Using classification tree analysis to reveal causes of mortality in an insect population

Abstract

Invasive species pose significant threats to native and managed ecosystems. How- ever, it may not always be possible to perform rigorous, long-term studies on invaders to determine the factors that influence their population dynamics, particu- larly when time and resources are limited. We applied a novel approach to determine factors associated with mortality in larvae of the sawfly Profenusa thomsoni Konow, a leafminer of birch, and a relatively recent invader of urban and rural birch forests in Alaska. Classification tree analysis was applied to reveal relationships between qualitative and quantitative predictor variables and categorical response variables in a large data set of larval mortality observations. 2 We determined the state (living or dead) of sawfly larvae in samples of individual leaves. Each leaf was scored for variables reflecting the intensity of intra-specific competition and leaf quality for leafminers, year of collection and degree-days accumulated were recorded for each sample. We explored the association of these variables with larval state using classification tree analysis. 3 Leafminer mortality was best explained by a combination of competition and resource exhaustion and our analysis revealed a possible advantage to group feeding in young larvae that may explain previously observed patterns of resource over- exploitation in this species. Dead larvae were disproportionately found in smaller leaves, which highlights the potential effect of competition on mortality and suggests that smaller-leaved species of birch will better able to resist leafminer damage. 4 We show that classification tree analysis may be useful in situations where urgency and/or limited resources prohibit traditional life-table studies.