Methods for censusing and studying birds of prey within large expanses of tropical forest are poorly developed. We put a great deal of effort into comparing, testing, and refining such methods.

Under the best of conditions, determining the abundance of raptors in the wild is a challenge. Census methods normally used for other birds are not often applicable, because of raptors' low population densities, hence low detection rates. For example, among the most widely used counting methods for small songbirds is the "point count," in which one tabulates all individuals detected within a certain radius of a point during a specified amount of time. Usually a time span of 3 to 10 minutes is used in such counts. Brief time spans are advantageous because they minimize the problem of bird movement during the count (ideally one wants a "snap shot in time"), and they permit one to sample more different points with a given amount of time and effort.

However, the odds of detecting even a single raptor during a 10-minute point count are extremely low! Hence, for starters, a method directed toward raptors must allow for a longer detection period, in order to generate any detections. In mature tropical forest, one has the added difficulty of simply detecting the raptors amidst the dense vegetation. Methods of enumerating tropical forest raptors were pioneered by Jean-Marc Thiollay, a French researcher who has sampled relative abundance patterns of tropical raptors at many sites around the world. Thiollay (1989) worked at a unique site in French Guiana, where rock outcrops or "inselbergs" emerge above the forest. Thiollay found that many tropical raptors have the habit of soaring above the canopy for a period on sunny days, apparently in territorial display. Sitting atop these inselbergs, Thiollay attempted to map out the territories of birds observed soaring. Thiollay also walked transects through the forest understory, noting all raptors detected in this manner.

Using Thiollay's efforts as a point of departure, we have experimented further with these and additional techniques, have formally compared their efficacy, and progressively refined the methods we consider most useful. Though there is room for additional experimentation, we have settled upon a set of techniques that, when used together, do an excellent job of detecting most of the forest raptors occurring at our study site. These methods, described in our "gray literature" progress reports, have been emulated at a number of tropical forest sites around the world, including Honduras, Costa Rica, Peru, Brazil, and Madagascar.

Our initial comparisons of raptor census methods at Tikal were conducted mainly by Charles Turley, with the involvement of William Burnham, J. Peter Jenny, and Elbert Cleaveland. Chuck compared the following methods: foot survey (walking transects through the forest, pausing periodically to listen); road surveys (driving a road through the forest, watching for raptors to the sides); canopy-emergent point counts (observing over the canopy from tree-tops or other vantage points); and acoustical luring (playback of recordings of raptor vocalizations or distress calls of prey species).

Among these methods, the clear winner was the canopy-emergent point count. In a similar amount of effort devoted to each method, the emergent point counts detected 24 species, including eight uncommon and one rare species. In comparison, walking transects and road surveys each detected 15 species, and acoustical luring detected 13 species. The latter two methods did a poor job of detecting the more uncommon species.

While canopy-emergent point counts, both in our trials and in Thiollay's work, were successful for detecting the many raptor species that habitually make display flights over the forest, the owls do not make such flights, and neither do certain other species, for example the odd "forest-falcons" of the genus Micrastur. These forest-falcons, however, are almost owl-like in behavior, becoming active well before the first light of dawn, voicing their loud, far-carrying calls during this still, pre-dawn hour. The Laughing Falcon also issues its loud call at this time of day. Hence, we found it natural to develop an early-morning listening-based point count for detecting this group of species.

We conducted some further experimentation with acoustical attraction via play-back of distressed prey calls, but found this method to have a low rate of return in comparison to the pre-dawn listening calls and canopy-emergent visual point counts. Acoustical luring, however, may well prove useful for specific objectives, for example, detecting certain focal species. Our tests and further refinements of these methods are reported in Whitacre and Turley (1990) and Whitacre et al. (1992).

The set of methods we have adopted and used widely in large areas of continuous tropical forest, and which we recommend for use elsewhere, is as follows.

1. Canopy-emergent visual point count

The basic element of this count method is use of a vantage point allowing an unobstructed view over the forest canopy. We have often used emergent tree tops, but use of existing high points or structures, or in more open or deforested environments, hill tops, can save the large effort required to find, climb, and prepare trees with adequate views. Having selected adequate view points, our procedure is to conduct a point count somewhere between two and four hours in duration, though clearly this should be standardized within a study. We feel it is advisable to begin the count at a fixed time in relation to official sunrise. After experimentation with different beginning times and durations, we elected to begin point counts one hour after sunrise, and used a duration of four hours. We found that those raptors that habitually soared over the forest did so most reliably during this period. Most soaring began once morning fog had broken up and the sun began to warm the forest, creating rising thermals and good soaring conditions. Although the timing of peak soaring activity often differed from day to day depending on weather, the period of peak soaring activity was virtually certain to fall within the period 1-5 hours after sunrise.

We used a bounded point count, defined by a 120� view angle extending out to a radius of 1 km from the point. This wedge-shaped plot, 1.05 km² in area, was subdivided into 10 numbered zones to facilitate noting where a bird was located within the plot. We found that use of a wider view angle was impractical, as few trees allow a wider angle, and even a 120� angle often demanded some trimming of obstructing foliage from the tree-top.

