Space-borne formaldehyde (HCHO) column measurements from the Ozone Monitoring Instrument (OMI), with 13 × 24 km2 nadir footprint and daily global coverage, provide new constraints on the spatial distribution of biogenic isoprene emission from North America. OMI HCHO columns for June-August 2006 are consistent with measurements from the earlier GOME satellite sensor (1996-2001) but OMI is 2-14% lower. The spatial distribution of OMI HCHO columns follows that of isoprene emission; anthropogenic hydrocarbon emissions are undetectable except in Houston. We develop updated relationships between HCHO columns and isoprene emission from a chemical transport model (GEOS-Chem), and use these to infer top-down constraints on isoprene emissions from the OMI data. We compare the OMI-derived emissions to a state-of-science bottom-up isoprene emission inventory (MEGAN) driven by two land cover databases, and use the results to optimize the MEGAN emission factors (EFs) for broadleaf trees (the main isoprene source). The OMI-derived isoprene emissions in North America (June-August 2006) with 1° × 1° resolution are spatially consistent with MEGAN (R2 = 0.48-0.68) but are lower (by 4-25% on average). MEGAN overestimates emissions in the Ozarks and the Upper South. A better fit to OMI (R2 = 0.73) is obtained in MEGAN by using a uniform isoprene EF from broadleaf trees rather than variable EFs. Thus MEGAN may overestimate emissions in areas where it specifies particularly high EFs. Within-canopy isoprene oxidation may also lead to significant differences between the effective isoprene emission to the atmosphere seen by OMI and the actual isoprene emission determined by MEGAN.