1. JMA AHI Rayleigh Scattering with MODIS
Yusuke YOGO and Masaya TAKAHASHI
Meteorological Satellite Center
Japan Meteorological Agency
I'll talk about Rayleigh scattering method for JMA's Himawari-8 and 9 AHI visible and near infrared bands

2. Aim / Contents Aim Contents
Himawari-8, 9/AHI VNIR vicarious calibration has some underestimation problems -> I'd like to ask how I can improve this method
Contents
AHI VNIR Vicarious Calibration using Radiative Transfer Calculation Method
AHI Rayleigh Scattering
Differences with CNES' Rayleigh Scattering
Summary and Future Plans
I was taken over this job from Arata Okuyama in the last April, and I'm facing on problems that there're some underestimations about Himawari-8, 9 AHI VNIR vicarious calibration
so I'd like to ask how I can improve this method
22 Mar 2018
2018 GSICS annual meeting

3. AHI VNIR Vicarious Calibration
Radiative transfer calculation using the model "RSTAR" for several types of targets
Inputs: MODIS L1B, Aura/OMI Ozone L3, JRA-55 (JMA's re-analysis)
Targets in use
Sea surface for darker scene or zero radiance
currently not used in bands 1-2 ( μm) due to underestimation
Water cloud for brighter scene
Targets currently not in use
Land surface, especially desert for middle scene
No ideal deserts (e.g. Sahara) over AHI observation area
DCC (RT calculation) for the brightest scene, < 1 μm
To be implemented
Result of MTSAT-2 in Feb. 2011
Simulated reflectivity
Observed reflectivity
DCC
Water cloud
Land
Sea
Firstly, I introduce the VNIR vicarious calibration method in JMA
calculation results by the radiative transfer model are used
Input data are MODIS L1B radiances to estimate aerosols, ozone amount and re-analysis to make radiative transfer calculation more accurately
# We think this can be called intercalibration between AHI and MODIS
Several types of targets are used in this method
cloud-free ocean surface for darker scene and water clouds for brighter scene are in use, but results of bands 1-2, around 0.5 micro meter, are not used due to a underestimation problem
We don't use cloud-free land surface calculation results, because no ideal deserts over AHI area
DCC calculation for the brightest scene of visible bands needs information about reflection by non-spherical ice crystals, so it is under implementation
22 Mar 2018
2018 GSICS annual meeting

4. AHI Rayleigh Scattering
RT calculation of ToA radiance over cloud-free ocean
limitations
JMA AHI
CNES (IOCCG, 2013)
RT model
RSTAR (Nakajima and Tanaka 1986)
6S (Vermote et al., 1997)
surface wind speed
< 7.0 m/s
< 5.0 m/s
aerosol optical depth
< 500 nm
< 865 nm
Sun zenith angle
< 60 deg. (AHI & MODIS)
< 60 deg.
Viewing zenith angle
sunglint
"sunglint angle" > 40 deg.
(AHI & MODIS)
cone angle > 60 deg.
region
no limitation
oligotrophic ocean only
other limitations
relative azimuth angle < 90 deg.
> 10px distant from cloud
RT calculation of cloud-free ocean is similar to the Rayleigh scattering method, which suggested by some agencies
this time I chose CNES' method for comparison, because it has been well discussed in gsics
# CNES' Rayleigh scattering method and it was revealed that JMA method is not as strict as CNES method
22 Mar 2018
2018 GSICS annual meeting

5. Vicarious calibration using water clouds
Period: 30 days, Jan 2017
Good agreement with VIIRS "ray-matching" method -> ☺ well estimated
0.0 < simulation > 1.0
0.0 < simulation > 1.0
band 1 band 2
(0.47 μm) (0.51 μm)
band 3 band 4
(0.64 μm) (0.86 μm)
band 5 band 6
(1.6 μm) (2.3 μm)
0.0 < observation > 1.0
0.0 < observation > 1.0
0.0 < simulation > 1.0
0.0 < simulation > 1.0
viirs比較図
0.0 < observation > 1.0
0.0 < observation > 1.0
I show the results of water clouds before the results of AHI Rayleigh scattering
Every band shows a good agreement with the result by SuomiNPP VIIRS ray-matching method and 1:1 diagonal lines
So these seem to be well estimated
0.0 < simulation > 1.0
0.0 < simulation > 1.0
0.0 < observation > 1.0
0.0 < observation > 1.0
22 Mar 2018
2018 GSICS annual meeting

6. Vicarious calibration using sea surface
(ranges of axes differ)
Bands 1-2 ( μm)
simulations < observations
undetestimation caused by ocean color and polarization?
Bands 3-4 ( μm)
simulations ~ observations
☺ show good agreements
Bands 5-6 ( μm)
Instrument's noise?
The estimation may be insufficient in NIR?
0.0 < simulation > 0.2
0.0 < simulation > 0.2
band 1 band 2
(0.47 μm) (0.51 μm)
band 3 band 4
(0.64 μm) (0.86 μm)
band 5 band 6
(1.6 μm) (2.3 μm)
0.0 < observation > 0.2
0.0 < observation > 0.2
0.0 < simulation > 0.2
0.0 < simulation > 0.05
0.0 < observation > 0.2
0.0 < observation > 0.05
Next, the results of sea surface
in bands 1-2, simulated reflectances are slightly smaller than observations
It is because of ocean color and polarization
in bands 3-4, simulations and observations show good agreements
in bands 5-6, reflectances are almost 0, and simulations are smaller than observations
I'm not confident of the cause, it may be instrument's noise, or the estimation is insufficient
0.0 < simulation > 0.05
0.0 < simulation > 0.05
0.0 < observation > 0.05
0.0 < observation > 0.05
22 Mar 2018
2018 GSICS annual meeting