With respect to the 1 km radius we used, certainly an unbounded plot could be used, but many raptors are not reliably detected or identified beyond 1 km. Ideally one would test whether the species of concern are equally well detected out to the chosen plot boundary; some smaller raptors may not be reliably sampled out to 1 km. Also, it may be worth experimenting with use of variable-radius point count methods. We initially experimented with such methods but abandoned the effort, feeling that our distance estimates were not accurate enough to justify that approach. This conclusion likely resulted in part from the poor quality of the range-finders we had. With higher quality range-finders such an approach may be practical, and would have the advantage of potentially allowing estimation of species densities rather than simply indices of relative abundance.

During each 5-minute period we noted, for each species, the number of distinct individuals detected within the plot. Birds detected outside the plot were noted in a fashion allowing their distinction from detections within the plot.

Data resulting from such a count are of several types, mainly: the number of 5-minute periods during which species X was detected, the number of individuals times the number of 5-minute periods within which they were detected, and the number of distinct individuals known to be present at some point during the count. The latter is usually equivalent to the largest number of distinct individuals of species X detected simultaneously during some 5-minute period. This latter item is the datum we have used in all analyses, and which we generally recommend for use. It is a simple datum to extract from census data sheets. Moreover, this datum is largely immune to the effects of weather and details of point count timing, and largely cancels out any differences in time of peak soaring activity both between days and between species.

2. Pre-dawn auditory count

At the base of each tree used for canopy-emergent visual counts, we also conducted a pre-dawn listening count. This count began one hour prior to official sunrise time, and ended at official sunrise. This count was very effective for detecting Micrastur forest-falcons, owls, and Laughing Falcons (Herpetotheres cachinnans), though the visual count generally turned up more Laughing Falcon detections. In addition, this count was highly effective for detecting a number of non-raptorial birds, namely guans, curassows, chachalacas, Ocellated Turkeys (Meleagris ocellata), various nightjars, motmots, tinamous, and other birds that begin calling before dawn. In addition, howler monkeys (Alouatta pigra) were very often detected, and spider monkeys (Ateles geoffroyi) somewhat less frequently.

We used a 1-hour count period, recording the number of individuals of each species detected during each 5-minute period. We normally used two persons, both listening, and one recording data on a data sheet. This was an un-bounded point count, with detections at any distance being recorded. Because different species are audible at vastly different distances, this method is appropriate for comparing detection rates of a given species across sites, but not for directly comparing the abundance of different species. In addition, different vegetations are known to differ with respect to sound propagation, hence even within a species, comparisons across habitats may be affected by differences in the distance of audibility. Hence, interpretation of point count results must be cautiously made.

One might use a shorter count period, which would allow more than one point to be counted per morning, per count team. However, there is a problem with such an approach. During the hour prior to official sunrise, calling frequency of the species treated here changed rapidly, with different species reaching peak calling frequency at different times. Hence, for almost all species treated here, no two 15-minute periods, nor even 30-minute periods within this hour, had equivalent calling rates. As a result, counts conducted within different subsets of this 1-hour period would not be comparable. Mainly for this reason, we recommend using a 1-hour count period (or perhaps somewhat shorter, but at any rate, only one count per team per morning), despite the fact that this limits each team to one point count per morning.

This count method has the advantage that it does not require any special site preparation. Indeed, sites need not be pre-selected, and one can simply pace a certain distance along a transect, for example, to select a count site. For this reason, one can easily sample a different point in space each morning, and this is the approach we generally recommend, based on the rationale that a 1-hour count at a given point is a substantial count effort, and more information is probably often gained by sampling additional points in space rather than sampling a given point repeatedly. An exception would occur in cases in which maximum confidence is desired regarding the species occurring at a limited number of points in space, in which case repeated counts at one point might be useful, especially if distributed seasonally so as to detect species that have differing seasonal patterns of calling frequency.

The types of data resulting from these pre-dawn counts are the same as those resulting from the canopy-emergent visual counts. Again, the datum we chose to analyze was the maximum number of distinct individuals detected--normally the largest 5-minute total for each species, no matter when during the count it occurred. Again this approach has the distinct advantage of canceling out the effect of differences between species in the timing of peak calling activity.

Efficacy of combining these two count methods

At Tikal, this pair of methods did a good job of sampling most raptor species, and we suspect this will prove true in many tropical localities. The canopy-emergent visual count detected most diurnal raptors at Tikal, because most of them frequently soar over the canopy in apparent territorial display. The pre-dawn listening count was effective for owls and other species that are active at dawn and that use loud vocalizations as their primary spacing mechanism.

Two species at Tikal, however, were infrequently detected by these methods: the Bicolored Hawk (Accipiter bicolor) and the Crested Eagle (Morphnus guianensis). Both of these species rarely if ever soared above the canopy at Tikal, and neither used loud vocalizations as spacing mechanisms. Acoustical luring, using play-back of conspecific vocalizations or recordings of distress calls of prey species, may be a useful adjunct in efforts to sample these species, or to fully document the raptor fauna of a given site. In other localities, no doubt additional raptor species may be difficult to detect and sampling efforts would likewise benefit from acoustical luring.

22

Maya Project Introduction | The Selva Maya | Tropical Deforestation | Maya Project

Home / Introduction / World Center for Birds of Prey / Visitors Center / Conservation Projects /
Notes from the Field / Explore Raptors / How You Can Help / Shop On-line / Archives of American Falconry /
Press Room / Research Library / What's New

THE PEREGRINE FUND IS A MEMBER OF EARTH SHARE

This section has been developed for The Peregrine Fund by
Asana Draper
Copyright � 2002 The Peregrine Fund. All Rights Reserved.
Last Revised:  12/10/08