High-Resolution Gridded Temperature and Salinity Fields From Argo Floats Based on a Spatiotemporal Four-Dimensional Multigrid Analysis Method

The shortage of in situ ocean observations greatly limits the development of gridded ocean subsurface environmental analysis products. Fully using the correlated information of the observations in time and space is an appropriate approach to this problem. Here we adopt a method of spatiotemporal four-dimensional multigrid analysis. A weekly gridded Argo product is also constructed in the Northwest Pacific Ocean (0 degrees-40 degrees N, 120 degrees-152 degrees E) from 2005 to 2020, with a horizontal resolution of 0.5 degrees x 0.5 degrees and 65 layers in the upper 1,500 m. Then, the gridded Argo product is evaluated by using in situ Argo floats, satellite altimeter data, the Tropical Atmosphere Ocean/Triangle Trans-Ocean Buoy Network (TAO/TRITON), high-resolution shipboard observations and some global gridded Argo products. Compared with in situ Argo floats, the reconstructed fields retain the spatiotemporal information of Argo floats well. The maximum vertical root mean square errors of temperature and salinity between the gridded Argo product and in situ Argo floats are less than 0.47 degrees C and 0.06 psu respectively. In addition, the product agrees well with independent observations of satellite altimeter data, TAO/TRITON. Compared with shipboard observations, the product can well reflect the structure of mesoscale eddies and detect some small-scale water masses. Compared with global gridded Argo products, our product retains more mesoscale features and has spatiotemporal correlation. Plain Language Summary Constructing subsurface environmental analysis products from in situ ocean observations is an important issue in oceanic research. Many previous works were conducted without considering the information of the time dimension. Moreover, less attention has been given to the reconstruction of extracting spatial-temporal information from observations in at multiple scales. In this paper, we focus on overcoming these problems and fully use spatial-temporal correlation in observations to construct a high-resolution gridded product.