COMS Visible Calibration
The meteorological instrument MI(Meteorological Imager) of COMS has one visible channel of the center wavelength at 0.675㎛ with 1km spatial resolution.
Channels | Wavelength | Spatial Resolution |
Applications |
---|---|---|---|
Visible | 0.67 | 1Km | Weekly cloud images, Asian dust, forest fires, fog observation, atmospheric motion vector |
※ The COMS MI channel characteristic
No on-board calibration
There is no on-board calibration target for COMS MI visible channel such as solar diffuser. Thus, ground calibration is needed for calibration and monitoring of visible channel degradation with time. However, in infrared channels, the on-board calibration is conducted from blackbody.
Ground calibration
NMSC has vicarious calibration system for visible channel. Vicarious calibration is the method using the corresponded theoretical reflectances with the observation point and time. In the calibration system, the theoretical reflectances for four targets: ocean, desert, water cloud(WC) and deep convective cloud(DCC) are simulated and utilized as reference values.
Algorithm
NMSC performs the vicarious calibration by using ocean, desert, WC and DCC. These algorithms are based on the development of meteorological data processing system for communication, ocean and meteorological satellite ATBD.
NMSC choose the four targets, whose properties are well-known for calculating TOA(Top-Of-Atmosphere) reflectance, for having uniformly distributed reflectances between 0 and 1. And NMSC selected the suitable models for each target to improve calibration accuracy.
The thresholds of each target and the models are followings :
Targets | Applications | |
---|---|---|
Ocean | Thresholds of Target |
|
Model | 6S | |
Desert | Thresholds of Target |
|
Model | 6S | |
Water Cloud | Thresholds of Target |
|
Model | SBDART | |
DCC | Thresholds of Target |
|
Model | SBDART |
※ 6S : Second Simulation of the Satellite Signal in the Solar Spectrum
SBDART : Santa Barbara DISORT Atmospheric Radiative Transfer
The characteristics of each target are followings:
Ocean
The ocean targets are usually homogeneous and dark but should be considered only a clear pixel without a cloud for sensor calibration. The reflectance of ocean is affected by aerosol, because the surface reflectance is so small. Therefore, AOT(Aerosol Optical Thickness) has to be considered as an ancillary data in radiative transfer modeling. In addition, it has the advantage that radiance can be calculated without BRDF (bidirectional reflectance distribution function) because the surface reflectance of the ocean is nearly homogeneous.
Desert
KMA conducted desert calibration for 11 fixed desert targets among Simpson Desert in Australia. There is no need to use AOT as ancillary data from other satellites, because of the low effect of aerosol by high surface reflectivity of desert. However, accurate surface BRDF needs to be used for model input, since the variation of surface BRDF is low with time. The surface BRDF can be updated with longer period than other targets.
Water Cloud
The atmosphere and surface effects can be minimized in the simulation of TOA radiance over the WC targets due to strong reflection of cloud layer. According to the actual test for WC reflectances sensitivity, TOA reflectance is more dependent on cloud optical properties than surface and atmospheric effects. In addition, the reflectance is more sensitive with cloud optical thickness(COT) than particle sizes of cloud since the absorption rate of visible rays into cloud particle is low. Thus, the cloud optical properties are necessary as the model input for more accurate result.
DCC
DCC target has the largest reflectance among the targets: ocean, desert, WC and DCC. The vertical distribution of DCCs is assumed from 1km to 15km. COT and particle size of the DCC are assumed to be 200 and 20µm, respectively. The pixel of which the brightness temperature (TB) of IR1(10.8µm) is less than 190K are selected as DCCs for calibration.
Algorithm
NMSC provides the three types of outcomes in this web page. There are the two time series of regression coefficients and ratios for each target, and the one scatter plot.
Regression coefficients between the observed and simulated reflectances
The time series of regression slopes(C1) and intercepts(C0) are computed between the observed and simulated reflectances. The C1 and C0 in the web page represent the 30days moving averaged values.
Reflectance(Observed)= C1 * Reflectance(Simulated) + C0
Ratios of the observed reflectance to simulated reflectance for each target
The time series of ratios of the observed reflectance to simulated reflectance represents the 30days moving averaged value.
Ratio=Reflectance(Observed)/Reflectance(Simulated)
Scatter plot between observed and simulated reflectance
The comparison of reflectance between observation and simulation is provided in the scatter plot with the regression line.
Reference
KMA, 2009: Development of Meteorological Data Processing System for Communication, Ocean and Meteorological Satellite(ATBD).
Chun, H. W., & Sohn, B. J., 2006: Calibration for the solar channel using MODIS-derived BRDF parameters over Australian desert targets. Target, 4(6), 8.
Ham S. H. 2006: Improvement of cloud's spectral reflection simulation using MODIS cloud products. 12th Conference on Atmospheric Radiation.
Sohn, B-J., Seung-Hee Ham, Ping Yang, 2009: Possibility of the Visible-Channel Calibration Using Deep Convective Clouds Overshooting the TTL. J. Appl. Meteor. Climatol., 48, 2271-2283.