Theses and Dissertations at Montana State University (MSU)
Permanent URI for this collectionhttps://scholarworks.montana.edu/handle/1/733
Browse
1 results
Search Results
Item Spatiotemporal mapping of mountain pine beetle infestation severity and probability of new infestation in the central U.S. Rocky Mountains(Montana State University - Bozeman, College of Agriculture, 2018) Bode, Emma Taylor; Chairperson, Graduate Committee: Rick L. LawrenceSynchronous, widespread, and severe mountain pine beetle (MPB; Dendroctonus ponderosae) outbreaks impacted forests of western North America at unprecedented levels in recent decades. Severe MPB epidemics can degrade ecosystem services and socio-economic assets. Mapping outbreak progression informs mitigation efforts and enables analysis of MPB attack processes on a landscape scale. Existing time-series methods for mapping disturbance focus on extent rather than severity. Infestation severity, expressed as within-pixel mortality percentage, is more robust for answering a variety of ecological questions. Our objectives were to: (1) map infestation severity from 2005-2015 using a time-series regression; and (2) analyze MPB attack processes by modeling new infestation probability using spatial and environmental variables in the central U.S. Rocky Mountains. We used spectral data from all available Landsat images, topographic data, and data from U.S. Forest Service aerial detection survey (ADS) polygons to model infestation severity. We collected reference data by interpreting National Agricultural Imagery Program images. We then employed logistic regression model-based recursive partitioning (MOB) to determine: (a) to what degree nearby infestation severity increased probability of new infestation; (b) the degree of variation in probability across space and time with respect to other spatial and environmental risk factors; and (c) the extent to which these effects were directional relative to prevailing winds. Validation of our infestation severity model against a randomly selected subset of the data resulted in no statistical difference between predicted and observed severity. Our raster maps allowed us to identify lower severity infestation not recorded by the ADS. The final MOB model obtained 72.1% accuracy in predicting new infestation. Nearby infestation severity strongly influenced the probability of new infestations. This effect varied with elevation, aspect, temperature, phase of the outbreak, and spatial location. Variation in probability of infestation was highest when surrounding infestation severity was low. Use of wind-informed directional effects did not improve the model. This analysis establishes the efficacy of mapping an infestation severity time series and demonstrates that severity maps facilitate novel analyses of MPB attack processes. The processes developed here can support management decisions with timely maps of MPB infestation severity and probability of new infestation.