Snow targets of Dome C and Greenland in the south and north poles are the calibration tracking targets, which are internationally applied. However, due to the existence of polar night phenomenon, it is difficult to continuously calibrate and track remote sensing instruments in a single polar ice and snow target. In response to this problem, a method of continuous tracking and monitoring of the radiation performance of the FY-3A/medium resolution spectral imager (MERSI) solar reflectance band using a combination of two polar ice targets is proposed. Firstly, using the relatively stable observation data of FY-3A/MERSI in late life stage, two bidirectional distribution function models of apparent reflectivity of ice and snow target are established. Then, the model is used to normalize the apparent reflectivity sequence of ice and snow target, eliminate the change information caused by the target non-Lambertian, obtain the parameters that reflect the change of the radiation performance of the remote sensor, and achieve the fusion of different ice and snow target data series. Finally, using the quadratic polynomial fitting model, the trend of the data after the fusion is analyzed, and it is compared with other related research results. The results show that the uncertainty of the tracking model is better than 3%, especially for the short wavelength of less than 550 nm, in which the uncertainty is better than 2%. Attenuation rate results show that the blue band attenuation is significant, especially in the band 8 (413 nm), the annual average attenuation rate is close to 7%. The red and the near infrared band (excluding the water vapor channel) are the most stable wave bands, and the annual average attenuation rate is within ±0.5%. The verification results show that the proposed method is in good agreement with other methods, and the deviation of the total attenuation rate of many years in most bands is less than 3%.