Infrared Limb Sounding MIPAS and HIRDLS - ECMWF 2 Atmospheric, Oceanic & Planetary Physics, University of Oxford A Dudhia ECMWF Workshop 23-26 June 2003 Infrared Limb Sounding MIPAS

Page 1

Atmospheric, Oceanic& Planetary Physics,University of Oxford

ECMWF Workshop 23-26 June 2003A Dudhia

Infrared Limb SoundingMIPAS and HIRDLS


Anu DudhiaAtmospheric Oceanic and Planetary Physics

University of Oxford

Page 2




MIPASMichelson Interferometer for Passive Atmospheric Sounding

Fourier Transform InterferometerLaunched on Envisat 1st March 2002~1 000 profiles per day

HIRDLSHigh Resolution Dynamics Limb Sounder

Filter RadiometerTo be launched on Aura early 2004 ~10 000 profiles per day

Page 3




IntroductionLimb Viewing GeometryChannel SelectionRadiative Transfer ModellingRetrieval Schemes

Page 4




Limb-sounders compared to nadir-sounders

Better vertical resolution – 3km MIPAS, 1km HIRDLSBetter upper vertical range – mesosphereWorse horizontal resolution – 100s kmMore sensitive to cloud – fewer gaps in limb view

Also: radiative transfer & retrieval more complicated (2D)

Page 5




(satellite moving clockwise in both figures)Limb Viewing Geometry

HIRDLS – fast limb scanTangent point slopes upwardstowards satellite

MIPAS – slow limb scan Tangent point slopesdownwards following satellite

Page 6




MIPAS Nominal Scan

-1000 -500 0 500 1000Distance along-track [km]

0

20

40

60

80

100

Alti

tude

[km

]

Page 7




HIRDLS Scan (single azimuth track)

-1000 -500 0 500 1000Distance along-track [km]

0

20

40

60

80

100

Alti

tude

[km

]

Page 8




Spectral CoverageMIPAS685-2410cm-1 at 0.025cm-1 spacing, 5 bands2-7 Microwindows up to 3cm-1 width selected for each species

HIRDLS21 filter channels from 570-1600cm-1

1-4 Filters up to 120cm-1 width assigned to specific species

Page 9




Radiance Contributions for 21km Tangent Height

500 1000 1500 2000 2500

0.1

1.0

10.0

100.0

1000.0

10000.0

Rad

ianc

e [n

W/(

cm2 s

r cm

-1)]

Total Radiance

CO2 H2OO3 CH4N2O NO2HNO3

500 1000 1500 2000 2500Wavenumber [cm-1]

0

20

40

60

80

Alti

tude

[km

]

A AB B C D

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

Page 10




Radiance Contributions for 21km Tangent Height

500 1000 1500 2000 2500

0.1

1.0

10.0

100.0

1000.0

10000.0

Rad

ianc

e [n

W/(

cm2 s

r cm

-1)]

Total Radiance

F11 F12 CLONO2N2O5 CONO

Aerosol

500 1000 1500 2000 2500Wavenumber [cm-1]

0

20

40

60

80

Alti

tude

[km

]

A AB B C D

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

MIPAS Microwindows

HIRDLS Channels

Page 11




MIPAS Microwindows

Selected to minimise total retrieval error for given CPU cost

Use spectral masks to exclude measurements

Computer, rather than human, selection

Also results in systematic error budget

Page 12




PT__0004 Microwindow

Radiance spectrum

Temperature Jacobian

Pressure Jacobian

H2O Interference

Spectral masks

Page 13




Page 14




0 500 1000 1500 2000dI/dT [nW/(m-2 sr cm-1) K-1]

0

20

40

60

Alti

tude

[km

]

Ch5 Ch4 Ch3 Ch2

5 km

10 km

15 km

20 km

25 km

30 km

35 km

40 km

45 km

50 km

HIRDLS pT ChannelsLarge area = good S/N

Sharp peak = good resolution

Require 4 CO2 Channels to cover complete vertical profile

Temperature Jacobians

Page 15




Radiance Modelling

Page 16




Ray TracingNo analytical solution for limb geometry- numerical integration required

