The Condor93279-285 0 The cooper olnithologicalsociety 199 1

THE EFFECT OF FLORAL ABUNDANCE ON FEEDER CENSUSES OF HUMMINGBIRD POPULATIONS

DAVID W. INOUYE~ Departmentsof Zoology and Botany, Universityof Maryland, CollegePark, MD 20742

WILLIAM A. CALDER~ Department of Ecology and Evolutionary Biology, Universityof Arizona, Tucson,AZ 85721

NICKOLAS M. WASER~ Department of Biology, Universityof Cahfornia, Riverside,CA 92521

Abstract. Numbersof Broad-tailedHummingbirds (Sehzsphorous platycercus) captured eachsummer from 1979-1989 at the Rocky Mountain BiologicalLaboratory were quite variable, rangingfrom 115 (1981) to 348 (1989) with new birds usuallyoutnumbering returning(previously banded) birds. Capturenumbers were negatively correlated with the abundanceof four speciesof flowersthey visited,Erythronium grandiflorum, Delphinium nelsonii, Zpomopsisaggregata and Delphinium barbeyc flower numberswere also highly variable during the study period. Since most of the captureswere at feeders,these data suggestthat in yearswith high floral abundancefeeders are lessattractive, while in years with low floralabundance hummingbirds with nestsor territoriesat greaterdistances increase their use of the feeders.This interpretationis supportedby seasonalvariation in use of feeders,which is highestduring the beginningand end of the seasonwhen floral abundance is lowest.Estimates of hummingbirddensity based on activityat feedersmay thusbe affected by the availability of floral food resources. Key words: Banding; census;floral abundance;Broad-tailed Hummingbird; population size; resources,

INTRODUCTION ticides. While these perceptions lack the rigor of Many bird populations vary in size from year to scientific methodology, casual observations by year (e.g., Mulvibill and Lcberman 1987), or even numerous individuals nevertheless suggestthat within years. In some casesthis variation can be these are plausible explanations. Accurate esti- ascribed to particular environmental or demo- mates of hummingbird population size would graphic variables. For example, droughts may permit more careful consideration of the signif- affect reproduction and survivorship through an icance of this variation. effect on food resources(Gibbs and Grant 1987, A variety of techniques can be used to assess Grant and Grant 1989), cold temperatures may population size. Banding offersan advantageover result in decreasedwinter swivorship, or small techniques based on visual sightings, or counts population sizes may result in local extinctions. of song or flight noise that may be conducted However, variation in population size usually more easily, in that individuals are identified and remains unexplained. counted only once. We have carried out long- Hummingbird numbers seem to vary consid- term studies using two methods, banding and erably, as perceived by numbers at feeders. Ap- flight noise, on a resident population of Broad- parent lows are popularly attributed to environ- tailed Hummingbirds (Selusphorus platycercus mental catastropheson the wintering grounds for Swainson). During our 18-year study we have many speciesin Central America, such as vol- observed substantial year-to-year variability in canic eruptions, forest fires, and the use of pes- abundance of Broad-tailed Hummingbirds in Colorado. There is also significant annual vari- ation in the availability of floral resourcesused by these birds. Here we examine the relationship 1Received 14 May 1990. Final acceptance26 De- cember1990. between the abundance of the hummingbirds, as * Rocky Mountain BiologicalLaboratory;P.O. Box indicated .by banding studies, and floral re- 5 19, CrestedButte, CO 8 1224. sources.

W91’ 280 D. W. INOUYE, W. A. CALDER AND N. M. WASER

TABLE 1. Mean dates of first and last flowering by the four flower species, 1982-1988. Datesof lastflowering werecalculated by addingthe mean lengthof the floweringperiod to the mean dateof first flowering.Data are only shownif floweringoccurred in two or more plots.Data for this table are from a subsetof 23 of the 29 plotsmonitored for anotherstudy (the plotsin whichthese four speciesoccurred). Zpomopsis did not flowerin this subsetof the plotsin 1982 or 1983.

