1. Patterns of fire refugia across climate gradients in northern Rocky Mountain landscapes IALE 2015 Portland, Oregon USA

2. Our Research Team Jonathan Coop Assistant Professor Western State Colorado University Biology, Environment & Sustainability Ellen Whitman Spatial Analyst Simon Fraser University Sandra Haire Landscape Ecologist Haire Laboratory for Landscape Ecology Geneva Chong Research Ecologist US Geological Survey Carol Miller Research Ecologist— Wilderness Fire Aldo Leopold Wilderness Research Institute US Forest Service Meg Krawchuk Assistant Professor Simon Fraser University Landscape and Conservation Science Research Group Enric Batllori-Presas Research Ecologist InForest JRU (CSIC-CTFC- CREAF), Barcelona, Spain Marc-André Parisien Research Scientist Canadian Forest Service

3. Acknowledgements and thanks to our partners, home institutions, and agencies for their support The R Project for Statistical Computing

4. What are Refugia?  Places where suitable habitats persist as surrounding landscapes change  Keppel et al. 2012.  Habitats buffered against disturbance Background

5. What are Fire Refugia?  Fire mosaic strongly related to environmental heterogeneity  Climate, topography  Ephemeral “refuges” vs. persistent “refugia”  Stochastic and short-term vs. predictable and buffered against future disturbance Background

6. Research Question: Can we identify fire refugia? – Relationship to environment – Persistence through time – Ecological function

7. Background Research Question: Can we identify fire refugia?  Regional scale How do patterns of refugia correspond to variability in broad-scale gradients?  Intermediate- to fine- scale (within burn mosaic) How do patterns of refugia correspond to topographic heterogeneity? Environmental Heterogeneity Climate Topography Fire Refugium

8. Great Northern LCC Study Region US & Canada ~1.2 x10 6 km 2 Elev: 8 to 3966 m Temp: -6.1 to 13.7⁰C Precip: 214 to 4808 mm Fire: 1897 large fires between 1984-2011

9. Study Design Climate Space Framework 12 climate variables PCA (rotated) Latitudinal gradient: temp & summer precip Longitudinal gradient: continentality & winter precip 2 Climate Gradients Whitman et al. 2015. J. Biogeogr.

10. 10 Geographic space Study Design 2 Climate Gradients 73 Climate Domains Data space Climate Space Framework

11. Study Design Sample Fires 52 forest fires 1985-2010 400 - 229,000 ha (median 8050 ha) 25 climate domains 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 3 3 3 9 4 4 4 4 # fires sampled ~160,000 points randomly sampled 1 per 10 ha

12. Methods Response variable PLAND: percent of landscape classified as refugia in a 1 km 2 moving window Observed refugia: low values of burn severity metric  -200 < ∆NBR < 200

13. Broad-scale gradients Regional scale  How do patterns of refugia correspond to variability in: Climatic gradients (RC1, RC2) Roadless volume (Watts et al. 2007) Percent non-fuel (rock, water, barren) Methods Climate (2000-m) Roadless volume (1000-m) Percent non-fuel 1000-m Fire shadows

14. Broad-scale gradients Regional scale  How do patterns of refugia correspond to variability in: Climatic gradients (RC1, RC2) Roadless volume (Watts et al. 2007) Percent non-fuel (rock, water, barren) Methods Climate (2000-m) Roadless volume (1000-m) Percent non-fuel 1000-m  Graphical analysis in Climate Space  Exploratory data analysis using random subsamples from fires (GAM, GLM, OLS, QR)

15. LowHigh Cool & summer wet Warm & summer dry Maritime & winter wet Continental & winter dry PLAND Observed Refugia (90 th quantile) mapped for Climate Domains in geographic space (left) and data space (right) Broad-scale gradients – climate Results

16. Broad-scale gradients Results Analyses (OLS, QR, GLM, GAM) suggest that refugia: increase with RC1 & RC2 and Roadless Volume nonlinear response to Percent Non-Fuel Random effects of Climate Domain and Fire ID: large effect Random effects

17. Broad-scale gradients Summary  Some trends across broad-scale gradients  Results of random effects point to need to examine sources of variation within climate domain and within individual fire events.

18. Topographic Heterogeneity Methods Bridge Fire, 2007, ID Intermediate- to fine-scale  How do patterns of refugia correspond to: Catchment height Catchment slope Relative position Topographic wetness Topographic convergence (30-m resolution)  GLM modeling  PLAND ~β 1 x + β 2 x + … +β 5 x + ε  Graphical analysis of model statistics in Climate Space

19. Topographic Heterogeneity LowHigh Results Cool & summer wet Warm & summer dry Maritime & winter wet Continental & winter dry Mean Deviance Explained by Variability in Terrain (GLM) summarized across Climate Domains  Climate space as a coarse filter for variability within domains and within fires  Terrain a strong predictor in some (but not all) fires/environments.  Other factors dominating in certain places (weather, legacies, operations) 0.43 0.01

20. Topographic Heterogeneity GLM modeling  Strong links to terrain identified: greater potential for persistent refugia  Significant factors varied among fires Mt. Shanks: Larger, more highly clustered refugia in wetter areas, steeper slopes, deeper basins. Less abundant and more scattered in drier areas with less relief. Results Mt. Shanks Fire, 2001, BC catchmt height catchmt slope topo wetness dev expl = 0.41

21. Climate Space as a coarse filter to explore fire regime attributes, including refugia patterns. More potential for refugia in certain environments. Broad-scale gradients influential, but variability within climate domains and within fires is key. Conclusions

22. Fires where refugia can be linked to deterministic factors provide models for understanding how topography forms a “template for resilience” useful to management. Moritz et al. 2014. Learning to co- exist with fire. Nature 515: 58-66.

23. THANK YOU Questions? 23