Skip to main content
Dryad logo

Data from: Widespread bird species show idiosyncratic responses in residual body mass to selective logging and edge effects in the Colombian Western Andes


Jones, Harrison; Colorado, Gabriel; Robinson, Scott (2022), Data from: Widespread bird species show idiosyncratic responses in residual body mass to selective logging and edge effects in the Colombian Western Andes, Dryad, Dataset,


This dataset consists of banding data (N = 1589 captures, 129 bird species), including morphological and breeding biology measures, collected from understory birds in subtropical cloudforest at roughly 2000 m.a.s.l. in the municipality of El Cairo in Colombia's Western Andes (Serrania de los Paraguas range). Roughly 8,350 net hours were divided into two three-month field seasons (June-August 2017 and January-March 2018), both corresponding to local dry seasons. Birds were banded along 500-meter transects in forest interior across a gradient of forest fragment patch sizes (N = 8 fragments, area range = 10-173 ha) and a private ~750 ha reserve (Reserva Natural Cerro El Ingles) in the same landscape. Each non-hummingbird capture was fitted with an aluminum leg band with a unique combination, and we collected data on mass, tarsus length, wing chord, pectoral muscle score, cloacal protuberance and brood patch score, and age and sex (where possible due to plumage differences). We captured 101 species (including 6 boreal migrants) in 844 captures (53%) during the January-March sampling period and 102 species in 745 captures during the June-August sampling period; each site was sampled during both sampling periods. The most captured bird families were Trochilidae (24 spp., 36% of total captures), Thraupidae (17 spp.), Tyrannidae (16 spp.), Furnariidae (14 spp.), and Turdidae (6 spp.).

A subset of these data (uploaded as a separate file) were used to calculate body condition indices for 20 species of commonly captured birds (N = 984 captures) that occured across much of the patch size gradient (at least 8 out of 14 transects), and were related to fragment area and measures of edge and selective logging effects (see Jones et al. 2022). In this file, we include all predictor variables necessary to run the generalized linear mixed models in Jones et al. (2022), including proportion forest cover within 1 km (a proxy for patch size), edge density, average distance to forest edge from the netting transect, average canopy cover and height along the transect, multivariate measures of understory plant density, large-diameter tree density, and vertical vegetation structure, elevation of each transect, and average yearly rainfall for each transect.


Study sites

Study sites were located on the east slope of the Serrania de los Paraguas range in the Colombian Western Andes, specifically in the municipality of El Cairo (4°45′39″ N, 76°13′21″ W), Valle del Cauca department. The landscape consisted of a typical patchwork of forest fragments, cattle pasture, and shade coffee plantations, with ongoing forest clearing for the latter activities. Natural forest cover in this region is subtropical Andean forest, characterized by over 200 tree species, abundant epiphytes, and ~20-m canopies. Within this landscape, we identified all isolated forest fragments surrounded by cattle pasture and in the 1900-2300 m.a.s.l. elevational band using satellite images in Google Earth. We stratified this subset of fragments into large (≥ 100 ha), medium (30-50 ha), and small (≤ 20 ha) size categories, before selecting at least two of each to survey in detail (N = 8 fragments, range = 10-173 ha). Selected fragments were mid- to late-successional forest patches, which we defined as having >10 m canopies, trees of >10 cm diameter at breast height (DBH), a closed canopy, and a diversity of tree age classes. We only sampled two small-sized fragments because fragment size tended to increase with elevation in the focal landscape and because we wished to spend approximately equal sample effort at each site (limiting the number of surveys possible within the field season). We additionally sampled a private forest reserve within the same landscape (RNC Cerro El Inglés, ~750 ha) as a reference site. This site is connected to thousands of hectares of continuous forest along the spine of the Western Andes and in the Chocó lowlands to the west. Forest fragments were private lands (sampled with the help of a local NGO: Serraniagua) which varied in their land-use histories, particularly the intensity of selective logging, a common practice in the Colombian Andes (Aubad et al. 2008).  

Quantification of Patch Size, Edge Effects, and Rainfall

We used a buffer analysis to quantify landscape composition and configuration within a km of each sampling transect. This scale was selected because it affected the occupancy of tropical bird communities in lowland studies (Carrara et al. 2015) and landscape composition at this scale was highly correlated with the composition at other buffer scales at our sites (e.g., 500 meters). All landscape analyses were conducted using ArcGIS (ArcMap 10.3.1; Esri, Redlands, CA), specifically the ‘isectpolyrst’ tool in the Geospatial Modelling Environment (version; Beyer 2015). Buffers were centered on the full length of the transect, resulting in a non-circular buffer shape. We used the proportion of forest cover within the buffer region as a proxy for the patch size of the fragment; we did not use patch size measurements directly because our continuous forest reference site had no patch size. We quantified fragment configuration (and edge effects) using the ‘edge density’ measure from Carrara et al. (2015), which is defined as the density of forest edge habitat within the 1 km buffer, measured in meters per hectare. We also calculated a straight-line distance to forest edge for each transect, which was the average of five measurements taken from the center point of each 100-meter transect segment to the nearest forest edge. We measured proportion of forest cover and edge density using a land-cover use map for our study area from the departmental conservation authority (Corporación Autónoma Regional del Valle del Cauca), which we converted to a 25-m cell-size raster. Yearly rainfall data for each transect from the WorldClim2 monthly data set (Fick and Hijmans 2017) were collected at ~1km2 resolution. Sampling transects were superimposed on the global raster of rainfall data. We summed the mean monthly historical rainfall data (1970-2000) for each transect to calculate a yearly average; where a transect overlapped multiple raster squares, monthly values for each were averaged. 

