1. Understanding the Influence of Biomass Burning on Tropospheric Ozone through Assimilation of TES data Jennifer Logan Harvard University Dylan Jones, Mark Parrington University of Toronto Kevin Bowman, Helen Worden, John Worden, Greg Osterman Jet Propulsion Laboratory California Institute of Technology

2. Impact of Biomass Burning on Tropospheric O 3 TES CO, 421 mb: Nov 4-17, 2004 TES O 3, 421 mb: Nov 5-17, 2004 GEOS-Chem CO 421 mb: Nov 4-17, 2004 GEOS-Chem O 3, 421 mb: Nov 4-17, 2004 Climatological emission inventory in the model underestimates the impact biomass burning on CO and O 3 in the southern hemisphere ppb Objective: Assess the potential of TES data to improve O 3 in the model in a chemical data assimilation framework

3. Impact of Assimilation on CO and O 3 (using a sequential sub-optimal Kalman filter with TES O 3 and CO profile retrievals for Nov. 4-17, 2004) Change in O 3 at 7 km (assim. - without assim.) 24-hr averaged assimilated O 3 at 7 km on Nov. 17 (ppb CO) percent Assimilation increases CO throughout the southern hemisphere Largest increases in O 3 (20-50%) are over the Indian Ocean and the Indonesian/Australian region (ppb O3) percent 24-hr averaged assimilated CO at 7 km on Nov. 17 Change in CO at 7 km (assim. - without assim.)

4. Assimilation of TES O 3 for 1 July 2005 - 1 Jan. 2006 Assimilation Free running model O 3 difference: assimilation - free running model data gaps Mean GEOS-Chem O 3 at 8 km between 20°S-equator and 180°W-180°E In early Sept 2005 the assimilation increases O 3 by about 20% in upper troposphere During the 2 week data gap in September the analysis reverts to the state of the free running model

5. Comparison with Ozonesonde Data at La Reunion Island (21°S, 55°E) 12 Oct 2005 17 Oct 2005 28 Oct 2005 2 Nov 2005 assimilation free running model sonde The ozone tropopause in GEOS-Chem is too low in Austral spring 2005 compared to the sonde data Assimilation of TES data reduces the bias in the model

6. Comparison of the O 3 Analysis with TES Observations (350 mb) During October the assimilation reduces the bias in the model by about a factor of 2 Despite the reduction in the bias, the residuals for the OmA are still large Obs minus Forecast Obs minus Analysis

7. Comparison of the CO Analysis with TES Observations (350 mb) Following the warm-up of the TES optical bench in Dec. 2005, the assimilation significantly reduced the bias in CO in the model In contrast to the O 3 analysis, the CO OmA residuals are small, reflecting the longer lifetime of CO Obs minus Forecast Obs minus Analysis TES OB warm-up Assimilation extended through 1 Sept 2006

8. Latitudinal Dependence of the O 3 Analysis Residuals (350 mb) The assimilation has less impact in summer 2006 because we are propagating the forecast error variance without accounting for forecast error growth  by summer 2006 the forecast error is about 15% in the tropics and subtropics, compared to the assumed 50% error in July 2005 Larger OmA residuals in the tropics, reflecting the shorter O 3 lifetime and a lower density of TES data 20°S-0° 30°N-60°N Obs minus Forecast Obs minus Analysis

9. Conclusions Assimilation of TES O 3 data produces a much improved distribution of O 3 in the model, which provides greater constraints on model parameters such as the lifetime of NOx  a better constraint on the NOx lifetime will result in improved estimates of NOx emissions from lightning and of the export of NOy from continental source regions In contrast to the CO analysis, the residuals in the O 3 assimilation are large, especially in the tropics, reflecting the shorter lifetime of O 3 (and the low density of the TES data)  assimilating trace gases that are more chemical active than CO will be a challenge A better approach for exploiting the satellite data would be to optimize the model parameters, such as the emissions, using adjoint techniques  work is needed to characterize the forecast errors across the range of chemical timescales in the model