Background/Question/Methods
Remote sensing (RS) data are increasingly used to study community composition and the distribution of biological diversity. These data are especially important when vegetation structure is known to influence patterns of species co-occurrence, as is the case for some forest bird communities. Here we evaluate the correlation between remote-sensed and climate data and composition of 189 hummingbird communities in Ecuador, a region with strong environmental gradients in moisture and elevation. Given variation in habitat requirements and specialization on flowers across hummingbird clades we expect some clades of hummingbirds to respond more directly to vegetation changes (captured by RS). Specifically, brilliants and hermits tend to require forest habitat and generally have more specialized flower requirements then other groups of hummingbirds. We evaluated the relationship between the environment and different aspects of community composition, including composition and richness of all hummingbirds and specific clades using both traditional and phylogenetically based metrics using general linear models. To describe the environment, we used remote-sensing variables relating to canopy roughness and surface moisture including annual horizontal and vertical means, and their corresponding standard deviations from microwave data sent from QSCAT (Quick Scatterometer); variables related to vegetation cover such as annual maximum and mean leaf area index, and the vegetation continuous field (related to percent tree cover; PTC) derived from the 1-km resolution global 8-day MODIS (Moderate Resolution Imaging Spectroradiometer); and climate data from WORLDCLIM.
Results/Conclusions
We found that horizontal scatter from QSCAT was an important correlate of overall richness and composition of hummingbird communities. Further, this variable came out as an important predictor for several clades, potentially because QSCAT data captures aspects of vegetation structure and may provide useful information in areas of persistent cloud cover that are common in some parts of Ecuador. Radar data, such as QSCAT, are increasingly identified as important in modeling bird species distributions, especially in forested areas. As predicted the two most specialized hummingbird groups, hermits and brilliants, were influenced by percent tree cover. For the high elevation generalist hummingbird clade, the coquettes, no RS data made an important contribution to models. In contrast, emeralds, also a clade with many generalist species, were influenced by QSCAT. These results indicate that for some species, RS data are essential descriptors of the environment and are therefore useful for both modeling and species management. Further, knowledge of species ecology can help determine when RS data will likely be most important.
