Predictability study using the Environment Canada Chemical Data Assimilation System

1. Predictability study using the Environment Canada Chemical Data Assimilation System Jean de Grandpré Yves J. Rochon Richard Ménard Air Quality Research Division WWOSC conference, Montréal August 18th 2014

2. Outline • Global/Regional Chemical Data Assimilation • Ozone predictability and radiative coupling • Results from CDA cycles with ozone assimilation • Summary

3. CDA for improving the Air Quality operational system (RAQDPS) • GEM-MACH as the core model • Comprehensive on-line tropospheric chemistry • Chemical Data Assimilation: 3D-Var/Envar • Assimilation of O3, NO2, CO, AOD … • NRT measurements: GOME-2, SBUV/2, IASI, OMPS, MODIS and surface observations (O3, PM2.5, NO2…) Comprehensive regional CDA system :

4. CDA for improving the Global NWP system (GDPS) Simplified and integrated Global CDA system : • Model : On-line linearized stratospheric chemistry (GEM-LINOZ) • Assimilation of ozone, AOD and GHGs • Chemical Data Assimilation : 3D-Var/Envar • NRT measurements (GOME-2, SBUV/2, IASI, OMPS…) • Radiatively coupled model (ozone heating) • Use of ozone analyses in the NWP DA system • Produce UV-index forecasting (see poster by Y. Rochon)

5. The Global Chemical Data Assimilation system chem Obs Met Analysis chem Obs Met Analysis chem Obs O3 Analysis O3 Analysis O3 Analysis 6-hr forecast 6-hr forecast 6-hr forecast Multi-day Forecast Multi-day Forecast Model: GEM-LINOZ Assimilated observations: GOME-2, SBUV/2, MLS 3D-Var Data Assimilation Independant measurements: ACE-FTS, MIPAS,OSIRIS, OMI, …

6. Assimilation of ozone from MLS • GEM-Global (80 levels, lid=.1 hPa, 33km resolution) • Linearized stratospheric chemistry • 2 months assimilation cycle [winter 2009] • 3D-var Microwave Limb Sounder (EOS-AURA) Day/night measurements ~3500 profiles per day ~ 2.5 km in the vertical Vertical range : [215 - .02 hPa] V2.2 retrievals

7. Anomaly correlation , : Forecast and analysis values : Climatology - ) : ( over the verification area

8. Ozone predictability

9. Column Ozone predictability

10. Ozone radiative coupling

11. NRT ozone measurements 6 hr sample (centered about 0 UTC) on 25 July 2008 Nadir UV-visible Spectrometer (MetOp-A) Total column amounts Day only and cloud free v8 (level-2) retrievals ~80 x 40km resolution ~18 000 measurements per day Nadir Solar Backscatter UV instrument (NOAA-17-18) 20 partial column layers ~3.2km thickness v8 (level-2) retrievals

12. Assimilation of Total Column Ozone Background error standard deviations δQ = (HBHT + R)-1 (z – Hxb) δx = BHTδQ Q : Total column ozone analysis increment at the observation locations xb : ozone mixing ratio z : total column ozone measurements

13. Evaluation of ozone analyses against ozone sondes: O-A (%)[January-February]MLS vs GOME-2

14. MLS vs GOME-2

15. MLS vs GOME-2

16. Evaluation of ozone analyses against ozone sondes: O-A (%)[January-February]GOME-2 vs SBUV/2

17. GOME-2 vs SBUV/2

18. SBUV/2 Partial column retrievals Sample SBUV/2 averaging kernels at ~45 degrees V8 Partial column retrievals “y” δx = K (y – Hxb) Xb: ozone mixing ratio (80 levels) y : partial column ozone (DU) (20 levels) H : vertical integrator New partial column retrievals “z” δx = K (z – AHxb) z : partial column ozone without a priori (DU) (20 levels) A : Averaging kernels matrix (20 levels)

19. Evaluation of SBUV/2 retrievals against ozone sondes: O-A (%)[January-February]With/Withouta priori

20. O-A : SBUV/2 retrievals with/withouta priori

21. SUMMARY/CONCLUSIONS • Anomaly correlation diagnostic based on total column is a useful metric for evaluating ozone analyses system. • CDA cycles using GOME-2 total column measurements and MLS observations have been compared. In the NH, O-A and O-F results are generally within 5%. The column ozone predictability for GOME-2 after 10-days is larger by ~½ day. • CDA cycles using SBUV/2 partial column measurements and GOME-2 have been compared. Results are similar in the NH but significantly worst for SBUV/2 in the SH. • The impact of using different SBUV/2 retrievals on ozone forecasts is negligible.

22. Ozone Column (DU) July, 2008 February, 2009 Observation LINOZ - Observation

23. Evaluation of ozone forecast against ozone sondes: O-F(10-days)[January-February]MLS vs GOME-2

24. Ozone Column (DU) July, 2008 February, 2009 SBUV/2 - Observation LINOZ - Observation

25. Assessment of ozone analyses/forecasts • Total column ozone (July, 2008) • Relative to OMI With SBUV/2 assimilation With GOME-2 and SBUV/2

28. Sample ozone observation distributionTangent point orbit tracks for a 6 hour period (centered about 0 UTC) on 25 July 2008 584 1748 165-300 km along track ~ 2.5 km in the vertical (NRT: 0.2 to 68 hPa) 20 usable partial column layers with ~5 ‘no-impact’ tropo. layers ~3.2 km layers Day only Total column amounts Thinning: 1 degree separation Day only cloud free points 5502

29. July average ozone error standard deviations (%)(before and after adjustment via Desroziers approach and 2Jo/N consideration)MLS SBUV/2 (NOAA 17) GOME-2: 1% applied SBUV/2: A priori removed before assimilation. Averaging kernels applied in assimilation. Sample SBUV/2 averaging kernels at ~45 degrees

30. Background error standard deviations • Prescribed 6 hr ozone background error covariances • Initial values set to 5% of ozone climatology (vmr). • Adjustments to ~3-15% (of vmr) based on the Desroziers approach above =0.7 (from assimilation of MLS and using 30 degree bands). • Below =0.7: Constant extrapolation in absolute uncertainty up to a maximum of 30%. Winter Summer (ppmv) (ppmv) 0.4 0.4 0.6 0.6 0.2 0.2