Assimilative Mapping : TS18 Model Error Covariance

Specification of the ionospheric convection electric field over the entire high-latitude region using only measurements is not possible. Instead, a fitting or assimilative technique can be used to combine information from sparse measurements and a background model. The Super Dual Auroral Radar Network (SuperDARN) Assimilative Mapping technique (SAM) is one such technique [Cousins et al., 2013a]. SAM combines the available line-of-sight velocity measurements from SuperDARN with a statistical or climatological model of the high-latitude convection and produces a global solution of the electrostatic potential for a given period of interest. The method of combination used in SAM is known as optimal interpolation, achieved by weighting the model and observations by their respective errors. Statistical background model errors are normally derived from a large dataset and provided in the form of a covariance matrix. While the SAM technique currently uses a relatively recent climatological model derived using measurements from high-latitude SuperDARN radars, referred to as the CS10 model [Cousins and Shepherd, 2010], it does not use the latest model, referred to as the TS18 model [Thomas and Shepherd, 2018], which uses data also from radars located at middle and polar latitudes. In order to obtain the error covariance matrix for the TS18 model we have characterized the dominant modes of variability as a set of empirical orthogonal functions (EOFs) by fitting a basis functions to residual (error) vectors between a large set of SuperDARN observations and corresponding model-produced vectors. We compare the resulting EOFs to those obtained from the CS10 model and investigate differences obtained from the SAM procedure for selected periods