Evaluation on Radiometric Capability of Chinese Optical Satellite Sensors

The radiometric capability of on-orbit sensors should be updated on time due to changes induced by space environmental factors and instrument aging. Some sensors, such as Moderate Resolution Imaging Spectroradiometer (MODIS), have onboard calibrators, which enable real-time calibration. However, most Chinese remote sensing satellite sensors lack onboard calibrators. Their radiometric calibrations have been updated once a year based on a vicarious calibration procedure, which has affected the applications of the data. Therefore, a full evaluation of the sensors’ radiometric capabilities is essential before quantitative applications can be made. In this study, a comprehensive procedure for evaluating the radiometric capability of several Chinese optical satellite sensors is proposed. In this procedure, long-term radiometric stability and radiometric accuracy are the two major indicators for radiometric evaluation. The radiometric temporal stability is analyzed by the tendency of long-term top-of-atmosphere (TOA) reflectance variation; the radiometric accuracy is determined by comparison with the TOA reflectance from MODIS after spectrally matching. Three Chinese sensors including the Charge-Coupled Device (CCD) camera onboard Huan Jing 1 satellite (HJ-1), as well as the Visible and Infrared Radiometer (VIRR) and Medium-Resolution Spectral Imager (MERSI) onboard the Feng Yun 3 satellite (FY-3) are evaluated in reflective bands based on this procedure. The results are reasonable, and thus can provide reliable reference for the sensors’ application, and as such will promote the development of Chinese satellite data.


Introduction
The stability and consistency of satellite sensors are critical for providing consistent data and products for the monitoring of Earth's resources and dynamics [1]. Remote sensing imagery is an effective tool for monitoring global environment and climate changes because of its long-term observations [2]. The quantitative applications using remote sensing are sensitive to the sensor's radiometric capability [3]. Discerning secular trends in geophysical processes is one goal of earth observation research, thus long-term data from remote sensing instruments with good performance is required [4].
Frequent calibration for on-orbit sensors is required. As Wyatt [5] pointed out, radiometric calibration is a key radiometric characteristic required for understanding an instrument's performance. Sensors' radiometric capability can be expected to change during launch and on-orbit operations because of the space environmental varying and instrument aging [6]. In order to correct the changes, regular and reliable onboard absolute calibration are required to assure a sensor's radiometric capability Third, the site has a typical continental arid climate, and the atmosphere is always dry and clean (low aerosol).

Data
As a typical representative of moderate-high resolution remote sensing sensors, the CCD cameras are VNIR imaging sensors onboard the Chinese environment and disaster monitoring and forecasting satellite constellation, which is called Huan Jing 1 in Chinese and abbreviated as HJ-1 (hereafter the CCD camera onboard HJ-1 satellite is written as HJ-1/CCD). The HJ shoulders heavy responsibility of environmental monitoring, disaster prevention and reduction in China, and has been in orbit for more than six years and remains in operation, despite a two-year design life. However, due to lack of an onboard calibration system to track HJ-1 sensors' on-orbit behavior throughout the life of the mission, vicarious calibration needs to be done every other time and the frequency is usually once a year and will be increased with the aging of instruments. Since the CCD cameras are not state-of-the-art instruments, the radiometric capability is not so stable that it does not change for the whole year. In order to obtain HJ-1/CCD data which are accurately and continuously calibrated, CRESDA performs the vicarious calibration measurements using the reflectance-based method at the Dunhuang calibration site, and releases the calibration coefficients once a year through its website at http://www.cresda.com.
The FengYun-3 satellite constellation, the second generation of Chinese polar-orbiting meteorological satellites with moderate-low resolution, is composed of three satellites: FY-3A, FY-3B, and FY-3C, respectively. FY-3A and FY-3C were launched on 27 May 2008 and 23 September 2013 by China National Satellite Meteorological Center (NASC), separately, in a sun-synchronous morning orbit with a local equator-crossing time of 10:30 a.m. in descending node, while FY-3B was launched on 5 November 2010 by NASC, in an afternoon orbit with an equator-crossing time of 1:30 p.m. in ascending node [15]. The MERSI and VIRR are the keystone payloads onboard FY-3. They are designed to enhance Chinese three-dimensional atmospheric sounding capability and global data acquisition capability. The MERSI images the Earth with 20 spectral bands covering from visible to long-wave infrared. The applications of MERSI include retrieval of true color image; monitoring cloud, ice, and snow coverage; monitoring ocean color; and detecting disasters such as flooding and wildfires [16].The VIRR is a 10-channel VIS/IR multi-purpose imaging radiometer, with a spectral range from 0.43 μm to 12.5 μm. VIRR is the only instrument on the current FY-3 platform that can provide high spatial resolution and narrow-band infrared window measurements, which are vital for cloud or fog detection and quantitative retrieval of cloud properties, surface temperature, and the

