1. Top-down constraints on emissions of biogenic trace gases from North America:
Mapping isoprene emissions from space
Dylan Millet
with
D.J. Jacob and K.F. Boersma
Atmospheric Chemistry Modeling Group, Harvard University
T.P. Kurosu and K. Chance
Harvard-Smithsonian Center for Astrophysics
C. Heald (UC Berkeley), A. Guenther (NCAR), A. Fried (NCAR), B. Heikes (URI), D. Blake (UCI), and H. Singh (NASA-Ames)
IGAC-WMO-CACGP Symposium
Cape Town, South Africa
September 17-22, 2006

2. Biogenic Emissions Affect Atmospheric Composition and Climate
HCHO O3
SOA …
OH, h, O3
Air Quality
VOC
NOx, VOC, SO2
Tropospheric chemistry
Climate
Isoprene
Most important biogenic NMVOC
~ 6x anthropogenic VOC emissions

3. Mapping Isoprene Emissions from Space
HCHO vertical columns measured by OMI
(K. Chance, T.P. Kurosu et al.)
VOCs
HCHO
OH, h
ki, Yi
kHCHO
VOC source
Distance
downwind
ΩHCHO
Isoprene
a-pinene
propane
100 km
detection
limit
Local ΩHCHO-Ei Relationship
Palmer et al., JGR (2003,2006).

4. Ratio between HCHO along light path and the vertical column amount
Testing the Approach: Errors in satellite HCHO measurements
ΩHCHO = SEisoprene+ B
HCHO
GOME/OMI sensitivity
Main Sources of Error
Fitting uncertainty
~ 4 x 1015 molecules cm-2
Ratio between HCHO along light path and the vertical column amount
HCHO vertical profile
scattering by air molecules, aerosols, clouds
surface albedo
Use INTEX-A aircraft data & GEOS-Chem model to test errors in HCHO measured from space
Clouds: primary source of error
1σ error in HCHO satellite measurements: 25–31%
Recommended cloud cutoff: 50%
Millet et al., JGR (in press).

5. Testing the Approach: Relating isoprene emission to HCHO column
ΩHCHO = SEisoprene+ B
What drives variability in column HCHO?
Test model HCHO yield
M = 3.5
M = 3.6
Observed
GEOS-Chem
Measured HCHO production rate vs. column amount
PHCHO (1012 molec cm-2 s-1)
INTEX-A
ΩHCHO (1016 molec cm-2)
ΩHCHO (1016 molec cm-2)
Isoprene dominant source when ΩHCHO is high
Other VOCs give rise to a relatively stable background ΩHCHO
 Not to variability detectable from space
ΩISOP (1016 molec cm-2)
HCHO yield from isoprene:
Y = 1.6 ± 0.5
ΩHCHO variability over N. America driven by isoprene
Millet et al., JGR (in press).

6. Using OMI HCHO to Define Spatial Distribution of Eisoprene
HCHO columns measured with the OMI satellite instrument (summer 2005)
Isoprene emissions from the MEGAN biogenic emission inventory (summer 2005) ?
Comparison between emission inventory and HCHO columns from OMI indicates mismatch in hotspot locations
Implications for O3, SOA production

7. Model of Emissions of Gases and Aerosols from Nature
Guenther et al., Atmos. Chem. Phys., 6, 3181–3210, 2006.
Vegetation-specific baseline emission factors
Environmental drivers (T, h, LAI, leaf age, …)
Land cover database
MEGAN Isoprene emissions

8. Similarity in broad pattern (r2 = 0.80) … but fine-scale discrepancies
OMI vs. GEOS-Chem with MEGAN Emissions
OMI 44% lower
Similarity in broad pattern (r2 = 0.80)
… but fine-scale discrepancies

9. ΩHCHO-Eisoprene relationship ΩHCHO-Eisoprene relationship
Relating HCHO Columns to Isoprene Emissions
ΩHCHO = SEisoprene+ B
Domain-wide
ΩHCHO-Eisoprene relationship
Local
ΩHCHO-Eisoprene relationship

10. Spatial Patterns in Isoprene Emissions
MEGAN w/ Community Land Model (CLM)
Domain-wide
ΩHCHO-Eisoprene relationship
Local
ΩHCHO-Eisoprene relationship
Normalized OMI - MEGAN
Normalized OMI - MEGAN

11. Spatial Patterns in Isoprene Emissions
MEGAN w/ CLM Land Cover
MEGAN w/ Olson Land Cover
Normalized OMI – MEGAN
July-August, 2005
Scale up OMI to remove overall bias
Drive MEGAN with 2 land cover databases
Olson [2001]
Community Land Model (CLM)
Large sensitivity to surface database used
MEGAN higher than OMI over ‘hotspots’ such as the Ozarks, lower over deep South & Atlantic coast

12. Emissions Overestimated in Ozarks & Other ‘Hotspots’
MEGAN w/ CLM Land Cover
MEGAN w/ Olson Land Cover
Normalized OMI – MEGAN
July-August, 2005
Bottom-up emissions are too high in Ozarks, Virginia
Large emissions driven by oak tree cover, high temperatures
OMI comparison suggests broadleaf tree emissions are overestimated
Olson Broadleaf Trees

13. Emissions Underestimated in Deep South & Atlantic Coast
Bottom-up emissions are too low in deep South, Atlantic coast
MEGAN w/ CLM Land Cover
MEGAN w/ Olson Land Cover
Normalized OMI – MEGAN
July-August, 2005
Underestimate of pine emissions in Southeast?
Errors in vegetation cover?
Underestimate of regional crop emissions also possible? (cotton, peanuts, tobacco)
CLM Fineleaf Evergreen Trees
CLM Crops

14. Conclusions
OMI’s small footprint (13 x 24 km) allows us to define surface fluxes of trace gases with unprecedented spatial detail
OMI HCHO columns are broadly consistent with state-of-the-art bottom-up emission inventories (R2 = 0.80)
… but with important spatial differences!
Bottom-up isoprene emission estimates are too high in the Ozarks and other ‘hotspots’
Overestimate of broadleaf tree emissions?
Bottom-up isoprene emission estimates are too low over the deep South and along the Atlantic coast
Underestimate of pine (possibly crop) emissions?
Regional broadleaf tree coverage underestimated?

15. Acknowledgements
NOAA Postdoctoral Program in Climate and Global Change
NASA/ACMAP
OMI science team
B. Yantosca, P. Palmer (now at Leeds), M. Fu, and other coworkers at Harvard
The INTEX-A science team