Point counts of landbirds in bottomland hardwood forests of the Mississippi Alluvial Valley: How long and how many?

Abstract

To quantify efficacy of point count sampling in bottomland hardwood forests, we examined the influence of point count duration on corresponding estimates of number of individuals and species recorded. To accomplish this we conducted a totalof 82 point counts 7 May-16 May 1992distributed among three habitats (Wet, Mesic, Dry) in each of three regions within the lower Mississippi Alluvial Valley (MAV). Each point count consisted of recording the number of individual birds (all species) seen or heard during the initial three minutes and per each minute thereafter for a period totaling ten minutes. In addition, we included 384 point counts recorded during an 8-week period in each of 3 years (1985-1987) among 56 randomly-selected forest patches within the bottomlands of western Tennessee. Each point count consisted of recording the number of individuals (excluding migrating species) during each of four, 5 minute intervals for a period totaling 20 minutes. To estimate minimum sample size, we determined sampling variation at each level (region, habitat, and locality) with the 82 point counts from the lower (MAV) and applied the procedures of Neter and Wasserman (1974:493; Applied linear statistical models). Neither the cumulative number of individuals nor number of species per sampling interval attained an asymptote after 10 or 20 minutes of sampling. For western Tennessee bottomlands, total individual and species counts relative to point count duration were similar among years and comparable to the pattern observed throughout the lower MAV. Across the MAV, we recorded a total of 1,62 1 birds distributed among 52 species with the majority (8721/1621) representing 8 species. More birds were recorded within 25-50 m than in either of the other distance categories. There was significant variation in numbers of individuals and species among point counts. For both, significant differences between region and patch (nested within region) occurred; neither habitat nor interaction between habitat and region was significant. For = 0.05 and L3 = 0.10, minimum sample size estimates (per factor level) varied by orders of magnitude depending upon the observed or specified range of desired detectable difference. For observed regional variation, 20 and 40 point counts were required to accommodate variability in total birds (MSE = 9.28) and species (MSE = 3.79), respectively; 25 percent of the mean could be achieved with 5 counts per factor level. Corresponding sample sizes required to detect differences of rarer species (e.g., Wood Thrush) were 500; for common species (e.g., Northern Cardinal) this same level of precision could be achieved with 100 counts.