Fire influences carbon dynamics from local to global scales, but many uncertainties remain regarding the remote detection and simulation of heterogeneous fire effects. This study integrates Landsat-based remote sensing and Biome-BGC process modeling to simulate the effects of high-, moderate-, and low-severity fire on pyrogenic emissions, tree mortality, and net ecosystem production. The simulation area (244,600 ha) encompasses four fires that burned approximately 50,000 ha in 2002-2003 across the Metolius Watershed, Oregon, USA, as well as in situ measurements of postfire carbon pools and fluxes that we use for model evaluation. Simulated total pyrogenic emissions were 0.732 Tg C (2.4% of equivalent statewide anthropogenic carbon emissions over the same 2-year period). The simulated total carbon transfer due to tree mortality was fourfold higher than pyrogenic carbon emissions, but dead wood decomposition will occur over decades. Immediately postfire, burned areas were a simulated carbon source (net C exchange: -0.076 Tg C y-1; mean ± SD: -142 ± 121 g C m-2 y-1). As expected, high-severity, stand-replacement fire had disproportionate carbon impacts. The per-unit area effects of moderate-severity fire were substantial, however, and the extent of low-severity fire merits its inclusion in landscape-scale analyses. These results demonstrate the potential to reduce uncertainties in landscape to regional carbon budgets by leveraging Landsat-based fire products that account for both stand-replacement and partial disturbance. © 2011 Springer Science+Business Media LLC. Abstract reproduced by permission.