|
|
|
|
Methods
Given the aims of the research, we focused our attentions on the 319 bird species that are associated with the interior or edge of lowland rainforest habitats (Terborgh et al. 1990). Twenty families of bird species associated with other habitat types, e.g. oxbow lakes, river edge habitats, swamps, arboreal dwellers, etc. were judged to be not associated with lowland rainforest habitat, and were thus excluded from the principle surveys in this study (e.g. Ardeidae, Ciconiidae, Scolopacidae, Apodidae, Alcedinidae, Hirundinidae, etc). These species were only recorded during general observations particularly if they were rare or new sightings, to complete the bird species lists for each lodge. These records are presented in the section ‘Species Accounts’, and in Appendix 2. Bird Census Methods A combination of different census methods must be used to achieve the best possible accuracy in counts of species that possess different social and territorial systems (Terborgh et al. 1990). In the few population surveys that have been conducted in lowland rainforest habitats in Madre de Dios, the principle method used (in conjunction with other methods) has been Spot-Mapping (Kendeigh, 1944; Terborgh et al. 1990; Kratter, 1995a, 1995b, 1997). These spot-mapping surveys have usually been conducted during the peak in the breeding season of the lowland forest bird species (based on the respective authors personal observations), and they produce a density of number of breeding pairs per unit area of habitat. However, our study is a continuous survey over a 21-month period, and is not restricted to this corresponding timing of the peak in the breeding season. In fact we make no assumptions as to whether there is a peak in the breeding season. We aim to calculate absolute densities of the number of individuals per unit area of habitat and not number of breeding adult pairs. Thus the spot-mapping method was not chosen for this study. The methods that were implemented were based on Distance Sampling Methods (Buckland et al. 1993). Distance sampling incorporates estimates of distance between bird contacts and the observer at fixed point count stations in the forest to estimate densities of species (Lloyd et al. 1998). These density estimates take account of the fact that some birds are detectable over much greater distances than others, or that species are more easily detected in one habitat type than another (Lloyd et al. 1998). This distance data also allows us to make comparisons between species and between the same species in different habitats that are not possible with other relative density estimation methods (Lloyd et al. 1998). There are four basic assumptions of distance sampling that have to be adhered to, in order to obtain unbiased results: 1) points and transects are placed irrespective of prior conceptions of bird density; 2) objects (birds) directly over each point count station or on a transect line are always detected; 3) objects are detected at their initial location prior to natural movement or movement in response to the observers; 4) distances are measured accurately or at least estimated with small and random error (Lloyd et al. 1998). Our principle distance sampling method we used to survey the local diurnal bird species of each lodge was the Variable Circular Plot method (VCP)(Reynolds et al. 1980), this is also based on point transect methodology (Buckland et al. 1993; Jones et al. 1995: Marsden et al. 1997; Marsden and Jones, 1997; Marsden in press; and Lloyd et al. 1998). At three of the sites, ECO (type B forest), SACHA (type F forest), and EI (type H forest), 39 census stations were located along transect routes. For the other two smaller areas, 26 census stations were established in type E/F forest at TRC while 24 stations were established in type B/F forest at the CAPH site. Each census station was located 200 m apart, while each transect was also located at a minimum distance of 200 m apart. At all sites transects were situated either perpendicular across existing trails, along existing trails, or they were established in areas were there were no trail systems. Transects located perpendicular across trails had a central census station placed directly on the trail, with the remainder of the transect continuing up to 400 m either side of the trail. The number of census stations and length of each transect was dependant on the amount of available habitat at each site. During January and September of each year, VCP surveys began at 0530 hrs and were concluded by 0830 hrs after which time vocal activity decreases significantly. During the months of September to December, dawn arrives earlier. Thus VCP surveys had to begin at 0445 hrs in order to adequately census the bird species that normally only sing during the predawn chorus (see Terborgh et al. 1990). Two observers spent 10 minutes at each census station recording all bird contacts. These contacts were assigned to one of three categories; seen, heard or seen and heard. Observers noted the time of the contact, species, and the number of individuals. Each bird contact was also assigned to one of five height categories: 1 = ground level (< 1 metre); 2 = understorey (1 – 5 m); 3 = mid-canopy (5 – 15 m); 4 = canopy; 5 = flying above the canopy. Observers then accurately estimated horizontal distance from the centre of the census station to each individual contact. For species that habitually occur in monotypic flocks, such as the parrot species, estimations of distance were made from the point count to the centre of the flock. Similarly, for the canopy mixed species flocks, it proved impossible to identify flock members during the predawn hours, high up in the canopy. Distance estimations were made from the point count to the geometric centre of each flock. The number of species and individuals was later obtained by searching for the same flock during later hours of the day. Two repeats of each transect were made per visit to each site, with the number of visits to each site governed by permission from the lodge owners (see Results). The direction of the surveys along each transect was rotated to counter the bias of bird activity and the time of day. Distance data was then analysed using the program Distance, version 3.5 (Buckland, 1998). For nocturnal species, principally members of the families Psophiidae (not strictly nocturnal, but most vocally active during the night), Strigidae, Nyctibiidae, and Caprimulgidae along with certain species of the families Tinamidae, Rallidae, and Cracidae, a second distance sampling method was used. The Variable-Distance Line Transect method (VDLT) is similar to the VCP method except that this method is based on line transect methodology (Buckland et al. 1993). This method is more suited to these larger, mobile species which are assumed to occur at lower densities (Jones, 1998). Nocturnal surveys were only conducted at three of the five sites, due to time and resources available. Between September 1997 and September 1998 each census