Data
As a typical representative of moderate-high resolution remote sensing sensors, the CCD cameras are VNIR imaging sensors onboard the Chinese environment and disaster monitoring and forecasting satellite constellation, which is called Huan Jing 1 in Chinese and abbreviated as HJ-1 (hereafter the CCD camera onboard HJ-1 satellite is written as HJ-1/CCD). The HJ shoulders heavy responsibility of environmental monitoring, disaster prevention and reduction in China, and has been in orbit for more than six years and remains in operation, despite a two-year design life. However, due to lack of an onboard calibration system to track HJ-1 sensors' on-orbit behavior throughout the life of the mission, vicarious calibration needs to be done every other time and the frequency is usually once a year and will be increased with the aging of instruments. Since the CCD cameras are not state-of-the-art instruments, the radiometric capability is not so stable that it does not change for the whole year. In order to obtain HJ-1/CCD data which are accurately and continuously calibrated, CRESDA performs the vicarious calibration measurements using the reflectance-based method at the Dunhuang calibration site, and releases the calibration coefficients once a year through its website at http://www.cresda.com.
The Feng Yun-3 satellite constellation, the second generation of Chinese polar-orbiting meteorological satellites with moderate-low resolution, is composed of three satellites: FY-3A, FY-3B, and FY-3C, respectively. FY-3A and FY-3C were launched on 27 May 2008 and 23 September 2013 by China National Satellite Meteorological Center (NASC), separately, in a sun-synchronous morning orbit with a local equator-crossing time of 10:30 a.m. in descending node, while FY-3B was launched on 5 November 2010 by NASC, in an afternoon orbit with an equator-crossing time of 1:30 p.m. in ascending node [15]. The MERSI and VIRR are the keystone payloads onboard FY-3. They are designed to enhance Chinese three-dimensional atmospheric sounding capability and global data acquisition capability. The MERSI images the Earth with 20 spectral bands covering from visible to long-wave infrared. The applications of MERSI include retrieval of true color image; monitoring cloud, ice, and snow coverage; monitoring ocean color; and detecting disasters such as flooding and wildfires [16]. The VIRR is a 10-channel VIS/IR multi-purpose imaging radiometer, with a spectral range from 0.43 µm to 12.5 µm. VIRR is the only instrument on the current FY-3 platform that can provide high spatial resolution and narrow-band infrared window measurements, which are vital for cloud or fog detection and quantitative retrieval of cloud properties, surface temperature, and the earth radiation budget [17]. Annual field campaigns have been routinely carried out at China Radiometric Calibration Site (CRCS) since September 2008, and calibration coefficients is released or updated through its website at http://www.nsmc.cma.gov.cn [18].
The MODIS sensor was launched on 18 December 1999 aboard the Terra satellite and has been in operation for nearly 17 years, providing global data that are used to monitor long-term changes in the Earth system. The absolute calibration of MODIS reflective solar bands is based on the regular measurements from the Solar Diffuser and a Solar Diffuser Stability Monitor [19][20][21]. Another onboard calibrator, the Spectroradiometric Calibration Assembly, is used to track the VNIR band's spectral response shifts and bandwidth changes on-orbit [20]. MODIS also periodically views the moon through its space view port, which provides an additional point reference for the calibration [21]. The calibration uncertainties of the MODIS TOA reflectance products are ±2%, whereas a 5% uncertainty requirement is specified for the reflective solar bands at sensor spectral radiance calibration [22]. The MODIS instrument onboard the Earth Observing System's (EOS) Terra and Aqua platforms have a state-of-the-art onboard calibration system with an absolute accuracy better than 2% [23], which has calibrated the solar reflective bands to reflectance units. Therefore, MODIS as a well-calibrated instrument is very practical for radiometric cross-calibrating those sensors without a good onboard calibration system by using targets of known reflectance [24,25]. Furthermore, as one of the most stable sensors, MODIS has been widely used for radiometric capability evaluation of other sensors [6,11,25]. In this paper, the MODIS level 1b data including MOD02 (Level-1B Calibrated Geolocation Data Set) and MOD03 (Geolocation Data Set is chosen as the reference data to evaluate the radiometric capability of the three sensors.
The primary characteristics of the CCDs, MERSIs, VIRRs and MODIS are listed in Table 1.

