High-Resolution Seamless Daily Sea Surface Temperature Based on Satellite Data Fusion and Machine Learning over Kuroshio Extension

Sea SurfaceTemperature (SST) is a critical parameter for monitoring the marine environment and understanding various ocean phenomena. While SST can be regularly retrieved from satellite data, it often suffers missing data due to reasons including cloud contamination. In this study, we proposed novel two-step fusion framework generating high-resolution seamless daily multi-satellite sources. The approach consists of (1) reconstruction based on Data Interpolate Convolutional AutoEncoder (DINCAE) using SSTs derived two sensors (i.e., Moderate Resolution Imaging Spectroradiometer (MODIS) Advanced Microwave Scanning Radiometer 2(AMSR2)), (2) improvement through random forest consistency with in situ measurements schemes scheme 1 reconstructed MODIS variables 2 both AMSR2 variables). was evaluated over Kuroshio Extension Northwest Pacific, where highly dynamic pattern found, 2015 2019. results showed that DINCAE yielded very good performance Root Mean Square Errors (RMSEs) 0.85 0.60 °C Absolute (MAEs) 0.59 0.45 °C, respectively. second step produced RMSEs 0.75 0.98 MAEs 0.53 0.68 respectively, compared measurements, which proved superiority Scheme also comparable or even better than operational products similar spatial resolution. particular, at simulating features fine resolution (~50 km). promising study area, producing high quality feature