All chemical transport models require an estimation of the vertical distribution of smoke particles near the source. This study quantitatively examines the strengths and weaknesses of several fire products for characterizing plume buoyancy and injection heights in the North American boreal forest during [2004][2005]. Observations from the Multiangle Imaging Spectroradiometer show that 21% of smoke plumes are injected more than 500 m above the boundary layer (BL 500 ) and 8% exceed 2.5 km above ground level. Corresponding observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) show that probability of injection above the BL 500 exceeds 60% for pixel-based fire radiative power (FRP p ) values abovẽ 2500 MW. Increasing values of subpixel-retrieved fire area and temperature also correspond to higher injections but only after removing fire pixels with a weak 11 μm fire signal and clustering. The probability of injection above the BL 500 reaches 50% when the subpixel radiant flux (FRP f flux) exceeds 20 kW/m 2 , highlighting its potential for estimating plume buoyancy. However, these data have limitations similar to FRP p , where the highest probability of injection corresponds to a small percentage of the data set (5-18%), and many high-altitude injections occur with lower values. Examinations of individual smoke plumes highlight the importance of combining pixel-level and subpixel outputs and show that plume injection is also sensitive to the fire pixel spatial distribution and meteorology. Therefore, an optimal method for predicting high-altitude injections will require some combination of injection climatology, FRP p , FRP f flux, and meteorology, but each variable's importance will depend on fire event characteristics.