Methods
The procedure in this paper for evaluating radiometric capability of Chinese optical satellite sensors using MODIS is shown in Figure 2 and it includes three parts: (1) calculating the TOA reflectance; (2) spectral matching to eliminate the influence of different spectral response; (3) evaluating the radiometric capability of the sensors using the long-term series TOA reflectance between sensors and MODIS after spectral matching.  The HJ-1/CCD data used in this study are derived from the Level 2 products, and the ground observation station of Chinese resources satellites has conducted radiation correction on these data to correct for the radiation deviation of the sensors. The TOA reflectance of the CCD can be calculated using Equation (1)    The HJ-1/CCD data used in this study are derived from the Level 2 products, and the ground observation station of Chinese resources satellites has conducted radiation correction on these data to correct for the radiation deviation of the sensors. The TOA reflectance of the CCD can be calculated using Equation (1).
where ρ λ is the TOA reflectance; L λ is the TOA radiance; d is the Earth-Sun distance in astronomical units; θ SE is the solar elevation; ESUN λ is the solar irradiance at the top of the atmosphere; and L λ can be calculated using Equation (2)   The TOA reflectance of the FY-3A/MERSI can be calculated using Equation (3).
where k 0 , k 1 and k 2 is the calibration coefficients, which can be read from the data source file. For the record, the calibration coefficients are not always the same because the vicarious calibration is routinely carried out annually, and updated through the China National Satellite Meteorological Center (NSMC) website (http://www.nsmc.cma.gov.cn/). The calibration coefficients of FY-3A/MERSI published by NSMC have experienced three updates since the launch. That is, four stages of calibration can be divided from launch until 2015. All the calibration coefficients are listed in Table 3. The radiometric calibration of the FY-3B/MERSI has experienced two stages. In the first stage, the TOA reflectance of the FY-3B/MERSI data from 18 October 2010 to 6 March 2013 can be calculated using Equation (4) using coefficients in Table 4. After that, the organization structure of the FY-3B/MERSI data source file has been updated and calibration equation is changed. The TOA reflectance of the FY-3B/MERSI data from 6 March 2013 until now can be calculated using Equation (4) where Re f Factor is the factor of reflectance, and can be read from the data source file.
where S is the calibration slope, I is the calibration intercept, and C E is the channel count value. S, I and C E are calibration coefficients and can be obtained from the data source file. Both of them have two stages of calibration coefficients, which are listed in Table 5. Unlike the CCDs, MERSIs and VIRRs, whose calibrations are performed relying on the site calibration annually, the calibration of MODIS in VNIR bands are based on the regular measurements from the onboard calibrators, so the calibration coefficients of MODIS data are different according to the data acquisition time and satellite attitude, etc. The TOA reflectance of the MODIS can be calculated using Equation (6).
where ρ λ is the TOA reflectance; θ SZ is the solar zenith angle; re f lectance_scale and re f lectance_o f f sets are the gain and offset of the reflectance, respectively; DN is the channel count value.

