Burned landscapes present several challenges to quantifying landscape carbon balance. Fire scars are composed of a mosaic of patches that differ in burn severity, which may influence postfire carbon budgets through damage to vegetation and carbon stocks. We deployed three eddy covariance towers along a burn severity gradient (i.e., severely burned, moderately burned, and unburned tundra) to monitor postfire net ecosystem exchange of CO2 (NEE) within the large 2007 Anaktuvuk River fire scar in Alaska, USA, during the summer of 2008. Remote sensing data from the MODerate resolution Imaging Spectroradiometer (MODIS) was used to assess the spatial representativeness of the tower sites and parameterize a NEE model that was used to scale tower measurements to the landscape. The tower sites had similar vegetation and reflectance properties prior to the Anaktuvuk River fire and represented the range of surface conditions observed within the fire scar during the 2008 summer. Burn severity influenced a variety of surface properties, including residual organic matter, plant mortality, and vegetation recovery, which in turn determined postfire NEE. Carbon sequestration decreased with increased burn severity and was largely controlled by decreases in canopy photosynthesis. The MODIS two-band enhanced vegetation index (EVI2) monitored the seasonal course of surface greenness and explained 86% of the variability in NEE across the burn severity gradient. We demonstrate that understanding the relationship between burn severity, surface reflectance, and NEE is critical for estimating the overall postfire carbon balance of the Anaktuvuk River fire scar.