ozone data assimilation

45
Ozone Data Assimilation K. Wargan S. Pawson M. Olsen A. Douglass P.K. Bhartia J. Witte bal Modeling and Assimilation Office

Upload: irving

Post on 24-Feb-2016

61 views

Category:

Documents


0 download

DESCRIPTION

Ozone Data Assimilation. K. Wargan. S. Pawson M. Olsen A. Douglass P.K. Bhartia J. Witte. Global Modeling and Assimilation Office. Motivation and challenges. Ozone is important Controls temperature in the stratosphere – long range forecasts should start from realistic ozone - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Ozone Data Assimilation

Ozone Data AssimilationK. Wargan S. Pawson

M. OlsenA. DouglassP.K. BhartiaJ. Witte

Global Modeling and Assimilation Office

Page 2: Ozone Data Assimilation

2

Motivation and challenges Ozone is important

Controls temperature in the stratosphere – long range forecasts should start from realistic ozone

Pollution in the troposphere, greenhouse gas Modern reanalyses have ozone

Not easy to assimilate Complex, variable structure Limited data coverage Limited vertical resolution

What we do to make it work Fine tuning of background errors More data sources (MLS, radiances)

Page 3: Ozone Data Assimilation

3

Plan of talk

• Description of the ozone data and the GEOS-5 data assimilation system

• The impact of assimilating averaging kernels• Background error covariance modeling• Assimilation of the Microwave Limb Sounder

(MLS) ozone• Towards direct assimilation of MLS radiances

Page 4: Ozone Data Assimilation

4

Ozone Data

• Solar Backscatter Ultraviolet Radiometer 2 (SBUV/2)

• Ozone Monitoring Instrument on EOS Aura (OMI)

• Microwave Limb Sounder on EOS Aura (MLS); currently only in research analyses

Page 5: Ozone Data Assimilation

5

Solar Backscatter Ultraviolet Radiometer 2 (SBUV/2)

SBUV measures backscattered UV radiation.Ozone is retrieved in 21 layers, ~3 km thick, compare to 72 GEOS-5 layers.Footprint: 170 × km 340 kmVery good agreement with independent data.Long data record (1970s to present)Small structures are not resolved – poor resolution below the ozone maximum.

NOAA 17, 18, 19

700

100

10

1

0.1

90S 60S 30S EQ 30N 60N

[DU]

Partial columns / Dobson Units

12 Z

Page 6: Ozone Data Assimilation

6

Ozone Monitoring Instrument (OMI)

On EOS Aura Nadir viewing geometry Measures radiance in visible and ultraviolet Footprint 13 km × 24 km Retrieved species: ozone, NO2, SO2, BrO, OClO, aerosols

Observation locations and ozone total column, 10/08/2012 12 Z

Page 7: Ozone Data Assimilation

7

All ozone data

Data coverage, 12Z

OMI Total ozone column

SBUV – vertical information

SBUV and OMI observe sunlit atmosphere

Page 8: Ozone Data Assimilation

8

All ozone data

Radiance data - meteorology

All observations within a 6h window

~2,500,000

Ozone observations

~20,000

Data coverage, 12Z

OMI Total ozone column

SBUV – vertical information

SBUV and OMI observe sunlit atmosphere

Page 9: Ozone Data Assimilation

9

The GEOS-5 Data Assimilation System• Atmospheric General Circulation Model:

• Horizontal resolution: flexible - 2.5° to ¼°• 72 layers from the surface to 0.01 hPa• Parameterized ozone chemistry (stratospheric P&L; dry deposition)

• 3D-Var analysis: • Gridpoint Statistical Interpolation• developed in collaboration with NCEP

• Observations: • Conventional (surface, sondes, radar, aircraft, MODIS-derived winds,…)• Satellite radiance data (TOVS/ATOVS, AIRS, IASI, SSM/I, GOES, GPS)• Ozone data

Page 10: Ozone Data Assimilation

10

A priori profile

efficiencyfactor

[DU]

1000

10

100

Pres

sure

[hPa

]1

1000

10

100

1

8.4S, 130.5W

0 20 40 60 80 0 1.0 2.0 3.0

OMI – efficiency factors (averaging kernels)

Reduced sensitivity near the surface

The retrieved ozone column is a combination of the true signal and a priori climatology.The a priori is removed in the assimilation

