We present an inverse model analysis to quantify the emissions of wildfires in Alaska and Canada in the summer of 2004 using carbon monoxide (CO) data from the Measurements of Pollution in the Troposphere (MOPITT) remote sensing instrument together with the chemistry transport model MOZART (Model for Ozone and Related Chemical Tracers). We use data assimilation outside the region of the fires to optimally constrain the CO background level and the transport into that region. Inverse modeling is applied locally to quantify the fire emissions. Our a posteriori estimate of the wildfire emissions gives a total of 30 ¿ 5 Tg CO emitted during June--August 2004 which is of comparable order to the anthropogenic emissions for the continental US. The simulated CO fields have been evaluated by comparison with MOPITT and independent aircraft data.