• Chen, Cen
  University of Maine, School of Forest Resources
  
• Weiskittel, Aaron
  University of Maine, School of Forest Resources
  
• Bataineh, Mohammad
  University of Arkansas at Monticello
  
• MacLean, David A.
  University of New Brunswick
  

Insect defoliation reduces the growth and survival of trees. Evaluating these effects on trees requires understandings of the variation and dynamics of defoliation, which has been limited by the coarseness of analytical scales. This is especially the case for defoliation caused by spruce budworm (SBW; Choristoneura fumiferana (Clem.)), the primary forest defoliator in North America. In this study, we developed Bayesian models based on a Markov chain Monte Carlo technique to evaluate patterns of SBW defoliation by predicting individual tree defoliation using stand-level measurements of defoliation that are potentially more efficient to obtain through remote sensing. Additionally, the temporal development of individual tree defoliation was also analysed by the same modelling approach. Data containing over 47 000 observations of individual tree defoliation collected during the last SBW outbreak in the 1970s–1980s from an extensive network of permanent sample plots in Maine, US and New Brunswick, Canada were used in model development. Our results demonstrated that the variation in individual tree defoliation was predominantly dependent on species, while all the other examined tree-, stand- and site-level characteristics had a more limited influence individually or even in combination. Despite significant species-specific differences in magnitude, defoliation of both balsam fir (Abies balsamea L.) and red/black spruce (Picea rubens Sarg., Picea mariana (Mill.) B.S.P.) developed towards their respective converged trajectories regardless of differences in initial defoliation, and other examined tree-, stand- and site-level characteristics. These findings were consistent between Maine and New Brunswick despite varying forest management history and species composition. Overall, the results highlight the high variability in SBW defoliation, while the developed modelling framework should be extendable to other regions and other forms of defoliation.