began at 0700 hrs and concluded at 1030 hrs, with brief observations also made during the hours of 0330 hrs and 0530 hrs. Two transects were surveyed at ECO (transect length = 3,700 m and 1,115m respectively), while three were established at SACHA (transect length 2,400m, 1,775m and 2,275m respectively), and two at the TRC (transect length = 2,850m and 1,600m respectively). Each transect was marked at 25 m intervals and were located either along existing trails at the lodges, or were situated in forest away from the trail system. Transects were walked at a slow but steady pace. The observer recorded the time of each contact, the species, number of individuals, the distance of the nearest transect marker, and an estimation of distance from the bird contact perpendicular to the transect. For Psophia leucoptera, distance estimations were made to the geometric centre of each group, perpendicular to the transect. One to three repeats of each transect were made per visit, depending on suitable weather conditions. As for the VCP data, the data obtained during VDLT surveys was also analysed using program Distance version 3.5 (Buckland, 1998). For the more furtive, cryptic understorey species that do not vocalise frequently, and are usually under-represented in song-based census methods, Mist-netting was employed. This technique has been the subject of debate by numerous authors (e.g. Karr, 1976, 1979; Parker and Remsen, 1987; Terborgh and Weske, 1975; Lovejoy et al. 1986). On every morning at each of the five sites, nine mist nets were located at three census stations, with three nets erected at each station in a straight line. Each net was 2 m high and 12 m long, consisting of four or five trammels, and were placed at ground level. Each station was only sampled once per visit, at stations that were not being sampled during VCP surveys that same morning. Nets were opened at 0600 hrs and closed at 1100 hrs, at which time they were moved to the next three census stations, erected and left furled in preparation for the next day’s survey. For each individual bird captured, the following variables were recorded: date, time, census station, net number, species, sex, age (adult or juvenile) wing length, tail length, culmen length, tarsus length, and weight. Recaptures were re-measured and compared to when and where they were first captured. During September 1997 and September 1998, we also recorded whether or not each individual was moulting. Direct counts were reserved for several species of diurnal raptors associated with forest habitat. These species are notoriously difficult to census accurately (Thiollay, 1989; Robinson, 1994). Observers are confronted by the lack of standardised methodology, species low population densities, high diversity, large territory sizes, and inconspicuous behaviour (Robinson, 1994). Forest dwelling raptors rarely vocalise apart from during the predawn chorus whereby some species of the Falconidae family are most regularly heard (Terborgh et al. 1990; Lloyd, per obs). During VCP surveys whenever a raptor was encountered at any of the census stations, the species and number of individuals was noted. The abundance of these species was then referred to as either present or absent (see Results). Direct counts of territorial pairs were also conducted on two species of bamboo specialists; Celeus spectabilis and Cecromacra manu. During May and July 1998 one observer spent many hours of general observations, searching for the territories of these two species, following no standardised methodology. These two species were located on the periphery of the study site, and represent an important component of the bamboo bird community (see Noteworthy Records for the conservation status of both species). The criteria for classifying indicator bird species follow that of Stotz et al. (1996). Broadly defined they: (1) occur in one or a very few habitats; (2) they are relatively common; (3) they can be detected easily; (4) they show high sensitivity to habitat disturbance (i.e. they become rare or disappear, in habitats that are altered, over-hunted or fragmented). The complete lists of indicator species that are applicable to the forest habitats of the Tambopata region are shown in Appendix 1. Habitat Surveys Habitat surveys were conducted at 30 of the census stations located at ECO, SACHA and TRC, and conducted at all of the census stations at the two remaining sites. We estimated the vegetation cover at each census station using a sighting tube (Bibby et al. 1993). Four height categories were used; ground level (<1m), understorey level (1-5m), mid-canopy level (5-15m) and canopy level. At each census station, and within a 30m radius, the ten nearest trees with a dbh at breast height of > 0.2 m were selected and labelled with aluminium tags. For each tree the following variables were recorded: estimation of height, dbh (using a dbh tape measure), distance to the centre of the station, presence or absence of viney tangles and liana’s, whether the tree was a species of palm tree or non-palm species. Features of tree architecture were also recorded following the methods described by Torquebiau (1986); Jones et al. (1995) and summarised by Bibby et al. (1998). Each of the ten trees were allocated to one of the following groups; crown of the tree branches above half its height (trees which have developed under the closed canopy of primary forest tend to have the first major branching of the crown well above half their height); branching below half their height (trees which have developed in more open canopies usually have the major branching of the crown below half their height); and presence or absence of scars (trees which have major scars from dropped branches tend to be characteristic signs of regeneration). From September 1997 through until September 1998, for each of the ten trees selected, observers recorded the presence or absence of fruit and/or flowers. For each subsequent visit the same trees were re-examined for the presence of fruit or flowers. This would help us determine the timing of fruiting and flowering for the larger trees over the year. The number of subsequent visits was unfortunately dependent on permission of the lodge owners, and some sites were visited more than others (see Results). Tree density at each site was calculated using the equation 5.1, where; D = tree density per hectare, P = pi, dmax = distance to the furthest of the 10 trees (m). Equation 5.1: D = 100 000/ P . (dmax) 2 At SACHA further habitat details were recorded in order to determine habitat relationships of the bamboo specialists located there. Within the 30 m radius of each of the 30 census stations, an 8m x 8m quadrat was randomly located. A team of four observers then walked through each quadrat, recording the following variables: total number of bamboo stems; the number of live and dead bamboo stems; the total number of non-bamboo stems > 0.02m in dbh; the number of live and dead non-bamboo stems; and the dbh of all bamboo stems (live and dead). |