Medium Resolution Spectral Imager (MERSI) is a keystone instrument onboard Fengyun-3 (FY-3), the second generation of polar-orbiting meteorological satellites in China. The first unit still in operation is FY-3A which was launched on May 27, 2008 in a sun-synchronous morning orbit with a local equator-crossing time of 10:30 AM in descending node. The second unit still in operation is FY-3B which was launched on November 5, 2010, in an afternoon orbit with an equator-crossing time of 1:30 PM in ascending node. FY-3 MERSI provides global coverage on top-of-atmosphere (TOA) radiances used for a broad range of scientific studies of the Earth’s system. Nineteen of the 20 MERSI spectral bands are reflective solar bands (RSBs) from 412 NM to 2130 nm, which cannot be absolutely calibrated onboard. The long-term on-orbit response changes of FY-3A/B MERSI are relatively large at visible bands. A multisite calibration tracking method has been developed to monitor the RSB radiometric response variation, revealing that the overall degradation for 412 nm of FY-3A MERSI is about 43% until June 2014. A daily calibration updating model is developed to recalibrate FY-3A/B MERSI, and the data quality is monitored using SNO targets against Aqua MODIS. This paper demonstrates the radiometric performance of FY-3A/B MERSI RSBs after recalibration accounting for the temporal variation of radiometric response. The recalibrated MERSI shows good agreement with MODIS. For FY-3B MERSI band 1 (470nm), the overall percentage difference (Mean±Std) is within 4%.
Medium Resolution Spectral Imager (MERSI) is the key imaging sensor on board Fengyun-3 (FY-3), the second generation polar-orbiting meteorological satellites in China, currently operating on both FY-3A, FY-3B and FY-3C satellites. It has 20 spectral bands, including 19 reflective solar bands (RSBs) with center wavelengths from 0.41μm to 2.1μm and 1 thermal emissive band (TEB) with center wavelength 12μm, making observations at two spatial resolutions: 250 m (bands 1-5) and 1km (bands 6-20). The FY-3C has been launched in 23, Sept., 2013. The MERSI doesn't carry on-board calibration standards. To obtain RSBs radiometric responses, pre-launched field radiometric calibration test which is called Solar Radiation Based Calibration(SRBC) was taken in Dali in 27, Feb. to 2, Mar., 2013. For the SRBC measurement which the sun was the source of irradiance, MERSI viewed the reflected solar irradiance from a set of the sixteen reference spectral on panels with different reflective level. The uniformity, reflectivity and BRDF (Bidirectional reflection of distribution function) of sixteen reference panels were tested in advance. There are two kinds of calibration coefficient generation methods used in SRBC. One is similar as the Sea-WiFS pre-launch calibration method by Langley calibration. Besides this, we use a portable spectrometers produced by Analytical Spectral Devices inc. (ASD inc.) named FieldSpec 3 to measure the absolutely reflected radiance simultaneously. The calibrated spectrometers measured radiance could be as the reference radiance and the the calibration coefficient of the MERSI can be calculated. We called this method Calibration Based on Reference Instrument(CBRI). The results of these two methods are comparable. The CBRI results are less then 6% difference with Langley calibration method in most channels except water-vapor channels and channel 15. An non-linear feature of the most FY-3C/MERSI detectors was found for the first time. This phenomenon is even more obvious for the water-vapor channels. The second order coefficient determined by pre-launched calibration is quite useful to improve the on-obit calibration accuracy.
An inter-calibration method of infrared channels of FY-3C MERSI and VIRR using NPP/CrIS and Metop/IASI as the hyperspectral reference sensor is introduced. Based on FY-3C SNO collocated samples with CrIS and IASI, on early orbit, we analyze the calibration biases of infrared bands of MERSI and VIRR. The results show that the brightness temperatures (BT) from the MERSI observation and CrIS have a good consistency, and the BT biases present an approximately normal distribution and the mean BT bias is about -0.18K with standard deviation of 0.83K. When the scene BT is lower than 250 K, the result of MERSI is higher than that of CrIS, while the result of MERSI is lower at the more than 250K scene. The BT from VIRR shows significant systematic bias with respect to CrIS and the mean BT bias is about -0.65 K (channel 4) and -0.72 K (channel 5) at 250K scene with standard deviation of 0.15 K and 0.12 K, respectively. Long term monitoring analysis demonstrates the above biases are stable in the early 6 months. The inter-calibration results using different hyperspectral sensors IASI and CrIS indicate the MERSI/VIRR biases with respect to two reference sensors have a good consistency and this further verifies the reliability of the method. It provides significant information to further correct the calibration biases of MERSI and VIRR.
FY-3C/MERSI has some remarkable improvements compared to the previous MERSIs including better spectral response function (SRF) consistency of different detectors within one band, increasing the capability of lunar observation by space view (SV) and the improvement of radiometric response stability of solar bands. During the In-orbit verification (IOV) commissioning phase, early results that indicate the MERSI representative performance were derived, including the signal noise ratio (SNR), dynamic range, MTF, B2B registration, calibration bias and instrument stability. The SNRs at the solar bands (Bands 1–4 and 6-20) was largely beyond the specifications except for two NIR bands. The in-flight calibration and verification for these bands are also heavily relied on the vicarious techniques such as China radiometric calibration sites(CRCS), cross-calibration, lunar calibration, DCC calibration, stability monitoring using Pseudo Invariant Calibration Sites (PICS) and multi-site radiance simulation. This paper will give the results of the above several calibration methods and monitoring the instrument degradation in early on-orbit time.
