Forest insect and disease pests alter the appearance of trees, thereby impacting visual resources. Because of the complexity of most forest landscapes, the degree of visual impact of pest-infested forest stands is difficult to quantify. This paper describes a method of measuring visual impacts of pest-infested forest stands. Photographs of healthy Ponderosa pine trees were entered into a computer video-image-processing system. Using this system, images of trees were altered to simulate different degrees of infestation by limb rust, a forest pathogen. The altered and unaltered images were shown to groups of observers who rated the scenes in terms of "scenic beauty." The great majority of individuals were able to detect a change in the appearance of trees infected with limb rust disease even when only small sections of a tree were altered. There was also general agreement within the groups of observers that the presence of limb rust disease had a detrimental effect on the visual quality of the forest scenes. The tests also suggested that the location of infestation in the tree crown, the amount of crown mortality, and the number of infected Ponderosa pine in a stand influenced the degree to which visual quality was impacted.

This dissertation is concerned with effects dispersal and forest structure on forest insect population dynamics, and with identifying generating processes by comparing observed patterns to model predictions. In chapter 2, we investigated effects of changing forest landscape patterns on integro-difference models of host-parasitoid population dynamics. We demonstrated that removing habitat can increase herbivore density when herbivores don't disperse far, and parasitoids disperse further, due to differences in dispersal success between trophic levels. This is a novel potential explanation for why forest fragmentation increases the duration of forest tent caterpillar outbreaks. To better understand spatial model behaviour, we proposed a new local variation of the dispersal success approximation. The approximation successfully predicts effects of habitat loss and fragmentation on realistically complex landscapes, except when outbreak cycle amplitude is very large. Local dispersal success is useful in part because parameters can be estimated from widely available habitat data. In chapter 3, we investigated how well a discretized integro-difference model of mountain pine beetle population dynamics predicted the occurrence of new infestations in British Columbia. We found that a model with a large dispersal kernel, and high emigration from new, low severity infestations yielded the best predictions. However, we do not believe this to be convincing evidence that many beetles disperse from new, low severity infestations. Rather, we argued that differences in habitat quality, detection errors, and Moran effects can all confound dispersal patterns, making it difficult to infer dispersal parameters from observed infestation patterns. Nonetheless, predicting infestation risk is useful, and large kernels improve predictions. In chapter 4, we used generalized linear mixed models to characterize spatial and temporal variation in the propensity of jack pine trees to produce pollen cones, and account for confounding effects on the relationship between pollen cone production and previous defoliation by jack pine budworm. We found effects of stand age, and synchronous variation in pollen cone production among years. Accounting for background patterns in pollen cone production clarified that pollen cone production declines in with previous defoliation, as expected.

This dissertation is concerned with effects dispersal and forest structure on forest insect population dynamics, and with identifying generating processes by comparing observed patterns to model predictions. In chapter 2, we investigated effects of changing forest landscape patterns on integro-difference models of host-parasitoid population dynamics. We demonstrated that removing habitat can increase herbivore density when herbivores don't disperse far, and parasitoids disperse further, due to differences in dispersal success between trophic levels. This is a novel potential explanation for why forest fragmentation increases the duration of forest tent caterpillar outbreaks. To better understand spatial model behaviour, we proposed a new local variation of the dispersal success approximation. The approximation successfully predicts effects of habitat loss and fragmentation on realistically complex landscapes, except when outbreak cycle amplitude is very large. Local dispersal success is useful in part because parameters can be estimated from widely available habitat data. In chapter 3, we investigated how well a discretized integro-difference model of mountain pine beetle population dynamics predicted the occurrence of new infestations in British Columbia. We found that a model with a large dispersal kernel, and high emigration from new, low severity infestations yielded the best predictions. However, we do not believe this to be convincing evidence that many beetles disperse from new, low severity infestations. Rather, we argued that differences in habitat quality, detection errors, and Moran effects can all confound dispersal patterns, making it difficult to infer dispersal parameters from observed infestation patterns. Nonetheless, predicting infestation risk is useful, and large kernels improve predictions. In chapter 4, we used generalized linear mixed models to characterize spatial and temporal variation in the propensity of jack pine trees to produce pollen cones, and account for confounding effects on the relationship between pollen cone production and previous defoliation by jack pine budworm. We found effects of stand age, and synchronous variation in pollen cone production among years. Accounting for background patterns in pollen cone production clarified that pollen cone production declines in with previous defoliation, as expected.