TOWARDS REGIONAL FIRE OCCURRENCE MODELING IN ALASKA: DISENTANGLING IMPACTS OF METEOROLOGICAL AND ENVIRONMENTAL DRIVERS ACROSS BOREAL AND TUNDRA ECOSYSTEMS

Abstract

Wildfire is the dominant natural disturbance shaping ecological, biogeochemical, and climatological processes across the high northern latitudes (HNL). Although these fires rarely occur near population centers, they exert substantial influence on ecosystem succession, permafrost stability, carbon emissions, and regional climate forcing. The HNL have warmed at more than twice the global average over the past two decades, contributing to increases in lightning activity, fire occurrence, fire season length, and the prevalence of large and extreme fire events. Yet the mechanisms linking atmospheric conditions, cloud-to-ground lightning, ignition, and sustained fire growth remain insufficiently resolved, particularly when comparing the fuel-rich boreal forests with the colder, drier, and more heterogeneous tundra ecosystems of Alaska. Existing wildfire prediction tools often operate at coarse spatial and temporal scales or rely on aggregate fire weather indices, limiting their ability to capture the biome-specific dynamics that govern lightning occurrence and ignition on daily and sub-regional scales.My goal through this dissertation is to improve the understanding and predictability of cloud-to-ground lightning and fire ignition across Alaska’s boreal and tundra biomes. To accomplish this, I address the following three science questions: (1) to what extent is cloud-to-ground lightning occurrence predictable across Alaska at 5 km spatial resolution and lead times up to 10 days using an ensemble-based artificial intelligence modeling framework; (2) what meteorological and environmental conditions drive cloud-to-ground lightning occurrence across Alaska, and how do these drivers differ between tundra and boreal forest ecosystems; and (3) what meteorological and environmental conditions drive fire ignition occurrence across Alaska, and how do these drivers differ between tundra and boreal forest ecosystems. To answer these questions, I develop an integrated, physics-informed, data-driven forecasting framework capable of generating up to 10-day predictions for two operationally relevant products: Cloud-to-Ground Lightning Potential, a probabilistic assessment of future lightning activity; and Risk of Ignition, an estimate of ignition likelihood that incorporates lightning occurrence and ignition-conditioning environmental factors. This framework synthesizes satellite observations, numerical weather outputs, lightning detection networks, land-surface datasets, and advanced machine learning methods to quantify the relative importance of convective thermodynamics, moisture availability, fuel and surface conditions, atmospheric stability, and synoptic-scale forcing across Alaska’s biomes. Through this dissertation, I quantify biome-specific drivers of cloud-to-ground lightning and wildfire ignition across the HNL. I also develop a unified modeling framework that delivers fine-scale, daily probabilistic predictions, providing new insights into early-stage wildfire development and supporting operational fire management tools for Alaska’s rapidly evolving climate.