Boreal forest fires are highly variable in space and time and also have variable vertical injection properties. We compared a University of Maryland Chemistry and Transport Model (UMD-CTM) simulation of boreal forest fire CO in the summer of 2000 to surface observations from the NOAA Cooperative Air Sampling Network and satellite observations of CO from the Measurement of Pollutants in the Troposphere (MOPITT) instrument to investigate the sensitivity of these measurements to injection height and to evaluate the bulk injection properties of the boreal fire source. Our results show that emissions at the surface produce more than twice the signal in surface CO measurements compared with emissions injected into the upper troposphere. Surface injection yielded the best agreement with surface observations, but high-altitude injection resulted in very small variations at the surface, and so the statistical comparison with surface observations was inconclusive. Because of the vertical sensitivity of MOPITT, estimated total CO burden north of 30 degrees N was 10% higher for upper tropospheric injection of boreal forest fire CO compared to surface release. We used a contrast filter to select the MOPITT retrievals most sensitive to boreal forest fire injection height and found that the best agreement between simulation results and MOPITT observations was obtained with midtropospheric injection of emissions and with pressure-weighted distribution of emissions through the tropospheric column. Appendix A uses CTM output to examine quantitatively the bias and errors in calculations of total column CO and total CO burden using MOPITT CO retrievals.