Statistical Analysis of Bioindicator Value of Epiphytic Lichens

Abstract

Generalized linear modelling was applied in analysing the bioindicator value of epiphytic lichens for air pollution around an industrial centre in the Siberian mountains using data published previously by Zobel (1988a). The occurrence of taxa on tree boles was systematically related to the distance from the pollution source and to the altitude from sea level. Binomial distribution was postulated to describe the random variation and maximum likelihood and quasi-likelihood methods were used. Zobel had pooled the original data by grouping distance and altitude into three qualitative levels. The model for pooled data with qualitative factors generally had a poor fit. We assigned quantitative values, linear and quadratic contrasts, to the altitude and distance levels and a linear by linear interaction term between altitude and distance. In most taxa the interaction appeared significant, and these models had generally a better fit and smaller overdispersion than qualitative models. Finally, more effective analysis was performed on the original plotwise data utilizing unpublished quantitative information for distance and altitude. Most species (16/19) had a significant and usually positively monotone response to distance. In the majority of these (10/16) the effect of distance varied between the altitude levels most often so that the lichens were scarcer at low altitude and their increase was faster at high altitude.