Tom Swystun's research while affiliated with Natural Resources Canada and other places

Communities interspersed throughout the Canadian wildland are threatened by fires that have become bigger and more frequent in some parts of the country in recent decades. Identifying the fireshed (source area) and pathways from which wildland fire may ignite and spread from the landscape to a community is crucial for risk‐reduction strategy and planning. We used outputs from a fire simulation model, including fire polygons and rate of spread, to map firesheds, fire pathways and corridors and spread distances for 1980 communities in the forested areas of Canada. We found fireshed sizes are larger in the north, where the mean distances between ecumene and fireshed perimeters were greater than 10 km. The Rayleigh Z test indicated that simulated fires around a large proportion of communities show significant directional trends, and these trends are stronger in the Boreal Plains and Shields than in the Rocky Mountain area. The average distance from which fire, when spreading at the maximum simulated rate, could reach the community perimeter was approximately 5, 12 and 18 km in 1, 2 and 3 days, respectively. The average daily spread distances increased latitudinally, from south to north. Spread distances were the shortest in the Pacific Maritime, Atlantic Maritime and Boreal Plains Ecozones, implying lower rates of spread compared to the rest of the country. The fire corridors generated from random ignitions and from ignitions predicted from local fire history differ, indicating that factors other than fuel (e.g. fire weather, ignition pattern) play a significant role in determining the direction that fires burn into a community.

A spread day is defined as a day in which fires grow a substantial amount of area; such days usually occur during high or extreme fire weather conditions. The identification and prediction of a spread day based on fire weather conditions could help both our understanding of fire regimes as well as forecasting and managing fires operationally. This study explores the relationships between fire weather and spread days in the forested areas of Canada by spatially and temporally matching a daily fire growth database to a daily gridded fire weather database that spans from 2001 to 2019. By examining the correlations between spread day fire weather conditions and location, conifer coverage (%), and elevation, we found that a spread day happens under less severe fire weather conditions as latitude increases for the entire study area and as conifer coverage increases within non-mountainous study areas. In the western mountain areas, however, with increasing conifer coverage more severe fire weather conditions are required for a spread day to occur. Using two modeling approaches, we were able to identify spread day indicators (generalized additive model) and to predict the occurrence of spread days (semi-binomial regression model) by Canadian Ecozones both annually and seasonally. Overall, Fine Fuel Moisture Code (FFMC), Initial Spread Index (ISI), and Fire Weather Index (FWI) performed the best in all models built for spread day identification and prediction but varied depending on the conditions mentioned above. FFMC was the most consistent across all spatial and temporal scales.

Great efforts have been made to understand the impacts of a changing climate on fire activity; however, a reliable approach with high prediction confidence has yet to be found. By establishing linkages between the longest duration of fire-conducive weather spell and fire activity parameters, this study projected annual area burned (AAB), annual number of fires (ANF), and annual maximum fire size (MFS) into the future. We found that even though the rates of change differ, the spatial pattern of changes for all three parameters are similar by Canadian ecozone. Areas with the lowest fire activity may see higher rates of change in comparison to high fire activity areas. By end of the century, the changes of AAB and MFS for the study area are projected to be about four and five times that of the baseline respectively, while ANF could almost double.

Understanding climate change impacts on wildland fire activity has been constrained by the high uncertainty embedded in the prediction of fire size (FS), annual number of fires (ANF), and annual area burned (AAB). While there has been a sustained effort to make connections between fire weather and fire activity, most studies have focused on individual parameters instead of treating them as a connected group. This study explores the intrinsic relationships among the major parameters of fire activity and how they relate to fire-conducive weather conditions to determine optimal prediction models. We found maximum number of fire spread days and maximum FS best predict ANF and AAB, respectively. Assessing the robustness of these relationships across Canada’s ecozones showed they are stronger in the Cordillera than in the Shields and Plains and more universal for AAB than for ANF. We also found skewness of FS distributions may play an important role in relationship strength. These relationships provide a unique way to model future fire activities under changing climate conditions.

