An Improvement on Assimilation of Satellite-derived Sea Surface Temperature for NEMOVAR with the use of MODIS data

Abstract

SST is important in coupling the ocean and atmosphere through exchanges of heat, momentum, moisture, and gases. SST data plays a major role as a bottom boundary condition in numerical weather prediction (NWP) systems. Although in-situ measurements are considered to be true values, they have temporal and spatial constraints. A crucial issue is how to implement satellite-derived SST data as boundary conditions for ocean models as well as NWP models. The operational administrations and organizations have conducted the system for real time global SST products. But, the retrieving high quality and operational ocean initial SST data is still challenging subject. The objective of this thesis is to implement satellite-derived SST measurements into an ocean data assimilation system to improve the ocean-state forecasts. The Global Ocean Data Assimilation Experiment High-Resolution SST Pilot Project (GHRSST-PP) is a well-organized system providing the near-operational high-resolution satellite SST GHRSST Level 2 Pre-processed data used in this thesis. Advanced Along-track Scanning Radiometer, Advanced Microwave Scanning Radiometer-EOS, Advanced High-Resolution Radiometer, and Moderate Resolution Imaging Spectrometer (MODIS) data are collected and assimilated on the NEMOVAR ocean data assimilation system. The selected datasets have different characteristics and can yield a good effect when used together. According to the hindcast results, most SST satellite data are well processed. Their RMSEs are less than 0.76°C globally in all seasons (summer, autumn, and winter). However, the RMSEs for SST data from all instruments other than MODIS are approximately 0.5°C globally, even in the summer. Improving the impact of MODIS on the global dataset will require additional statistics or a quality control (QC) method. A multiple assimilation method based on a variational quality control (VarQC) is proposed. The impacts of additional QC predict that RMSEs will be reduced to less than 40% (average, maximum is 57.45% in Agulhas Return region in June). This type of approach has the advantage of using the scheme itself for quality control. This method can advance the application of satellite data to assimilation systems. Incidentally, sensitivities to spatial density of GHRSST are acceptable in that the resolution of the satellite-derived SST is close to the model grid rather than denser than the model grid.