Hummingbirds and Pollen Flow in a Neotropical Agricultural Mosaic Amy Strieter Department of English, Kenyon College ABSTRACT Land transformation in the Tropics leads to greater landscape and forest fragmentation, creating land use mosaics. Some spe cies adjust to new habitats while others go extinct. Adaptability is more complicated, however, when a species is involved in a mutualistic relationship. If host plants are spatially isolated, their pollinators may or may not be able to continue linking th at subpopulation to the metapopulation; the pollinator population in turn may suffer due to decreased food supply, habitat, etc. This study examines hummingbird responses to a tropical agricultural/forest mosaic and tracks artificial pollen flow between fe eders in different habitats. Hummingbird presence was monitored at feeders in forest, pasture, coffee, and banana crops, and four intermediate edges. Most hummingbird species were not found exclusively at their predicted habitats, and total species richnes s for each habitat type were not statistically significantly different Chi square = 5.762, df = 3, p = 0.12. Habitat edges did not appear preferred or avoided by hummingbirds, and diversity was essentially even across the land use mosaic. Additionally, h ummingbird feeders were set with fluorescent dye powder as a Â€pollenÂ donor and adhesive tape as a collector. Pollen incidence and ranked pollen counts suggest that pollen from every habitat is moving to every habitat, with the exception of pasture pollen. Results indicate that having agricultural edges, crop variety, and crops in proximity to one another is more useful to several hummingbird species than homogeneous crop areas. RESUMEN La transformaciÃ³n de la tierra en los trÃ³picos conduce a mÃ¡s fragmen taciÃ³n del paisaje y del bosque, creando mosaicos de diferentes usos de la tierra; algunas especies se ajustan a los nuevos hÃ¡bitats en tanto que otras se extinguen. Sin embargo, la adaptabilidad se hace mÃ¡s complicada cuando una especie estÃ¡ envuelta en u na relaciÃ³n mutualistica . Si las plantas que donan polen estÃ¡n aisladas, sus polinizadores pueden continuar o no con el flujo genÃ©tico entre esta subpoblaciÃ³n y la metapoblaciÃ³n; a su vez, la poblaciÃ³n de polinizadores puede sufrir una disminuciÃ³n en la p rovisiÃ³n de comida, hÃ¡bitat, etc. Este estudio examina la respuesta de los colibrÃes a un mosaico tropical de agricultura y bosque, y tambiÃ©n sigue el transporte de polen artificial entre comederos en hÃ¡bitats diferentes. Se estudiaron las visitas de coli brÃes a comederos en un bosque, un potrero, un cafetal, un bananal y en cuatro orillas de por medio. La mayorÃa de las especies de colibrÃes no se encontraron exclusivamente en los hÃ¡bitats predichos y las riquezas totales de especies para cada tipo de h Ã¡bitat no fueron estadÃsticamente diferentes Chi cuadrado = 5.762, df = 3, p = 0.12. Las orillas de los hÃ¡bitats no fueron ni preferidas ni evitadas por los colibrÃes y la diversidad fue prÃ¡cticamente uniforme en todo el mosaico. Adicionalmente, se ofrec ieron comederos con pintura fluorescente en polvo como donadora de polen y cinta adhesiva como recolectora. La incidencia de polen y los conteos de rangos del polen sugieren que el polen de cada hÃ¡bitat es transportado a todos los demÃ¡s hÃ¡bitats, con la e xcepciÃ³n del polen del potrero. Los resultados indican que la presencia de fronteras agrÃcolas y la variedad y cercanÃa de las cosechas son mÃ¡s Ãºtiles para algunas especies de colibrÃes que la presencia de Ã¡reas de cosechas homogÃ©neas. INTRODUCTION Human land use is causing worldwide biodiversity decline Vitousek et al. 1997. As much as 10 to 15% of the EarthÂ‚s surface is covered in row crop agriculture or urban industrial occupation, and 6 to 8% is currently pastureland. Including degraded but fallow l and, the majority of EarthÂ‚s
vegetated land has been degraded by humans Vitousek et al. 1997. In the Tropics, land transformation is creating an increasingly fragmented landscape. Costa Rica has lost more than 50% of its primary forest in the last 50 yea rs, which originally covered about 85% of its landscape Borgella et al. 2001. As humans are transforming the Tropics, land use is becoming more diverse and forests are becoming more fragmented. This dynamic creates a mosaic of land use types that set the stage for the future of Tropical biodiversity Aldrich and Hamrick 1998, Daily et al. 2001. Although tropical agricultural landscapes contain a dramatically reduced number of forest species, in the last few decade s ecologists have found that some land u se areas may support Â€forestÂ species, especially if the land is used in multiple ways Hughes et al. 2002. Thus, it is important to understand how plant and animal species are reacting to a habitat matrix because their ability to adjust to specific chang es will