1. Intercontinental Source Attribution of Ozone Pollution at Western U. S
Intercontinental Source Attribution of Ozone Pollution at Western U.S. Sites Using an Adjoint Method
Lin Zhang, Daniel J. Jacob, Monika Kopacz
Harvard University
Daven K. Henze
University of Colorado at Boulder
Kumaresh Singh
Virginia Polytechnic Institute of State University
Dan A. Jaffe
University of Washington-Bothell
The 8th Annual CMAS Conference (October 19, 2009)

2. Intercontinental influence of ozone
Ozone pollution is not just a local and regional problem, but can be a hemispheric problem.
Ozone once lifted in the free troposphere, has a lifetime of days to months, so that can be transported to downwind continent.
Two transport modes: direct episodic transport vs. increase of hemispheric background
Large uncertainties in estimating the influences: emission, evolution during transport, and subsidence.

3. Asian 2006 NOx emissions constrained from OMI
OMI NO2 retrieval from KNMI
Asian NOx anthropogenic emissions for the year 2000 from Streets et al. [2003]
2 x 2000 Asian NOx emissions
Much interest has focus on whether the rising Asian anthropogenic emissions would impact the air quality over United States.
Here shows OMI observations of tropospheric NO2 column in April and May 2006 over eastern Asia averaged over model resolution.
Using the year 2000 emissions, we see it captures similar spatial distribution, but is largely underestimated.
We found that we need to increase the 2000 NOx emissions by a factor of 2 as shown in the third column.
Comparison of GEOS-Chem simulation (Global CTM; 2ºx2.5º) for spring 2006 with OMI NO2 tropospheric column observations indicates a factor of 2 increase of Asian anthropogenic NOx emissions from 2000 to 2006
Some of the corrections (e.g. Japan and Korea) may reflect an underestimation in the prior emissions inventory.
Zhang et al., 2008

4. TES and AIRS observations of a transpacific plume
CO columns
TES
on the Aura satellite
GEOS-Chem
Global CTM (2ºx2.5º)
AIRS
on the Aqua satellite
May 5
May 6
May 7
May 8
May 9
2006 B A
CO has a lifetime of two months and is a good tracer of long-range pollution transport.
We see that the model simulation clearly shows an Asian pollution lifted and transported across the Pacific. It breaks into two branches over the northeast Pacific.
AIRS has fairly good global coverage. It captures the progression of the transpacific plume. TES daily observations (averaged over model resolution), although sparse, show these hot spots transport across the Pacific.
Zhang et al.,
2008

5. Impact of Asian emissions on U.S. surface ozone
Data from Dan Jaffe
O3 measurements at Mt. Bachelor
(44oN, 122oW, 2.7 km, in central Oregon)
Surface O3 enhancements from Asian anthropogenic emissions for spring 2006
54 ± 10
53 ± 8
9.4 ± 2.5
1 – 5 ppbv
How would the resulting transpacific ozone pollution affect the US surface ozone air quality?
We have measurements from Mt Bachelor. It is a mountain site in the central Oregon. Here we use it to link the free troposphere to the ground level.
The left panel show the ozone measurements at MBO during INTEX-B. Model simulation agrees very well with the measurements. The blue line is the Asian ozone enhancements. The variability is very small reflecting a background enhancement rather than episodic transport. On average 9 ppbv ozone enhancement from Asian emissions is not negligible.
The purple line shows simulated increases of ozone concentrations at MBO from 2000 to 2006 in the model. The simulated Asian ozone enhancements increase by 2.4 ppbv on average and up to 5 ppbv during high pollution episodes.
The right panel shows the Asian ozone enhancement over the US surface. I find 5-7 ppbv Asian ozone enhancements in the western United States. Previous literature has reported 4-5 ppbv Asian ozone in the west. The rising Asian emissions increased the surface ozone in the west by 1-2 ppbv.
rise in Asian emissions increased U.S. surface ozone
by 1-2 ppbv in the West.
Zhang et al., 2008

6. Forward vs. adjoint sensitivity analysis for source attribution
Forward Model (source-oriented)
Changes of concentration
winds
Perturbation at source region t0 tn
Adjoint Model (receptor-oriented)
Concentration at the receptor
A standard way to estimate the source contribution based on forward analysis. We perturb a source, and run transport, chemistry, convection, mixing and so on in the model, to get changes of concentrations in the receptor region. The computational time would largely increases if we need to look at multiple source regions.
The adjoint provides a very efficient method to calculate sensitivities.
the adjoint perturbs the concentration at the receptor site, and the adjoint run backwards in time to get the all the source contribution.
Here I show you how the adjoint model can be used to examine source-receptor relationship.
winds
area of
possible origin t0 tn

7. Ozone observations at the western U.S sites
For the period of April 17 - May 15, 2006
MBO (2.7 km)
54 ± 10
TH (sea level)
53 ± 9
13 ± 3.6
41 ± 7
observation
Both sites measure North America inflow air masses. The model simulation In red are in good agreement with the observations in black.
MBO frequently observes air masses in the free troposphere, and thus has a larger Asian influences. It observed a large Asian episode on May 1 as indicated by the arrow, and some small ones around May 10.
At the sea-level site (TH), the variability of Asian ozone simulated by the model is small. The Asian ozone pollution over the United States in the model mostly reflects a hemispheric-scale enhancement rather than the direct transport of pollution plumes from Asia to North America.
GEOS-Chem
43 ± 5
8.4 ±1.4
Tagged ozone produced over Asia
Tagged ozone simulation uses a linearized chemistry driven by ozone production rates and loss frequencies saved from the standard simulation.

