The FY-3D medium resolution spectral imager (MERSI-II) has been in orbit for more than five years. Quantitative application inversion shows that the radiometric response of some channels has deteriorated significantly, seriously affecting the accuracy of satellite quantitative remote sensing products. Therefore, it is necessary to evaluate the radiometric response of the MERSI-II and update the operation calibration according to the evaluation results. Compared with traditional targets such as deserts and glaciers, deep convective cloud (DCC) features a higher signal-to-noise ratio, isotropic reflection characteristics close to Lambertian, and minimum water vapor absorption. In-orbit calibration methods based on DCC target are widely employed to monitor the radiation performance of satellite sensors. However, the uncertainty factors in the DCC calibration method will affect the accuracy and stability, such as DCC bidirectional reflectance distribution function (BRDF) correction model, reflectivity probability density function (PDF) eigenvalues, and seasonal fluctuations. In this study, the basic DCC calibration method is optimized. Firstly, the BRDF correction effects of the Hu model and CERES thick ice cloud model are compared, as well as the stability of PDF mode and the mean of DCC reflectivity. Secondly, a deseasonal method based on reflectivity fitting residual and the moving average is proposed as the BRDF correction model of DDC is invalid in the SWIR band. Finally, the radiation performance of FY-3D/MERSI-II reflected solar bands is quantitatively evaluated by the optimization method. The evaluation results will be adopted as an important basis for the service calibration update of MERSI-II reflected solar bands.

We utilize DCC targets to evaluate the trend of radiometric response changes in most of the MERSI-II reflected solar bands. Firstly, the DCC target pixels are identified according to the infrared 10.8 μm channel brightness temperature threshold, observation geometric conditions (latitude, solar zenith angle, observation zenith angle), and spatial homogeneity conditions. Secondly, the solar zenith angle and earth-sun distance corrections are performed on the identified reflectivity of each DCC pixel to obtain the apparent reflectivity of the DCC. Then, two DCC BRDF models are leveraged to correct the anisotropy of apparent reflectance and compare the correction effects. Thirdly, the monthly DCC reflectivity PDF is constructed. Finally, the radiometric response of the MERSI-II sensor is evaluated by tracking the monthly DCC reflectivity PDF mean or mode sequence. In this section, we investigate the seasonal characteristics of DCC and propose a deseasonal method to reduce the fluctuations of the reflectivity PDF mean or modal sequence.

For the VIS/NIR bands, the correction effect of the CERES thick ice cloud model is better than the Hu model (Fig. 2), the RSD of the reflectivity is reduced by 15%-40%, and the fluctuations are reduced by about 10%-30% (Fig. 3). However, for the SWIR band, the two BRDF models have no obvious correction effect. For the VIS/NIR band, the monthly DCC reflectivity PDF mode is more stable than the mean, while the mean is more stable in the SWIR band (Table 2). The DDC distribution area migrates from north to south seasonally, and the same month in each year has high similarity, while the distribution of different months in each year is different (Fig. 4). The annual variation range of monthly DCC reflectance in VIS/NIR and SWIR bands is about 1.5% and 6.4% respectively, and SWIR band has higher seasonal sensitivity (Fig. 5). The proposed deseasonal method has yielded significant results in the SWIR band (Fig. 7). The reflectance RSD of 1.38 μm, 1.64 μm, and 2.13 μm channels decreases by about 22.4%, 22.0%, and 23.9% respectively, and the fluctuations decrease by about 52.9%, 51.2%, and 54.5% respectively, which compensates for the inefficiency of the BRDF model in the SWIR band (Table 3). The optimized DCC calibration method is employed to quantitatively evaluate the radiation degradation of the MERSI-II reflected solar bands from 2018 to 2022 (Fig. 8). The annual average degradation rate of 0.65 μm channel is only 0.02826% and the 0.47 μm blue channel is greater than 1.3%. The annual degradation rate of the three water vapor absorption channels in the NIR band is between 0.5% and 1%. This reveals a law that the degradation rate increases with the wavelength, and the fluctuation index is less than 3%, proving the superiority of the DCC calibration method in the water vapor absorption channel. The SWIR band has the most significant degradation, of which 1.38 μm has the largest degradation and the strongest fluctuation. The annual decay rates of 1.64 μm and 2.13 μm channels are 3.114 % and 2.134 %, respectively (Table 5).

We adopt the DCC target to evaluate the radiometric response trend of the MERSI-II reflected solar bands, and optimize the basic DCC calibration method. Firstly, the DCC BRDF correction model and the selection of monthly PDF eigenvalues (mean/mode) are optimized. Secondly, the seasonal characteristics of DCC are investigated, and a deseasonal method based on reflectivity fitting residuals and moving averages is proposed to compensate for the ineffectiveness of Hu and CERES thick ice cloud BRDF models in SWIR bands. This method significantly mitigates the RSD and fluctuations of monthly DCC reflectance in the SWIR band from 2018 to 2022, which are reduced by about 23% and 53%, respectively. This method is not only applicable to MERSI-II but also provides references for the radiometric calibration of other satellite optical remote sensors. Finally, the optimized DCC calibration method is utilized to quantitatively evaluate the attenuation of the radiometric response of the FY-3D/MERSI-II reflected solar bands. The results show that the blue channel and the four SWIR channels have significant degradation.