Spectral Matching
Figures 3-5 show the relative spectral responses of the CCDs, MERSIs, VIRRs and MODISs in blue, green, red, and NIR bands. These relative spectral response profiles differ in shape, bandwidth, peak value, location of the central wavelength, and degree of overlap between channels. Since the similar principle of design, the relative spectral response profiles of the sensors are almost unanimous. Both FY-3A/VIRR and FY-3B/VIRR data can be calibrated using Equation (5) 2 () 100 cos( ) where S is the calibration slope, I is the calibration intercept, and E C is the channel count value. S , I and E C are calibration coefficients and can be obtained from the data source file. Both of them have two stages of calibration coefficients, which are listed in Table 5. Unlike the CCDs, MERSIs and VIRRs, whose calibrations are performed relying on the site calibration annually, the calibration of MODIS in VNIR bands are based on the regular measurements from the onboard calibrators, so the calibration coefficients of MODIS data are different according to the data acquisition time and satellite attitude, etc. The TOA reflectance of the MODIS can be calculated using Equation (6).
where   is the TOA reflectance; SZ  is the solar zenith angle; _ reflectance scale and _ reflectance offsets are the gain and offset of the reflectance, respectively; DN is the channel count value.

Spectral Matching
Figures 3-5 show the relative spectral responses of the CCDs, MERSIs, VIRRs and MODISs in blue, green, red, and NIR bands. These relative spectral response profiles differ in shape, bandwidth, peak value, location of the central wavelength, and degree of overlap between channels. Since the similar principle of design, the relative spectral response profiles of the sensors are almost unanimous.       Due to the atmospheric transmittance and the atmosphere outside solar irradiance varying with wavelength, different responses will affect the sensors' reflectances at the top of the atmosphere. Thus, spectral matching is performed to compare radiation characteristics between the sensors to be evaluated and MODIS. Given the spectral behavior of the site (the ground feature is desert) and the spectral response functions of the instruments, the spectral difference between MODIS and other sensors can be expressed as a linear translation equation of the form where ()  is the spectral range of other sensor; 3  − 4  is the spectral range of MODIS. The ground-measured spectrum of the calibration site we used in this paper comes from the measurement in the Badain Jaran Desert [14]. Based on the definition of the spectral matching factor, the spectral matching factors between the sensors and the MODIS in reflective bands are calculated and listed in Table 6. Radiometric stability and accuracy are the two most important indicators to evaluate a sensor's radiometric capability. Radiometric stability is defined as follows: given a calibration site with very stable radiometric characterizations, such as Badain Jaran Desert in this study, the radiance measured by the sensor to be evaluated is stable temporally. Ideally, it would be permanent, but it is usually defined as the sensor's lifetime. In this study, the radiometric stability of a sensor is analyzed by using the tendency of long-term top-of-atmosphere (TOA) reflectance at the calibration site. Radiometric accuracy is defined as how the radiance measured from the sensor is close to the actual radiance Due to the atmospheric transmittance and the atmosphere outside solar irradiance varying with wavelength, different responses will affect the sensors' reflectances at the top of the atmosphere. Thus, spectral matching is performed to compare radiation characteristics between the sensors to be evaluated and MODIS. Given the spectral behavior of the site (the ground feature is desert) and the spectral response functions of the instruments, the spectral difference between MODIS and other sensors can be expressed as a linear translation equation of the form ρ sensor,i = a i · ρ modis,j , where i represents band number of the other sensor and j is the corresponding band number of MODIS. The spectral difference coefficient a i is calculated based on the Equation (7) [14,26,27].
where f sensor (λ) and f modis (λ) are the relative spectral response functions for other sensors and MODIS, respectively. λ 1 −λ 2 is the spectral range of other sensor; λ 3 −λ 4 is the spectral range of MODIS. The ground-measured spectrum of the calibration site we used in this paper comes from the measurement in the Badain Jaran Desert [14]. Based on the definition of the spectral matching factor, the spectral matching factors between the sensors and the MODIS in reflective bands are calculated and listed in Table 6.

