Constrained Fire Projections Across the Continental United States

To constrain Coupled Model Intercomparison Project Phase 6 (CMIP6) fire projections, we primarily used the observational constraints from six satellite-based fire carbon dioxide emissions inventories based on three methods: machine learning (ML), Bayesian model averaging (BMA), and emergent constraints (EC). The satellite fire emissions inventories include GFAS (Global Fire Assimilation System Version 1.2), FEER (Fire Energetics and Emissions Research Version 1.0), QFED (Quick Fire Emissions Dataset Version 2.4), FINN (Fire Inventory from NCAR Version 2.5; MODIS-based), GFED5 (Global Fire Emissions Database Version 5 Beta), and GFED41s (Global Fire Emissions Database Version 4.1s). In the constrain process, we also used a few auxiliary datasets to represent surrounding environmental conditions (e.g., air temperature), fuel availability (leaf area index and land use), and ignition sources (flash rate and population density).