determine the long term health of their populations. Species involved in mutualistic interactions may be more sensitive to changes in land use Aldrich and Hamrick, 1998. Tight co evolution, such as that found in many plant pollinator relationsh ips, may serve as a restriction on adaptability if, as in some cases, one member in the relationship cannot function in an agricultural landscape Borgella et al. 2001. Even if one or both members are able to function sub optimally, changes within inter s pecies relationships will still result. Pollination studies in tropical America show that plant populations in separated habitats may stay connected temporarily via wide range pollinators, but eventually the strain from lacking sufficient populations of es sential pollinators will eventually cause population crashes Roubik 2000. Studies on fig fig wasp mutualism have indicated that increasing interaction between forest fragments has strong positive effects on regional population stability Mawdsley et al. 1998. If pollinator populations are insufficient or absent, their plants will essentially be isolated Borgella et al. 2001. Even if plants are able to maintain populations in fragmented habitats or establish populations in unfamiliar habitat types, the degree to which subpopulations are isolated will affect the genetic future of the metapopulation Aldrich and Hamrick 1998. Fragmented populations may experience genetic bottlenecks or inbreeding, and the resulting reduced genetic variation increases the chance for population crash in the event of disease or unusual predation. Acceptable habitat for hummingbirds differ per species, but two major requirements are the level of vertically structured vegetation and certain species of year round flowering nect ar sources Borgella et al. 2001. If vegetation structure or flower sources Âƒ both of which may change with agricultural development Âƒ are inadequate, the birds may cease visiting those subpopulations and subsequently sever them genetically from the rest of t he metapopulation. Hummingbirds in mosaic areas in southern Costa Rica have been observed visiting flowers in pastures and farmland, including banana crops Borgella et al. 2001, Buono 2005. Buono 2005 found in a CaÃ±itas land mosaic that forest showed t he lowest hummingbird diversity of forest, banana, and pasture. Her study also found that most hummingbird and pollen movement occurred between pasture and its edge, as well as banana and its edge. This study measured hummingbird presence and pollen flow between different land use types. By recording bird activity and tracking pollen exchange, I discerned movement between major land use types, including forest, pasture, and coffee and banana crops.
METHODS Study Sites The study was conducted at La Finca SantamarÃa in CaÃ±itas of Monteverde, Costa Rica Figure 1. This organic agriculture property contains several land use types; those specifically of interest were intact forest, pasture, and coffee and banana cropland. Figure 1 : Treatment sites i n land use mosaic on La Finca SantamarÃa in CaÃ±itas in Monteverde, Costa Rica. Four land use sites in Forest, Banana, Coffee, and Pasture habitats marked with XÂ‚s in center of habitats. Four edge sites marked on habitat borders. Solid lines indicate edge g radients: thick = Forest Banana gradient, medium = Banana Coffee gradient, thin = Coffee Pasture gradient. Note: the Forest site was actually 30m from the forest edge. All other sites are approximately to scale. Treatments Eight treatment sites were se t up over the land mosaic: one in each of the land use habitats described above, and four edge sites between the land uses Figure 1. There was an area within the mosaic that was not studied because the crops it contained were all close to the ground and there were no trees within the area on which to set up a feeder. Feeders were set up at each of the other sites, including edges between Forest Banana, Banana Crop, Crop Coffee, and Coffee Pasture. Two sites Forest Banana Edge and Crop Coffee Edge had tw o feeders each, because
territoriality was observed early in the study at these sites and it was necessary to prevent territorial individuals from guarding the availability of an entire site. Hummingbird feeders were modified for tracking artificial poll ination: on the donor feeders in the land use habitats, above two of the Â€flowersÂ were fixed cotton with fluorescent paint powder dye; on the other two Â€flowersÂ there was clear tape, adhesive side exposed, to collect dye Jones and Little, 1983. One pain t color corresponded to each of the four land use types. The edge feeders only had receptor tape. All feeders were refilled as necessary with a 30% concentrate sucrose solution Buono 2005, McMahon 2005. On each of ten observation days, all sites were o bserved. Each site observation period lasted 20 minutes. To