Fire is a key ecological process affecting vegetation dynamics and land cover. The characteristic frequency, size, and intensity of fire are driven by interactions between top-down climate-driven and bottom-up fuel-related processes. Disentangling climatic from non-climatic drivers of past fire regimes is a grand challenge in Earth systems science, and a topic where both paleoecology and ecological modeling have made substantial contributions. In this manuscript, we (1) review the use of sedimentary charcoal as a fire proxy and the methods used in charcoal-based fire history reconstructions; (2) identify existing techniques for paleoecological modeling; and (3) evaluate opportunities for coupling of paleoecological and ecological modeling approaches to better understand the causes and consequences of past, present, and future fire activity.

We developed a new climate-sensitive vegetation state-and-transition simulation model (CV-STSM) to simulate future vegetation at a fine spatial grain commensurate with the scales of human land-use decisions, and under the joint influences of changing climate, site productivity, and disturbance. CV-STSM integrates outputs from four different modeling systems. Successional changes in tree species composition and stand structure were represented as transition probabilities and organized into a state-and-transition simulation model. States were characterized based on assessments of both current vegetation and of projected future vegetation from a dynamic global vegetation model (DGVM). State definitions included sufficient detail to support the integration of CV-STSM with an agent-based model of land-use decisions and a mechanistic model of fire behavior and spread. Transition probabilities were parameterized using output from a stand biometric model run across a wide range of site productivities. Biogeographic and biogeochemical projections from the DGVM were used to adjust the transition probabilities to account for the impacts of climate change on site productivity and potential vegetation type. We conducted experimental simulations in the Willamette Valley, Oregon, USA. Our simulation landscape incorporated detailed new assessments of critically imperiled Oregon white oak (Quercus garryana) savanna and prairie habitats among the suite of existing and future vegetation types. The experimental design fully crossed four future climate scenarios with three disturbance scenarios. CV-STSM showed strong interactions between climate and disturbance scenarios. All disturbance scenarios increased the abundance of oak savanna habitat, but an interaction between the most intense disturbance and climate-change scenarios also increased the abundance of subtropical tree species. Even so, subtropical tree species were far less abundant at the end of simulations in CV-STSM than in the dynamic global vegetation model simulations. Our results indicate that dynamic global vegetation models may overestimate future rates of vegetation change, especially in the absence of stand-replacing disturbances. Modeling tools such as CV-STSM that simulate rates and direction of vegetation change affected by interactions and feedbacks between climate and land-use change can help policy makers, land managers, and society as a whole develop effective plans to adapt to rapidly changing climate.

Western Tasmania, Australia contains some of the highest levels of biological endemism of any temperate region in the world, including vegetation types that are conservation priorities: fire-sensitive rainforest dominated by endemic conifer species in the genus Athrotaxis; and firetolerant buttongrass moorlands. Current management focuses on fire suppression, but increasingly there are calls for the use of prescribed fire in flammable vegetation types to manage these ecosystems. The long-term effects of climate and alternative management strategies on the vegetated landscape are unknown. To help identify controls over successional trajectories, we parameterized a spatially explicit landscapescale model of vegetation and fire (FireBGCv2) for a study area in Cradle Mountain-Lake St Clair National Park in western Tasmania using new data on fine-scale topography, plant communities, and fuels loads. Our parameterized model displays a high level of agreement with previous empirical and modeling studies for the region. The model was experimentally tested for three different levels of ignition suppression (0, 50, and 90 %); simulations ran for 1000 years and were replicated 10 times. The different scenarios yielded distinct fire return intervals, with cascading effects on successional dynamics and vegetation composition. Model results indicate that fire-sensitive endemic conifer rainforest will be restricted to upland refugia that total far less area than its present distribution, even under maximal ignition suppression. Because the distribution of vegetation types was unstable temporally and across stochastic replicates, present distributions may be a legacy of previous climate, Aboriginal fire management, or both.