Point counts outperform line transects when sampling birds along routes in South African protected areas
Bird atlases have become an important source of distribution data for broad-scale analyses in ecology, biogeography and conservation. However, national bird atlases are undertaken in different ways in different countries, usually with little formal assessment of alternatives. Existing research suggests that the differences in accuracy between line and point transects are influenced by context, habitat and other details of the sampling protocol. To determine the best approach to use for southern Africa, where collection of data for atlases must cope with a wide range of field conditions, we compared data from a point-count approach to data from line transects on 302 routes in 19 South African National Parks. Data were analysed using paired sample tests, linear models, qualitative assessment and a Bayesian multispecies occupancy model to estimate species richness, while accounting for imperfect detection. On average, five more bird species were recorded along a route using point counts, even though the line transects covered three times more area. There was no obvious difference in the kinds of species recorded. Occupancy models showed that point counts consistently and substantially outperformed transects in estimating species richness. In transects, low detection probability leads to highly biased and imprecise estimates. For almost identical observer effort, standardised point counts along a route provided significantly higher quality data than transects.
Keywords: sampling, atlas, citizen science, amateur, ornithology, occupancy model