1982 1983 1984 1985 1986 1987 1988 1989

Erythronium grandiflorum 6106 6/14 6/12 6/01 6/01 5/20 5/25 5123 6/17 6/21 6/30 6/09 6/10 6/01 6/04 5/31 Delphinium nelsonii 6/16 6/23 6/25 6/09 603 6/06 6/07 6/03 7108 7/12 7/14 6/26 7/01 6/22 6/20 6/20 Zpomopsisaggregata 7/13 7/7 7105 6/29 6/29 7/04 9/05 8/30 8/28 8/07 8/14 8/21 Delphinium barbeyi 7/27 7/25 7/25 7/11 7/17 7106 7/06 7/06 8/18 8/16 8/21 8/04 8/13 7/27 7/19 7/20 Number of daysof overlan 2 0 36 26 28 21 14 15

MATERIALS AND METHODS most of the growing season (early May to mid- This study was conducted at the Rocky Moun- September), all flowers in the plots were counted. tain Biological Laboratory (RMBL),Crested Plots were scatteredamong habitats including a Butte, CO, at an elevation of approximately 2,900 dry rocky meadow (7 plots), aspenforest (2 plots), m. Broad-tailed Hummingbirds, the only resi- wet meadow (9 plots), dry (but not rocky) mead- dent breeding speciesof hummingbird at RMBL, ow (3 plots), aspen-meadow interface (3 plots) have been studied extensively at RMBL (e.g., and willow-meadow interface (5 plots). For each Calder 1975, 1981, 1984, 1985; Calder and speciesin every plot, the peak number of flowers Booser 1973; Calder et al. 1983; Waser 1978, produced during a given year was determined. 1988; Waser and Inouye 1977; Waser and Price For analyses in this study, the peak numbers for 1983). Hummingbirds were captured and band- each specieswere added acrossplots for each year ed each year at two to three different sites sep- to produce a single value of peak number of flow- arated by about 500 m. Most birds were captured ers. Not all speciesoccurred in all plots; for ex- at feeders at cabins using cages, mist nets, or ample, Ipomopsis was only found in l-7 of the butterfly nets, with some additional mist net cap- plots in any given year. However, representation tures in meadows around RMBL. Although the of speciesin the plots approximated their abun- banding program began in 197 1, it was intensi- dance at the study site. fied beginning in 1979 to the point where we Data from four species of flowers that are think that almost all resident birds were banded; among those most commonly visited by hum- by the end of 1979 and subsequentsummers the mingbirds at the study site were used for this continued netting and trapping yielded primarily analysis. They were Erythronium grandiforum recapturesofbirds already recordedfor that year. (glacier lily; Liliaceae), Delphinium nelsonii (Nel- In addition to data from banding, a potential son’s larkspur; Ranunculaceae), Delphinium bar- correlate of population size was also measured: beyi (tall larkspur; Ranunculaceae), and Ipomop- the numbers of male Broad-tailed Humming- sis aggregata (scarletgilia; Polemoniaceae). Only birds heard during 10 min of listening from a the latter fits the classical characterization of a fixed location in a meadow at RMBL at a stan- hummingbird-pollinated plant (long, red, tubu- dard time of day (noon). Such censuseswere con- lar flowers). Data were not included for other ducted every two to four days throughout each flowersused by hummingbirds, suchas Castilleja summer. Males of this speciesmake a mechan- spp. (Scrophulariaceae), or Aquilegia elegantula ical wing whistle that is audible up to 100 m (Ranunculaceae)because they were relatively rare. away (Miller and Inouye 1983). The four flower speciesused for this study span Data on flower abundance were collected from much of the flowering season (Table 1; Fig. 1). 29 2 x 2-m plots, located within 500-1,000 m of Erythronium, with large, yellow, pendant lily banding sites.Approximately every other day for flowers, is one of the first speciesto flower at the FEEDER CENSUSESOF HUMMINGBIRD POPULATIONS 281

