Merging Sea Surface Temperature Observed By Satellite Infrared And Microwave Radiometers Using Kalman Filter

Sea Surface Temperature (SST) is one of important parameters in the global ocean-atmosphere system. The study of marine environment, global climate changes, and the disaster prevention and reduction, etc. is very important. Currently, there are a number of satellites on orbit to provide multi-sensor sea surface temperature observations, and the infrared channel of polar-orbiting satellites radiometer with high spatial resolution characteristics to measure SST, it of geosynchronous satellites with high time resolution to measure SST. However, the infrared sensors can not penetrate clouds, could not provide SST observations when there exist clouds. The microwave radiometers can penetrate clouds, but its spatial resolution much lower. As the limitations of single satellite sensor, to merge the infrared SST data and microwave SST data is meaningful. The Kalman filter is used to merge multi-sensor SST data in this paper in order to generate all-weather and high spatial resolution SST data.Fist of all, AVHRR and AMSR-E SST data during the period of 2008 are compared with in situ SST data respectively. While the AVHRR data come from the Ocean University of China satellite ground station, and the AMSR-E SST(Advanced Microwave Scanning Radiometer-Earth Observing System, AMSR-E)data from Remote Sensing Systems, and the buoy data from the NEAR-GOOS Delayed Mode Database. To select the Northwest Pacific region as the study area in the paper: 10N-50N,105E-145E. The mean bias between in situ data and AVHRR SST data in daytime is 0.23℃, and in -0.01℃in nighttime; The standard deviation of them is 0.93℃in daytime and 0.87℃in nighttime. The mean bias between in situ data and AMSR-E SST data in daytime is 0.11℃, and in 0.06℃in nighttime; The standard deviation of them is 0.72℃in daytime and 0.79℃in nighttime. The deviations between AVHRR data and measured data are mainly due to the missed detected clouds in the AVHRR data. Then we used the Kalman filter to merge the AVHRR SST and the AMSR-E SST. The space resolution of merged SST data is about 2 km, and the time resolution is 24 hours. The previous day's SST data as the initial value of filter when carrying out Kalman filter to merge the SSTs. At last, to compare the merged SST data and in situ SST data, the mean bias of merged SST data is 0.18℃, and the standard deviation of it is 1.05℃. The spatial coverage from 31% of the AVHRR and 49% of the AMSR-E increased to 97%. We also analyzed the variance, entropy and definition of the three datasets to evaluate the quality. The variance of AVHRR SST,AMSR-E SST and merged SST in the March of 2008 were 4.67℃,4.57℃,4.67℃, and the definition of them were 0.36℃,0.05℃,0.31℃, and the entropy were 3.30℃,4.17℃,4.39℃. The results show that in the case of the merged data and raw data with the same deviation from the mean, the merged data kept most details of AVHRR data, and carried the most information. This research showed that the Kalman filter to merge the satellite SSTs improves the spatial and temporal resolution of SST data, keeps the more details and the accuracy deviate from original SST data less.