28
Analysis scheme for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team • Value of Aura data for data fusion studies; Resolution Kernels and Scale-consistent DA schemes First UTLS ozone analysis results with HIRDLS data (V13); Quality Control of HIRDLS O3 data. Cross isentropic filaments in the middle and upper stratosphere Current studies and future plans

Analysis scheme for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

  • Upload
    jania

  • View
    41

  • Download
    0

Embed Size (px)

DESCRIPTION

Analysis scheme for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team. Value of Aura data for data fusion studies; Resolution Kernels and Scale-consistent DA schemes First UTLS ozone analysis results with HIRDLS data (V13); Quality Control of HIRDLS O3 data. - PowerPoint PPT Presentation

Citation preview

Page 1: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Analysis scheme for HIRDLS/Aura retrievals

Valery Yudin and HIRDLS Science Team

• Value of Aura data for data fusion studies; Resolution Kernels and Scale-consistent DA schemes

• First UTLS ozone analysis results with HIRDLS data (V13); Quality Control of HIRDLS O3 data.

• Cross isentropic filaments in the middle and upper stratosphere

• Current studies and future plans

Page 2: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

01.23.2006: OMI data and GEOS-5.01 ozone columns

Page 3: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

HIRDLS and MLS capability to observe the UTLS ozone structures at ~20N, 155 W

24-01-2006Dashed lines are vertical

resolutions:

HIRDLS

MLS

GEOS-5

Page 4: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

GEOS-5.01 PV: ozonesonde location and HIRDLS orbit for 24.01.2006

Page 5: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

MLS and HIRDLS orbits

HIRDLS spatial sampling along the orbits (dx~1 deg) is consistent to the HIRDLS vertical resolution of retrievals (dz~0.7 km spacing).

For extra-tropics:

dx/dz ~ N/f

Scale-consistent sampling gives opportunity to study baroclinic disturbances and low-frequency waves in the UT and stratosphere.

Page 6: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

UTLS observed and analyzed Ozone and PV structures along satellite orbits: 01/23/2006

Page 7: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Resolution Ozone Kernels of N-V instruments /DFS ~ 0.7 – 3.5/ and L-V (DFS ~N vertical levels)

OMI-TOMS Layer Efficiency Factors

TES Kernels

Page 8: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

HIRDLS O3streamers and some “GW” O3signatures cannot be fully supported by

analyzed winds

Vertical Mapping of data: HzXo =>

Projecting HIRDLS Data => GEOS5 grid

e.g. GW –control (averaging or digital filters......)

Horizontal Mapping of forecast: HsXf =>

Binning “high-resolution” forecast to HIRDLS footprint

Joint Observation-Forecast Space:

Jobs = (HzXo – HsX)Wdd(HzXo – HsX)T

Page 9: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Specifics of trial HIRDLS ozone data analysis

• Model –> Tracer version of MOZART models, 3D daily mean ozone production and loss terms; IC –January monthly averaged WACCM ozone.

• GEOS-5.01 – Transport (72 levels with degraded to 2ox2o hor. resolution, “LAPTOP” version).

• DA –scheme: sub-optimal Kalman Filter with state-dependent stochastic model error growth to improve data weight in large spatial gradients of O3

• Quality Control: PV-O3 correlations + Large OmF are discarded Tropical UTLS ozone (P > 100 hPa) is not assimilated Additional Control using 12.1 k aerosol cloud flag Only data with relative accuracy of ~25% or better => to analysis

• Experimental Data Version of HIRDLS retrievals: V2.04.13 (with 2 km-altitude shift.....) processed only for Jan 2006 /for UTLS, released V2.04.09 => high O3, experimental versions V2.04.13-16 can fix some issues for ozone DA studies./

Page 10: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

01.23.2006:Orbital (HIRDLS) PV-structures GEOS-5.01, GEOS-4, GFS and GEOS-5.1.0 (not

available)

Page 11: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

PV-O3 correlations along MLS and HIRDLS orbits (20o-70o N)

Page 12: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Jan 23 2006: Two versions of HIRDLS O3

Page 13: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

UTLS O3 : Analysis and CTM forecast driven by GEOS-5.01 winds

Good News, CTM-

forecast supports

HIRDLS/MLS ozone

streamers

Page 14: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Data and DA results: linear (unbiased) correction of O3-forecast by data is performed

Quality control

Issue ?

Page 15: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Data assimilation of HIRDLS O3 retrievals in the UTLS

• Current GEOS-5.01 O3 analyses of column/sub-column ozone satellite data have some issues with resolution of vertical structures e.g. representation of ozone laminaes, and intrusions of air masses across and along the tropopause.

• HIRDLS UTLS ozone retrievals can be scale-consistently assimilated in the CTM driven by GEOS-5 transport.

• Consistency in the horizontal (along the orbit) and vertical resolution of UTLS retrievals is unique feature of HIRDLS for data assimilation studies.

• Some filtering of GW signatures in the retrieved O3 unsupported by GEOS-5.01 dynamics should be performed to achieve optimal constrain of ozone forecast by HIRDLS data. Revision of retrieval errors are also expected.

Page 16: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

23/01/2006: Stratospheric O3, HIRDLS and MLS (relatively small orbital data-data differences)

Page 17: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Examples of stratospheric O3-analyses

Page 18: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Weighted with density PV-field (MPV) and

HIRDLS O3

Discussions on the MPV conservation laws in Lait 1994,

Muller & Gunther, 2003

Page 19: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Stratospheric vertical filaments seen by MLS O3 and N2O retrievals: 01/23/2006

Page 20: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Current Studies & Future Plans

• We plan to proceed in O3 N2O and HNO3 multi-instrumental (MLS/HIRDLS) DA studies in WACCM-GEOS5 stressing on UTLS-region.

