Study on Bayesian hierarchal model-based SST data fusion methods

Sea surface temperature (SST) is one of the most important variables related to the global ocean-atmosphere system, which play an important role in studies of air-sea heat exchange, upper ocean processes, and weather forecast. SST data are routinely measured from ships, buoys and offshore platforms. In this paper, the weekly 4 km resolution AVHRR SST data (1985-2006), the weekly 4 km resolution MODIS SST data (2002-2007) and the daily 25 km resolution AMSR-E SST data (2002-2007) are chosen for merging. These SST data are derived from different Remote Sensors with different spatial and temporal resolution. By merging these SST data, we can get a new SST product and obtain more information. The bayesian hierarchical model using Markov Chain Monte Carlo (MCMC) simulation methods was used to merging the thermal infrared MODIS SST data and passive microwave AMSR-E SST data. The results show that merged SST data have a better completeness than MODIS SST and AMSR-E SST products. Comparing merged SST data with drift buoy SST, the validation result shows that the bias is 0.32118K and RMSE is 0.8026K.