Rodriguez el at. • VENEZUELAN BIRD SURVEYS 
231 
Caribbean Sea 
Colombia 
NorthCoast 
| NWCordilleras 
Llanos 
[ | Guayana 
Brazil 
FIG. I. Venezuela showing location and identification number of the 27 study cells. NeoMaps numbers, names, and 
ecoregions are in Table 1. 
behaviors (e.g., Karr 1971, 1981: Terborgh el al. 
1990: Wunderle 1994; Casagrande and Beissinger 
1997; Poulsen el al. 1997; Herzog cl al. 2002; 
Rompre el al. 2007). These techniques lend lo be 
designed for monitoring populations intensively at 
particular locations with high costs in terms of 
time, effort, and funds, making them impractical 
for monitoring over large tropical regions (Mar- 
gules and Redhead 1995). Past and recent efforts 
in Venezuela have included bird inventories, but 
most have focused either on a single location or a 
unique taxon or guild (e.g., Casagrande and 
Beissinger 1997, Terborgh et al. 1997, CJiner F 
2001. Lasso el al. 2006. Martinez 2008, Seharis et 
al. 2008, Lentino et al. 2009, Sanz et al. 2010, 
Vilella el al. 2010), and none has been conducted 
at the country scale. 
Rodriguez et al. (2007) used three different 
combinations of numbers of points and count 
duration as part of the Neotropical Biodiversity 
Mapping Initiative, NeoMaps (Rodriguez and 
Sharpe 2002, Ferrer-Paris et al. 2011) with the 
goal of testing and adapting the BBS protocol to 
the tropics; they concluded the combination of 50 
3-min point counts also appeared to be the most 
efficient in the tropics. Our objective was to 
develop and test the performance of a large-scale 
survey in characterizing the avifauna of Vene¬ 
zuela necessary for a systematic, quick and 
low-cost method for obtaining reliable data on 
bird richness and abundance at the country level. 
METHODS 
Sampling Design. —Our sampling design was 
based on the stratified spatial sampling design 
proposed by the NeoMaps Initiative (Rodriguez 
and Sharpe 2002, Ferrer-Paris et al. 2011). We 
first superimposed u 0.5 X 0.5 degree grid over a 
map of Venezuela (~5() km on a side, or 
—2.500 knr/cell). The sampling universe was 
defined as 177 from a total of 377 cells that had a 
minimum of 90 knr accessible by secondary 
roads (at least by 4-whecl drive vehicles). We 
applied a principal component analysis (PCA) 
within this sampling universe, based on 14 
environmental variables to define three orthogo¬ 
nal environmental axes (physical-climatic, forest 
cover, and drought intensity). A subset of 27 cells 
was selected to represent the range of environ¬ 
mental variation in the PCA within the major 
biogeographical regions in the country (Rodriguez 
2003; JRF, unpubi. data) (Fig. I ). 
Sampling Transects.—We identified a 40-km 
section of a secondary or tertiary road in each of 
27 cells. We tried to avoid urban centers and roads 
in poor condition, which would make driving 
during the surveys difficult or slow. This was not 
completely possible, as in some cells there was 
