Representation Error of Oceanic Observations for Data Assimilation

Abstract

A simple approach to the estimation of representation error (RE) of sea level (?), temperature (T), and salinity (S) observations for ocean data assimilation is described. It is assumed that the main source of RE is due to unresolved processes and scales in the model. Therefore, RE is calculated as a function of model resolution. The methods described here exploit the availability of mapped sea level anomalies (mSLAs) and along-track sea level anomalies (atSLAs). The mSLA fields or atSLA observations are regarded as the true ocean state. Here, they are averaged according to the resolution of the model grid, and the averaged field is taken as a representation of the true state on the given grid. The difference between the original data and the averaged field is then regarded as the RE for ?. Subsequently, the RE is projected for ? over depth using a standard technique, giving an estimate of the RE for T and S. Examples of RE estimates for an intermediate- and high-resolution global grid are presented. It is found that there is significant spatial variability in the RE for ?, T, and S, with values that are typically greater than or comparable to measurement error, particularly in regions of strong mesoscale variability.