• Move DA studies to 1ox1o (0.5x.05) CTM resolutions.

• Optimize data analysis scheme for chemically-active regions (include diurnal cycles in P-D terms)

• Plans to look at the residual tropospheric O3 columns using OMI data.

• CO: radiance data assimilation schemes for CO (MOPITT) and bias-corrected MLS CO retrievals

• Demonstrate a power of Aura scale-consistent chemical observations to constrain transport ?

• Participate in campaigns to provide the UTLS tracer forecast constrained by Aura chemicals.

Page 21: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

CO across the tropopause: MOPITT and MLS,

model and assimilation of MOPITT in CTM 01/2005: CO, MLS-MOPITT

-40 -20 0 20 40 60 80 Latitude

0

5

10

15

20

25

30

35 01/2006: CO, MLS-MOPITT

-40 -20 0 20 40 60 80 Latitude

0

5

10

15

20

25

30

35

Jan: WACCM-CO, ppbv

-40 -20 0 20 40 60 80 Latitude

0

5

10

15

20

25

30

35 Jan: Assim-CO, ppbv

-40 -20 0 20 40 60 80 Latitude

0

5

10

15

20

25

30

35

Page 22: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

IR CO Retrievals in the tropopshere =>Assimilating Profiles or Partial Columns

MOZART CO-MODELMOPITT CO, May 2000

Profiles AssimAssimilating Partial CO Sub-columns:

1) Use the Total Column Kernel Vector to evaluate Data minus Forecast CO column deficit

2) Update Partial Model Columns according to standard statistical estimation

3) The Vertical Structure of CO analysis is less damaged by extra-smoothing from retrieved profiles

Page 23: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Orbital plots: GEOS-5.01 O3 analysis and NV TES O3 retrievals

• Possible explanations why O3-analysis cannot resemble PV-structures may be addressed to the analysis schemes of column-based data that can degrade thin low-ozone streamers.

• For example, TES O3 retrievals tend to produce the low-O3 values between 20o-40o N. However, assimilating TES “smoothed” profiles can degrade ozone streamers seen by HIRDLS and MLS.

• To prevent ozone streamers algorithms should adjust only observable scales. Additional tracer forecast, PV-O3 correlations may serve to identify shortcomings of analysis schemes that work with the sub-column ozone data

Page 24: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Biases in DA and inverse estimation studies /example of wavy T-biases in the stratosphere/

Attractive feature of these explorative for DA:

Before assimilation of data they diagnose and attempt to suppress the large model biases operating with model physics and persistent OmF differences.

Dee, 2005

NP SPSTRAT: AMSU-A rad-es

Page 25: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Scatter plots Trop. O3 column estimations with various definitions of the tropopause boundaries

Page 26: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Total and Trop. Columns Estimations: DAS vs OMI

Page 27: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Math summary for scale-consistent and rank-deficient computations of analysis increments for Dw/Lc >> 1

• Scale-consistent analysis

• For deep layer sensitive channels Dw/Lc >>1 V-shapes W-shapes, e.g. Gaussian shapes. <dT>-increment is not affected by “wavy” vertical

correlations.

• Rank-deficient analysis schemes are close to direct use of OE formulae or linear filters that ignore consistency of scales => extra-sensitvity

• K =WCff’[WCffW’+Cbb]-1

<dT> = K<dTb>, DFS = tr(KW) ~.5-2

For DFS~[0.5-2]. K-vector can be modulated by the forecast errors on scales invisible to the instrument.

Adjustment of fine-scale structures and errors by deep-layer sensitive channels is a signature of extra-sensitivity of the inverse projection from data space => forecast.

• Scale-inconsistent Error Analysis:

[Can ]-1= [Cff]-1 + W[Cbb]-1WT

Mixture 2km 10 km of scales Cor. Length Width of W

• Wavy structure of analysis <dT> initiates spurious “DA” temperature waves

• SVD of W provides natural tapering of vertical correlations and fine structures in T-variances invisible for AMSU radiances.

look at SVD of Jacobians

( )

&

e.g. Y-point <=>X-point

,

b

T

T Tb

T Tb

svd

dT WdT

W USV

U dT S V dT

dX V dT dY U dT

dY SdX

dX K dY dT V dX

Page 28: Analysis scheme  for HIRDLS/Aura retrievals Valery Yudin and HIRDLS Science Team

Challenges in the MA data assimilation• DA of radiances from deep-layer

sensitive channels (AMSU-10:14) in SMLT /Dee, Polavarapu et al., 2005/.

• Two scales of inverse solution: vertical width of Jacobians (Dw) and vertical correlation lengths (Lc): Dw/Lc >>1.

• In rank-deficient schemes( Dw/Lc >>1) initiates “wavy” T-increments that are not bounded by W, AMSU Jacobians;

• In areas of high-density data insertion, analysis can be damaged by persistent errors related to scale-inconsistent projections of radiance misfits onto model levels (polar DA waves).

• In DA of AMSU data dT-analysis increments adjust layer averaged values rather than T-profiles.

• dT-anal spreads between model levels due to wide width of W-Jacobian and should be insensitive to short-scale T-correlations and variances.

W-Jacobian

T-corr.

T-Var

dT-anal

Dw

Lc

DA wave

Dw/Lc >>1 <=> DFS << N-levels