Observation impacts on the Mid-Atlantic Bight front and cross-shelf transport in 4D-Var ocean state estimates: Part I — Multiplatform analysis

A nested configuration of the Regional Ocean Modeling System (ROMS) comprising three grids was used in conjunction with a 4-dimensional variational (4D-Var) data assimilation system to compute ocean state estimates of the Mid-Atlantic Bight (MAB). The three nested grids have a horizontal resolution ranging from ∼7 km to ∼0.8 km and capture circulation regimes that span the Gulf Stream western boundary current, through the mesoscale eddy field, and down to the rapidly evolving and energetic sub-mesoscale. All of these circulation regimes are challenging for any data assimilation system, yet the 4D-Var system was found to perform well across this range of space- and time-scales. The observational data used to constrain the ocean state estimates comes from a wide range of remote sensing, in situ, and mobile platforms. An adjoint-based procedure was used to compute the impact of each observing platform on several different indexes that describe the position of the MAB front, stratification, and associated cross-shelf exchange processes in the vicinity of the U.S. National Science Foundation’s Ocean Observatories Initiative Pioneer Array. The impact of observations from each observing platform on the chosen indexes varies across the three grids. It is a function of several factors that include the nature of the background circulation and the level of error assumed for the background ocean state and the observations. The geographic distribution of the observation impacts is remarkably robust across the various indexes and the three grids. In addition, observations that are both local to and remote from the target regions that define each index can exert a significant influence on the circulation. Variations in the observation impacts through time can be used to identify observations that exert unexpectedly large influence on the 4D-Var analyses (i.e, outliers), and routine monitoring of observation impacts is a useful indicator of the efficacy of different components of the observing system. Also, the observation impacts were found to be a useful performance indicator for the data assimilation system.