Evaluation Method
Radiometric stability and accuracy are the two most important indicators to evaluate a sensor's radiometric capability. Radiometric stability is defined as follows: given a calibration site with very stable radiometric characterizations, such as Badain Jaran Desert in this study, the radiance measured by the sensor to be evaluated is stable temporally. Ideally, it would be permanent, but it is usually defined as the sensor's lifetime. In this study, the radiometric stability of a sensor is analyzed by using the tendency of long-term top-of-atmosphere (TOA) reflectance at the calibration site. Radiometric accuracy is defined as how the radiance measured from the sensor is close to the actual radiance measured using ground instruments with high precision. In this study, the MODIS is used as the standard instruments with excellent performance on radiometric capability and the radiometric accuracy of a sensor is determined by comparing with the TOA reflectance from MODIS after spectrally matching.
In order to evaluate the radiometric stability and accuracy for a given sensor, a procedure including three steps is worked out and the procedure is as follows: Firstly, the long-term TOA reflectance of MODIS is plotted and the stability of MODIS is analyzed. Furthermore, it can be compared to the to-be-evaluated Chinese sensors visually.
Secondly, the long-term TOA reflectance of each reflective band for every sensor is plotted, which can illustrate the radiometric stability very intuitively. If the stability is good enough, a linear fitting of the long-term radiometric tendency can be done and the stability can be quantitatively described using the slope of the fitted line. If the stability is not good, the possible reasons can be analyzed based on the plot.
Thirdly, compare the TOA reflectance between the sensor to be evaluated and the MODIS to get the radiometric accuracy of the to-be-evaluated sensors. By employing some indices including maximum, minimum, mean, and standard deviation of TOA reflectance, the accuracy of the sensor can be quantified.
The indices for evaluating a sensor's radiometric capability are summarized as follows: (1) Slope: the slope of the fitted line for long-term tendency of the TOA reflectance at the calibration site indicates the radiometric stability of a sensor. When a slope value is close to 0, it means the TOA reflectance remains stable. A positive slope means an increasing tendency of TOA reflectance and a negative one means a decreasing tendency. (4) Standard deviation. The standard deviation is a measure that is used to quantify the amount of variation or dispersion of a set. A low standard deviation indicates that the elements tend to be close to the mean of the set, while a high standard deviation indicates that the elements are spread out over a wider range of values. The standard deviation can be calculated using the Equation (8): where i represents serial number of the selected images, ρ abs is the mean value of the TOA reflectance, and ρ abs(i) is the value of the TOA reflectance difference.

