Evaluating risk for current and future Bromus tectorum invasion and large wildfires at multiple spatial scales in Colorado and Wyoming, USA

Abstract

The Western United States is experiencing rapid ecologic change. These changes are driven largely by anthropogenic factors including introduction of alien invasive species, wildfire ignition and suppression, climate change, and feedbacks between these occurrences. Average temperatures in some areas of the Western U.S. increased as much as 1.1 °C between 2000 and 2006. The advancement of spring also provides evidence for climate change in the region; earlier snowmelt and runoff has been documented in recent decades for areas of the Intermountain West. These rapid changes will certainly affect the distribution of the alien invasive B. tectorum and large wildfires in Colorado and Wyoming as well as their associated feedbacks and cascading ecosystem effects. Prompted and inspired by natural resource manager concerns, this research evaluates these ecological phenomena at three spatial scales: Rocky Mountain National Park, Colorado; a wildfire disturbance in Medicine Bow National Forest, Wyoming; and the area encompassed by these two states. The products from this research are maps that can be incorporated into decision support systems for land management and vulnerability assessments for climate change preparedness. An evaluation of the current and future suitable habitat for B. tectorum in Rocky Mountain National Park was conducted at a 90 m² spatial resolution using a MaxEnt model fit with climatic, vegetation cover, and anthropogenic covariates (i.e. distance to roads as a surrogate for propagule pressure). One of the important considerations of this research was spatial scale; 250 m² and 1 km² resolution climate surfaces cannot capture climate refugia in a small area such as Rocky Mountain National Park (1,076 km²) with high topographic heterogeneity (2,300 m to 4,345 m elevation). Based on model results, the suitable habitat for B. tectorum in the Park increases more than 150 km2 through the year 2050. Four multi-temporal and multiscale species distribution models were developed for B. tectorum in the Squirrel Creek Wildfire post-burn area of Medicine Bow National Forest using eight spectral indices derived from five months of 30 m² Landsat 8 imagery corresponding to changes in species phenology and time of field data collection. These models were improved using an iterative approach in which a threshold for abundance (i.e. ≥40% foliar cover) was established from an independent dataset, and produced highly accurate maps of current B. tectorum distribution in Squirrel Creek burn with independent AUC values of 0.95 to 0.97. The most plausible model based on field observations showed the distribution of B. tectorum has increased 30% from pre-disturbance observations in the area. This model was incorporated in a habitat suitability model for B. tectorum in the same area using topographic covariates with inclusion of propagule dispersal limitations to provide an estimate of future potential distribution. Three historic (years 1991 – 2000) environmental suitability models for large wildfires (i.e. > 400 ha) in Colorado and Wyoming were developed at a 1 km² spatial resolution and tested using an independent dataset of large wildfire occurrence in the same area from the subsequent decade (years 2001 – 2010). The historic models classified points of known fire occurrence exceptionally well using decadal climate averages corresponding to the temporal resolution of wildfire occurrence and topographic covariates. When applied to an independent dataset, the test sensitivity was 0.91 for the best model (i.e. MaxEnt). We then applied the model to future climate space for the 2020s (years 2010-2039) and 2050s (years 2040-2069) using two future climate ensembles (i.e. two representative concentration pathways; RCP 4.5 and RCP 8.5 with ensemble average projections from 15 global circulation models) to rank areas for large wildfire risk in the future.