Using data assimilation to combine TOC data sets
Using data assimilation to combine TOC data sets: The Multi-Sensor Reanalysis (MSR) ozone data record Ronald van der A, Marc Allaart, Henk Eskes, Michiel van Weele Royal Netherlands Meteorological Institute (KNMI)
Constructing the Multi-Sensor Reanalysis of ozone
Multi Sensor Reanalysis (MSR) of ozone Assumption: • The ground observations are on average a good approximation for the truth. Procedure: • All UV-VIS satellite data (TOMS, SBUV, GOME, SCIAMACHY, OMI, GOME-2) in the period 19782008 is used. • Step 1 : Correct satellite data to avoid biases. The reference data that is chosen are ground data observations from reliable WOUDC stations. • Step 2 : Satellite data is assimilated in a chemical-transport model to achieve complete global and temporal coverage. • Multi Sensor Re-analysis (MSR), version 1, available at www. temis. nl • Published as: R. J. van der A , M. A. F. Allaart and H. J. Eskes, Multi sensor reanalysis of total ozone, Atm. Chem. Phys. , 2010, 11277 -11294.
Reference data set: • From WOUDC 91 ground stations are selected with a long and reliable dataset (Fioletov et al. , 2008) • Dobson, Brewer(3, 4)–instruments (no filter-instruments used) • Dobson corrected for temperature dependence (Kerr et al. , 2002)
Corrections satellite data Expected dependencies of satellite data: Parameter Physical mechanism Solar zenith angle Light path Viewing zenith angle Scan mirror Effective temperature O 3 cross-section Time (trend) Instrument degradation Offset Calibration • Generate time series of the 14 satellite data sets for all stations. • Fit all time series as function of the 5 parameters. • Apply corrections as function of the fit parameters to construct the Multi-Sensor Reanalysis (MSR) level 2 data
Correction of level 2 data After correction Before correction Satellite minus Brewer observations for the Uccle ground station
Data assimilation of the MSR level 2 data • Level 2 data: – Days without observations (longest data gap is 4 days) – Data is on satellite footprint – Measurements on several points during the day • => Therefore, data assimilation is used to create a homogene data record Note: because of the continuously feeding of observations, the long-term (> few days) performance of the model is not affecting the results.
MSR (level 4) results - Total ozone field every 6 hours - Spatial resolution is 1 x 1. 5 degree - Daily local time ozone field at noon (for UV index) Fitted offset (DU) between MSR level 4 and ground observations
Comparison to the SPARC/AC&C ozone database
Comparison of MSR and AC&C/SPARC Historical ozone (and UV) 1980 -2010: • Ozone observations: MSR extended with Sciamachy – Sciamachy corrected with MSR method and assimilated • Ozone climatology from AC&C/SPARC (for CMIP 5) – No dynamics included – Zonal averaged stratosphere Future ozone (and UV): scenario in period 2010 -2020: • RCP 4. 5 • SCIAMACHY 2010 -2011
Intercomparison for De Bilt
AC&C SPARC ozone versus MSR (annual zonal mean)
Development for a new version MSR 2 data set
Improvements for version 2 Level 2 data correction: • Extending data until 2012 • Adding BUV data to start in 1970 • Latest versions of level 2 data sets • Multiplicative correction added Data assimilation: • Use of ERA-interim • New chemistry parametrisation for TM
MSR-2 Multi sensor reanalysis of total ozone 1970 -2012 Ronald, Henk, Michiel, Marc Step 1: collection of 18 datasets Step 2: calculation of correction coefficients Step 3: generation of MSR-2 level-2 dataset Step 4: data assimilation Step 5: analysis of the results Correction to WOUDC (Brewer) data For each of the 18 datasets Xcor = A 1*Xsat + A 2*year + A 3*VZA + A 4*SZA + A 5*Teff New chemical parameterization of TM (including time-dependence) Cariolle v 2. 9 or COPCAT
MSR-2 preliminary corrections DU per [year, degree, DU]
Summary 1979 1984 1989 1994 1999 2004 2009 Multi Sensor Reanalysis of total ozone: • 18 total ozone data sets from TOMS, SBUV, GOME, SCIAMACHY and OMI are corrected by comparison with Brewer and Dobson data (WOUDC) • These corrected data sets are assimilated with TM-DAM on a regular grid. Comparison to AC&C/SPARC: • In general good agreement. • Zonal differences mid-latitudes related to high UTLS variability • Polar differences related due sparse satellite data and data assimilation in polar night
Outlook • Version 2 of the MSR for the period 1970 -2012 is in development. • Is it possible to apply this method to ozone profiles ? – Bias correction is much more complicated – The algorithm for data assimilation of profiles is available (O 3_cci), … –. . . but the vertical distribution still has to be validated especially for long assimilation runs.
- Slides: 18