Presentation is loading. Please wait.

Presentation is loading. Please wait.

Infrared Sounding Data in the GMAO Data Assimilation System JCSDA Infrared Sounding Working Group (ISWG) 30 January 2009.

Similar presentations


Presentation on theme: "Infrared Sounding Data in the GMAO Data Assimilation System JCSDA Infrared Sounding Working Group (ISWG) 30 January 2009."— Presentation transcript:

1 Infrared Sounding Data in the GMAO Data Assimilation System JCSDA Infrared Sounding Working Group (ISWG) 30 January 2009

2 Outline  GMAO atmospheric data assimilation system --- current & next generation  Lessons learned from MERRA (reanalysis)  Observation sensitivity study --- introduction & application to AIRS  Plans Emily, H. C. Liu

3 GMAO Atmospheric Data Assimilation System Next System (3D/4D VAR)  AGCM  Model Adjoint: FV core with simple physics  Analysis: GSI-based 4D-VAR (Todling & Tremolet)  Can switch to GSI-based analysis with IAU procedure (3D-VAR)  Options for minimization algorithm (Conjugate Gradient, Quasi-Newton, Lanczos)  Computation of time- dependent departures (OmF’s)  Observation windowing flexibility  Preliminary version of model- analysis interface Current System  AGCM  Finite-volume dynamic core  Bacmiester moist physics  Physics integrated under the Earth System Modeling Framwork (ESMF)  Catchment land surface model  Prescribed aerosols  Interactive ozone  Analysis  Grid Point Statistical Interpolation (GSI)  Apply Incremental Analysis Increments (IAU) to reduce shock of data insertion NASA GMAO

4 Lessons Leaned from MERRA --- I  The current scheme for radiance bias correction in GSI is variational bias correction (VBC) as described in Derber and Wu (1998). The advantage of VBC is that the bias estimate is adaptive and consistent with all components in the analysis. However, VBC does not work well in regions where observations are sparse. It is prone to include systematic errors from the forecast model.  The analysis can be serious affected by the inclusion of systematic background errors into the bias estimation for observations.  Scan bias and airs-mass dependent bias for each sensor/channel are estimated. y=h(x)+b scan +b air (x)+e inst b scan = b scan (scan position) b airs = β o + ∑β i P i (x); i=1,2, …, N e inst = random instrument error Emily, H. C. Liu

5  The bias estimations for infrared surface channels are misled mostly by biases in the forecast model over land.  The scan angle bias correction is more vulnerable to biases in the forecast model.  The current fix for the problem ---> exclude the surface sensitivity infrared data over land from the analysis  The bias estimations for microwave surface channels do not significantly affected by systematic background error. The relative larger number of microwave surface observations maybe helpful in producing more objective estimations  Need to identify the sources of bias from land surface in GEOS-5  May try to estimate the scan angle bias correction within the analysis. Emily, H. C. Liu

6 Lessons Learned from MERRA --- II  MLS is a limb sounding instrument. It can provide detailed temperature structure from upper troposphere (316 hPa) up to the Mesosphere (0.001 hPa) with 3.5- 14 km vertical resolution.  To obtain statistically significant validation, one month of MLS data were collocated with GEOS-5 analyses. Approximately 95,000 collocations were found and used in the monthly statistics. daily coverage6-hour coverage  AMSU-A channel 14 peaks near stratopause where observations are sparse. The analysis may be benefit from turning VBC off for this channel. VBC ON VBC OFF Difference  In between 10 hPa and 1 hPa, the agreement between MLS temperatures and GEOS-5 analyses is significantly improved for experiments with VBC turned off for AMSU-A channel 14 Emily, H. C. Liu

7 More Evidence --- Validation with AIRS  AIRS channels 73-86 peaking in the stratosphere are not assimilated (passive). Therefore, OB-AN of brightness temperature for these passive AIRS channels can be used as a validation metric.  Biases are reduced significantly indicating the analyzed temperatures in the stratosphere agree better with observed AIRS radiances when VBC is off. VBC OFF VBC ON VBC OFF Emily, H. C. Liu

8 Adjoint Tools for Observation Impact Studies  The difference e a − e b = ∆e is due entirely to the assimilation of observations at 00Z. → measures the impact of the observations  ∆e < 0 indicates that the error of the forecast started from x a is less than that started from x b → the observations are beneficial  ∆e can be estimated as a sum of contributions from individual observations using information from the model adjoint, analysis adjoint, and the innovations  The impact of all or any arbitrary subset of observations (e.g. instrument type, channel, location) can be easily quantified by summing only the terms involving the desired elements of the innovation vector  Estimation of observation impact can be useful in improving data quality control and selection (e.g. data from hyperspectral instrument) summed observation impact analysis adjointmodel adjoint forecast error measure (dry energy, troposphere) analysis equation xbxb xaxa ebeb eaea xvxv observations assimilated background forecast analysis forecast 00Zt+24h t-6h Error Ron Gelaro & Yanqiu Zhu

9 (J/kg) Global Tropics SH NH Total 24hr Forecast Error Reduction due to Observations GEOS-5 Adjoint Data Assimilation System July 2005 00UTC Ron Gelaro & Yanqiu Zhu

10 Observation Impact on GEOS-5 24h Forecast for a Single Case Impact of 500mb Radiosonde Temps  Observations that degraded the 24h forecast  Observations that improved the 24h forecast  Observations that had small impact on 24h forecast Error Reduction Error Increase Impact of AIRS Ch.221 Radiances Error Reduction Error Increase (Reduction in energy-based global error measure for 00UTC 10 July 2005 ) …the statistical aspects of data assimilation imply that there will be a mixture of positive and negative impacts, even for ‘good’ observations. Ron Gelaro & Yanqiu Zhu

11 The Impact of AIRS by Channels  The observation impact study indicates that the some of the AIRS moisture channels have negative impact on the forecast skills  The observation system experiments also indicate that the forecast skills are increased when moisture channels from AIRS were not included Control Control without AIRS moisture channels Positive impact on forecast error reduction Negative impact on forecast reduction Ron Gelaro, Yanqiu Zhu, & Emily Liu

12 Localized examination of AIRS impacts July 2005 00UTC AIRS impact map H20H20 degrade AIRS impact by channel (All Channels) (20-50N, 0-80E) degrade improve Ron Gelaro & Yanqiu Zhu

13 Plans Revisit channel selection and quality control for AIRS using GEOS-5 adjoint tool. Refine bias correction for Infrared radiances IASI assimilation + adjoint tool Literature search --- looking for directions in using cloudy Infrared radiances in GEOS-5 Assess the quality of AIRS cloud cleared radiances with GEOS-5 background --- experimental Radiance monitoring Emily, H. C. Liu


Download ppt "Infrared Sounding Data in the GMAO Data Assimilation System JCSDA Infrared Sounding Working Group (ISWG) 30 January 2009."

Similar presentations


Ads by Google