (6.55) k L.4401 L.5892 1.0132 x 0.9500 1.0400 1.2680 ^2 0.6869 r.44L9 1.5337 r.4482 0.9112 " NB = Negative binomial distribution.

Table 3. Comparisonof goodnessof fit test of damselfly nymphal distributions in one rice field during 11 weeksof sampling to poisson and negative_binomial(NB) distribution.

Week i S2 Sr/i 52 df s2 df P I 0.5300 0.7365 1.389 8.9440 0.0300 I.5278 2 0.4658 2 0.4900 0.5757 1.175 r.5582 2 0.4588 0.0730 1 0.7870 J 0.9500 r.4419 1.528 2.2655 J 0.5192 1.0988 2 0.5773 t 1.0400 1.5337 r.475 4.2361 J 0.237L 1.3788 2 0.5019 1.2680 7.4482 L.t42 1.6672 d 0.6443 2.8642 2 0.2388 6 1.9700 3.4637 1.758 7.7820 4 0.0999 3.9663 o 0.2651 7 1.6364 2.3562 t.440 12.8896 5 0.0244 1.6665 4 0.7968 8 1.7000 2.6364 1.551 8.8571 5 0.1149 3.4972 4 0.4703 I 1.3400 2.1863 r.632 16.3934 0.0025 6.0154 o 0.1109 10 0.8687 1.1356 1.307 8.7332 3 0.0331 8.9841 2 0.0112 11 0.6869 0.9112 1.326 7.2197 0.0652 1.3091 0.5197 494 JounNer,oF THEAunRrceN Mosqurro Cournol Assocrnrrox VoL.6,No.3

larvae and aquatic beetle larvae, as well as for 2-5 the damselfly nymphs. The results of analysis to determine the rela- Irx 6 ,.o tionship between damselfly nymph density and = biomass of submergent vegetation (Nojos sp. o and Chara sp.) showedno relationship between < 1-5 o biomassof submergedvegetation and density of G o nymphs (r = 0.11,P > 0.1). Basedupon these z 1.o results, we have concludedthat the absenceand uJ presenceof submergentvegetation carpeting the o.5 bare bottom may be an important factor for habitat selection by damselfly nymphs but the biomass of the submergedvegetation has little effect on damselfly nymph density. o.5 t.o t.5 2.o MEANDENSITY(i)/DIP Fig. 3. Iwao's patchiness regression showing the ACKNOWLEDGMENTS relationship between mean crowding (i) and mean This study was funded, in part, by USDA (i) density for the distribution of damselfly nymphs CooperativeAgreement No. 82 CRSR2-1010, in Fresno Westsiderice fields. andby a California State Appropriation for mos- quito control research.We thank M. J. Pitcairn, Table 4. Habitat selectionby damselfly nymphs in University of California, Davis, for helpful crit- rice fields. Number of nymphs collected from 9 different rice plots (6.1x 6.1 m). icism of the manuscript. Replicates REFERENCES CITED Vegetation Anscombe,F. J. 1948.The transformationofpoisson, Normal rice stand with 35 52 37 41.33B binomial and negative-binomial data. Biometrika submergedvegetation 35:246-254. Normal rice stand with t225 6.33A Bliss,C. I. and A. R. G. Owen.1958. Negative binomial bare bottom distributions with a comrnon k. Biometrika 45: Denserice stand with 910 8.00A 37-58. bare bottom Cochran, W. G. 1953. Sampling techniques. Wiley, 'Means followedby the sameletter are not signif- New York. icantly different(a : 0.05,Duncan's multiple range Cochran,W. G. and G. M. Cox. 1957.Experimental test). designs.2nd Ed. Wiley, New York. Downing,J. A. 1986.A regressiontechnique for the of epiphytic invertebrate populations. "model (mean estimation of randomly distributed colonies FreshWater Biol. 16:161-173. colonysize fixed)." Iwao, S. 1968.A new regressionmethod for anaiyzing Vegetational association:Table 4 shows the pattern of populations. Res. Pop. Ecol. 10: results obtained from the vegetationassociation 1-20. study. The data that were analyzedby a 3 x 3 Merrill, R. J. and D. M. Johnson. 1984.Dietary niche Latin squaremethod showedthat only the treat- overlap and mutual predation among coexistinglar- ment effect (i.e., vegetation) was highly signifi- val Anisoptera.Odonatologia 13:387-406. population cant (F : 20.69.df = 2.2: P < 0.05). Position Miura, T. 1980. Estimation of mosquito in confinednatural breedingsites. Proc. Calif. and subsoil of the plots might have influenced size Mosq.Vector Control Assoc.48:108-112. vegetationalgrowth, but no significant effect on Miura. T.. R. J. Stewart and R. M. Takahashi.1982. the density of nymphal population was detected. Application of a capture-recapturemethod for esti- Results from Duncan's multiple range test mating densitiesof mosquitofish in rice fields. Proc. showedthat more damselfly nymphs were cap- Calif. Mosq. Vector Control Assoc.50:54-57. tured in the plots with submergent vegetation Miura, T. and R. M. Takahashi. 1988. A laboratory than those without. In addition, it was found study of predation by damselfly nymphs, Enallagma that the density of rice stands also had little ciuile tpon mosquito larvae, Culex tarsalis. J. Am. 4:129-131. effect on damselfly density. The rice fields with Mosq.Control Assoc. T.. M. Takahashiand W. H. Wilder. 1981. ecological Miura. R. submergent vegetation created an The selectedaquatic fauna of a rice field ecosystem community entirely different from that of the with notes on their abundances,seasonal distribu- bare bottom rice field. The submergedvegeta- tions and trophic relationships. Proc. Calif. Mosq. tion can be consideredas a carpet on the bottom Vector Control Assoc. 49:68-72. and it provided ideal resting and hunting places Mogi, M., I. Miyagi and T. Okazawa.1984. Population for many otganisms, such as chironomid midge estimation by removal methodsand observationson SEPTEMBER1990 Duvrspr,rr,y NyMpH Selrpr,rnc rn Rrcn 495

