During the setting up of the experiment in 1994, a control transplantation was made at the site of lichen collection, Skånberget, Ramsjö, in the province of Hälsingland, in south boreal Sweden, ca 300 km north of the experimental area. On the north and south sides of 20 trees material of two types was mounted, such that had been frozen for more than one month, i.e. resembling the treatment in the experiment, and also fresh material, in total amounting to 80 transplants. The survival and vitality of these transplants were re-assessed in August 2008. Generalized linear mixed models (GLMMs) with logit link

functions and Laplace approximation (Bolker et al., 2009) were first applied to test the effect of tree retention, aspect, and transplantation time for transplant survival and vitality

in 2008, check details and R428 research buy second to assess if there was a significant difference in the variables that described survival and vitality in both survey years. The effect of tree retention was tested in two different models, one testing if transplant survival and vitality differed between trees in the forest and clearcut, and the second one testing if there was a difference in transplant survival and vitality between grouped and scattered retention trees. The following binary response variables (1/0) were used: survival was defined as the transplanted thallus being present (1) or absent (0), and vitality as ⩾50% of the thallus being vital (1) or <50% of the thallus being vital (0). The global start model for the data of 2008 included forest stand and tree as random factors, and aspect (north or south), forest type (forest or clearcut) or clearcut type (grouped or scattered retention trees), tree diameter (measured in 1996), and transplantation time (spring 1994 or autumn 1994) as fixed effect variables. In the second model, survey year (1996 or 2008) was used as an additional fixed effect variable.

Tree diameter was not used in this model since we were not interested if the effect of tree diameter had changed between both survey years. For better comparison between the two survey years we also tested a third model, including only the data of 1996, but running the model in the same way as described for the data of 2008. This was done since the data analysis SPTLC1 in Hazell and Gustafsson (1999) used a different statistical approach. Biological meaningful interaction terms were added and all fixed explanatory variables in the interaction terms were centered and scaled (in the case of tree diameter) in order to achieve biologically interpretable estimates (Schielzeth, 2010). Akaike’s Information Criterion (AIC, or AICc for small sample sizes) and Akaike weights were used to assess the relative strength of support for all biologically considerable models, given the chosen explanatory variables (Akaike, 1974 and Burnham and Anderson, 2002).