Research On Sea Surface Salinity Fusion And 3D Dynamic Field Reconstruction Method And Their Application

It's an important task of data assimilation and reconstruction to take advantage of high-resolution satellite date to make up for the lack of in-situ observations and to produce 3D or even 4D element fields.Due to the noises in sea surface salinity(SSS)data and the shortness in the surface-subsurface projection methods,there remains problems in the assimilation and noise elimination of SSS.This article focuses on the problem of surface sea salinity fusion and assimilation from both technical and theoretical aspects.Technically,the noise filtering and blending algorithm of surface salinity data and the technical flow to diagnose the results of satellite data retrieval and assimilation are studied.Theoretically,the dynamic theory of surface-subsurface projection and the effects of SSS noises on reconstruction results are investigated.The main conclusions are as follows:(1)Non-uniform weight function and anisotropic weight function are raised to improve the original multi-fractal fusion algorithm,of which the parameters are determined by the dynamic-related information such as the deformation radius and surface geostrophic flow.And SST,SSH and climatology data are made full use of to synergize the filtered SSS product.Compared to the unfiltered SSS,the fused SSS has clearer fractal structures,lower meso-scale and micro-scale noise,less errors and higher correlation coefficients.Non-uniform weight function scheme and anisotropic weight function scheme perform better in RMSE and correlation coefficient compared with the original scheme,indicating the necessity of taking the heterogeneity and anisotropy of ocean fields into consideration when assimilating or fusing ocean data.Besides,except for the strong current,the anisotropic scheme is quite similar to the non-uniform scheme,while the latter takes effect in global range.(2)This article comes up with a set of quantitative indexes to describe the vertical modes,reconstruction results and the factors that affects them.Based on the reanalysis data,this article analyses the structure of surface-subsurface vertical projection and its possible influence on satellite data assimilation in different dynamic environments.Surface density signal and height signal penetrates into subsurface ocean through SQG mode and the barotropic and baroclinic modes of interior solution,respectively.The former is related to the strength of surface buoyancy signal and the depth and intensity of thermocline,while the latter is also confined by the thermocline.Surface geostrophic deviation index(DIsur)is the principal index to evaluate the result of vertical mode projection and isQG retrieval;In the areas with large DIsur,vertical projection is untrue and the retrieval result is unsatisfactory.If DIsur is rather small but SQG solution is strong(strong surface buoyancy signal and weak thermocline),the vertical projection of surface buoyancy signal makes up a quite large part of the subsurface signal,which accounts for the small retrieval bias even if there exist subsurface intensified eddies causing structural deviations;In these areas,assimilation of SSS or SST is enough to obtain good simulation of eddies.For those areas with small SQG solution but strong isQG solution(weak surface buoyancy signal and weak thermocline),if's rather hard for the weak surface buoyancy signal to affect subsurface ocean,while SSH signal can make amends for the defect of SQG mode through interior modes;In these areas,the retrieval error is small except for the depth where there is subsurface intensified structures,and the reconstruction result of SST and SSS is bad unless SSH is also assimilated.For the areas with small DIsur,weak SQG and isQG solution,neither surface buoyancy signal or SSH signal can penetrate the thermocline effectively and retrieval result is mainly decided by the strength of the subsurface intensified eddies.(3)Based on the SSS fused by different weight function schemes and combined with other observations,this article uses isQG method to investigate the impact of the error of SSS on subsurface dynamic retrieval.The noise contained in SSS leads to the false pattern of SSD and is further delivered to subsurface ocean by vertical mode,bringing on large deviations of density retrieval upon 250 m depth;Instead,the blended SSS can filter the surface noise and deduce retrieval deviation in the upper ocean,which indicates the essential of filtration and fusion in SSS assimilation.The density retrieved by different fused SSS is merely distinguished in magnitude but differences still exist in structure;while the velocity retrievals are different both in magnitude or direction.Generally speaking,the non-uniform scheme and anisotropic scheme embodies a better performance although the improvement is not prominent in the time-independent retrieval experiment.