Projections of future fire activity, derived from statistical models, are a powerful tool for anticipating 21st-century fire regimes. In previous work, we developed a set of statistical models that captures fire-climate relationships at 30-yr timescales in Alaskan boreal forest and tundra ecosystems1. These models reveal that fire-climate relationships are strongly nonlinear, exhibiting distinct climatic thresholds to fire occurrence. Informing these models with future climate projections further highlighted the potential for fire-regime shifts to occur as thresholds are crossed under climate warming, with tundra and forest-tundra regions particularly vulnerable to climate-induced increases in fire activity. These projections are also accompanied with significant sources of uncertainty, particularly related to the calibration of statistical models using datasets that span the relatively short observational period (e.g., 1950-2009). To evaluate our ability to predict fire activity outside the observational period, we compared model predictions to fire-history records derived from lake sediments, spanning the past ca. 1000 yr. The statistical models were driven with downscaled Global Climate Model (GCM) data for 850-1850 CE, and predictions were compared to 29 paleo-fire-history reconstructions from seven Alaskan ecoregions. Model-paleodata comparisons for 850-1850 CE highlighted varying levels of prediction error among ecoregions, with this variability strongly related to ecoregion proximity to the previously estimated summer temperature threshold to fire. Regions closer to this threshold exhibited significantly larger prediction errors, while predictions were relatively accurate in the most- and least-flammable ecoregions of Alaska, located further away from this climatic threshold. Furthermore, the accuracy of predictions was sensitive to even slight modifications to original fire-climate relationships, implying that future projections will be sensitive to uncertainties in GCM data and/or to changing fire-climate relationships. Our findings imply uncertainty of future projections will vary spatially and temporally, as different regions cross climatic thresholds at varying time periods during the 21st century. In Alaska, significant uncertainty is likely to accompany near-term projections (i.e., 2010-2039) of fire across the majority of the state, as approximately 50-70% Alaska is estimated to lie near (i.e., within 2 °C) the identified temperature threshold. This research highlights how the nature of fire-climate relationships (i.e., nonlinear, threshold responses) may lead to varying levels of predictive power under conditions outside the observational range, important to consider when using future projections to anticipate 21st-century ecological changes.