Skip to content

National Meteorological Satellite Center

Welcome you to the National Meteorological Satellite Center.

HOME > Activities(미사용) > GSICS > Infrared

GSICS / Infrared

COMS Infrared Calibration

COMS MI (Meteorological Imager) has 4 infrared channels and its radiometric calibration is conducted in stage of Level 1A data processing by measuring the radiances of warm and cold targets, blackbody onboard COMS and space-looking, respectively. In addition to onboard calibration, NMSC has examined the inter-calibration as a part of GSICS (Global Space-based Inter-Calibration System) activities. It is helpful to check the quality of operational measurements for COMS and sensor degradation with time. GSICS inter-calibration system is to compare two observation values measured by an instrument we want to calibrate and reference instrument which is known as relatively accurate.


The inter-calibration for COMS MI uses two hyper-spectral infrared sounders of polar-orbit satellites for reference measurement. One is AIRS (Atmospheric Infrared Sounder) onboard Aqua satellite launched by NASA and the other is IASI (Infrared Atmospheric Sounding Interferometer) onboard Metop-A launched by EUMETSAT. The inter-calibration is conducted by comparing the true measurements of COMS MI and simulated values from the measurement of reference sensors, in the radiance or brightness temperature (TB) format. Outcomes of the inter-calibration, especially provided in linear regression, are able to be used for correction of COMS observations. This correction might be important process for the retrieval of Essential Climate Variables (ECVs) such as Sea Surface Temperature (SST), Atmospheric motion vector (AMV) and so on.

Algorithm (ATBD)

The inter-calibration system for COMS infrared channels is thoroughly based on the GSICS Coordinate Center (GCC) Algorithm Theoretical Basis Document (ATBD). In order to conduct calibration system, first we should make collocation dataset for measured pixels of COMS and IASI. Then, the several channel measurements of hyper-spectral sounder is converted into a simulated radiance according to spectral response function (SRF) of COMS. It is called as constraint method, generating a super channel consisting of combination of a lot of channels to imitate a broadband channel. More details of the methods for collocation and spectral simulation are given in below paragraph.

Collocation method

Collocation algorithm used in NMSC is determined by GSICS research working group, and specific threshold values to filter out unsuitable pixels for inter-calibration are referred to JMA’s ATBD of GSICS. In following equations, subscript of GEO and LEO mean COMS and AIRS/IASI, respectively.

Collocated in space

Choose GEO pixel geometrically closest to the LEO pixel

Concurrent in time

|TimeLEO – TimeGEO| < 300 s (= 5 min)

Aligned in line-of-sight

|cos(VZALEO) / cos(VZAGEO)− 1| < ε1

Uniformity test for environmental pixels

To reduce the difference of two satellite measurements occurring due to different observational condition, especially time, navigation, optical path and cloud advection, only measurements over the uniform scene are selected for inter-calibration. Uniformity test for environmental pixels is conducted over 9x9 area (called as ENV box) of GEO.


Normality test

Because the observational field-of-view for AIRS/IASI is about 13 km, 3 times larger than COMS MI, the average of COMS measurement for 3x3 pixels (called as FOV box) is used for comparison with AIRS/IASI. COMS FOV box should be representative of environment passed uniformity test, therefore normality test is added.


Collocation algorithm used in NMSC is determined by GSICS research working group, and specific threshold values to filter out unsuitable pixels for inter-calibration are referred to JMA’s ATBD of GSICS. In following equations, subscript of GEO and LEO mean COMS and AIRS/IASI, respectively.

Normality test
clear cloudy clear cloudy all clear cloudy
ε1 0.01 0.03 0.01 0.03 0.01 0.01 0.03
ε2 1.65 3.31 1.82 3.64 0.311 0.0151 0.0302
ε3 2 2 2 2 1 2 2
Spectral simulation: Constraint method

In order to compare COMS MI measurement and hyper-spectral measurement, simulated radiance consisting of several channels measurements should be produced according to SRF of COMS MI. NMSC adopts a constraint method developed by Tahara (2008).

Is = ∑wiIi ∑wi / Sb(υ) ≈ Sb(υ) = ∑wiSi(υ)

Constraint method computes radiance of the super channel (Is) using super channel weight (wi) and measured radiance hyper-spectral sensor (Ii). Super channels weights are determined by minimizing difference between the SRF of super channel (Ss) and SRF of COMS broadband (Sb) considering the original SRF of hyper-spectral instrument (Si). This method is known as well simulating broadband radiance with high accuracy in case the spectrum of hyper-spectral sensor fully covers the spectral band of the broadband channel.

However, COMS MI channels are not fully covered with AIRS/IASI spectrum. So, Compensation process for gap and missing spectral range should be applied. It is performed by simulating the virtual radiance of gap channel for 8 kind of ideal atmospheric profiles including clear and cloudy conditions. More details of it are shown in the paper of Tahara and Kato (2009).


NMSC has provides the three types of outcomes for infrared inter-calibration in this web page, one Scatter plot and two time-series plots. The time-series is about regression coefficients (i.e. slope and intercept from linear regression) and mean TB bias, which is represented with monthly and daily values. Scatter plot includes both MI radiance vs. radiance from the reference sensors and difference of TBs vs. MI TB.

Change of regression coefficients with time

Regression coefficients such as slope (C1) and intercept (C0) are produced by comparing the COMS MI measurement and simulated measurements form AIRS/IASI. The C1 and C0 represented here are calculated using accumulated collocation datasets for 1 day for daily values and for 1 month for monthly values.

Radiance(AIRS/IASI) = C0+ C1ⅹRadiance(MI)

Moreover, NMSC also provides the number of collocation points used for inter-calibration.

Change of mean TB bias with time

Mean TB bias is the average of TB differences (COMS measurement minus simulated value from reference sensor) for 1 day and/or 1 month. Mean TB bias and its standard deviation are provided at 3 standard temperatures of 290 K, 250 K, and 220 K.

TB bias = TB(MI) – TB(AIRS/IASI)

Scatter plots of radiance and difference of TB

The scatter plots of radiance and difference of TB are updated once in a day, but data represented in plots are accumulated in a month. Past outcomes of scatter plots are only provided monthly. From these scatter plots, NMSC obtains the regression coefficients and mean TB bias used for the time-series plots.


EUMETSAT, 2010: ATBD for Prototype GSICS SEVIRI-IASI Inter-Calibration. Doc. No. EUM/MET/TEN/09/0774.

Hiromi, Owada (JMA), 2010: Theoreticcal Basis for MTSAT-AIRS/IASI Inter-calibration Algorithm for GSICS. Version 2010-07-14.

Tahara, Yoshihiko, 2008: New Approach to Intercalibration Using High Spectral Resolution Sounder. Meteorological Satellite Center Technical Note, No. 50, 1-14.

Tahara, Yoshihiko and Koji Kato, 2009: New Spectral Compensation Method for Intercalibration Using High Spectral Resolution Sounder, Meteorological Satellite Center Technical Note, No. 52, 1-37

Quick Link
  • COMS
  • Activities
  • FY-2D
  • Terra/Aqua
  • NOAA
  • WindSat