Quantifying sources of variation in the frequency of fungi associated with spruce beetles: Implications for hypothesis testing and sampling methodology in bark beetle-symbiont relationships

The spruce beetle, Dendroctonus rufipennis (Kirby), causes landscape level mortality to mature spruce (Picea spp.) throughout western and northern North America. As with other bark beetles, this beetle is associated with a variety of fungi, whose ecological functions are largely unknown. It has been proposed that the relative frequencies of specific fungi associated with bark beetles may vary with ecological factors such as host species, climate, or beetle population phase. We collected approximate to,1000 adult spruce beetles in south-central Alaska from 1999 to 2001. We employed a variety of insect collection and microbial isolation techniques during year 1 to devise optimal conditions. In the latter 2 years, we sampled live adults excavated from overwintering galleries, and isolated fungi by dragging beetles across malt agar amended with gentamicin. We obtained 10 fungal species. We used a multilevel generalized linear mixed model to (a) develop estimates of the prevalence of each isolated fungal species in the beetle population; (b) explore factors that might explain the frequencies of association of specific fungi with adult spruce beetles, such as insect population phase or positive or negative associations with other fungi, and (c) partition the relative sources of variation in beetle-fungal associations between the hierarchical, random effect variables (i.e., nested individual insect, collecting vessel, tree, and site variables). We implemented this model using three procedures with different computational algorithms within the commonly used software packages R and SAS, and compared the results. The most prevalent fungus was Leptographium abietinum, which was recovered from approximately 80% of beetles. The frequency of association of both L abietinum and a Pesotum species varied with insect population phase in 2001. The presence of Pesotunt spp. on spruce beetles was negatively associated with the presence of L. abietinum. The variance components models indicated that there was significant variation among beetle-fungal associations within trees and trees within sites. Site was generally the smallest source of variation, but estimates were not very precise. We demonstrate how these estimates of variation can be used to design practical sampling protocols to test other hypotheses. This analysis also highlights how attention must be given to understanding experimental design and analysis, and how conflicting conclusions can emerge from hypothesis testing of ecological factors. Our model can be extended as a general analytical approach for future hypothesis testing of bark beetle-fungal associations. (c) 2005 Elsevier B.V. All rights reserved.