On the characterization of patterning in spruce budworm time-series data

I outline the “definitional problem” in forest insect outbreak analytics and show how it is related to the “counting problem” in dendroentomology and the “forecasting problem” in forest insect population dynamics, through the ubiquitous presence of non-stationary complex periodicity. Using real-world examples from the spruce budworm ( Choristoneura fumiferana Clem.) system, I show that regardless how outbreak patterning is characterized—whether by peak impact, cycle frequency, interval duration, or interval severity—the distribution in pattern attributes appears to be extremely variable, regardless how the data are processed through definitional filters. I show that this extreme variability is an unavoidable and key feature of the system’s dynamics and argue that it needs to be viewed as an object of study, instead of a nuisance problem to be swept under the rug. The single biggest opportunity for rapid gains in spruce budworm predictive ecology is determining the environmental and ecological factors that separate high-intensity from low-intensity outbreak cycling.