Reconstruction Of Satellite-retrieved Sea Surface Temperature Based On Its Diurnal Variation

Sea Surface Temperature (SST) is a basic variable of ocean observations, which is critical for marine and atmospheric study. Nowadays, satellite remote sensing has become the major source of SST. However, a single satellite remote sensing is a snapshot of limited areas, which usually leads to inconsistency and gaps in the SST products. Therefore, a reconstruction of sea surface temperature based on the diurnal variation (DV) analysis was carried out in this paper.First, satellite remote sensing and numerical model are used to investigate the characteristics of diurnal variation in the study area (115-130°E,22-42°N). Considering the physical mechanism of DV (such as the role of air-sea heat exchange, mixing and diffusion), a one-dimensional mixed layer ocean model (General Ocean Turbulence model, GOTM) is selected to model the DV at different meteorological conditions. Then multi-source satellites SST are normalized to the same observed depth and local time.Secondly, based on the spatial coverage of the normalized SST, the study area is divided into the observation areas and the gaps regions. The multi-source SST is blended by Markov estimation in the observation areas. The gaps are filled by optimal interpolation (01). The spatial and temporal correlation model and background fields for hourly SST are constructed for OI. The main conclusions are as follows:(1) Based on the MODIS, AMSR-E and MTSAT observations, large diurnal warming (the maximum amplitude of up to ～5℃) are identified in the study area. It frequently occurs in spring and summer. SST DV is mainly controlled by wind and solar radiation. The lower wind speed and higher solar radiation usually leads to larger DV. In addition, the turbidity of the water would also promote the SST DV amplitude.(2) The DV observed by MTSAT are used to verify the capacity of DV modeling by two empirical models (CG03 and ASM) and a physical model (GOTM). It is found that GOTM shows better accuracy with mean bias～ -0.01℃ and RMSE 0.3-0.5℃ in simulating the DV process.(3) The normalization of multi-source SST can also effectively improved the spatial coverage up to about 75% by filling orbital gaps and overcoming the obstacles of clouds, rain and other adverse weather conditions on satellite remote sensing.(4) The SST spatial and temporal correlation model is constructed. The correlation time length scales is determined as ±3 hours for the analysis time within ±2 days and space length scales is determined as ～85 km on the zonal direction and～ 100km on the longitude direction. Finally the hourly blended SST in 2007 is validated against drifting buoy SST. The mean bias is about -0.14℃ and RMSE is about 0.57℃, this accuracy is comparable to the satellite observations.The main innovations of this paper are as follows:(1) The characteristic of SST DV in the study area is derived from the satellite remote sensing, and then GOTM, which based on the heat exchange and momentum process, is selected to model SST diurnal signal. Multi-tempolar satellite remote sensing SST are then normalized to the same local time by GOTM.(2) Markov estimation is used to blended multi-tempolar satellite SST in the observation covered area. It shows a better accuracy than any observations or least square method. Then the residual gaps are filled by OI with much more elaborate background and spatial and temporal correlation model.(3)Base on point (1) and (2), the time-space consistent with gap free series of blended SST can be generated multi-source of satellite remote sensing SST. This method provides mportant technical support to reconstruction of time-space gap free fileds from other multi-temporal and sources remote sensing products reconstruction.