a low of 115 in 1981 to a high of 348 in 1989. From 1980-1987 the number of recaptured (probably resident) birds was more constant than the number of newly captured (probably migrant) birds. The number of birds captured was highest in 1988 and 1989, when the numbers of flowers were lower than in any other year. These were the only two consecutive years of low flower numbers during the study, and this sequencewas apparently responsible for the large numbers of new birds (2 14) banded in 1988, and birds re- FIGURE 1. Flowering curves for the four flowerspe- captured in 1989; many of the recaptures were ciesfor 1986. first banded in 1988. There was also a large num- ber of new hummingbirds in 1985. Data from the census of wing-whistle noises study site, and is also visited by bumblebees and are also shown in Table 2; there was no signifi- other bees. The two blue-flowered Delphinium cant correlation between these census data and species are also visited frequently by bumble- the numbers of banded males (v < 0.100, it = bees, while hummingbirds are the primary vis- 11, P > 0.05), or between the number of flowers itors to Zpomopsis(Waser 1982) although bum- and the number of males heard in the censusof blebees commonly rob the nectar of Zpomopsis flight noises (r = 0.142, P > 0.05). flowers in some years (Inouye 1980). Despite the There was much variation among years in the variation among years in date of first flowering number of flowers produced by each of the four (Table l), the relative sequence of flowering is plant species(Table 3). The reasons for this are generally maintained, and the arrival and nesting not completely understood, although two factors of the birds appear to be synchronized with flow- appear to be important. First, there is a signifi- ering. cant correlation between the number of flowers of both Delphinium speciesproduced each sum- RESULTS mer and the amount of snowfall the previous Table 2 summarizes the numbers of Broad-tailed winter (Inouye 1989); there is no suchcorrelation Hummingbirds captured at RMBL from 1979- for the other two species.Second, every four to 1989. The sex ratio among captures was biased five years (1976, 1981, 1985, 1989) since 1973 in favor of males in 1980 and females in 1979 (when flower data were first collected from these and 1988 (binomial probability test, P < 0.05). plots) there have been late (mid-June) hard frosts The total number of birds captured ranged from (sometimes accompanied by snow) that have

TABLE 2. Numbers of Broad-tailed Hummingbirds captured at the Rocky Mountain Biological Laboratory. Numbers are reported separatelyfor returning birds (previously banded) and new birds (unbanded);data from 1979 were not available in the same resolution as in subsequentyears. Data for a censusof flight whistles of male Broad-tailed Hummingbirds are also shown (see Methods); numbers reported are mean number of flights heard, with sample sizes (number of replicate censusesduring the summer) in parentheses.

Ret. males New males Total males Ret. females New females Total females Total for both Broad-tailed census

1979 48 69 117 8.8 (24) 1980 11 94 105 45 ;: 76 181 5.9 (17) 1981 16 43 2: 30 56 115 7.2 (13) 1982 15 51 27 64 130 6.7 (17) 1983 16 54 70 3367 37 73 143 5.6 (9) 1984 14 57 71 33 37 70 141 9.7 (21) 1985 24 93 117 36 65 101 218 8.2 (17) 1986 24 34 58 25 34 117 4.1(20) 1987 15 64 79 28 65 9’; 172 4.1 (18) 1988 18 98 116 64 116 180 296 8.1 (17) 1989 32 128 160 89 99 188 348 6.2 (19) 282 D. W. INOUYE, W. A. CALDER AND N. M. WASER

TABLE 3. Peak number of flowers recorded for each of the four species. Numbers in parentheses represent the number of plots in which the species flowered.

Delphinium Ipomopsis barbeyi aggwglua TOtal

1979 82 (4) 992 (13) 1,891 (11) 71 (3) 3,036 1980 91 (5) 980 (11) 2,285 (12) 36 (2) 3,392 1981 50 (5) 468 (9) 374 (10) 135 (4) 1,027 1982 111 (5) 795 (13) 1,102 (11) 113 (4) 2,121 1983 124 (5) 962 (11) 1,181 (11) 77 (1) 2,344 1984 113 (4) 686 (12) 2,375 (12) 84 (4) 3,258 1985 116 (5) 523 (11) 432 (7) 234 (5) 1,305 1986 178 (5) 478 (13) 1,557 (12) 220 (7) 2,433 1987 157 (5) 356 (9) 541 (9) 181 (5) 1,235 1988 57 (5) 187 (9) 136 (5j 73 (3) 446 1989 133 (5) 203 (9) 361 (7) 87 (6) 784

