Abstract. Isoprene emissions from vegetation have a large effect on
atmospheric chemistry and air quality. “Bottom-up” isoprene emission
inventories used in atmospheric models are based on limited vegetation
information and uncertain land cover data, leading to potentially large
errors. Satellite observations of atmospheric formaldehyde (HCHO), a
high-yield isoprene oxidation product, provide “top-down” information to
evaluate isoprene emission inventories through inverse analyses. Past inverse
analyses have however been hampered by uncertainty in the HCHO satellite
data, uncertainty in the time- and NOx-dependent yield of HCHO from
isoprene oxidation, and coarse resolution of the atmospheric models used for
the inversion. Here we demonstrate the ability to use HCHO satellite data
from OMI in a high-resolution inversion to constrain isoprene emissions on
ecosystem-relevant scales. The inversion uses the adjoint of the GEOS-Chem
chemical transport model at 0.25∘ × 0.3125∘
horizontal resolution to interpret observations over the southeast US in
August–September 2013. It takes advantage of concurrent NASA SEAC4RS
aircraft observations of isoprene and its oxidation products including HCHO
to validate the OMI HCHO data over the region, test the GEOS-Chem isoprene
oxidation mechanism and NOx environment, and independently evaluate the
inversion. This evaluation shows in particular that local model errors in
NOx concentrations propagate to biases in inferring isoprene emissions
from HCHO data. It is thus essential to correct model NOx biases, which
was done here using SEAC4RS observations but can be done more generally
using satellite NO2 data concurrently with HCHO. We find in our
inversion that isoprene emissions from the widely used MEGAN v2.1 inventory
are biased high over the southeast US by 40 % on average, although the
broad-scale distributions are correct including maximum emissions in
Arkansas/Louisiana and high base emission factors in the oak-covered Ozarks
of southeast Missouri. A particularly large discrepancy is in the Edwards
Plateau of central Texas where MEGAN v2.1 is too high by a factor of 3,
possibly reflecting errors in land cover. The lower isoprene emissions
inferred from our inversion, when implemented into GEOS-Chem, decrease
surface ozone over the southeast US by 1–3 ppb and decrease the isoprene
contribution to organic aerosol from 40 to 20 %.