Page 11: Ozone Data Assimilation

11

ERRORxxFO

xxxxh

ERRORxxxy

forecasttruei

ii

priorforecasti

ii

priorforecast

priortruei

ii

priorobs

)(

)()(

)(

11

1

11

1

11

1

OMI – efficiency factors (averaging kernels)

The retrieved ozone column is a combination of the true signal and a priori climatology.The a priori is removed in the assimilation

Page 12: Ozone Data Assimilation

12

Generally reduced sensitivity near the surface. Efficiency factors can take values >1 due to multiple scattering

Page 13: Ozone Data Assimilation

13

Analysis without efficiency factorsAnalysis with efficiency factors

90S 60S 30S EQ 30N 60N 90N

90S 60S 30S EQ 30N 60N 90N

-0.2

0.0

0.2

0.4

0.6

0.8

0.0

-0.5

0.5

[ppb

v/da

y][p

pbv/

day]

Tendencies introduced by analysis

Reduced impact of ozone data near the surface

The system’s response to efficiency factorsZonal mean analysis tendency near 900 hPa

Zonal mean analysis tendency near 1000 hPa

Page 14: Ozone Data Assimilation

14

Tropospheric column relative difference: With minus without efficiency factors February 2007

[%]

OMI efficiency factors – the impact on assimilated tropospheric column

Page 15: Ozone Data Assimilation

15

The model – ozone chemistry

• Parameterized ozone production and loss rates in the stratosphere

• No chemistry is implemented below the tropopause – observations are the only source of information on ozone distribution in the free troposphere

• Dry deposition mechanism removes surface ozone

Accurate representation of ozone sources near the surface (e.g. anthropogenic ozone precursors) is needed to compensate for limited sensitivity of observations

Page 16: Ozone Data Assimilation

16

SOME RESULTS

Page 17: Ozone Data Assimilation

17

2010

2011

Springtime maximum (2009)

Springtime maximum (2011)

Antarctic ozone hole (2010)

The annual cycle of total ozone

northsouth[DU]

Springtime maximum (2010)

Antarctic ozone hole (2009)

2009

Page 18: Ozone Data Assimilation

18

y=0.94x+4.44R = 0.98RMS diff = 12.78 %Bias = 0.5 %

0

50

100

150

200

250

Assim

ilatio

n [D

U]

0 50 100 150 200 150Ozone sondes [DU]

Ozone integrated between the tropopause and 50 hPa

Comparisons with ozone sondes

Very good agreement with ozone sonde data in the lower stratosphere.

Important for climate forcing by ozone and stratosphere – troposphere exchange

Page 19: Ozone Data Assimilation

19

Transport using assimilated winds leads to realistic ozone profile structure in the UTLS

Assimilation of SBUV in GEOS-5 does not show this vertical structure: smoothing from the assimilation process

[ppmv]

Pres

sure

[hPa

]Pr

essu

re [h

Pa]

PV contours (white) and ozone (shaded) from the GEOS-5-driven CTM

HIRDLS retrievalsCTMSBUV analysis

Pres

sure

[hPa

]

Ozone [mPa]

Assimilated ozone from GEOS-5 using SBUV/2 data

Page 20: Ozone Data Assimilation

20

Background errorsGEOS-5/MERRA• NMC Method• 2-D lookup table of variances

and correlation length scales• High ozone gradients in the

UTLS are not resolved• Statistics-based; real-time

dynamics not accounted forBackgroundMean O3

tropopause

Ozone mixing ratioAl

titud

e

Page 21: Ozone Data Assimilation

21

Background errorsGEOS-5/MERRA• NMC Method• 2-D lookup table of variances

and correlation length scales• High ozone gradients in the

UTLS are not resolved• Statistics-based; real-time

dynamics not accounted forObservedMean O3

BackgroundMean O3

tropopause

Ozone mixing ratioAl

titud

e

Page 22: Ozone Data Assimilation

22

Background errorsGEOS-5/MERRA• NMC Method• 2-D lookup table of variances

and correlation length scales• High ozone gradients in the

UTLS are not resolved• Statistics-based; real-time

dynamics not accounted forObservedMean O3

BackgroundMean O3

tropopause

Ozone mixing ratioAl

titud

e

Background error correlation, vertical extent

Page 23: Ozone Data Assimilation

23

Background errorsGEOS-5/MERRA• NMC Method• 2-D lookup table of variances

and correlation length scales• High ozone gradients in the

UTLS are not resolved• Statistics-based; real-time

dynamics not accounted forObservedMean O3

BackgroundMean O3

tropopause

Ozone mixing ratioAl

titud

e

After assimilation: the analysis increment “leaks” through the tropopause, vertical gradient is reduced

Background error correlation, vertical extent

Page 24: Ozone Data Assimilation

How can we take advantage of data without damaging transport-

induced small-scale structures?

