Search for a command to run...
Abstract The most important source of information constraining the Navy’s operational global ocean forecasting system is sea surface height anomaly (SSHA) as measured by satellite altimetry. These surface observations inform a one-dimensional (1D) variational analysis to create synthetic profiles of temperature and salinity, which approximate the subsurface ocean structure associated with the observed SSHA that is assimilated in a three-dimensional variational analysis. The 1D analysis requires vertical error covariances that relate the differences in values between temperature and salinity at different depths. These vertical covariances are computed empirically from historical in situ observation profiles of temperature and salinity. The approach ensures that the assimilated profiles have realistic structure without drifting. A shortcoming of this approach is the availability of in situ observations extending at least 1000 m deep. Observations are sparser at high latitudes, often do not include salinity, and reach relatively shallow depths. We wish to use model data to address these limitations. Here we show that using a global 30-year model run to compute vertical covariances solves sampling issues while continuing to maintain accuracy. While the covariances derived from the model generally compare well with the observed ones, in some areas of the ocean, the numerical ocean model has different vertical covariances. A new method for determining where synthetics are most valuable is presented. The implication of having model derived covariances is the ability to extend covariance information at high latitude where in situ observations are sparce or have sampling anomalies. Results also suggest that salinity, if observed, would provide substantial improvement to the system.