Radiometric Capability Evaluation of Moderate Resolution Imaging Spectroradiometer (MODIS)
The Terra/MODIS acquired from 2008 to 2015 and the Aqua/MODIS acquired from 2010 to 2015 have been collected in order to evaluate the temporal radiometric capability of the two sensors. Figure 6 shows that the stability of MODIS from both Terra and Aqua are visually excellent. Subsequently, in order to evaluate the radiometric capability of MODIS quantitatively, a linear fitting to the TOA reflectance data was carried out and all the statistics are listed in Table 7. Table 7 summarizes the slopes and intercepts of the trend lines, and the maximum, minimum, mean and standard deviation values of the TOA reflectances in reflective bands including blue, green, red, and NIR. Several things about the table need to be clarified.
(1) The difference between maximum and minimum values indicates the range of TOA reflectance variation. The blue band usually has a larger variation than the other three bands and the variation becomes smaller and smaller with an increasing wavelength, because the TOA reflectance at a shorter wavelength incorporates more scattering radiation from the atmosphere, and the atmospheric effect becomes stronger with the increasing of aerosols. For example, the difference of TOA reflectance at the blue band for Terra is 0.185 − 0.106 = 0.079; compared to the mean TOA reflectance of 0.14, the variation reaches 0.079/0.14 = 56.43%. The variations of the other three bands are 40.96%, 27.80%, and 23.72% respectively. In addition, the minimum, maximum, and the values close to the minimum and the maximum appear for most of the years, which clarifies that the maximum and minimum are not abnormal values but rather the normal ones, and that the aerosol condition varies day by day, though the variation is very similar year by year. (2) Besides the atmospheric effect, the BRDF also contributes to the variation of the TOA reflectance. Figure 7 gives an example of the directional effect of the red band of Terra/MODIS. The reflectance varies with relative azimuth and solar zenith angle for each bin of view zenith angle, which shows systematic variation that is due to directional effects. The directional effect is about 15%. (3) The mean of TOA reflectance at the same band are usually a little bit different, especially at the blue and red bands. The radiometric accuracy of Aqua is slightly better than that of Terra [28].
Although several factors can induce the variation of the TOA reflectance, the long-term tendency of the TOA reflectance remains consistent and the slope values of the fitted lines for different bands range from 10 −7 to 10 −5 , which indicates a very small variation trending. The intercepts of the fitted lines are close to the corresponding means of TOA reflectance. The standard deviations of the TOA reflectance are within 0.02. All of the above show that the radiometric capability of both Terra/MODIS and the Aqua/MODIS are stable, and can be used to evaluate other sensors as the reference data. (1) The difference between maximum and minimum values indicates the range of TOA reflectance variation. The blue band usually has a larger variation than the other three bands and the variation becomes smaller and smaller with an increasing wavelength, because the TOA reflectance at a shorter wavelength incorporates more scattering radiation from the atmosphere, and the atmospheric effect becomes stronger with the increasing of aerosols. For example, the difference of TOA reflectance at the blue band for Terra is 0.185 − 0.106 = 0.079; compared to the mean TOA reflectance of 0.14, the variation reaches 0.079/0.14 = 56.43%. The variations of the other three bands are 40.96%, 27.80%, and 23.72% respectively. In addition, the minimum, maximum, and the values close to the minimum and the maximum appear for most of the years, which clarifies that the maximum and minimum are not abnormal values but rather the normal ones, and that the aerosol condition varies day by day, though the variation is very similar year by year. (2) Besides the atmospheric effect, the BRDF also contributes to the variation of the TOA reflectance. Figure 7 gives an example of the directional effect of the red band of Terra/MODIS. The reflectance varies with relative azimuth and solar zenith angle for each bin of view zenith angle, which shows systematic variation that is due to directional effects. The directional effect is about 15%. (3) The mean of TOA reflectance at the same band are usually a little bit different, especially at the blue and red bands. The radiometric accuracy of Aqua is slightly better than that of Terra [28].
Although several factors can induce the variation of the TOA reflectance, the long-term tendency of the TOA reflectance remains consistent and the slope values of the fitted lines for different bands range from 10 −7 to 10 −5 , which indicates a very small variation trending. The intercepts of the fitted lines are close to the corresponding means of TOA reflectance. The standard deviations of the TOA reflectance are within 0.02. All of the above show that the radiometric capability of both Terra/MODIS and the Aqua/MODIS are stable, and can be used to evaluate other sensors as the reference data.

Radiometric Capability Evaluation of Huan Jing 1/Charge-Coupled Device (HJ-1/CCD)
The same plots of the HJ-1/CCDs as the MODISs are presented in Figure 8. The temporal range of HJ-1/CCDs is 2009~2012 when the sensors are operational and in excellent condition. In Figure 8, it can be observed that the stability of the four CCDs is not as good as the MODIS for pronounced calibration drifts. The TOA reflectance of HJ-1/CCD varies periodically and the turning points are exactly the same as the calibration time, which is done once a year based on a vicarious calibration, thereby demonstrating that the procedure of the calibration may have some issues. However, as we have not involved the calibration, this is just an assumption.

