Disentangling complex fine‐scale ecological patterns by path modelling using GLMM and GIS

Abstract

AbstractQuestion: How can statistical modelling tools (GLMM) and GIS be used as an aid in understanding complex ecological patterns? This general question was approached by using bryophyte demography data as an example. More specifically, we asked what is the contribution of terrain shape to explaining the performance and fate of plant individuals, controlling for all other known relationships?Location: Norway.Methods: Information on demography was obtained for 140 populations of the perennial clonal bryophyte Hylocomium splendens in Norway spruce forests during an 11‐year period (1992‐2002). Performance (size and branching pattern) was recorded for mature segments and fate was recorded for growing points. Positions of each of the more than 30 000 recorded bryophyte ramets were coupled with (micro‐) topographic characteristics (slope and convexity) derived from fine‐scale digital elevation models in a GIS framework. Carefully planned sequences of generalised linear mixed models (GLMM) were performed to test predictions from a conceptual path model.Results: We demonstrate strong dependence of size on branching, fate and on vertical position in the bryophyte carpet, and an effect of vertical position on branching pattern. Micro‐topography contributed to explaining plant performance by four different mechanisms: (1) a direct effect of slope on the segment's vertical position in the carpet; (2‐3) direct effects of both slope and convexity on fates of individuals via controls on risk of burial; and (4) an indirect effect of convexity on branching pattern via a direct effect on size. No indication of a direct effect of terrain on branching was found.Conclusions: Our study exemplifies the usefulness of GLMM for disentangling complex ecological relationships. Specifically, we recognise micro‐topography as a potentially important factor for plant demography in general and for performance and fate of individuals in particular.