Assimilation of Diurnal Satellite Retrieval of Sea Surface Temperature with Convolutional Neural Network

The Sea Surface Temperature (SST) is one of the most examined and monitored ocean feature within the satellite era. A variety of SST datasets have been produced in recent years, each one of them with almost unique characteristics that differ from the SST variable of ocean general circulation models (OGCM), so that assimilating such datasets in an optimal way requires a mapping to the first model level. However, this projection is non-trivial and depends on the specific characteristics of the dataset. In this work we employ a data-driven approach to construct the projection operator with machine learning (ML). The benefit is that it can be used over different SST dataset by only re-training the network. We consider different convolutional neural networks, namely a U-Net, a pix2pix, and a random forest model for comparison. We train the ML models with L3 global diurnal subskin SST derived from AVHRR’s infrared channels on MetOp satellites to reproduce the ESA SST CCI taken as ground truth. The pix2pix is the most effective operator and we use it to produce several global one-year-long reanalysis-like experiments at 1/4° that assimilate the SST in different ways, e.g. direct satellite observations, unbiased approach, as observation operator. We find that the ML-unbiased approach improves the RMSE up to 10% w.r.t. direct assimilation in the global water column, while the observation operator is beneficial mainly in the tropics but degrades in the areas of strong mesoscale activities that probably requires higher resolution or longer training.