Currently, forecasts produced by the Oregon-Washington (OR-WA) Coastal Ocean Forecast System are constrained by assimilation of only surface observations. The 4-dimensional variational (4DVAR) data assimilation (DA) algorithm is utilized to combine the model and the data, with the time-independent forecast ("background'') error covariance B. In this study, two possible improvements are explored: (i) assimilation of subsurface observations from autonomous underwater gliders, using the present 4DVAR DA system, and (ii) estimation of B from an ensemble of 4DVAR runs (En4DVar).
We find that assimilation of glider observations alone creates erroneous energetic eddies in the vicinity of the glider transect. Assimilation of surface and subsurface observations in combination prevents these features from forming and reduces the forecast error.
The ocean state in the OR-WA region is highly variable in space and time. Therefore, it is unlikely that the current DA system, using a static B with a horizontally uniform covariance, can fully detect and correct the background errors. The En4DVar necessitated the development of: (a) the new, computationally-efficient Monte-Carlo localization method; (b) a Bayesian Hierarchical Model to obtain realistic estimates for the ensemble wind perturbations from scatterometer observations; and (c) the cluster search method which can cut the wall time needed to calculate the DA correction by 30% compared to conventional 4DVAR.
Compared to the current static B, the ensemble B exhibits stronger covariances in frontal areas that change their location from one assimilation window to another. The En4DVar yields a solution with improved temperature and salinity properties on the shelf compared to 4DVAR with the static B. However, assimilation of biased SST data amplified by the strong ensemble covariance in the area of the Columbia River plume yields strong unphysical changes in the surface salinity, which is not constrained by observations. 4DVAR was modified to reduce this effect.
Funding Statement (additional comments about funding)
National Oceanic and Atmospheric Administration (NOAA) Coastal Ocean Modeling Testbed (COMT) grant NA13NOS0120139, the NOAA Quantitative Observing System Assessment Program (QOSAP), National Science Foundation (NSF) grants OCE-0527168 and OCE-0961999, Integrated Ocean Observing System / Northwest Association of Networked Ocean Observing Systems (IOOS/NANOOS) grant NA16NOS0120019, the National Aeronautics and Space Administration (NASA) SWOT Science Definition Team project grant NNX13AD89G. This work used the Extreme Science and Engineering Discovery Environment (XSEDE) under allocation TG-OCE160001, which is supported by NSF grant number ACI-1548562.