Usually adequate to consider atmosphere locally spherical- radius of curvature varies by ±0.5% around orbit

Refraction is significant- deflection 70m at 25km, increases in proportion to density

Page 17




Transmittance CalculationsLimb viewing requires high spectral resolution calculations- typically 0.0005cm-1 to capture Doppler broadened lines

1650 1651 1652 1653Wavenumber [cm-1]

0

100

200

300

400

Rad

ianc

e [n

W/(

cm2 s

r cm

-1)]

15km30km47km

1650 1651 1652 1653Wavenumber [cm-1]

0

100

200

300

400

500

600

Rad

ianc

e [n

W/(

cm2 s

r cm

-1)]

15km30km47km

MIPAS measurement Atmospheric spectrum

Page 18




Irregular Grids

1652.65 1652.70 1652.75 1652.80 1652.85 1652.90 1652.95Wavenumber [cm-1]

0.1

1.0

10.0

100.0

Rad

ianc

e [n

W/(

cm2 s

r cm

-1)]

15km30km47km

MIPAS solution is to compute radiative transfer only at a subset of fine gridpoints, determined from spectral structure in eachmicrowindow

Page 19




Look Up TablesMIPAS uses look-up tables of absorption coefficient k(ν,p,T) for transmittance of each cell, containing absorber amount u

τc = exp ( - k u )

Path transmittance is then product of cell transmittances (Beer’s Law) τ = ∏ τc

Finally radiance is convolved with instrument line shape ΨR = ∫ ∫ B Ψ dτ d ν

Page 20




Band ModelHIRDLS solution is to use channel-averaged transmittances

∫ ∫ B Ψ dτ dν ≈ ∫B dτ

where τ = ∫ Ψ τ dν and B = ∫ B Ψ dν, Ψ is channel spectral response

Band transmittances are pretabulated τ(u,p,T) where p,T arepath-averaged quantities (Beer’s Law not valid for ∏τc )

This removes the spectral integration from the retrieval forward model, but has limited accuracy.

Page 21




Pressure-Temperature RetrievalInfrared radiances are highly sensitive to temperature (4%/K)and pressure (1-2% / %)

Nadir sounders retrieve T(p) directly, but limb-viewinggives T(z) or p(z)

Always start with joint pT retrieval for profiles p(z),T(z),

Use CO2 channels since concentration is known

Page 22




Retrieving two parameters requiresat least two independent pieces ofinformation.

Difficult at high/low altitudes

Usually use some relative pointingknowledge and hydrostatic constraint

d lnp = -(gM/RT) dz180 200 220 240 260 280 300

Temperature [K]

1000.000

100.000

10.000

1.000

0.100

0.010

0.001

Pre

ssur

e [m

b]

4.0

8.0 9.0

4.0

8.0 9.

0

HIRDLS Ch3 Ch4Hydrostatic Constraint

Page 23




Current MIPAS Retrieval

Least Squares Fit – no explicit a priori or regularisation

One dimensional – assume no horizontal gradients

Sequential – pT, then H2O, O3, HNO3, CH4, N2O, NO2

Page 24




Baseline HIRDLS Retrieval

Optimal Estimation – loose climatological a priori constraint

1D retrieval, 2D forward model – 2 passes, model gradients in 2nd pass

Sequential/Joint – pT, Aerosol, H2O, O3, NO2, (HNO3+F11+F12),

(CH4+N2O+ClONO2+N2O5)

Page 25




Future MIPAS Retrieval ?

Optimal Estimation – otherwise too many indeterminate parameters!

2D forward model – tomographic retrieval of complete orbit

Joint – pT+all species, including additional gases

Page 26




SummaryInfrared limb sounding provides good verticalresolution and measurements of additional species

Main disadvantages are poor horizontal resolution and cloud sensitivity

Limb viewing geometry generally adds another dimension to the complexity of radiance modelling

Someday someone will assimilate limb radiances …

Documents

Infrared Limb Sounding MIPAS and HIRDLS - ECMWF 2 Atmospheric, Oceanic & Planetary Physics, University of Oxford A Dudhia ECMWF Workshop 23-26 June 2003 Infrared Limb Sounding MIPAS