Radiometric Capability Evaluation of Huan Jing 1/Charge-Coupled Device (HJ-1/CCD)
The same plots of the HJ-1/CCDs as the MODISs are presented in Figure 8. The temporal range of HJ-1/CCDs is 2009~2012 when the sensors are operational and in excellent condition. In Figure 8, it can be observed that the stability of the four CCDs is not as good as the MODIS for pronounced calibration drifts. The TOA reflectance of HJ-1/CCD varies periodically and the turning points are exactly the same as the calibration time, which is done once a year based on a vicarious calibration, thereby demonstrating that the procedure of the calibration may have some issues. However, as we have not involved the calibration, this is just an assumption.

Radiometric Capability Evaluation of Huan Jing 1/Charge-Coupled Device (HJ-1/CCD)
The same plots of the HJ-1/CCDs as the MODISs are presented in Figure 8. The temporal range of HJ-1/CCDs is 2009~2012 when the sensors are operational and in excellent condition. In Figure 8, it can be observed that the stability of the four CCDs is not as good as the MODIS for pronounced calibration drifts. The TOA reflectance of HJ-1/CCD varies periodically and the turning points are exactly the same as the calibration time, which is done once a year based on a vicarious calibration, thereby demonstrating that the procedure of the calibration may have some issues. However, as we have not involved the calibration, this is just an assumption. (1) For HJ-1A/CCD1 (Table 8)  (2) For HJ-1A/CCD2 (Table 9)  (1) For HJ-1A/CCD1 (Table 8) is maybe enough to ensure the radiometric capability; however, it cannot guarantee the radiometric capability after the beginning stage of the operation. Subsequently, some new and highly frequent calibration methods need to be proposed and executed by the instruments' surveillance department.

Radiometric Capability Evaluation of Feng Yun 3/Medium-Resolution Spectral Imager (FY-3/MERSI)
The same plots of the FY-3/MERSIs as the MODISs are presented in Figure 9. The temporal range of FY-3A/MERSI is 2008~2015 and that of FY-3B/MERSI is 2010~2015. The radiometric characteristics of the MERSIs are as follows: (1) The stability of FY-3A/MERSI is very good at all times except the duration from February to December of 2012, when a strongly increasing trend occurred. Although calibration coefficients have been updated four times, the variation of the TOA reflectance remains consistent. The reasons inducing the abnormal increasing within stage 3 is unknown and we surmise that it may be caused by an instrument malfunction. After it was found by the surveillance department, a manual adjustment was made and the instrument suddenly became normal. The statistics for FY-3A/MERSI are calculated, excluding the instrument malfunction duration, and listed in Table 12. The slope values of the fitted lines for different bands are at the 10 −5 level, which indicates a very small variation trending. The highest variation of TOA reflectance occurs in the blue band, up to 62.72%. The variations of the other three bands are 33.74%, 24.86%, and 25.52% respectively. The differences in four bands between the mean TOA reflectance after spectral matching and corresponding bands of MODIS are all less than 0.015. All the standard deviations of the TOA reflectance are within 0.02. After the malfunction duration is removed, the stability and accuracy are very close to that of MODIS. (2) The stability of FY-3B/MERSI has a two-stage mode. At the first stage, three sub-stages are very obvious. An obvious over-estimation of the TOA reflectance appears at the first sub-stage; the second sub-stage has a close estimation; the third sub-stage has an under-estimation. The three sub-stages changed suddenly, so manual adjustments of the instrument were executed by the surveillance department. However, the radiometric stability and accuracy have become better and very close to the MODIS's. Only the statistics of the second stage are calculated and listed in

Radiometric Capability Evaluation of FY-3/Visible and Infrared Radiometer (VIRR)
The same plots of the FY-3/VIRRs as the MODISs are presented in Figure 10. The temporal range of FY-3A/VIRR is 2008~2015 and that of FY-3B/VIRR is 2010~2015. The radiometric characteristics of the VIRRs are as follows: (1) Both VIRRs have a tendency toward obvious decreasing in their first calibration stages. The reason for the trend can probably be attributed to the degradation of the instruments. At the beginning of the operations of VIRRs, the TOA reflectance in the four bands are abnormally high, especially in the blue band and even higher in the green band. After further investigated and constant adjustment, the instruments became normal and stable in the second stage. The statistics are calculated and listed in Table 13. respectively. Only the difference in blue band between the mean TOA reflectance after spectral matching and MODIS is close to 0 (0.0004). The mean TOA reflectance of the FY-3B/VIRR after spectral matching in other three bands are less than that of MODIS and the differences are higher than 0.02 (0.023 for green band, 0.022 for red band and 0.028 for NIR band). All the standard deviations of the TOA reflectance are within 0.02. (4) Low frequency of calibrations is still a problem for VIRRS and a high frequent procedure for calibration could be a solution for the decreasing trend's instability.

