文章基本信息

标题：On determining initial conditions and parameters in a simple coupled atmosphere-ocean model by adjoint data assimilation
本地全文：下载
作者：Jingxi Lu ; William W. Hsieh
期刊名称：Tellus A: Dynamic Meteorology and Oceanography
电子版ISSN：1600-0870
出版年度：1998
卷号：50
期号：4
页码：534-544
DOI：10.3402/tellusa.v50i4.14531
摘要：The general problem of retrieving both initial conditions and model parameters for coupledatmosphere-ocean models by the adjoint data assimilation method was formulated. In particular,for a simple coupled equatorial model, where the atmosphere and the ocean were eachrepresented by a linear shallow water model, retrieval of 3 oceanic initial conditions (the sealevel height (SLH) and the 2 horizontal current components) together with 6 (damping andcoupling) parameters was performed. Identical twin experiments assimilating wind and SLHdata were conducted to test (i) the mutual influence of initialization and parameter estimation,(ii) the information transfer efficiency between the atmosphere and the ocean during retrieval,and (iii) the effect of initial guess on retrieving the initial conditions and parameters. By assimilatingthe wind and SLH data once per day at every grid point, retrieving both the parametersand initial conditions together was found to be more difficult than retrieving either of themseparately. Assimilating the same data once per day in the TAO array yielded a much poorerbut still acceptable retrieval, as the amount of information might be insufficient. The wind andSLH data were comparable in importance when retrieving the initial conditions, but not sowhen retrieving parameters, the wind data being more helpful for retrieving the atmosphericdamping coefficients, and the SLH data for the oceanic damping coefficients. Errors in theinitially guessed values of the parameters or the initial conditions generally affected the retrievalquality. The initial condition retrieval was more sensitive to errors in the guessed initial conditions,especially to phase errors, while the parameter estimation was more sensitive to errorsin the guessed parameters.