First-time observer effects in the North American Breeding Bird Survey

Currently the operational analysis of Breeding Bird Survey (BBS) data by the National Biological Service accounts for observer differences in estimating the trend for each route, but within-observer differences are not modeled. We tested for the existence of a form of within-observer differences in skill level, namely a change in ability to count birds of a given species after an observer's first year on a given route. An increase in ability could positively bias the trend estimate. Removal of an observer's first year of observation on each route for the period 1966 to 1991 resulted in lower average unweighted trend estimates for 415 of 459 species (90%). These reductions were statistically significant for 213 species (46%). The average reduction in trend was 1.8% change per year (SD = 5.4%). In route-regression analysis, route data are weighted by a measure of precision. Removing first-year observer counts reduced the weighted trend estimate for 275 of 416 species (66%), but differences generally were small.