control for order effects, I rotated the order of site observations so each was observed at several different times of day throughout the study. Observations consisted of recording when and which hummingbird species visited the feeders. Hummingbird species were identified according to a bird guide by Stiles and Skutch 1989. All observations were completed in the morning on average between 06:30 until 10:30. After completing observations, I added new dye, replaced the tape, and applied each used piece of tape directly to a microscope slide. On non observation days, I used a compound microscope and a black light to determine dye color present on each slide. RESULTS Habitat Use by Hummingbirds Am ong eight land use and edge habitats, the feeders were visited by seven hummingbird species. These are listed in Table 1 with their habitat preferences as reported in Stiles and Skutch 1989. All birds observed are described as forest interior or edge dwe llers that will enter nearby edges or gaps; the only exception is the Green Violet ear that prefers open habitat. The most forest restricted birds witnessed in the study were the Striped tailed, Coppery headed, and Green crowned Brilliant; these require fo rest and edge but will enter nearby secondary forests. The Violet Sabrewing, Magenta throated Woodstar, and Purple throated Mountain Gem are usually found in a wider range of habitats, among forest, secondary forests, and more open habitats. Table 1 : Hum mingbird species found in La Finca Santamaria in CaÃ±itas of Monteverde, Costa Rica. All in Family Trochilidae. Each speciesÂ‚ commonly sited and preferred habitats are taken from Stiles and Skutch 1989. There are fifteen species of hummingbirds repor ted for this altitude in Monteverde Fogden, 1993, ten of which are listed as Common seen or heard several times a day in moderate to large numbers or Fairly Common seen or heard daily or almost daily, usually in Species Preferred and Common Habitat Green-crowned Brilliant Middle understory up to canopy; gaps, edges, tall secondgrowth Striped-tailed Hummingbird Mainly canopy; shrub level in gaps, edges, adjacent semi-open or secondgrowth Magenta-throated Woodstar Forest edge, clearings, semi open or second growth, scrubby pastures Coppery-headed Emerald All levels at edges, gaps, semiopen areas; males frequently high in canopy Violet Sabrewing Understory, forest edge, forest patches in disturbed areas, banana plantations Purple-throated Mountain-gem Steep broken areas; edges, breaks, second growth adjacent to forest, semi-open areas Green Violet-Ear Open, brushy highlands; pastures, second growth, large forest gaps, forest edge
small numbers. All species observed in this study are considered Common or Fairly Common according to Fogden 1993, excluding the Green crowned Brilliant, which is reportedly absent here but more common above 1500m Stiles and Skutch, 1989. Species common to the area which I observed only ra rely or not at all were the Green Violet ear observed twice, Fortktailed Emerald not observed and Steely vented not observed. These three species prefer open and scrubby habitat, including pasture, and should have been common in the study. Patterns of Species Richness Within this mosaic, coffee hosted the highest species richness during the entire study period Figure 1a: all seven species observed in the study were observed in the coffee. Half of the coffee visits were by the Green crowned Brilli ant, which according to Stiles and Skutch 1989, is not present in this part of Monteverde. The Violet Sabrewing did not comprise a large number of visits to the coffee area, but it is reported to frequent coffee crops Stiles and Skutch, 1989. Pasture h ad the next highest richness with 6 species Figure 1b. It should be noted, however, that three of these visited feeders only on the last day of observation and were not common. Because there were few visits to the pasture before the last day, the visits by the Coppery headed Emerald, Striped tailed Hummingbirds, and Magenta throated Woodstar appear graphically to be significantly present but are not in reality. Banana and forest were visited by the same four species 1c,d. Although varied, total species richness for each habitat type was not statistically significantly different Chi square = 5.762, df = 3, p = 0.12. A Green-crowned Brilliant Coppery-headed Emerald Striped-tailed Hummingbird Violet Sabrewing Green Violet-Ear Purple-throated Mountaingem Magenta-throated Woodstar B
