MTSAT Visible Vicarious Calibration

The Meteorological Satellite Center examines vicarious calibration for geostationary satellite visible channel. This type of calibration is intended to rebuild visible calibration coefficients, and its approach relies on radiative transfer calculation. Reflectivity at each observed location is simulated under atmospheric and geometric conditions corresponding to those of satellite observation. Calibration coefficients are computed using regression analysis for an appropriate number of pairs of observed and simulated reflectivity values.

Radiative transfer calculation requires atmospheric profiles, aerosol and cloud optical parameters, and observation geometry information. The input data are created from JMA's Re-Analysis data, total column ozone L3 data, (TOMS: Total Ozone Mapping Spectrometer, OMI: Ozone Monitoring Instrument) and Moderate Resolution Imaging Spectroradiometer (MODIS) L1B data.

L3 ozone and MODIS L1B data are collected via the Internet. In the event of data download failure, vicarious calibration computation may be suspended.

Outcomes

The results of vicarious calibration consist of a set of regression parameters.

Linear regression coefficients (C0 and C1) to associate MTSAT nominal reflectivity with calculated values are computed. C0 and C1 are defined by:

Reflectivity (calculated) = C0 + C1 × Reflectivity (nominal)    (1)

"Reflectivity (nominal)" is the reflectivity converted from digital count data according to the calibration table in the header part of disseminated HRIT data. Reflectivity is expressed by a dimensionless number between zero and one (0 and 100%). Reflectivity and radiance are given by:

Reflectivity (calculated) = Radiance × π / F0    (2)

where F0 for MTSAT-2 is 1,518 [W/m2/um]. This is the solar incident irradiance for the MTSAT-2 imager observation wavelength. The radiance is a value over the Spectral Response Function (SRF) for the imager with dimension [W/m2/sr/um].

Methodology

Target Selection

To reduce the uncertainty of radiative transfer calculation, it should be performed for a spatially uniform and temporally stable area. The observed image is therefore separated into grids of 0.05 by 0.05 degrees, and simulation is examined only for suitable grids. Such grids are referred to as "targets" here, and should cover a wide range of radiances to allow the creation of a reliable regression line. This approach uses the three target types of cloud-free ocean surface, cloud-free land surface and spatially uniform liquid cloud top as dark, medium and bright targets, respectively. The simulation requires inputs depending on each type of target.

Simulation

  • Radiative Transfer Code

    RSTAR, developed by the Atmosphere and Ocean Research Institute at the University of Tokyo (Nakajima and Tanaka, 1988), is employed as the radiative transfer code. It requires information on atmospheric profiles, total column ozone amount, surface reflectivity, optical parameters of scattering particles and viewing geometry. The input for radiative transfer calculation should be independent of satellite observation. In this approach, therefore, the atmospheric profile and ozone amount are based on JMA NWP data and total column ozone data, respectively. Scattering objects are aerosols (for cloud-free ocean and cloud-free land targets) and cloud particles (for cloud-top targets). The optical parameters of scattering particles are obtained from MODIS L1B data.

  • Simulation for Cloud-Free Ocean Targets

    For cloud-free ocean targets, sea surface reflectivity and aerosol optical parameters are necessary in addition to atmospheric profile data and ozone total amount as input for radiative transfer calculation. RSTAR can retrieve sea surface reflectivity data from sea surface wind speed figures. This approach utilizes the sea surface wind speed included in the JMA NWP model data. Aerosol optical thickness and the Angstrom exponent are obtained from Terra/MODIS L1B data using the REAP aerosol analysis package (Higurashi and Nakajima 1999). REAP computes aerosol optical parameters from two channels of satellite data (visible and near-infrared). This approach utilizes MODIS bands 1 and 2. The aerosol absorption/scattering parameters are based on non-absorption characteristics (Higurashi and Nakajima 2002).

    NASA’s MODIS aerosol products (known as MOD04 and MYD04) are alternatives to REAP-retrieved aerosol as simulation input data. However, the radiative transfer code used to retrieve aerosol data from MODIS L1B and that used to calculate simulated reflectivity should be consistent to ensure accurate simulation. As the radiative transfer algorithm included in REAP is the same as that of RSTAR, this approach adopts REAP-retrieved aerosol data instead of NASA’s MODIS aerosol product.

    Input data for cloud-free ocean targets
    Fig. 1: Input data for radiative transfer calculation with cloud-free ocean targets. The aerosol optical parameters are retrieved from MODIS L1B.

