gome-2 cloud top height and optical depth retrieval using...
TRANSCRIPT
GOME-2 Cloud top height and opticaldepth retrieval using ROCINN V3.0
Gimeno García1, Sebastián1; Lutz, Ronny1; Romahn, Fabian1; Loyola, Diego1;Spurr, Robert2
1German Aerospace Center (DLR)Remote Sensing Technology Institute
Oberpfaffenhofen, GERMANY
2RT SOLUTIONS Inc.Cambridge, MA, USA
June 10, 2015
ESA ATMOS 2015, Heraklion, 08-12 June 2015 1
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 2
Motivation – 1
Cloud information from spaceborne atmospheric spectrometersprimarily needed for accurate trace gas retrievalCloud effects on gas retrieval (1D):
albedo effect shielding effect multiple scattering
+ other: e.g. multiple cloud layering, . . .
ESA ATMOS 2015, Heraklion, 08-12 June 2015 3
Motivation – 2
Cloud information from spaceborne atmospheric spectrometersprimarily needed for accurate trace gas retrievalCloud effects on gas retrieval (3D):
neighboring pixel effect in-pixel inhomogeneity effect
+ other: e.g. effect of scene variability on spectral calibration
ESA ATMOS 2015, Heraklion, 08-12 June 2015 4
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 5
Retrieval strategy
ROCINN
UVN
Cloud fraction
Cloud-top height
Cloud albedo
O2 A-Band
OCRA
GOME type
Optical thickness),,,,),((},,{ 0 fazsimINVaz cssRNNcc
iBGRi
iiCFiifc,,
2 ))()()(,0max()(
OCRA: Optical Cloud Recognition Algorithm ROCINN: Retrieval Of Cloud Information through Neural Networks
OCRA/ROCINN schematic
ESA ATMOS 2015, Heraklion, 08-12 June 2015 6
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 7
OCRA – in short
OCRA is a RGB color space approach
determine monthly cloud free reflectance maps
map actual reflectance measurements to cloud (radiometric) fraction
radiometric cloud fraction:GOME2-A and GOME2-B merged together
⇒ See poster #47, Lutz et al.:Cloud Fraction Determination for GOME-2 A/B with OCRA V3.0 (!)
ESA ATMOS 2015, Heraklion, 08-12 June 2015 8
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 9
ROCINN – CAL & CRB
Cloud fraction (CF) is retrieved using a RGB color space approach→ OCRA
Cloud Top Height and Albedo (CTH, CTA) and Cloud OpticalThickness (COT) are retrieved in the Oxygen A-band usingregularization theory→ ROCINN
CRB: Clouds are treated as Reflecting BoundariesI Lambertian Equivalent Reflectors (LER)I simple and (numerically) robustI popular in the atmospheric gas retrieval communityI retrieved cloud parameters can greatly deviate from reality
CAL: Clouds are treated As horizontal homogeneous LayersI photon cloud penetration is allowedI multiple scattering is accounted forI modeled radiance contains information below the cloud layerI retrieved CTH is closer to the actual CTH
ESA ATMOS 2015, Heraklion, 08-12 June 2015 10
ROCINN – CAL cloud model
KISS – Keep It Simple Strategy (G. Marsaglia – RNG)
One cloud layer
Liquid water droplets
Spherical particles – Mie scattering theory
Modified gamma particle size distribution (α=6, β=0.11)
⇒ cloud optical propertiesI Single scattering albedoI Scattering phase functionI Optical thickness (not important: to be fit)
ESA ATMOS 2015, Heraklion, 08-12 June 2015 11
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 12
ROCINN_CRB: UPAS GDP 4.8 – 01 Jul 2009