Figure 1 : Species distributions in land use habitats based on richness. A: Coffee, B: Pasture, C: Banana, D: Forest. Chi square = 5 .762, df = 3, p = 0.12, n = 53. Daily observations show a different pattern. Considering the number of species seen each day, the forest habitat was richer in species than all others, with a mean rank number of species seen per day of 3.14 Friedman Test , Chi square = 5.762, df = 3, p = 0.12. Mean rank of species observed daily for banana was 2.786, coffee was 2.43, and pasture was 1.643. Most hummingbird species were not found exclusively at their predicted habitats. The Violet Sabrewing is often foun d in banana crops Figure 1 but in this study was mostly seen in the forest Figure 2a. The Striped tailed Hummingbird was found mostly in forest and banana Figure 2b, both of which were to be expected Figure 1. The Coppery headed Emerald and Green c rowned Brilliant seemed least discriminatory in foraging areas and visited habitats more equally Figure 2c d; this is odd only because Green crowned Brilliants are unlikely to be in pastures or far from the forest edge Figure 1. The Purple throated Mou ntain gem and Magenta throated Woodstar were found only in pasture and banana areas, which followed predictions that they would be in more open areas Figure 1. Habitat selection was most surprising for the Green Violet ear, which was found exclusively in the banana habitat and not at all in pasture, even though it is the most pasture adapted hummingbird of those observed in the study Figure 1. C D A 11 5 1 3 Forest Banana Coffee Pasture B 7 8 2 2
Figure 2 : Distribution of species among all habitats according to number of visits. A: Violet S abrewing n = 20, B: Striped tailed Hummingbird n = 19, C: Coppery headed Emerald n = 17, D: Green crowned Brilliant n = 35, E: Purple throated Mountain gem n = 6, F: Magenta throated Woodstar n = 3, G: Green Violet Ear n = 2. C 4 6 5 2 D 5 8 16 6 E 3 3 F 2 1 G
Edge habitats were readily used by most hummingbird species. In considering edge gradients, daily average rankings usually placed the edge richness mid to high compared to the habitats it bordered. In the banana coffee gradient, the banana crop edge and crop coffee edge shared 3.1 as the highest value for mean rank in the gradient, while coffee ranked 2.10 and banana 1.70 Friedman Test, Chi square = 10.364, p = 0.015, df = 3, n = 9. In the coffee pasture gradient, the crop coffee edge again had a significantly high mea n rank richness of 3.75 Friedman Test, Chi square = 13.316, p = 0.004, df = 3, n = 5. The coffee pasture edge had the lowest daily mean species richness of all habitats, with a value of 1.33. Pasture had a mean rank of 1.91 and Coffee 3.00. The forest ba nana edge did not have a significant difference in mean rank richness from the forest or banana, but it held the middle rank, suggesting the edge hosted considerable activity. Except for coffee pasture, which was visited by two individuals total, all edges hosted at least four species, suggesting hummingbirds moved freely between habitats, neither preferring edge nor being inhibited by it. Patterns of Species Abundance Hummingbird abundance was highest in the banana habitat 3.286 according to mean ran k Friedman Test, Chi square = 11.3, p = 0.01, df = 3, n = 7. Pasture showed the lowest abundance of 1.214; coffee held a mean rank of 2.57 and forest a mean rank of 2.93. Overall, edges had higher abundance mean ranks than the habitats they bordered, w ith the exception of the coffee pasture edge. Patterns of edge abundance were similar to those of edge richness. The crop coffee edge held the highest mean rank abundance 3.83 within the coffee pasture habitat Friedman Test, Chi square = 14.5, p = 0.002 , df = 3, n = 6. This edge also has the highest mean rank of all edges and land use sites. In the banana coffee gradient, the two edges, banana crop and crop coffee, both showed significantly higher abundance than either the banana or coffee habitats adja cent to the edges Friedman Test, Chi square = 7.99, p = 0.046, df = 3, n = 10. Recalling that Banana had the highest abundance of the land use habitats, it is interesting that within this banana coffee gradient, banana has the lowest mean rank abundance of 1.75; coffee = 2.20, crop coffee = 2.85, and banana crop = 1.75. The mean rank within the forest banana gradient is, again, lowest for banana 1.78, but not significantly so Friedman Test, Chi square = 5.07, p = 0.08, df = 2, n = 8. Forest banana edg e held a rank of 2.56, and forest habitat a rank of 1.67. Patterns of Species Diversity Among all habitats, bird diversity was the highest in the pasture and lowest in the coffee pasture edge Table 2. It seems unexpected that this polarity would happe n to neighboring habitats and more so that the pasture seems less isolated than the edge preceding it. This edge has a diversity value HÂ‚ less than half the value of the other habitats in its gradient. Table 2 : Diversity values HÂ‚ for each habitat. H abitats were tested in pair wise comparisons using the Shannon Weiner Index. Parenthetical number indicates sample size for that habitat. Significant comparisons for diversity were: Banana/Pasture, Forest/Pasture, Banana/Banana Crop Edge, Banana/Crop Coffe e Edge.
