Adam Lee Mahood , Michael J Koontz, Jennifer Balch

A key challenge in ecology is understanding how multiple drivers interact to precipitate persistent vegetation state changes. These state changes may be both precipitated and maintained by disturbances, but predicting whether the state change is fleeting or persistent requires an understanding of the mechanisms by which disturbance affects the alternative communities. In the sagebrush shrublands of the western United States, widespread annual grass invasion has increased fuel connectivity, which increases the size and spatial contiguity of fires, leading to post-fire monocultures of introduced annual grasses (IAG). The novel grassland state can be persistent, and more likely to promote large fires than the shrubland it replaced. But the mechanisms by which pre-fire invasion and fire occurrence are linked to higher post-fire flammability are not fully understood. A natural experiment to explore these interactions presented itself when we arrived in northern Nevada immediately after a 50,000 ha wildfire was extinguished.

We hypothesized that the novel grassland state is maintained via a reinforcing feedback where higher fuel connectivity increases burn severity, which subsequently increases post-fire IAG dispersal, seed survivorship, and fuel connectivity. We used a Bayesian joint species distribution model and structural equation model framework to assess the strength of the support for each element in this feedback pathway. We found that pre-fire fuel connectivity increased burn severity and that higher burn severity had mostly positive effects on the occurrence of IAG and another non-native species, and mostly negative or neutral relationships with all other species. Finally, we found that the abundance of IAG seeds in the seedbank immediately post-fire had a positive effect on the fuel connectivity 3 years after fire, completing a positive feedback promoting IAG. These results demonstrate that the strength of the positive feedback is controlled by measurable characteristics of ecosystem structure, composition and disturbance. Further, each node in the loop is affected independently by multiple global change drivers. It is possible that these characteristics can be modeled to predict threshold behavior and inform management actions to mitigate or slow the establishment of the grass-fire cycle, perhaps via targeted restoration applications or pre-fire fuel treatments.