killed flower buds ofmany species(Inouye 1988). fus that year (Pleasants and Waser 1985). To- These two environmental events appear to be gether, these observations suggestthat there was responsible for much of the annual variation in something unusual about the hummingbird pop- flower number in Delphinium species and for ulations at our study site in 198 1. It appearsthat some of the other speciesin these meadows. although the number of birds recaptured in 198 1 There is a significant negative correlation be- was about that predicted by the regressionequa- tween the number of birds captured and the peak tion (Fig. 3), the number of new birds was sub- number of flowers (Fig. 2; r = -0.6 19, P < 0.05) stantially lower than expected (Fig. 4). for the period 1979-1989. If the capture data are broken down into numbers of recaptured birds DISCUSSION and new birds (Table 2; data only available since It is perhaps surprising that we found any sig- 1980) and then correlated with flower numbers, nificant relationship between flowers and bird the negative correlation for recaptured birds is abundance, given our crude index; only about not significant (Fig. 3; r = -0.503, 0.1 > P > 38% of the variation in numbers of banded birds 0.05), while the correlation for new birds is sig- is explained by variation in the four speciesof nificant only if the 198 1 datum is excluded (Fig. flowers. Although the flowers we used are sig- 4; r = -0.756, P < 0.05). nificant resourcesfor the birds, they are not the The year with the greatestdeviation from the only species of flowers used by the birds, and regression in Figure 4 is 198 1, with fewer birds they do vary in both temporal availability (Table captured than predicted by the regressionequa- 1) and in nectar production. The deviations from tion (and the lowest number of birds during the the regression lines may reflect fluctuations in 1 l-year study). This year was also unusual in background population levels of the birds, or in another regard: it was the only year in the past the reliability of our estimates of the floral en- 15 years in which there was legitimate (not nectar vironment from the hummingbird’s point of robbing) visitation by large numbers of bumble- view. As with any result based solely on corre- bee queens (Bombus appositus) to Ipomopsis lation, and not explored further with experimen- flowers (personal observations; Pleasants and tal studies, we can suggestcausative factors but Waser 1985). During 198 1 the standing crop nec- cannot be positive about the definitive interpre- tar volumes in Ipomopsis flowers were much tation. higher than in all but one year (1977) of the seven Another factor that may cloud the correlation in which they were measured from 1975-1984 between numbers of birds and flowers is the fact (Pleasantsand Waser 1985). This apparently per- that these birds are long-lived but are not resi- mitted bumblebee queens (which cannot nor- dent at the study site year round; they only spend mally reach it) to collect the nectar. There was about two to three months of the year there. The also an unusual asynchrony between the time of rest of the year is spent in migration, or over- peak flowering and the influx of Selasphorusru- wintering farther south (probably in southern FEEDER CENSUSES OF HUMMINGBIRD POPULATIONS 283

1980

- E 1966 1979 ij 50 1981 1966

g 50.- = 0 I 3 01 I 0 500 1000 1500 2000 2500 3000 3500 z 0 500 1000 ,500 2000 2500 3000 3500 PEAK NUMBER OF FLOWERS OF 4 SPECIES PEAK NUMBER OF FLOWERS OF 4 SPECIES FIGURE 4. The relationship between the number of FIGURE 2. The correlation between the total num- new Broad-tailed Hummingbirds caught and the peak ber of Broad-tail Hummingbirds capturedat the Rocky number of flowers of four species used by the hum- Mountain BiologicalLaboratory from 1979-1989, and mingbirds for nectar. The regressionline is shown for the peak number of flowers of four species used by the relationship excludingthe 1981 datum; the regres- hummingbirds for nectar (Erythronium grandiflorum; sion equation is Y = 211.80 - 0.04X (r = -0.756, P Delphinium nelsonii, Delphinium barbeyi, and Ipo- < 0.05). mopsis aggregata). The regression equation is Y = 270.04 - 0.05X (r = -0.619, P < 0.05). of banded birds is that during years with lower floral abundance, our population of banded birds Mexico). Thus the availability of food resources derives from a larger area. During these years, during the short breeding seasonmay not be the for example, some birds probably fly farther to most important factor regulating population size, the feedersthat we use to attract them for band- even though it may affect the numbers of birds ing, and there might not actually be any signifi- using feeders. Floral resources are also not the cant variation in resident population size; we only food resourcesused by the hummingbirds. may just be increasing the size of our effective Insects provide a major sourceof protein as well study area by drawing in more distant birds. The as replacement of ions that are lost in the urine fact that the correlation between recapturedbirds (Calder and Hiebert 1983). and number of flowers is not significant (i.e., Hummingbird populations may in fact be more numbers of recaptured birds are relatively in- constant than indicated by our data. This would variant, with the exception of 1988-1989, when be true if the explanation for’increased numbers flower numbers were very low; Fig. 3), while that between new birds and number of flowers is sig- nificant (without the 198 1 datum; Fig. 4) sup- ports the hypothesis that there is a core popu- 9 (r125- lation of resident birds and a variable number z l 1989 of birds from outlying areas that are drawn to

