Document Type


Publication Date



Marine Science


Ocean Science and Engineering


Most ocean data assimilation systems are tuned to process and assimilate observations to constrain features on the order of the mesoscale and larger. Typically this involves removal of observations or computing averaged observations. This procedure, while necessary, eliminates many observations from the analysis step and can reduce the overall effectiveness of a particular observing platform. Simply including these observations is not an option as doing so can produce an overdetermined, ill-conditioned problem that is more difficult to solve. An approach, presented here, aims to avoid such issues while at the same time increasing the number of observations within the assimilation. A two-step assimilation procedure with the four-dimensional variational data assimilation (4DVAR) system is adopted. The first step attempts to constrain the large-scale features by assimilating a set of super observations with appropriate background error correlation scales and error variances. The second step then attempts to correct smaller-scale features by assimilating the full observation set with shorter background error correlation scales and appropriate error variances; here the background state is taken as the analysis from the first step. Results using a real high-density observation set from underwater gliders in the region southeast of Iceland, collected during the 2017 Nordic Recognized Environmental Picture (NREP) experiment, will be shown using the Navy Coastal Ocean Model 4DVAR (NCOM-4DVAR).


© 2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license

Publication Title

Monthly Weather Review





Find in your library

Included in

Oceanography Commons