1. Tiger Team project:
Processes contributing to model differences in North American background ozone estimates
AQAST PIs: Arlene Fiore (Columbia/LDEO) and
Daniel Jacob (Harvard)
Co-I: Meiyun Lin (Princeton/GFDL)
Project personnel: Jacob Oberman (U Wisconsin)
Lin Zhang (Harvard)
AQ management contacts: Joe Pinto (EPA/NCEA)
Pat Dolwick (EPA/OAR/OAQPS)
NASA AQAST Meeting
University of Wisconsin-Madison
June 14, 2012

2. Objective: Improved error estimates of simulated
North American background O3 (NAB)
Problem: Poorly quantified errors in NAB distributions complicate NAAQS-setting and interpreting SIP attainment simulations
To date, EPA NAB estimates have been provided by one model.
Approach:
Compare GFDL AM3 and GEOS-Chem NAB (regional, seasonal, daily)
Process-oriented analysis of factors contributing to model differences
YEAR 2006
GEOS-Chem
GFDL AM3
Resolution
½°x⅔° (and 2°x2.5°)
~2°x2°
Meteorology
Offline (GEOS-5)
Coupled, nudged to NCEP U and V
Strat. O3 & STE
Parameterized (Linoz)
Full strat. chem & dynamics
Isoprene nitrate chemistry
18% yield w/ zero NOx recycling
8% yield w/ 40% NOx recycling (obs based; Horowitz et al, 2007)
Lightning NOx
tied to model convective clouds, scaled to obs. flash climat; higher NOx at N. mid-lat
tied to model convective clouds
Emissions
NEI fires (emitted at surface)
ACCMIP historical + RCP4.5 (2005, 2010); vert. dist. climatological fires
ALL DIFFERENT! 2

3. Seasonal mean North American background in 2006
(estimated by simulations with N. American anth. emissions set to zero)
North American background (MDA8) O3 in model surface layer
AM3 (~2°x2°)
GEOS-Chem (½°x⅔°)
AM3: More
O3-strat +
PBL-FT
exchange?
Spring (MAM)
GC: More
lightning NOx
(~10x over SWUS column)
+ spatial
differences
Different contributions from summertime Canadian wildfires?
(use of 2006 in GC vs climatology in AM3)
Summer (JJA)
ppb
J. Oberman 3

4. Space-based constraints on mid-trop O3?
Comparison with OMI & TES “500 hPa” in spring
Bias vs. N mid-latitude sondes
subtracted from
retrievals
Masked out
where products
disagree by
> 10 ppb
L. Zhang
Models bracket retrievals
Qualitative constraints where the retrievals agree in sign 4

5. Large differences in day-to-day and seasonal variability
of N. American background: Eastern USA, Mar-Aug 2006
Voyageurs NP, MN: 93W, 48N, 429m
GEOS-Chem ( ½°x⅔° ) AM3 (~2°x2°) OBS.
Mean(σ)
Total model O3
Model NAB O3
GC NAB declines into summer
AM3 NAB too high in summer:
Excessive fire influence?
Both models too high in summer
Similar correlations with obs
GC captures mean
AM3 +11 ppb bias: isop. chem.?
Georgia Station, GA: 84W, 33N, 270m
AM3 NAB declines in Jul/Aug
(when total O3 bias is worst)
GC NAB varies less than AM3
(total O3 has similar variability)
Does model horizontal resolution matter? 5

6. Horizontal resolution not a major source
of difference in model NAB estimates
Between
LARGEST DIFFERENCES OCCUR IN SUMMER at CASTNET SITES < 1.5 km (CONUS except CA)
GC Higher resolution broadens
distribution + shifts closer
to observed mean (lower)
GC 2°x2.5°
GC ½°x⅔°
GC High-res shows slight
shift towards higher NAB
(vertical eddies
[Wang et al., JGR, 2004])
GC NAB 2°x2.5°
GC NAB ½°x⅔°
AM3 ~2°x2°
AM3 NAB ~2°x2°
AM3 represents distribution
shape but biased high
OBS
SPRING (MAM) CASTNet sites above1.5 km
GC ½°x⅔°
similar to
GC 2°x2.5°
Much larger differences between AM3 and GC distributions (both total and NAB O3) than between the 2 GC resolutions 6

7. Large differences in day-to-day and seasonal variability
of N. American background: Western USA, Mar-Aug 2006
Gothic, CO: 107W, 39N, 2.9km
GEOS-Chem ( ½°x⅔° ) AM3 (~2°x2°) OBS.
Mean(σ)
Total model O3
Model NAB O3
Models bracket Obs.
AM3 larger σ than GC
(matches obs)
Mean NAB is similar
GC NAB ~2x smaller σ
than AM3
AM3 NAB > GC NAB
in MAM (strat. O3?);
reverses in JJA (lightning)
Fig 3-58 of O3 Integrated Science Assessment
Grand Canyon NP, AZ: 112W, 36N, 2.1km 7

8. How much does N. American background vary year-to-year?
NORTH AMERICAN BACKGROUND IN AM3 (ZERO N. Amer. emissions )
MEAN OVER 27 YEARS
STANDARD DEVIATION
Western CO experiences
largest year-to-year variability: What drives this?
ppb
ppb 8

9. Stratospheric O3: key driver of daily (+ inter-annual) variability, particularly late spring – e.g shown here
r2=0.44 (vs. obs)
r2=0.31 (vs. obs)
OBS
AM3
O3-strat
r2=0.45 (vs. obs)
r2=0.50 (vs. obs)
For GTH:
r(obs, total) = 0.67
r(obs, o3s)  = 0.71
r(total,o3s) = 0.85
For GRC:
r(obs,total) = 0.66
r(obs,o3s) = 0.56
r(total,o3s)= 0.81
Langford et al., 2009
Examine observational constraints on strat. influence (M. Lin)
M. Lin 9

10. Improved error estimates of simulated North American background O3 (NAB) that inform EPA analyses
AQ management outcomes:
Improved NAB error estimates to support:
ongoing review of ozone NAAQS (EPA ISA for O3),
SIP simulations focused on attaining NAAQS,
development of criteria for identifying exceptional events
Deliverables:
Report to EPA on confidence and errors in NAB estimates & key factors leading to model differences (peer-reviewed publication)
Guidance for future efforts to deliver estimates of sources contributing to U.S. surface O3
What next?
 satellite constraints: how quantitative?
 multi-model effort (more robust; error characterization)?
-- focus on specific components of NAB tied to multi-platform observations
-- choose a common study period (2008? )?
-- leverage AQAST IP + other TT projects where possible 10