So...

Page 25: Ozone Data Assimilation

25

Background errorsGEOS-5/MERRA• NMC Method• 2-D lookup table of variances

and correlation length scales• High ozone gradients in the

UTLS are not resolved• Statistics-based; real-time

dynamics not accounted for

Alternate approach• Proportional method• A possible candidate for σ2:

]/[32 ][ molmolO

[O3] – background ozoneα – specified parameter

• Process-based

Page 26: Ozone Data Assimilation

26

2007 May 9th, SBUV/2 ozone assimilated

Sonde profileNMC errorsNew errors

Stony Plain53.5N, 114W

0 5 10 15 20 25Ozone [mPa]

1000

10

100

Pres

sure

[hPa

]

Sonde ozoneAnalysis, new errors

0 5 10 15 20 25Ozone [mPa]

New vs. old background errors

Bratt’s Lake51N, 104W

Page 27: Ozone Data Assimilation

27

Latitude

Thet

a [K

]Th

eta

[K]

Thet

a [K

]Ozone field in the Upper Troposphere – Lower Stratosphere. Fine structures

The High Resolution Dynamics Limb Sounder (HIRDLS) detects small scale structures in the ozone field near the tropopause.

These structures are often absent in SBUV/2 & OMI analysis but do get captured with the new background error model

Page 28: Ozone Data Assimilation

28

Interpolate to theta (only above 260 hPa)

Average profiles in 2° latitude bands

Determine lamina bottom and top

Apply thickness and magnitude criteria

Lamina must be coherent across 3 mean profiles

Lamina Identification

Page 29: Ozone Data Assimilation

29

Number of profiles with Laminae April 2007

Total # 1131

Counting the laminae

Latitude

Thet

a [K

]

HIRDLS data

SBUV assim, NMC errors SBUV assim., New errors

Total # 12 Total # 283

Thet

a [K

]

Latitude Latitude

The new background errors lead to large improvements in the representation of the UTLS ozone

Page 30: Ozone Data Assimilation

30

Conclusions so farSBUV and OMI based ozone analyses produce ozone

fields which are in a reasonable agreement with ozone sonde data in terms of vertically integrated partial columns

Small vertical features near the tropopause are not always represented

Replacing NMC, static background error variances with a process dependent error model leads to a better representation of these features (as compared with independent data)

Page 31: Ozone Data Assimilation

31

Microwave Limb Sounder on EOS Aura

The EOS Aura satellite was launched on July 15th 2004

Page 32: Ozone Data Assimilation

32

Microwave Limb Sounder on EOS Aura

MLS measures atmospheric limb emissions in five spectral regions centered around 118 GHz , 190 GHz, 240 GHz, 640 GHz, and 2.5 THz

Designed to measure atmospheric temperature and composition including ozone, moisture, CO, ClO, SO2, cloud ice, ...

Ozone profiles are retrieved at relatively high vertical resolution (varies with version) between ~260 hPa and top of the atmosphere

Page 33: Ozone Data Assimilation

Why assimilate MLS ozone• Assimilation of limb sounder

data improves the representation of fine scale vertical structures

• Global day and night coverage• Possibility to do it in NRT if

either NRT MLS retrievals or MLS radiances are used

• A template to assimilate other point measurements (e.g. NPP OMPS-Limb Profiler)

33

Legionowo, 52.4N, 21E Apr 6th 2005

Sonde profileSBUV analysisMLS analysis

0 5 10 15 20 25Ozone [mPa]

10

100

1000

Pres

sure

[hPa

]

We can assimilate: Retrieved data – already implemented in GSI Radiance data

Page 34: Ozone Data Assimilation

34Latitude

Thet

a [K

]Th

eta

[K]