Radiometric Capability Evaluation of FY-3/Visible and Infrared Radiometer (VIRR)
The same plots of the FY-3/VIRRs as the MODISs are presented in Figure 10. The temporal range of FY-3A/VIRR is 2008~2015 and that of FY-3B/VIRR is 2010~2015. The radiometric characteristics of the VIRRs are as follows: (1) Both VIRRs have a tendency toward obvious decreasing in their first calibration stages. The reason for the trend can probably be attributed to the degradation of the instruments. At the beginning of the operations of VIRRs, the TOA reflectance in the four bands are abnormally high, especially in the blue band and even higher in the green band. After further investigated and constant adjustment, the instruments became normal and stable in the second stage. The statistics are calculated and listed in Table 13. respectively. Only the difference in blue band between the mean TOA reflectance after spectral matching and MODIS is close to 0 (0.0004). The mean TOA reflectance of the FY-3B/VIRR after spectral matching in other three bands are less than that of MODIS and the differences are higher than 0.02 (0.023 for green band, 0.022 for red band and 0.028 for NIR band). All the standard deviations of the TOA reflectance are within 0.02. (4) Low frequency of calibrations is still a problem for VIRRS and a high frequent procedure for calibration could be a solution for the decreasing trend's instability.

Conclusions
Since an excellent radiometric capability of satellite optical sensors is prerequisite, the radiometric capability needs to be evaluated first. Most of the new satellite optical sensors have onboard calibrators, which have the ability of real-time monitoring; therefore, an excellent radiometric capability can be expected. However, most of the Chinese satellite optical sensors lack onboard calibrators and the radiometric capability of these sensors needs to be evaluated often to ensure the quality of the data. Thus, this paper proposed an alternative procedure for monitoring the radiometric capability of Chinese optical satellite sensors using long time series data. In this procedure, two major indicators including radiometric stability and accuracy are employed for evaluation. Following an evaluation of the eight Chinese satellite optical sensors, the following conclusions can be made.
First of all, almost all of the Chinese satellite optical sensors are less stable than the Moderate Resolution Imaging Spectroradiometers (MODISs), and their radiometric accuracy is less than that of the MODISs.
Secondly, among all the evaluated sensors, the Visible and Infrared Radiometer (VIRRs) thus far have the best stability throughout their lifetime, although they have an obvious decreasing trend probably induced by instrument degradation; the Medium-Resolution Spectral Imagers (MERSIs) have the best radiometric performance in terms of both stability and accuracy at their later stages.
Thirdly, a vicarious calibration procedure carried out once a year or once every several years has been operated by the surveillance departments of the sensors, which has greatly contributed to improving the lower-quality radiometric capability of Chinese satellite optical sensors. However, a more frequent calibration procedure urgently needs to be developed and fulfilled in the future due to the lack of onboard calibrator. Moreover, in order to fully take advantage of the ample Chinese satellite data, the recalibration of the historical data also needs to be carried out in the future.
Fourthly, with the ongoing integrated global observations, more and more optical sensors onboard different satellites with global observing capability have been placed into orbit. Quantitative applications through multi-sensors require continued and consistent measurements from different instruments. This study provides reliable reference for the sensors' applications, and as such will promote the development of Chinese satellite data.
In the near future, we will evaluate the radiometric capability of other Chinese optical satellite sensors, such as the Gao Fen (GF) and Zi Yuan (ZY) series' satellites; subsequently, more abundant and reliable data from Chinese optical satellite sensors are expected, which will greatly contribute to research and applications.