On Measuring Bird Habitat: Influence of Observer Variability and Sample Size

Abstract

We studied the effects of observer variability when estimating vegetation characteristics at 75 0.04-ha bird plots. Observer estimates were significantly different for 31 of 49 variables. Multivariate analyses showed significant interobserver differences for five of the seven classes of variables studied. Variable classes included the height, number, and diameter of trees, height to the first live tree branch, height and number of shrubs, and composite variables. We then compared observer estimates with measurements of the same habitat variables. Univariate and multivariate comparisons of observer estimates with actual measurements revealed no clear pattern because estimates by each observer tended to deviate unpredictably from different measured values for 21 variables. Sample size requirements for selected variables ranged from 20 to 50 for measurements and from 20 to >75 for estimates. We noted significant differences in the point estimates and associated levels of precision between the two methods. Consequently, studies that rely on ocular estimates might sacrifice accuracy in lieu of potential time and cost savings.