Q lOO-- feeders during years of low flower abundance. There is circumstantial evidence to support this hypothesisof a relatively constant core pop- ulation. For example, the numbers of birds using feedersvaries seasonally. Use of feedersis high- est during the early part of the season,when flow- is 25.- ers are not yet available, and decreasesas the

fj O-1 I availability of floral resourcesincreases. At the _ 0 500 1000 1500 2000 2500 3000 3500 L end of the season,when floral abundance begins PEAK NUMBER OF FLOWERS OF 4 SPECIES to decline, or when flowers are preempted by the FIGURE 3. The relationship between the number of more aggressivemigrant Rufous Hummingbirds, Broad-tailed Hummingbirds that were recapturedfrom use of feeders again increases.At times like this previous years (probably local residents)and the peak individual birds will fly 2-6 km to our feeders number of flowers of four speciesused by the hum- mingbirds for nectar. The regressionequation is Y = (W. Calder, unpubl. data). 88.26 - 0.012X (r = -0.503, P > 0.05). The negative correlation between numbers of 284 D. W. INOUYE, W. A. CALDER AND N. M. WASER flowers and banded hummingbirds indicates one manuscript benefitted from suggestionsfrom A. D. difficulty with use of artificial feeders to attract McGuire, C. A. Keams, W. H. Baltosser, and two birds for population studies: there is potential anonymous reviewers. for competition from natural food sources. In LITERATURE CITED years with higher floral abundance, numbers of BALTOSSER,W. H. 1989. Nectar availability and hab- banded birds are lower, apparently reflecting de- itat selection by hummingbirds in Guadalupe creased use of the feeders by birds with nests or Canyon. Wilson Bull. 101:559-578. territories at greater distances. Thus, despite a CALD~~,W. A. 1975. Daylength and the humming- significant effort at capturing and banding birds birds’ use of time. Auk 92:8 l-92. each year, our results are biased by flower avail- CALD~~,W. A., III. 1981. Heat exchangeof nesting hummingbirds in the Rocky Mountains. National ability. GeographicSociety Research Reports 13:145-169. This pattern differs from that observed for nec- CALD~~,W. A., III. 1984. How long is a long-term tarivorous honeyeaters by Pyke (1988); he found study?Auk 101:893-894. that despite a consistent seasonalpattern of daily CALD~~,W. A., III. 1985. The comparative biology of longevity and lifetime energetics.Exp. Geron- production of nectar-energy per unit area, the tol. 20:161-170. seasonal pattern of honeyeater abundance was CALDER,W. A., ANDJ. BOOSER.1973. Nocturnal hy- not consistent between years, and did not cor- pothermia of Broad-tailed Hummingbirds during respond to the pattern of nectar-energy produc- incubation in nature with ecologicalcorrelations. tion. He concluded that honeyeater abundances Science 180:751-753. - CALDE~.W. A.. AND S. M. HIEBERT. 1983. Nectar must be determined by factors other than local feeding, dimesis, and electrolyte replacement of nectar production. However, Baltosser (1989) hummingbirds. Physiol. Zool.-56:325-334. found that nectar availability was important to CALDER.W. A.. S. M. HIEBERT.N. M. WASER.D. W. the organization of a guild of hummingbirds, and IN&JYE, & S. J. MILL&. 