8. Adjoint sensitivity for ozone at MBO on May 10
Sensitivity of ozone concentration at MBO on May 10, 2006 at 18 UT to ozone fields at earlier time steps (May 3 - May 10, 2006)
This movie illustrates an example of the sensitivities backwards in time.
we select the ozone pollution episode observed at MBO on May 10. This shows the sensitivity of this instantaneous ozone concentration to ozone fields at earlier time. In other words, it shows how ozone fields at earlier time steps impact the ozone concentration at MBO on May 10.
It looks similar to back-trajectory, but it include the full model transport processes as well as chemistry. The values show the amount of air masses transported to MBO at earlier time steps. If times the ozone production rates, we can get the sensitivity to ozone production rates that shows the amount of ozone transported to MBO.
Adjoint computes the sensitivity at the model resolution over the history of air parcels reaching the site.

9. Source attribution of ozone pollution episodes
Integrated sensitivity to ozone production
Time evolution of sensitivity
MBO: May 1, 2006
MBO: May 10, 2006
Here are the results. We run the adjoint backwards for two months. The left panels show the sensitivities to ozone production rates integrated over the two month period and over the tropospheric column depths at the model horizontal resolution. The unit is [ppbv/grid]. Summing these values globally over all grid squares approximates the ozone concentrations observed at the receptor sites -minus the stratosphere influence.
The right panels show the time evolution of sensitivities to production over Asia, North Pacific, North America, and Rest of World.
Here we show two pollution episodes. We can see most of the ozone production contributing to MBO ozone on those days took place over East Asia.
Both plumes show a secondary maximum of ozone production just off the west coast of United States. This is driven by PAN decomposition. PAN (peroxyacetylnitrate) is a thermo-unstable NOx reservoir species. It forms over the source continent and can be transported long distance. During subsidence, it decomposes and releases NOx, and drives further ozone production. The May1 plume took a more northerly and higher-altitude route than the May 10 plume, resulting in less ozone production over the Pacific.
The sensitivity spectra on the right panels show the transport timescales from production region to the receptor site. We see that ozone produced over North America had an immediate impact on MBO; this mostly reflects the decomposition of PAN in the subsiding air mass. The North American contribution also shows a weak secondary peak at 20 days that reflects ozone produced in the eastern United States and transported in the westerly atmospheric circulation.
Ozone production over Asia begins to impact MBO after a 6-day time lag and that maximum Asian influence for the two events is at time lags of 8-11 days. This is consistent with previous studies showing that Asian pollution plumes can be transported across the Pacific in 5-10 days
Decomposition of Asian PAN
The May1 plume took a more northerly and higher-altitude route than the May 10 plume, and thus had less production over the Pacific.
Maximum Asian influence for the two events occurs at time lags of 8-11 days

10. Mean conditions at MBO and TH
MBO: INTEX-B mean (Apr 17-May 15, 2006)
MBO (2.7 km)
TH (sea level)
TH: INTEX-B mean
This slide shows to you the source attribution for the mean ozone concentration at MBO and TH.
We can see the fine geographical structure that is hard to get from the forward sensitivity studies. The Influence of direct Asian transport is weaker. We can see maximum contributions from eastern China and Japan. The frontal passages are more frequent over Japan and thus increase the impact of Japanese emissions. We also see the background production distributed over the northern mid-latitudes belt (20ºN-60ºN).
The mean concentration at the TH shows similar source contribution but with more regional production influences.
As a concluding mark, this study attributed ozone pollution to its production regions. Using the adjoint for full chemistry can calculate the sensitivity to precursors’ emissions.
This study attributes ozone sources based on production regions instead of precursor emissions. Use of the adjoint of full chemistry mechanism can improve.
Zhang et al., 2009

11. Tropospheric ozone from TES and OMI on Aura
2006 tropospheric ozone data at 500 hPa averaged on 4ox5o resolution
TES (IR)
V003 data
OMI (UV)
Data from Xiong Liu (NASA/GSFC)
Here shows TES and OMI measurements of tropospheric ozone at 500 hPa where both instruments are relatively sensitive. This shows the year 2006 averaged over each season. Notice here all the data are reprocessed with a single fixed a priori, which means all the geographic and seasonal variations are obtained from the satellite information and differences between the instruments reflect the measurement itself.

12. Using GEOS-Chem for intercomparison between TES and OMI
2006 tropospheric ozone data at 500 hPa averaged on 4ox5o resolution
GEOS-Chem simulation
with TES vs. OMI averaging kernels
OMI (UV)
Data from Xiong Liu (NASA/GSFC)
TES (IR)
V003 data
Here shows TES and OMI measurements of tropospheric ozone at 500 hPa where both instruments are relatively sensitive. This shows the year 2006 averaged over each season. TES and OMI are showing very similar geographic features and seasonal variability.
The model simulation is in 4x5 resolution, and sampled along TES observations at the observing time, and then smoothed by corresponding averaging kernels.
This shows that some of the difference can be explained by different vertical sensitivities.
Zhang et al., in prep.

13. Using satellite observations to diagnose model errors
For 2006 and averaged on 4ox5o resolution
Regions with the bias between TES and OMI larger than 10 ppbv are masked as black.
Model is too low over the tropics, and too high in the extra-tropics.

14. Thank you!