Although there are differences in diversity among habitats, those differences are relatively small Figure 3, suggesting that diversity was more or less similar among the land use mosaic. Of the major land use habitats, forest did host the lo west diversity, surpassed by banana, coffee, and finally, pasture 3a. Banana and pasture had statistically significant different diversities Shannon Weiner, t = 2.58, df = 17.95, as did the forest and pasture Shannon Weiner, t = 2.67, df = 29.82, a lthough these differences are less than one half of 1 HÂ‚ value, and probably not very different in reality. In terms of diversity, the edges show a different pattern of species composition from those of richness and abundance. One half of the edges have a diversity value one half to one third that of the corresponding land use habitats in their gradient Figures 3b d. Figure 3 : Bird Diversity for different habitats using the Shannon Weiner Index. For HÂ‚and N values, see Table 3. A: Hummingbird Divers ity among Land Use Types. B: Bird Diversity in Habitat H'-Value Forest 27 1.311 Forest-Banana Edge 43 1.217 Banana 27 1.367 Banana-Crop Edge 50 1.647 Crop-Coffee Edge 45 1.649 Coffee 31 1.503 Coffee-Pasture Edge 2 0.693 Pasture 17 1.65 A 0 0.5 1 1.5 2 Forest Banana Coffee Pasture Habitat Type H'-Value B 0 0.5 1 1.5 2 Forest ForestBanana Edge Banana Habitat H' Value C 0 0.5 1 1.5 2 Banana BananaCrop Edge CropCoffee Edge Coffee Habitat H' Value D 0 0.5 1 1.5 2 CropCoffee Edge Coffee CoffeePasture Edge Pasture Habitat H' Value
Forest Banana Gradient. C: Bird Diversity in Banana Coffee Gradient. D Coffee Pasture Gradient. Patterns of Species Overlap between Sites Comparing similarity of species composition between the land use hab itats revealed that all habitats were overall rather similar Table 3; within this overarching similarity, the forest and banana sites were the most similar, and banana and pasture were least similar. The similarity comparison also shows that pasture had low similarity to most other habitats or < 0.5 since there is a wide disparity between values above and below 0.5, and every comparison to the coffee pasture edge had low similarity. The coffee pasture edge gradient as a whole tends to have lower similar ity than the forest banana or banana coffee edge gradients, suggesting that the pasture and coffee pasture edge are outliers in terms of species composition. Finally, land use habitats are more similar to their edges than each other except concerning past ure. Table 3 : Overlap in species diversity along edge gradients and between land use habitats. Pair wise comparisons using SorensonÂ‚s Indices for species overlap show land use habitats are more similar to their edges than each other. Sorenson's Calculations for Bird Species Diversity Overlap in Forest-Banana Gradient Forest v For-ban 0.686 Banana v For-ban 0.571 Banana v Forest 0.778 Diversity Overlap in Banana-Coffee Gradient Banana v Ban-Crop 1.158 Banana v Crop-Cof 1.5 Banana v Coffee 0.552 Ban-Crop v Crop-Cof 0.779 Ban-Crop v Coffee 0.765 Crop-Cof v Coffee 0.632 Diversity Overlap in Coffee-Pasture Gradient Crop-Cof v Coffee 0.632 Coffee v Cof-Past 0.121 Coffee v Pasture 0.625 Crop-Cof v Cof-Past 0.085 Crop-Cof v Pasture 0.538 Cof-Past v Pasture 0.211 Diversity Overlap in Land Use Habitats Forest v Banana 0.778 Forest v Coffee 0.414 Forest v Pasture 0.545 Banana v Pasture 0.005 Banana v Coffee 0.552 Coffee v Pasture 0.625
Pollen Flow between Habitats The fluorescent paint dye, its application on feeders and its transferred deposition onto other feeders, was an acceptable proxy for actual pollen. Hummingbirds readily came to feeders and fed from openings on the feeders near dye as often as from openings with transparent tape dye receptors. Contact was witnessed through binoculars between hummingbird head and the dye or receptor tape. Henceforth, I will refer to the fluorescent paint as pollen, understanding that it is only a proxy for a ctual pollen. Actual pollen counts were not conducted in this study. Incidence of Pollen Transfer between Habitats Incidence of pollen transfer was calculated as number of pollen receptors transparent tape transferred to microscope slides with a given pollen type color for each habitat. For the habitats with donor feeders i.e., land use habitats, the total possible number would be 10, for five days of pollen collection and two receptors per feeder. For edge habitats, the