1983. Site hdelity, longevity,and populationdynamics of Broad-tailed that they responded to reductions in nectar sup- Hummingbirds: a ten year study. Oecologia 56: plies. He also noted an effectof an unusual freeze 689-700. on the distribution and number of hummingbird GIBBS, H. L., AND P. R. GRANT. 1987. Ecological nestsin one year. In this case,however, the effect consequencesof an exceptionally strong El Niilo was apparently mediated by damage to potential event on Darwin’s finches.Ecology 68: 1735-l 746. GRANT,B. R., AND P. R. GRANT. 1989. Natural se- nest sites rather than floral resources. lection in a population of Darwin’s finches. Am. Despite the variation we have seenin numbers Nat. 133:377-393. of birds, there does not appear to be any partic- INOW, D. W. 1980. The terminology of floral lar- ular long-term trend in either numbers of birds ceny. Ecology 61:1251-1253. INOUYE,D. W. 1988. Natural variation in plant and or of flowers at our study site, indicating that animal populations, and its implications for stud- these populations may be at some type of equi- ies ofrecoveringecosystems,p. 39-50. In J. Cairns librium. The pattern we describe also suggests [ed.], Rehabilitating damaged ecosystems. CRC that there has been no negative effectof the band- Press, Boston. ing study on the hummingbird population (con- INOUYE, D. W. 1989. Environmental control of vari- ation in flowering by two Delphinium speciesin trary to fears raised by Rosenberg and Stejskal montane Colorado. Am. J. Bot. 76(6) supplement: 1988 about the effect of banding). It does, how- 108-109. ever, suggestthat banders who rely on resources MILLER,S. J., AND D. W. INOUYE. 1983. Roles of the suchas feedersto attract birds for banding should wing whistle in the territorial behaviour of male be aware of the interactions between natural food Broad-tailed Hummingbirds (Selasphorm pluty- cercus).Anim. Behav. 3 1:689-700. supplies and feeders. MULVIHILL, R. S., AND R. C. LEBERMAN.1987. Bird banding at Powdermill, 1985: with a summary of Ruby-throated Hummingbird banding data. Pow- ACKNOWLEDGMENTS dermill Nature Reserve Research Report #46. Carnegie Institute, Pittsburgh. This research was supported by grants from the Na- PLEASANTS,J. M., AND N. M. WASER. 1985. Bum- tional GeoaranhicSocietv (2665.2824.3064.3242 and blebee foraging at a “hummingbird” flower: re- 35 13; WAC),*Earthwatch ‘(DWI and NMW), and the ward economicsand floral choice. Am. Midl. Nat. National Science Foundation (DEB8 102774, 114:283-291. BSR8313522, BSR8905808: NMW; DEB75 15422, PUKE,G. H. 1988. Yearly variation in seasonalpat- DEB7807784, DEB81 108387:0. R. Taylorand DWI), terns of honeyeaterabundance, flower density and and a summer Research Award from the University nectarproduction in heathlandnear Sydney.Austr. of Maryland Graduate Research Board (DWI). The J. Ecol. 13:1-10. FEEDER CENSUSES OF HUMMINGBIRD POPULATIONS 285

ROSENBERG,G. H., ANDD. STWSKAL. 1988. Arizona. WASER, N. M., AND D. W. INOUYE. 1977. Implica- Amer. Birds 42~1325. tions of recapturesof Broad-tailed Hummingbirds WASER,N. M. 1978. Competition for hummingbird banded in Colorado. Auk 94:393-395. pollination and sequential flowering in two Col- WASER. N. M., AND M. V. PRICE. 1983. Pollinator brado wildflowers. Ecology 59:93&944. behaviour’and natural selection for flower colour WASER.N. M. 1988. Comnarative oollen and dve in Delvhinium nelsonii.Nature 3021422424. transfer by pollinators of Delphkum nelsoki. Fund. Ecol. 2:41-48.