베이지안 모형평균화를 이용한 멀티센서 해수면 온도 앙상블 산출

Abstract

Sea surface temperature(SST) is used with important parameter at climate system. So measuring accurate SST is demanded. But SST products derived from satellite-borne instruments have some different value because of different of retrieval algorithm and sensors. These products are retrieved by algorithm so they have uncertainty. Also, each product has different size of uncertainty. To reduce the uncertainty, data assimilation based on ensemble is needed. Bayesian model averaging is weighted averaging method using posterior probability as the weight. Using training data, the posterior probability distribution of each data are calculated and the mean and the variance are estimated by Expectation-Maximization(EM) algorithm. The estimated mean is used posterior probability as weight. The monthly SST products of the Moderate Resolution Imaging Spectroradiometer(MODIS) sensor and Advanced Microwave Scanning Radiometer for EOS(AMSR-E) sensor of Aqua and Advanced Very High Resolution Radiometer (AVHRR) sensor of National Oceanic and Atmospheric Administration(NOAA) are used as the ensemble member products. Advanced Along-Track Scanning Radiometer(AATSR) product of Environmental satellite(Envisat) is used for training data and validation. To create the ensemble monthly averaging data, AATSR monthly data is preferentially made. Using the BMA and EM algorithm that calculate the weight through repeat process and likelihood function, the weight and weighted averaging were created. To validate about the accuracy of the BMA ensemble, reducing RMSE than the existing mean and median method is confirmed using one-leave-out-cross validation method.