total possible number would be 20, for five days of pollen and four receptors per feeder. This number would be 40 for the forest banana edge and the crop coffee edge, where there were two feeders at each site to control for the observed territoriality of some individuals. Pollen i ncidence shows that all land use habitats received pollen from all other sties fairly frequently, with the exception of pasture pollen Figure 4a d. All ten pollen receptor slides from the forest had coffee pollen and 8/10 had banana pollen Figure 4a. T his trend was similar for other habitats, where all or nearly all slides had at least some pollen from the other sites, especially coffee and forest pollen. The obvious exception was pasture pollen, which was entirely absent in half the habitats. Chi squar e tests show that there are significant differences between the incidences of pollen types for a given habitat Forest Chi 2 , p = 0.05, df = 2; Banana Chi 2 , p = 0.05, df = 2. The Chi square differences are all due to the absence or very low presence of pas ture pollen in these samples. 0 1 2 3 4 5 6 7 8 9 10 Banana Pollen Coffee Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen A 0 1 2 3 4 5 6 7 8 9 10 Forest Pollen Coffee Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen B
Figure 4 : Pre sence of dye as found on number of slides from different habitats. A: Forest Habitat Chi square, p = 0.05, df = 2, n = 18; B: Banana Habitat Chi square, p = 0.05, df = 2, n = 17; C: Forest Banana Edge Chi square, p = 0.05, df = 3, n = 115; D: Banana Crop Edge Chi square, p = 0.05, df = 3, n = 59; E: Crop Coffee Edge Chi square, p = 0.05, df = 3, n = 105; F: Coffee Habitat Chi square, p > 0.05, df = 2, n = 21; G: Coffee Pasture Edge Chi square, p = 0.05, df = 3,n = 55; H: Pasture Habitat Chi s quare, p > 0.05, df = 2, n = 18. Pollen incidence in edge habitats shows a similar pattern, but pasture pollen is present Figure 4e h. All edges have many slides with pollen from forest, banana, and coffee. While pasture is less common, it is present in all edges and can be common there as was the case for the coffee pasture edge, in which nearly one third of the 20 slides had pasture pollen Figure 4g. Chi square tests show that there are also differences between the incidence of pollen types for edg e habitat Forest Banana Edge Chi square, p = 0.05, df = 3; Banana Crop Edge Chi square, p = 0.05, df = 3; Crop Coffee Edge Chi sq u are, p = 0.05, df = 3; Coffee Pasture Edge Chi square, p = 0.05, df = 3. Again, significant differences between pollen types for edges reflect a reduced number of slides containing pasture pollen and, for the coffee/pasture edge, a small number of 0 5 10 15 20 25 30 35 40 Forest Pollen Banana Pollen Coffee Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen C 0 2 4 6 8 10 12 14 16 18 20 Forest Pollen Banana Pollen Coffee Pollen Pasture Pollen Presence of Dye Number of Slides Transferred Pollen D 0 5 10 15 20 25 30 35 40 Forest Pollen Banana Pollen Coffee Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen E 0 1 2 3 4 5 6 7 8 9 10 Forest Pollen Banana Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen F 0 2 4 6 8 10 12 14 16 18 20 Forest Pollen Banana Pollen Coffee Pollen Pasture Pollen Presence of Dye number of slides Transferred Pollen G 0 1 2 3 4 5 6 7 8 9 10 Forest Pollen Banana Pollen Coffee Pollen Presence of Dye number of slides Transferred Pollen H
slides with banana or pasture pollen only 8 or 9 slides of 20, compared to 19 of 20 for forest and coffee pollen. Pollen Counts an d Habitat Type Artificial pollen was counted in five unit increments up to 65, at which point it was considered a Â€megadeposit.Â Here, I present pollen counts as ranks of ordinal categories 0 5; 6 10; etc. and compare mean total pollen counts for each habitat on a per slide basis Figure 5. Again, land use habitats had 10 slides total, banana crop and coffee pasture edges had 20, and forest banana and crop coffee edges had 40. In a separate analysis, I count only the number of slides with megadeposits, assuming that megadeposits either indicate a more direct flight between feeders or more frequent flights between feeders and, therefore, better pollen transfer. Figure 5 : Average abundances of pollen per slide found in each land use habitat. Abundance of each pollen present on slides collected at one habitat, counted in five unit increments. Pollen counts set as ranks of ordinal categories 