7. Differences with CNES' Method
Region Selection
VZA < 60 deg. is the only condition for target area selection
much wider than CNES' method, which uses only certain oligotrophic (= less nutrient and plankton) areas
Why NIR bands are calculated
Almost no reflection nor Rayleigh scattering
Instrument's noise could be validated
Any ideas/applications in other agencies?
Why we don't utilize in bands 1-2 ( μm)
Underestimation problems remain
How to resolve
consider polarization (handed over from Arata, under investigation)
consider Chlorophyll-A (same as above)
restrict to oligotrophic areas
JMA
CNES
There are some differences between JMA method and CNES' method
One of the major differences is about region selection,
CNES method chooses only certain oligotrophic areas
in JMA method, viewing zenith angle is smaller than 60 deg is the only condition
and we're applying this method to not only visible bands, but also NIR bands
practically, almost no reflection by sea surface, nor Rayleigh scattering effects, so we acquire almost 0 reflectances
and instrument's noise could be validated by using this method
as I said before, currently we have not utilized the results of bands 1-2 as final results,
because underestimation problems remain
this can be resolved by considering polarization and ocean color
to consider ocean color, there are some ways, e.g. inputting chlorophyll-a, and restricting to oligotrophic area
22 Mar 2018
2018 GSICS annual meeting

8. Trial: restrict to oligotrophic areas
RT calculation over "ClimZOO" North-West of Pacific (Fougnie et al., 2002)
Lon: 10.0N to 22.7N, Lat: 139.5E to 165.6E - suitable geometry for AHI
"ClimZOO"
North-West Pacific
I tried to do the RT calculation over only oligotrophic area, which is one of suggested by CNES method
I chose ClimZOO north-west pacific area, that has suitable geometry for AHI
(cited from Fougnie et al., 2002)
22 Mar 2018
2018 GSICS annual meeting

9. Entire Region band 1 band 2 band 3 (0.47 μm) (0.51 μm) (0.64 μm)
0.0 < simulation > 0.2
0.0 < simulation > 0.2
0.0 < simulation > 0.2
band 1 band 2 band 3
(0.47 μm) (0.51 μm) (0.64 μm)
band 4 band 5 band 6
(0.86 μm) (1.6 μm) (2.3 μm)
0.0 < observation > 0.2
0.0 < observation > 0.2
0.0 < observation > 0.2
first, entire region with viewing zenith angle of less than 60 deg
0.0 < simulation >
0.0 < simulation >
0.0 < simulation >
0.0 < observation > 0.05
0.0 < observation > 0.05
0.0 < observation > 0.05
22 Mar 2018
2018 GSICS annual meeting

10. Limited to "ClimZOO" Region
seem to be very stable
In bands 1-2 and 5-6, still mismatched between obs and sim
Data inputting or RT cal may have some problems
0.0 < simulation > 0.2
0.0 < simulation > 0.2
0.0 < simulation > 0.2
band 1 band 2 band 3
(0.47 μm) (0.51 μm) (0.64 μm)
band 4 band 5 band 6
(0.86 μm) (1.6 μm) (2.3 μm)
0.0 < observation > 0.2
0.0 < observation > 0.2
0.0 < observation > 0.2
and these are the result with restriction to oligotrophic area
in all bands, results are well converged into narrow ranges,
but there're still mismatches between obs and sim
so I think that data input or radiative transfer calculation may have problems
0.0 < simulation >
0.0 < simulation >
0.0 < simulation >
0.0 < observation > 0.05
0.0 < observation > 0.05
0.0 < observation > 0.05
22 Mar 2018
2018 GSICS annual meeting

11. Summary / Future Plans Summary Future plans
JMA AHI Rayleigh scattering method has some issues
Bands 1-2 ( μm) are underestimated due to not considering ocean color and polarization enough
Bands 5-6 ( μm) also show mismatches, but apparently with a different reason
Future plans
Resolving underestimations in blue-green bands / NIR bands
considering polarization, Chlorophyll-A, ...
Transition to VIIRS
finally I summarize this talk
JMA's AHI Rayleigh scattering method has some issues, about mismatches in some bands
bands 1-2 are underestimated due to not considering ocean color and polarization enough
bands 5-6 NIR also show mismatches, but apparently with a different reason
I plan to resolve underestimations in blue-green bands / NIR bands, considering polarization, Chlorophyll-A,
and transit from MODIS to VIIRS
22 Mar 2018
2018 GSICS annual meeting

12. Thank you for your attention
22 Mar 2018
2018 GSICS annual meeting

13. References
Fougnie et al., 2002: Identification and Characterization of Stable Homogeneous Oceanic Zones : Climatology and Impact on In-flight Calibration of Space Sensor over Rayleigh Scattering, Ocean Optics XVI Proceedings.
IOCCG, 2013: In-flight Calibration of Satellite Ocean-Colour Sensors. Frouin, R. (ed.), Reports of the International Ocean-Colour Coordinating Group, No. 14.
Nakajima, T. and M. Tanaka, 1986: Matrix formulation for the transfer of solar radiation in a plane-parallel scattering atmosphere. J. Quant. Spectrosc. Radiat. Transfer, 35,
Vermote, E., D. Tanré, J. L. Deuzé, M. Herman, and J. J. Morcrette, Second Simulation on the Satellite Signal in the Solar Spectrum, 1997: 6S: An overview, IEEE Trans. Geosci. Remote Sens., 35,
22 Mar 2018
2018 GSICS annual meeting