Cloud Fraction Cloud Top Height
Cloud Fraction (CF) product nicely captures cloud systemmacroscopic structureCRB cloud top height (CTH) captures variability inside cloudsystem
ESA ATMOS 2015, Heraklion, 08-12 June 2015 13
ROCINN_CRB: UPAS GDP 4.8 – 01 Jul 2009
Cloud Top Albedo Cloud Top Height
CRB cloud top albedo (CTA) is less variable than CFCRB cloud top height (CTH) captures variability inside cloudsystems
ESA ATMOS 2015, Heraklion, 08-12 June 2015 14
ROCINN_CAL: UPAS GDP 4.8 – 01 Jul 2009
Cloud Optical Thickness Cloud Top Height
CAL cloud optical thickness (COT) captures more variability thanCRB_CTACAL cloud top height (CTH) higher than CRB_CTH
ESA ATMOS 2015, Heraklion, 08-12 June 2015 15
ROCINN_CAL: UPAS GDP 4.8 – 01 Jul 2009
Cloud Optical Thickness MODIS MOD08 COT
@ NASA GES DISC: data generated with Giovanni tool
CRB CAL_COT shows similar patterns than MODIS total COT
Equatorial crossing times:
I GOME-2 on EPS/MetOp-A: ∼ 9:30 AM local timeI MODIS on NASA-EOS/Terra: ∼ 10:30 AM local time
ESA ATMOS 2015, Heraklion, 08-12 June 2015 16
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 17
Conclusions
ROCINN3.0 CAL and CRB algorithms have been successfullyimplemented in UPAS (GDPv4.8 –GOME-2)
ROCINN3.0 CAL and CRB ready for operational processing
CAL and CRB cloud top height (CTH) captures the variabilityinside cloud systems
CRB delivers systematic lower CTH than CAL
CRB CTA is much smoother than CF
CAL COT shows much more variability than the related quantityCTA (due to the CAL cloud model)
A preliminary comparison with MODIS total COT shows a goodqualitative agreement
ESA ATMOS 2015, Heraklion, 08-12 June 2015 18
Outline
1 Motivation
2 Retrieval strategy
3 OCRA
4 ROCINN – CAL & CRB
5 Results: GOME2-A cloud products
6 Conclusions
7 Outlook
ESA ATMOS 2015, Heraklion, 08-12 June 2015 19
Outlook
Verification activities: compare OCRA/ROCINN with otherGOME-2 cloud products
I SACURA: CTH, COTI FRESCO: CF, CHI HICRU: CF
Compare OCRA/ROCINN with external cloud products:I AVHRR Patmos-XI MODIS MOD08I . . .
Reprocess GOME-2 on MetOp-A and MetOp-B
Setup UPAS for Sentinel-5 Precursor processing
ESA ATMOS 2015, Heraklion, 08-12 June 2015 20
MoCaRT
Thank you for your attention!
ESA ATMOS 2015, Heraklion, 08-12 June 2015 21
OCRA – method
OCRA is a RGB color space approachI OCRA RGB-bands for different sensors
determine monthly cloud free reflectance mapsmap actual reflectance measurements to cloud (radiometric)fraction
ESA ATMOS 2015, Heraklion, 08-12 June 2015 22
ROCINN – Forward model
CRB radiance forward model
ICRB = CF · Iclr (CH,CA;α) + (1− CF ) · Iclr (SH,SA;α)
CAL radiance forward model
ICAL = CF · Icld (CTH,COT ,CGT ,SH,SA;α) + (1−CF ) · Iclr (SH,SA;α)
whereI CF: cloud fractionI SH: surface heightI SA: surface albedoI C(T)H: cloud (top) heightI C(T)A: cloud (top) albedoI COT: cloud optical thicknessI CGT: cloud geometrical thicknessI α: auxiliary parameters (observation geometry, . . . )
ESA ATMOS 2015, Heraklion, 08-12 June 2015 23
ROCINN – Inversion
Tikhonov regularization
minx‖y− F(x)‖2 + ‖Γ(x− x0)‖2
whereI x: state vectorI x0: expected value of the state vectorI y: observationsI F: forward modelI Γ: Tikhonov regularization matrix
ESA ATMOS 2015, Heraklion, 08-12 June 2015 24