Changes in the distribution of species between two bird atlas surveys: the analysis of covariance after controlling for research activity

Abstract

Heterogeneous coverage ofdata is a serious problem in atlas mappings, because it may bias the interpretation of maps and cause difficulties in statistical tests of distribution changes of species between successive atlas surveys. The areal coverage of the Finnish bird atlases during the years 1974-1979 and 1986-1989 was evaluated by the research grade of 10 x 10-km atlas squares, using six categories (0 = no data, 1 = occasional observations, 2 = fair survey, 3 = satisfactory survey, 4 = well-surveyed, and 5 = thoroughly surveyed square) . The research grade of each atlas square was calculated by applying regression analysis on the sum of breeding evidences of species, coded 1 = possible, 2 = probable and 3 = confirmed breeding. Also, the amount of land area, latitude and the presence of coastal habitats in the square were taken into account. The analysis of covariance was applied to analyse changes in the distribution of species, using only such squares that were studied at least fairly well in both atlases . The squares were evened out to represent "equal" research activity by using the research grade as covariate. Statistical significance was calculated for (a) changes in the distribution of each species between the atlas surveys and (b) the areal homogeneity of the change . In 20-25% of Finnish breeding birds the increase in research activity between the atlas surveys biased the analysis . To keep the data comparable, conservatism is suggested for the reforms of atlas methodology.