population regulating factors of crab-hole mosquito of Culex forsolis larvae and pupae in rice fields. immatures (Diptera: Culicidae) in the Ryrrkyr-rs,Ja- Mosq. News 43:729-735. pan. J. Med. Entomol. 21:720-726. Stewart, R. J., C. H. Schaefer and T. Miura. 1983. Ostle,B. 1954.Statistics in researchbasic concepts Sanpling Culex tarsalis (Diptera: Culicidae) imma- and techniquesfor researchworkers. Iowa St. Coll. tures on rice fields treated with combinatitins of Press.Ames, IA. mosquitofish and Bacillus thuringiensisH-14 toxin. Sailer, R. I. and S. E. Lienk. 1954.Insect predators of J. Econ.Entomol. 76:91-95. mosquito larvae and pupae in Alaska. Mosq. News Wada, T. 1962a.Studies of the population estimation 14:14-76. for insects of medical importance. 1. A method of Service,M. W. 1976. Mosquito ecology.Appl. Sci. estimating the population sizeof mosquito larvae in Publ. Ltd., London. a fertilizer pit. Endemic Dis. Bull. Nagasaki Univ. Southwood, T. R. E. 19?8. Ecological methods with 4:22-3O. particular referenceto the study of insect popula- Wada, T. 1962b.Studies of the population estimation tions 2nd ed. Methuen and Co. Ltd.. London. for insects of medical importance. 2. A method of Stewart,R. J. and T. Miura. 1985.Density estimation estimating the population size of larvae of Aedes and population growth of mosquitofish (Gambusia togoi inthe tide-water rock pool. Endemic Dis. Bull. affinis) in rice fields. J. Am. Mosq. Control Assoc. NagasakiUniv. 4:141-156. 1:8-13. Zippin, C. 1956.An evaluation of the removal method Stewart, R. J. and C. H. Schaefer.1g83. The relation- of estimating animal populations. Biometrics ship betweendipper counts and the absolutedensitv 12:163-178.