0 5; 6 10; etc.. A: Ranked pollen count found in forest Chi square, p = 0.05, df = 2, n = 18; B: Ran ked pollen count found in banana Chi square, p > 0.05, df = 2, n = 18; C: Ranked pollen count in Chi square, p > 0.05, df = 2, n = 11; D: Ranked pollen count found in coffee Chi square, p > 0.05, df = 2, n = 21. Error bars represent standard deviatio n. Ranked pollen counts suggest that activity at the forest and coffee is higher than the pasture and banana habitats. Pollen type counts were different with statistical significance only for the forest samples Figure 5a; Chi square, p = 0.05, df = 2, n = 18. Although the other comparisons are not statistically significant, it is conspicuous that forest and coffee pollen are found with highest and second highest abundance in every comparison Figure 5 a d. However, the overall trend is that pollen fro m every habitat is moving to every habitat Âƒ with the exception, A -2 0 2 4 6 8 10 Banana Pollen Coffee Pollen Pasture Pollen Pollen From Other Habitats Average ranked abundance of pollen per slide B 0 2 4 6 8 10 Forest Pollen Coffee Pollen Pasture Pollen Pollen From Other Habitats Average ranked abundance of pollen per slide C -2 0 2 4 6 8 10 12 Forest Pollen Banana Pollen Pasture Pollen Pollen From Other Habitats Average ranked abundance of pollen per slide D -2 0 2 4 6 8 10 Forest Pollen Banana Pollen Coffee Pollen Pollen From Other Habitats Average ranked abundance of pollen per slide
once again, of pasture pollen. Interestingly enough, despite the deficient presence of pasture pollen among other habitats Figures 5a,b,d, all other habitat pollen is found in the pasture Fi gure 5c. Megadeposits and Habitats Artificial pollen was counted in five unit increments up to 65, at which point it was considered a Â€megadeposit.Â The megadeposit analysis was conducted under the assumption that large deposits indicate either a more direct flight between feeders or more frequent flights between feeders and, therefore, better pollen transfer. Results show that all land use habitats have 0 2 megadeposits Figure 6a, and that edges tend to have more megadeposits than habitats Figure 6b d. Figure 6 : Megadeposits per habitat. A: Megadeposits in different land use habitats n = 5. B: Megadeposits in forest banana gradient n = 26. C: Megadeposits in banana coffee gradient n = 18. D: Megadeposits in coffee pasture gradi ent n = 26. 0 2 4 6 8 10 12 14 Forest Banana Coffee Pasture Habitat Number of megadeposits A 0 2 4 6 8 10 12 14 Forest Forest-Banana Edge Banana Habitat Number of megadeposits Blue Forest Pink Coffee Yellow Banana Green Pasture B 0 2 4 6 8 10 12 14 Banana BananaCrop Edge Crop-Coffee Edge Coffee Habitat Number of megadeposits C 0 2 4 6 8 10 12 14 Coffee CropCoffee Edge CoffeePasture Edge Pasture Habitat Number of megadeposits D
DISCUSSION Hummingbirds are moving freely between habitats and through edges. Results concur with literature that reports that all these species visit edges or habitats near forest edges. Since pollinators are moving freely, it follows t hat pollen is flowing readily between habitats and across edges. The only exception is pasture pollen, which never reaches forest, banana, or coffee but does make it to edge habitats. Paradoxically, pasture receives pollen from those same three habitats. T his could be explained by disproportionately lower frequency of visits to the pasture. If the hummingbirds are spending most of their time in the rest of the mosaic and only occasionally go to the pasture, the likelihood of retaining pollen from one of the other habitats is high, and the likelihood of retaining and transferring pollen from the pasture is lower. The most likely pollen movers in this mosaic are those that were found in all four habitats: the Violet Sabrewing, Striped Tailed Hummingbird, Cop per headed Emerald, and Green crowned Brilliant. It is important that these are reported as forest dwellers that leave the forest only to nearby habitats, because it explains that although pollen is transferring frequently among a varied land mosaic, it is doing so because the mosaic is small scale and hummingbirds are able to access nearby non forest habitats. Therefore, pollen flow of hummingbird pollinated plants in this mosaic is extensive. The prime movers need nearby forest, however; it thus becomes necessary to keep sufficient forest area preserved and in proximity to the farm in order for the pollen