Atmospheric restoration is a prerequisite for sub-pixel fire mapping: A smoke-robust protocol across five fire regions

Wildfires exert profound impacts on global ecosystems; however, accurate satellite-based burn area extraction is frequently confounded by the inherent mixed-pixel issue and complex atmospheric conditions. Although sub-pixel mapping (SPM) can resolve the mixed pixel issue, the extent to which its extraction accuracy depends on initial atmospheric preprocessing under diverse haze and smoke conditions remains poorly quantified. Furthermore, the conditions under which such preprocessing remains effective, and the boundary of its operational envelope, are not well understood. This study introduces a Smoke-Robust Sub-pixel Mapping (SRSPM) protocol to systematically diagnose atmospheric preprocessing efficacy prior to SPM. Using both simulated and real atmospheric states across five distinct fire regions, we systematically evaluate the sensitivity of sub-pixel fire perimeter delineation to different levels of haze contamination. Our findings demonstrate that atmospheric scattering inherently degrades the spatial allocation capabilities of SPM, leading to severe omission errors. Restoring contaminated spectral signals is shown to be essential, rather than merely beneficial, for accurate mapping when haze contamination is substantial. Crucially, our quantitative diagnosis reveals the operational limits of the SRSPM protocol: the dehazing benefit grows with haze intensity and remains substantial up to the strongest simulated contamination ( w = 1); in one region it turns positive only beyond w ≈0.73, delineating the quantitative envelope of the SRSPM protocol. Overall, this work establishes the SRSPM protocol as a practical diagnostic tool for quantifying the interaction between atmospheric correction and sub-pixel spatial allocation, providing empirical guidance for operational disaster monitoring in complex atmospheric environments.