This project aims to develop fire-weather risk indices based on near-surface meteorological variables and to evaluate their reliability across the longest historical context possible. Using the NOAA-CIRES-DOE 20th Century Reanalysis version 3 (20CRv3), a 200-year, 80-member ensemble dataset that spans the time period of 1806-2015, we examine the climatological long-term averages (1951-2000), variability, and trends of key fire-weather indices across North America, including Potential Evaporation Rate (PEVPR), 2-meter Vapor Pressure Deficit (VPD2m), and other metrics related to fire danger. We find that the ensemble of the 20CRv3 enables quantitative, reliable uncertainty estimates that indicate the limits of their historical applicability. To illustrate the utility of these indices, we analyze atmospheric and terrestrial conditions preceding three of the largest wildfires in U.S. history that occurred in the Great Lakes region: the Great Michigan Fire of 1871, the Peshtigo Fire of 1871, and the Thumb Fire of 1881. The analysis suggests pronounced positive anomalies in PEVPR and VPD2m preceding these three events, indicating persistent dryness and elevated evaporative demand that likely preconditioned the landscape for extreme fire behavior. These results underscore the value of long-term reanalysis datasets in reconstructing and contextualizing fire-weather risk, even in the 19th century when direct observations were sparse. Our findings suggest that well-chosen fire-weather indices, interpreted within an ensemble framework, can provide insights into historical and modern wildfire danger. This work advances the use of reanalysis-derived fire weather indices for contextualizing variations and trends in historical wildfire activity and emphasizes the value of quantifying uncertainty in historical reconstructions.