CBL Seminar: Ekaterina Smirnova (VCU)

March 4, 2020 3:30pm iCal Google Calendar

Title: Spatial ecological analysis of Jeffrey pine beetle outbreak dynamics within the Lake Tahoe Basin

Abstract: From 1991 to 1996, Jeffrey pine beetles (Dendroctonus jeffreyi Hopkins) (JPB) caused tree mortality throughout the Lake Tahoe Basin during a severe drought. Census data were collected annually on 10,721 trees to assess patterns of JPB-caused mortality. This represents the most extensive tree-level, spatiotemporal dataset collected to-date documenting bark beetle activity. In the epidemic phase, JPB activity occurred in all topographic positions and caused mortality in spatially expanding clusters, with majority (92–96 %) of mass-attacked trees were within 30 m of a brood (JPB-attacked) tree. Our main goals were: 1) the assessment of characteristics associated with the probability of successful JPB mass-attack and group aggregation behavior that occurred during epidemic outbreak; and 2) the one-year horizon predictive modeling of the JPB outbreak expansion. To address these goals, we employed the combination of tree- and cluster- based spatial analyses. In particular, for the tree-based analysis, we used 1993 epidemic stage data to identify the best subset of forest characteristics variables in JPB-attack logistic regression model and examined the effects of spatial adjustment.  We further validated the proposed model one-year predictive accuracy on the following 1994 epidemic stage data. For the cluster- based analysis, we introduced a novel functional representation approach to describe the complex shape and characteristics of spatial clusters. By expressing the complex cluster regions as functions of the direction from the cluster center, we develop a method for modeling the association between the shape and size of clusters and various cluster attributes. This approach allows to use functional data modeling to quantify the directions of beetle expansion in our problem; similar tools can be used to quantify the directions of microbial concentrations using image density analysis. While results are applied to the forest ecological problem, methodological approaches are rather general and appear in multiple environmental, imaging and other application areas.