Intercalibration against a well-calibrated instrument at Low Earth Orbit (LEO) is a common method which has been widely used to assess the in-flight calibration of a new instrument. Different instruments on LEO spacecraft with similar spectral channels can be compared with each other using their simultaneous nadir observations (SNO). The postlaunch calibrations of Medium Resolution Spectral Imager (MERSI) and the Visible Infrared Radiometer (VIRR) in visible channels which are two major multi-spectral imaging radiometers onboard FY-3C are addressed based on SNO intercalibration method. Collection 6 reflectance products of AQUA MODIS are used as reference. The spectral difference impacts of matching channels are simulated and adjusted using GOME-2 hyperspectral measurements. As monitoring the stability of monthly forcing fits, it is found the linear fitting slopes of MERSI VIS channel 1~12 are scene reflectance dependence with relative differences greater than 20%, while the monthly forcing fits of VIRR show well agreement in VIS channels. This is proved to attribute to the nonlinear response of MERSI as the monthly measurements cover different dynamic ranges. A new radiometric calibration equation considering nonlinear correction is proposed based on an on orbit linear adjustment to prelaunch quadratic calibration. The new calibrations are more consistent with SNO samples, and greatly improve the performance over high reflective scene comparing with linear results verified by statistical measurements over Deep Convective Clouds targets. It is demonstrated that other reference is necessary in ocean color channels as MODIS reflectance is within 10% where the nonlinear feature is likely much serious. It is an invaluable lesson that the temporal variation of calibration slope not always indicates the detector’s degradation, but maybe is the valuable information that helps to expose undiscovered characters of instrument.
Using Global Space-based Inter-Calibration System (GSICS), the thermal infrared (TIR) channel was calibrated with high precision. During this procedure, a new calibration table was made, and the sensor non-linear effect was corrected. During the TIR channel image remapping and Level2 (L2) dataset generation procedure, the bilinear interpolation was widely used. Most of the L2 data was stored in D/N count with corresponding calibration table, which assumes D/N count is linear. But in the real world, the non-linear D/N count which comes from imprecise modeled A/D transformation of instrument sensor, will lead to the temperature bias on L2 dataset, even though the high precision calibration Look up Table (LUT) was regenerated. In this paper, D/N bias comes from the mapping process was diagnosed, with the consideration of the temperature difference between neighbor pixels.
BRDF has numerous applications in on-orbit satellites vicarious calibration. The 2013 Dunhuang Gobi surface directional reflectance measurements experiment were held during Aug. 20 to Aug. 28. In order to match the spatial resolution (0.25-1.25km) of meteorological satellites, 3*3 sample points were selected covering the 10*10km area. All the data were measured during (3 hours before and after) the noon without taking into account the large sun zenith angle because of the lack of the satellite passing through. Totally 9 groups of directional reflectance (DREF) were measured by the use of ASD (350-2500nm), standard reference board and a portable DREF measurement system. At each point, DREF were measured by different observation zenith angle (0, 20, 40 and 60 degree) and azimuth angle (0, 45, 90, 135, 180, 225, 270, 315 and 360 degree) in 30 minutes. Different BRDF models were selected such as Walthall, Sine Walthall, Hapke, Roujean and Ross-Li. The model coefficients were derived corresponding to the observed data. The relative differences (RD) of the models with respect to the measured values were calculated. The accuracy of MCD43 products in the Julian day of 233 and 241 were also validated. Results showed that Ross-Li model had the smallest RD. The RD between the DREF from MCD43 products and the measured values were 10.26%(233) and 8.96% (241)@550nm, respectively.
After January 13, 2012, FY-2F had successfully launched, the total number of the in orbit operating FengYun-2 geostationary meteorological satellites reached three. For accurate and efficient application of multi-satellite observation data, the study of the multi-satellites normalization of the visible detector was urgent. The method required to be non-rely on the in orbit calibration. So as to validate the calibration results before and after the launch; calculate day updating surface bidirectional reflectance distribution function (BRDF); at the same time track the long-term decay phenomenon of the detector's linearity and responsivity. By research of the typical BRDF model, the normalization method was designed. Which could effectively solute the interference of surface directional reflectance characteristics, non-rely on visible detector in orbit calibration. That was the Median Vertical Plane (MVP) method. The MVP method was based on the symmetry of principal plane, which were the directional reflective properties of the general surface targets. Two geostationary satellites were taken as the endpoint of a segment, targets on the intersecting line of the segment's MVP and the earth surface could be used as a normalization reference target (NRT). Observation on the NRT by two satellites at the moment the sun passing through the MVP brought the same observation zenith, solar zenith, and opposite relative direction angle. At that time, the linear regression coefficients of the satellite output data were the required normalization coefficients. The normalization coefficients between FY-2D, FY-2E and FY-2F were calculated, and the self-test method of the normalized results was designed and realized. The results showed the differences of the responsivity between satellites could up to 10.1%(FY-2E to FY-2F); the differences of the output reflectance calculated by the broadcast calibration look-up table could up to 21.1%(FY-2D to FY-2F); the differences of the output reflectance from FY-2D and FY-2E calculated by the site experiment results reduced to 2.9%(13.6% when using the broadcast table). The normalized relative error was also calculated by the self-test method, which was less than 0.2%.