  • Simulation for Cloud-Free Land Targets

    The land targets are bare-soil areas near Alice Springs and Tennant Creek in Australia. Key elements for cloud-free land targets are ground surface reflectivity and aerosol optical thickness. As retrieving aerosol optical thickness data from satellite observation on land areas generally involves uncertainty, climate data are adopted as aerosol optical parameters. To estimate ground surface reflectivity accurately, the Bidirectional Reflectivity Distribution Function (BRDF), which shows reflectivity corresponding to an incoming direction and an outgoing direction, is necessary. In this approach, NASA’s MODIS BRDF product is utilized to allow consideration of ground surface anisotropy. This product provides parameters to enable characterization of ground surface reflectivity depending on geometry information. It supplies parameters for the seven MODIS bands from visible to infrared and three broad bands. It is necessary to modify the parameters to adopt them for MTSAT because the SRF for MTSAT imager is different from that of MODIS. The ground surface reflectivity based on the MTSAT SRF is approximated based on interpolation of the parameters for the seven MODIS bands.

    Input data for cloud-free land targets
    Fig. 2: Input data for radiative transfer calculation with cloud-free land targets

  • Simulation for Liquid Cloud Targets

    Cloud parameters such as optical thickness and effective radius are dominant factors for liquid cloud targets. These parameters are obtained from Terra/MODIS L1B data using the CAPCOM cloud analysis package (Nakajima and Nakajima, 1995). CAPCOM computes cloud optical parameters from three channels of satellite data — visible, near-infrared and infrared. This approach utilizes MODIS bands 1, 20 and 31. The radiative transfer algorithm included in CAPCOM is consistent with that of RSTAR. It endorses accurate simulation (as mentioned above) for ocean targets. For this cloud target type, the scattering object consists of liquid cloud particles rather than aerosols.

    Input data for liquid cloud targets
    Fig. 3: Input data for radiative transfer calculation with liquid cloud targets. The cloud optical parameters are retrieved from MODIS L1B.

Regression

The simulated reflectivity is compared with the MTSAT nominal reflectivity. The linear regression coefficients are the calibration adjustment coefficients C0 and C1 in (1) and are obtained from 30 days data in this approach.

Radiative Transfer Calculation Accuracy

To estimate the accuracy of this approach, the simulated radiance for the three types of targets are compared not with MTSAT observed radiance but with Terra/MODIS radiance. MODIS carries onboard calibrators for visible bands, which makes the observation data well calibrated and reliable. The input factors for the radiative transfer code are the same as those for the simulation for MTSAT, e.g., aerosol and cloud optical parameters retrieved from MODIS L1B.

Figure 4 shows a match-up between the simulated radiance and the MODIS observed radiance. The error between the two is less than 1%, indicating that the methodology is sufficiently reliable.

Comparison of MODIS Band.1 observations with simulations
Fig. 4: Match-up between simulated radiance and Terra/MODIS observed radiance. The slope of the regression line is 1.00 and the bias is almost zero. The radiative transfer code can simulate MODIS observed radiance with an error margin of less than 1%.

Reference

  • Okuyama, A., Toru Hashimoto, Ryuichiro Nakayama, Yoshihiko Tahara, Toshiyuki Kurino, Hideaki Takenaka, Satoru Fukuda, Takashi Y. Nakajima, Akiko Higurashi, Miho Sekiguchi, Tamio Takamura, Teruyuki Nakajima, "Geostationary Imager Visible Channel Recalibration", 2009 EUMETSAT METEOROLOGICAL SATELLITE CONFERENCE, SESSION 6. Monitoring climate and understanding climate processes with satellites, ISBN 978-92-9110-086-6 (PDF, 302 kB)

  • (RSTAR)
    Nakajima, T., and M. Tanaka, (1988) Algorithms for radiative intensity calculations in moderately thick atmospheres using a truncation approximation., J. Quant. Spec. Rad. Trans., 40, pp 51-69

  • (REAP)
    Higurashi, A., and T. Nakajima, (1999) Development of a Two Channel Aerosol Retrieval Algorithm on Global Scale Using NOAA / AVHRR., J. Atmos. Sci., 56, pp 924-941
    Higurashi, A., and T. Nakajima, (2002) Detection of aerosol types over the East China sea near Japan from four-channel satellite data., Geophys. Res. Lett., 29, pp 1836-1839

  • (CAPCOM)
    Nakajima, T.Y., and T. Nakajima, (1995) Wide area determination of cloud microphysical properties from NOAA AVHRR measurement for FIRE and ASTEX region., J. Atmos. Sci., 52, pp 4043-4059

  • The RSTAR, REAP and CAPCOM tools are available on the Open CLASTR website of the Atmosphere and Ocean Research Institute at the University of Tokyo.