Data Quality Assessment Of Domestic Satellite FY-3D/MERSI ? And Its Application In Ocean Water Color Remote Sensing

The Medium Resolution Spectral Imager ? is the core optical imager onboard the second generation polar-orbiting meteorological satellite FY-3D of China.It has the characteristics of multiple spectral bands,wide coverage,and high measurement accuracy.Although many scholars have done a lot of research on FY-3D/MERSI ? sensors,few researches have been carried out for Marine applications.In order to promote the application of MERSI ? data in the ocean,the Visible Infrared Imaging Radiometer Suite(VIIRS),one of the mainstream water remote sensing sensors at present,was used as a reference to evaluate the quality of MERSI ? data from the aspects of signal-to-noise ratio(SNR).Then,based on the Neural Network in the machine learning algorithm,the MERSI ? topal-of-atmospheric reflectance(?_t)fitting model and Rayleigh correction reflectance(R_rc)inversion model for the open ocean were established by using the MERSI ? topal-of-atmospheric reflectance data.Finally,based on the neural network method,we used the MERSI ? R_rc data to establish the chlorophyll a concentration(Chl-a)and diffuse attenuation coefficient(Kd?490)inversion models for the ocean area,respectively,which lays a foundation for promoting the application of MERSI ? data in the ocean.The main findings are as follows:(1)The data quality of MERSI ? was evaluated from the aspects of signal-to-noise ratio,etc.The results show that the signal-to-noise ratio of MERSI ? in visible band was higher than300,and the maximum can reach 900,meeting the requirements of water monitoring.It was slightly lower than VIIRS in the visible band and near infrared band,and the SNR was not much different from VIIRS in the short-wave infrared band.The top of atmospheric reflectance data of MERSI ? in various bands were in good agreement with the data of VIIRS,with the coefficient of determination(R²)higher than 0.70 on the whole,and the R² of the blue band up to 0.95.The average absolute percentage error(MAPE)was 5%?50%.After a simple linear fitting,the MERSI ? top of atmospheric reflectance data were closer to the VIIRS data,and MAPE of each band was reduced to 2%?15%.These results indicate that the data quality of MERSI ? is comparable to the data quality of VIIRS,and can be used in Marine remote sensing applications.(2)Based on the quasi-real-time matching of 29-scene MERSI ? and VIIRS remote sensing image data,the MERSI ? top of atmospheric reflectance fitting model and R_rc inversion model for the oceanographic region were established by using the neural network method.The accuracy evaluation results show that the top of atmospheric reflectance fitting model has a good fitting results under different images,and the blue band R² can reach more than 0.95.The accuracy of the R_rc inversion model can reach more than 0.90.Through verification,it was found that the difference between MERSI ? R_rc retrievals by the R_rc inversion model and VIIRS R_rc was small,and the spatial distribution trend was consistent.(3)Using MERSI ? R_rc data,the inversion models of Chl-a and Kd?490 for the oceanic region were established.Through verification,it was found that the MERSI ? Chl-a products and VIIRS Chl-a products have a good consistency in spatial distribution,and the model accuracy was up to 0.90.The accuracy of Kd?490 inversion model was high,but its applicability needs to be improved.Through analysis,it was found that although the observation conditions or the sensor settings may have some impact on the accuracy of product inversion,the data of the two sensor are generally consistent,which preliminarily verifies the availability of MERSI ? data in the ocean.