A data-assimilative model reanalysis of the U.S. Mid Atlantic Bight and Gulf of Maine: Configuration and comparison to observations and global ocean models

•4D-Var data assimilation produces skillful 15-year coastal circulation analysis.•Climatological data assimilation corrects open boundary data biases.•Downscaling with data assimilation improves on global models in the coastal ocean.•Coastal satellite altimetry improves sea level variability across all time scales.•Bottom temperatures that impact fisheries are modeled well. A 15-year reanalysis (2007–2021) of circulation in the coastal ocean and adjacent deep sea of the northeast U.S. continental shelf is described. The analysis uses the Regional Ocean Modeling System (ROMS) and four-dimensional variational (4D-Var) data assimilation (DA) of observations from in situ platforms, coastal radars, and satellites. The reanalysis downscales open boundary information from the Copernicus Marine Environmental Monitoring Service (CMEMS) global analysis. The dynamic model is forced by regional meteorological analyses, observed daily river discharges, and harmonic tides that augment the open boundary conditions. A complementary analysis of the mean seasonal cycle of regional circulation, also computed using ROMS 4D-Var but with climatological mean observations and forcing, is used to reduce biases in the CMEMS boundary data and to provide a dynamically and kinematically constrained Mean Dynamic Topography to use in conjunction with the assimilation of satellite altimeter sea level anomaly observations. The configuration of ROMS 4D-Var used is described, presenting details of the comprehensive suite of observations assembled, data pre-processing and quality control procedures, and background and observation error hypotheses. Control variables of the DA are the initial conditions, surface forcing, and boundary conditions of a sequence of non-overlapping 3-day analysis cycles. Comparisons to a non-assimilative version of the same ROMS model configuration show the added skill brought by assimilation of local observations. The improvement that downscaling with assimilation achieves over ocean state estimates from CMEMS and the U.S. Naval Research Laboratory Global Ocean Forecast System (GOFS) is demonstrated by the reduction in residuals of the DA, and by comparison to independent (unassimilated) observations. Wherever data volumes allow, skill assessments are made with the respect to anomalies from the mean seasonal cycle to emphasize performance at the ocean mesoscale. To highlight the utility of the analysis to inform studies related to coastal sea level variability and