Quantification of Local Vegetation Structure 

For each transect, we measured local vegetation structure and density to quantify the effects of human disturbance, particularly selective logging. Vegetation measurements were taken from June-August 2017, though we observed little annual variation. We followed the sampling methodology of Stratford and Stouffer (2013), which we modified to be used along belt transects. The methodology was broadly comprised of two components: (1) the quantification of canopy height, percentage canopy cover, and foliage height diversity using point sampling situated every 10 meters along the transect and (2) the quantification of tree size category and understory vegetation density, respectively, using 3-meter-wide belt sampling. Because transects ran along trails, we measured vegetation at least three meters from the trail edge on a randomly selected side of the trail. For the point sampling, we measured variables at ten-meter intervals, for 50 points per transect. As a measure of foliage height diversity, we noted the presence or absence of live vegetation in five height bands: <0.5 m, >0.5–3 m, >3–10 m, >10–20 m, and >20 m. We used a laser rangefinder (Raider 550; Redfield, Beaverton, OR) to determine heights above 3 meters, sighting through a tube with crosshairs while straddling the point. For each point, we also recorded the highest canopy height, to the nearest meter. The proportion of canopy cover at each point was calculated to the nearest eighth of the field of view using a vertical densiometer (Densitometer; Geographic Resource Solutions, Arcata, CA). For each transect, we averaged values for both canopy height and canopy cover and calculated the proportion of points at which vegetation was present for each height category. To quantify foliage height diversity, we calculated the Shannon diversity index of the proportion of points with vegetation present in each of the five height bands for each transect. Foliage height diversity was highly correlated with canopy height at our sites (Pearson’s r = 0.90), so we retained foliage height diversity.

For the belt sampling, we surveyed all trees (woody vegetation > 2 m in height) on 1.5 meters to either side of the observer and measured their DBH. Trees were later categorized into six DBH size classes for analysis: 1-7 cm, 8-15 cm, 16-23 cm, 24-30 cm, 31-50 cm, and > 50 cm. We also recorded the largest tree DBH recorded on each transect. Because selective logging targets large, old-growth trees we consider this to be a proxy measure for current and historical logging pressure at each site. To capture differences in understory vegetation density, we also recorded the density of shrubs, ferns, lianas (vines), palms, and tree ferns along the same belt survey for each transect. To reduce redundancy and minimize correlation between variables, we used ordinated measures of the tree DBH and understory vegetation data for each transect from respective principal component analyses (PCAs), taken from Jones and Robinson (2020). In each case, we used the first principal component axis; greater values indicate higher densities of large-diameter trees and understory vegetation, respectively. We inverted the sign of the understory vegetation PC axis so that larger values would indicate higher densities for ease of interpretation of results. 

Mist Netting and Morphological Measurements

We surveyed understory bird communities along each transect using passive mist netting. Each site was sampled twice, from June-August 2017 and from January-March 2018; both correspond to a local dry season. In each case, we surveyed transects for two and a half consecutive field days, and staggered visits to small, medium, and large fragments to avoid any correlation between fragment size and day of year. We deployed twelve 12x3 m mist nets (38 mm mesh; Avinet Research Supplies, Portland, ME) along each transect, placed in locations likely to capture understory birds (i.e., with dense understory vegetation and outside of direct sunlight). Nets were open from dawn to dusk (~0700-1700 hrs.) and were operated by two or more technicians. We closed nets during periods of heavy rainfall and high winds. All captures were brought to a central banding station where they were identified to species and fitted with an aluminum leg band with a unique number before being released. We did not band hummingbirds (Trochilidae), and instead cut the tip off a unique rectrix for each capture at a site. For each unique capture, we measured the length of the right tarsus and wing chord to the nearest millimeter; tarsus length was not recorded for hummingbirds. We also recorded body mass for each capture. For the first field season, mass was measured in a bag using a spring scale (± 0.5 g.; Pesola Precision Scales, Schindellegi, Switzerland). In the second field season, we used a more precise digital scale (± 0.01 g.; American Weigh Scales, Cummingham, GA), with the focal individual placed in a measuring cone. We also evaluated the breeding condition of all non-hummingbird captures by scoring the presence and extent of a cloacal protuberance (0-3 scale) and brood patch (0-4 scale) according to the criteria in Pyle (1997). Finally, we identified juvenile individuals, where possible, using a combination of plumage differences, molt limits, and enlarged or colorful rictal flanges. Plumage differences and molt limits are poorly described for our focal species, however.


Animal Behavior Society

Tinker Foundation

Katherine Ordway Endowment for Ecosystem Conservation