Exploring invasibility with species distribution modeling: How does fire promote cheatgrass (Bromus tectorum) invasion within lower montane forests?

Aim: Cheatgrass (Bromus tectorum) is notorious for creating positive feedbacks that facilitate vegetation type conversion within sagebrush steppe ecosystems in the western United States. Similar dynamics may exist in adjacent lower montane forest. However, fire-forest-cheatgrass dynamics have not been examined. We used species distribution modeling to answer three questions about fire and invasibility in lower montane forests: (Q1) Does fire create more suitable habitat for cheatgrass? (Q2) If so, which site attributes are altered to increase site suitability? (Q3) Does fire increase connectivity among suitable habitat and enhance spread? Location: Shoshone National Forest, Wyoming, USA. Methods: We measured cheatgrass presence-absence in 93 plots within Interior Douglas-fir (Pseudotsuga menziesii var. glauca) forests. Random Forests predicted cheatgrass distribution with and without fire using nine site attributes: elevation, slope, aspect, solar radiation, annual precipitation, maximum temperature in July, minimum temperature in January, forest canopy cover and distance to nearest trail or road. Additionally, invasion pathways and spread were mapped using Circuitscape. Results: Cheatgrass distribution was controlled by topographic and climate variables in the absence of fire. In particular, cheatgrass was most likely to occur at low elevation along dry, south- and east-facing slopes. High-severity fire increased potential cheatgrass distribution when forest canopy cover was reduced to below 30%. This process created new invasion pathways, which enhanced cheatgrass spread when modelled in Circuitscape. Main conclusions: Our study showed that in the absence of fire, drier south- and east-facing slopes at low elevation are most susceptible to cheatgrass invasion. However, high-severity fire increased the total area susceptible to invasion-allowing cheatgrass to expand into previously unsuitable sites within lower montane forests in the western United States. These results are important for present day management and reflect that integrating responses to disturbance in species distribution models can be critical for making predictions about dynamically changing systems.