Forest understory vegetation is an important characteristic of the forest. Predicting and mapping understory is a critical need for forest management and conservation planning, but it has proved difficult with available methods to date. LiDAR has the potential to generate remotely sensed forest understory structure data, but this potential has yet to be fully validated. Our objective was to examine the capacity of LiDAR point cloud data to predict forest understory cover. We modeled ground-based observations of understory structure in three vertical strata (0.5 m to < 1.5 m, 1.5 m to < 2.5 m, 2.5 m to < 3.5 m) as a function of a variety of LiDAR metrics using both mixed-effects and Random Forest models. We compared four understory LiDAR metrics designed to control for the spatial heterogeneity of sampling density. The four metrics were highly correlated and they all produced high values of variance explained in mixed-effects models. The top-ranked model used a voxel-based understory metric along with vertical stratum (Akaike weight = 1, explained variance = 87%, cross-validation error = 15.6%). We found evidence of occlusion of LiDAR pulses in the lowest stratum but no evidence that the occlusion influenced the predictability of understory structure. The Random Forest model results were consistent with those of the mixed-effects models, in that all four understory LiDAR metrics were identified as important, along with vertical stratum. The Random Forest model explained 74.4% of the variance, but had a lower cross-validation error of 12.9%. We conclude that the best approach to predict understory structure is using the mixed-effects model with the voxel-based understory LiDAR metric along with vertical stratum, because it yielded the highest explained variance with the fewest number of variables. However, results show that other understory LiDAR metrics (fractional cover, normalized cover and leaf area density) would still be effective in mixed-effects and Random Forest modelling approaches.

Forest understory vegetation is an important feature of wildlife habitat among other things. Predicting and mapping understory is a critical need for forest management and conservation planning, but it has proved difficult. LiDAR has the potential to generate remotely sensed forest understory structure data, yet this potential has to be fully validated. Our objective was to examine the capacity of LiDAR point cloud data to predict forest understory cover. We modeled ground-based observations of understory structure in three vertical strata (0.5 m to < 1.5 m, 1.5 m to < 2.5 m, 2.5 m to < 3.5 m) as a function of a variety of LiDAR metrics using both mixed-effects and Random Forest models. We compared four understory LiDAR metrics designed to control for the spatial heterogeneity of sampling density. The four metrics were highly correlated and they all produced high values of variance explained in mixed-effects models. The top-ranked model used a voxel-based understory metric along with vertical stratum (Akaike weight = 1, explained variance = 87%, SMAPE=15.6%). We found evidence of occlusion of LiDAR pulses in the lowest stratum but no evidence that the occlusion influenced the predictability of understory structure. The Random Forest model results were consistent with those of the mixed-effects models, in that all four understory LiDAR metrics were identified as important, along with vertical stratum. The Random Forest model explained 74.4% of the variance, but had a lower cross-validation error of 12.9%. Based on these results, we conclude that the best approach to predict understory structure is using the mixed-effects model with the voxel-based understory LiDAR metric along with vertical stratum, but that other understory LiDAR metrics (fractional cover, normalized cover and leaf area density) would still be effective in mixed-effects and Random Forest modelling approaches.

Surveillance programs to detect alien invasive pests seek to find them as soon as possible, but also to minimize the cost of damage from invasion. To examine the trade-offs between these objectives, we developed an economic model that allocates survey sites to minimize either the expected mitigation costs or the expected time until first detection of an invasive alien pest subject to a budget constraint on surveillance costs. We also examined strategies preferred by ambiguity-averse decision makers that minimize the expected and worst-case outcomes of each performance measure. We applied the model to the problem of detecting Asian longhorned beetle (Anoplophora glabripennis) in the Greater Toronto Area, Canada, one of the most harmful invasive alien insects in North America. When minimizing expected mitigation costs or expected time to detection, the trade-off between these survey objectives was small. Strategies that minimize the worst-case mitigation costs differed sharply and surveyed sites with high host densities using high sampling intensities whereas strategies that minimize the worst detection times surveyed sites across the entire area using low sampling intensities. Our results suggest that preferences for minimizing mitigation costs or time to detection are more consequential for ambiguity-averse managers than they are for risk-neutral decision-makers.