flow to continue. Â€SufficiencyÂ is determined by the needs of the present hummingbird species. Insufficient area this study does not examine the parame ters for sufficient area will eventually favor the Green Violet ear, the only real open habitat hummingbird. This was not a common bird in my study and visited only the coffee habitat; this bird is supposedly only found above 1500m, so there are no common open habitat species to move pollen at this altitude if the partially forest dependent species are absent due to excessive deforestation and fragmentation. Results also suggest that diversity in agricultural edges may tend to be higher than main agricul ture use areas and that, given a larger scale mosaic, coffee agriculture may be better for more hummingbird species than banana areas, and certainly better than pasture. Data also indicate that including agricultural edges, crop variety, and crops in proxi mity to one another is more useful to a variety of hummingbird species than homogeneous crop areas. A crop such as coffee or similar vegetation with some proximity to forest may be ideal in terms of balancing agriculture with preserving biodiversity. ACK NOWLEDGMENTS Thanks that number the stars to Dr. Alan Masters for his endless insight, generosity with time and advice, and kindly advising. Many thanks also to Oliver Hyman and Maria Jost for their statistical genius, assistance in procuring supplies, and much appreciated encouragement. Thank you also, Dr. Karen Masters and Javier MÃ©ndez, for your continual encouragement, humor, and suggestions along the way. Additional thanks to Javier for his translating prowess. I am quite grateful to the SantamarÃa fam ily for allowing me to conduct this study on their farm, and particularly to Edwin and Norman SantamarÃa, who were always so accommodating and helped me acquire permission to use neighboring property. Thank you, Andrew Rodstrom, for being my third TA and f or doing more than I can detail herein. Lastly, many, many thanks to Kristen Becklund, Arden Piland and Kevin Loope for their crepuscular comradery and support in the Nerdery so many nights.
LITERATURE CITED Aldrich, Preston R., and J. L. Hamrick. 1998 . Reproductive Dominance of Pasture Trees in a Fragmented Tropical Forest Mosaic. Science 281: 103 105. Borgella, RenÃ© Jr., Allison A. Gavin, and Thomas A. Gavin. 2001. Species Richness and Pollen Loads of Hummingbirds Using Forest Fragments in Souther n Costa Rica. Biotropica 331: 90 109. Buono, Ashley E. 2005. Hummingbird Facilitated Pollen Flow between Transformed Neotropical Habitats. CIEE Fall 2005. 87 100. Daily, Gretchen C., Paul R. Ehrlich, and G. Arturo SÃ¡nchez Azofeifa. 2001. Countryside biogeography: use of human dominated habitats by the avifauna of southern Costa Rica. Ecological Applications 11:1 13. Fogden, Michael. An Annotated Checklist of the Birds of Monteverde and PeÃ±as Blancas. LitografÃa e Imprenta LIL, S.A.: San JosÃ©, Cost a Rica, 1993. Hughes, Jennifer B., Gretchen C. Daily and Paul R. Ehrlich. 2002. Conservation of tropical forest birds in countryside habitats. Ecology Letters, 5: 121 129. Jones, C. Eugene and Little, R. John, Editors. Handbook of Experimental Pollinatio n Biology. New York: Van Nostrand Reinhold Company Inc., 1983. Lamb, David, Peter D. Erskine, and John A. Parrotta. 2005. Restoration of Degraded Tropical Forest Landscapes. Science 310: 1628 1632. Mawdsley, Nick A., Stephen G. Compton and Robert J. Wh ittaker. 1998. Population Persistence, Pollination Mutualisms, and Figs in Fragmented Tropical Landscapes. Conservation Biology 12: 1416 1420. McMahon, Taegon A. 2005. Increased Agonistic Behavior in Hummingbirds Family Trochilidae in Montever de, Costa Rica with a Reduction of Food at Artificial Feeders. CIEE Fall 2005. Roubik, David W. 2000. Pollination System Stability in Tropical America. Conservation Biology 14: 1235 1236. Sekercioglu, Cagan H., Gretchen C. Daily, and Paul R. Ehrli ch. 2004. Ecosystem consequence of bird declines. PNAS 101: 18042 18047. Stiles, F. Gary and Skutch, Alexander F. A Guide to the Birds of Costa Rica . Cornell University Press: Ithaca, New York, 1989. Vitousek, Peter M., Harold A. Mooney, Jane Lubchenco, and Jerry M. Melillo. 1997. Human domination of the earthÂ‚s ecosystems. Science 277: 494 499. Warner, Matt. 2005. The role of primary forest on tropical avian communities in a mixed agricultural landscape. CIEE Fall 2005: 163 179.