Thet

a [K

]

High resolution data combined with state-dependent background error covariance model reproduces the structure of the UTLS ozone field

Vertical structure near the tropopause

Page 35: Ozone Data Assimilation

35

HIRDLS data

SBUV assim., New errors

Number Profiles with Laminae April 2007

• MLS assimilation has the most faithful representation of ozone structure

Total # 1131 Total # 821

Total # 283

UTLS ozone laminae in GEOS-5 – comparison with High Resolution Dynamics Limb Sounder

Latitude Latitude

Thet

a [K

]Th

eta

[K]

MLS assim. old errors

MLS assim. New errors

Total # 934

Page 36: Ozone Data Assimilation

36

MERRA (SBUV/2)MLS+SBUV, GEOS-5.6.1 Station data

400

300

200

100

0

Tota

l ozo

ne co

lum

n [D

U]

Sep Oct Nov 2010

The 2009 Antarctic ozone hole

Oct 18thMLS+ SBUV analysis agrees well with data from the South Pole balloon sondes.

SBUV analysis (without MLS) overestimates ozone over the South Pole by almost 100 DU in September. Good agreement in October and November

Page 37: Ozone Data Assimilation

37

MLS+SBUV analysis Total Ozone, Oct 15th, 6Z And SBUV coverage

MLS+SBUV analysis Total Ozone, Sep 3rd, 6Z And SBUV coverage

SBUV analysis Total Ozone, Sep 3rd, 6Z

SBUV analysis Total Ozone, Oct 15th, 6Z

The polar night region is not constrained by SBUV observations until October. The polar ozone is overestimated in September

MLS provides near global coverage

Page 38: Ozone Data Assimilation

38

ASSIMILATION OF MLS RADIANCES

Page 39: Ozone Data Assimilation

39

Why radiances?

Retrieved product is always affected by priorsIn case of MLS temperature the priors come

from GEOS-5 analysis – self-contamination of the system if these data are assimilated

By assimilating radiances we avoid additional errors resulting from the retrieval process

Page 40: Ozone Data Assimilation

40

The MLS viewing geometry

125 vertical scans per 25 s in the direction of motionEach scan records emissions at a number of distinct frequenciesBand 7 radiances, centered at 240 GHz are sensitive to ozone, 25 channels

Page 41: Ozone Data Assimilation

41

Information from MLS radiances• Contrast: ozone concentration• Breadth: tangent pressure• Position: baseline/extinction

We need all three pieces. In the current implementation

only the contrast is assimilated. We use previously retrieved

tangent pressure data.Implementation of online baseline

retrieval is underway

Observed minus simulated radiances

Page 42: Ozone Data Assimilation

MLS profile v3.3, 82SRadiance assimilation

Some results

42

Comparison with MLS V3.3 retrieved ozone. A single profile

Ozone [ppmv]

Pres

sure

[hPa

]

Ozone [ppmv]Pr

essu

re [h

Pa]

Reasonable agreement in mid- and upper stratosphere – sanity check passed

Mean MLS, 60N-90NRadiance assimilation

Page 43: Ozone Data Assimilation

43

Equator60S60N30S30N

Relative RMS difference [%]

Relative RMS difference: Assimilation of MLS v3.3 minus radiance assimilation

Agreement within 5% in mid to upper stratosphere except southern high latitudes.The lower stratosphere is expected to improve once the extinction retrieval is implemented

Page 44: Ozone Data Assimilation

44Ozone [mPa]

MLS v3.3 analysisRadiance analysis

The MLS v3.3 analysis (assimilation of retrieved MLS ozone) exhibits oscillations in the tropical lower stratosphere – even in the zonal mean.

These are not seen in radiance assimilation

Lower stratospheric zonal mean profile at the equator

Page 45: Ozone Data Assimilation

45

Summary New process-based background error covariance model

leads to significantly improved representation of fine scale structures in assimilated ozone

High vertical resolution ozone data (MLS) brings further improvements to the assimilated product – the structure of the ozone field in the Upper Troposphere – Lower Stratosphere layer are well represented

The impact of assimilating OMI efficiency factor has yet to be fully assessed

Direct assimilation of MLS radiance data (in ozone and temperature bands) is underway.