Modeling landscape dynamics and environmental association for spruce mortality

Abstract

This study addresses important issues related to mortality of spruce species (Picea sp.) associate with outbreaks of spruce beetle (Dendroctonus rufipennis Kirby) by 1) modeling large scale landscape dynamics of spruce mortality associated with long-term climate in Colorado and Alaska; 2) modeling environmental association between spruce mortality and small scale environmental covariates including climatic factors. In the first chapter, we review the ecology and etiology of spruce mortality in Colorado. In the second chapter, we evaluate landscape dynamics of spruce mortality at the regional scale of Colorado and Alaska. We used climate transition matrices (CTMs) as a method to assess the influence of climate on spruce forest extent and mortality. We quantify the probabilities of observing spruce forest, spruce mortality, and the mismatches between probabilities for the presence of host and mortality as indicated by differential effects. All values were calculated to populate elements of CTMs. The polynomial functions of ordinary regressive model and spatial autoregressive model were implemented to represent the association between climate zones and the responses. The results show us that there are influences of long-term precipitation and temperature on both probabilities. Presence of spruce forest in Colorado is associated with high precipitation at moderately low temperatures while probability of spruce mortality has a similar association. High probability of observing spruce forest in Alaska is associated with low to moderate precipitation while the probabilities of observing spruce mortality are positively associated with high precipitation at warmer temperatures. From the differential effects, there are mismatches of responses between host and mortality implying the advantageous of host associate with moderate temperatures and high precipitation in Colorado while healthy forest is found in the moderately low temperature and moderate precipitation in Alaska. In the third chapter, we describe associations between stand scale environmental conditions and spruce mortality. We modeled the association using zero-and-one inflated beta regression model based on hierarchical Bayesian framework. Two-stage Bernoulli logistic models were applied to indicate the occurrence of the extreme values represent presence and absence of mortality; continuous proportional responses were then addressed by beta regressive model. Multivariate Gaussian latent process was included in the function to express the exponential spatial errors term. The results indicate that spatial distribution of the occurrence and intensity of spruce mortality were both associated with the local stand covariates of temperature zone, precipitation zone, class of stand structure level, relative dominance class, and size class. The colder temperature zones have highly negative effects on both the probability of mortality occurrence and the probability of full mortality occurrence, while the warmer temperature zone is positively associated with the presence of full mortality. The results also indicate that stand characteristics are important factors associated with mortality. Mortality occurrence is positively associated with single-story stands with medium to large size classes. The higher-complexity stand structures have highly positive associations with the probability of entire stand mortality, while medium to high dominance classes have negative effects on full mortality. The largest size class and the highest dominance class have negative associations with the proportion of partial mortality.