Wildfire Refugia Within a Boreal Shield Peatland and Rock Barrens Landscape: Identification, Drivers, and Ecohydrological Indicators

Abstract

Fire refugia, defined as unburned, functionally intact patches of habitat within a fire footprint, play an important role in post-fire recovery and landscape resilience to fires. Increased fire activity in the Canadian boreal forest due to climate change highlights the need to properly identify and manage wildfire refugia to protect the natural resilience of boreal ecosystems. While previous fire refugia research has focused on western Canada, we present the first characterization of fire refugia, with a focus on peatland fire refugia, in Ontario. We use remotely sensed multispectral imagery and stereo-derived DEM data from the 2018 Parry Sound 33 wildfire in the Ontario Boreal Shield to determine the primary drivers of fire refugia formation on this landscape, and to develop a model to predict the occurrence of potential fire refugia based on these drivers. We found that the Normalized Difference Moisture Index (NDMI) and the Topographic Position Index (TPI, 200m radius neighbourhood) had the strongest control on wildfire refugia probability in the model, with a combined relative influence of 63.8%. Additionally, wildfire refugia tended to form in peat-filled depressions, valleys, and forested areas within the study area, whereas drier, open rock barrens were most susceptible to fire. Overall, the model had a high predictive accuracy, with a cross-validated AUC of 0.88, and a sensitivity of 81.2%. We conclude that local scale topography and simple flow accumulation models can act as a powerful tool in predicting fire refugia occurrence in this landscape. In the second part of this study, we examined the in-situ indicators of peatland fire refugia occurrence. We conducted vegetation surveys at eight peatland fire refugia and eight reference sites representative of the range of wetland types found on this landscape. We found that the peatland fire refugia had a significantly different understorey vegetation composition when compared to the reference sites. Environmental factors within the peatland fire refugia which significantly influenced this separation included median peat depth, pH, and specific conductance (SpC); where peatland fire refugia were deeper and had a lower pH and SpC when compared to the reference sites. While no vascular indicator species were identified within the peatland fire refugia, there were two bryophyte indicator species: Sphagnum rubellum and Sphagnum magellanicum which were significantly associated with the peatland fire refugia. We conclude that understorey vegetation composition, indicator species presence, peat depth, pH and SpC could be useful when distinguishing peatlands with a high refugia probability, however, further research is needed to understand how this may vary geographically and in response to top-down controls, such as fire weather. Overall, the preliminary characterization of fire refugia in the Ontario Boreal Shield will provide a basis for the identification and mapping of fire refugia within this ecozone for applications in conservation, restoration, and fire and land management.