An offshore subsurface thermal structure inversion method by coupling ensemble learning and tide model for the South Yellow Sea

Abstract

The South Yellow Sea Cold Water Mass (SYSCWM), which occurs in the South Yellow Sea (SYS) during summer, significantly impacts the hydrological characteristics and marine ecosystems but lacks fine interior data. With satellite observations, significant achievements have been made in reconstructing high-resolution ocean subsurface thermohaline structure based on machine learning. However, the accuracy of offshore subsurface parameter estimation will be affected due to the macro-tidal environment and fewer in situ observations. In this paper, we coupled the TPXO tide model and Light Gradient Boosting Machine algorithm to develop an inversion model of offshore subsurface thermal structure for the SYS using sea surface data and in situ observations. After light modelling, the subsurface temperature structure in the SYS is retrieved from sea surface parameters with a spatial resolution of 0.25° at depths of 0-55 m. Observation-based dataset (ARMOR3D) and in situ observations are used for model evaluation. According to the validation of the mooring buoy observations, the overall coefficient of determination (R2), which determines the percentage of variance in the dependent variable that can be explained by the independent variable, is more than 0.95. Furthermore, the R2 is improved by 12% due to coupling tide model below the thermocline during the maturity stage of SYSCWM, which is helpful for a better reconstruction of SYSCWM. Comparing with the cruise data, the average R2 of the proposed model is 0.927 which is slightly better than the accuracy of the observation-based ARMOR3D dataset. Since the R2 exceeds 0.8 in the most area of 121°E~123.5°E, 33°N~36°N, the reconstruction is reliable in this area. The method provides a new explorable direction for reconstructing the ocean thermal structure in offshore areas.