Regional Ocean Data Assimilation

Abstract

This article reviews the past 15 years of developments in regional ocean data assimilation. A variety of scientific, management, and safety-related objectives motivate marine scientists to characterize many ocean environments, including coastal regions. As in weather prediction, the accurate representation of physical, chemical, and/or biological properties in the ocean is challenging. Models and observations alone provide imperfect representations of the ocean state, but together they can offer improved estimates. Variational and sequential methods are among the most widely used in regional ocean systems, and there have been exciting recent advances in ensemble and four-dimensional variational approaches. These techniques are increasingly being tested and adapted for biogeochemical applications.

On the Benefit of Current and Future ALPS Data for Improving Arctic Coupled Ocean-Sea Ice State Estimation

“…One powerful way to integrate these vastly heterogeneous (in space and time) streams of in situ and remote-sensing observations is to produce an Arctic ocean-sea ice synthesis using existing ocean-ice data assimilation or state estimation frameworks (Wunsch and Heimbach, 2007). Edwards et al (2015) and Stammer et al (2016) provide lists of regional and global examples of ocean data assimilation, and discuss the advantages and disadvantages of filter-versus smoother-based assimilation frameworks, with a key difference being that in the smoother framework, the equations of motion, for example, conservation laws, are strictly obeyed (i.e., no mass, momentum, or heat can be artificially created or destroyed). Data assimilation focusing on sea ice in polar regions is covered at length in Carrieres et al (in press).…”

Modeling and Forecasting Ocean Acoustic Conditions

“…The reasons include incomplete dynamics in models (such as the absence of explicit, double-diffusive influences on intrusions known to affect sound (Colosi et al 2012a;Duda and Sellers 2016)), insufficient resolution of the dynamical models, and insufficient data to constrain data-driven models in the absence of special regional focus. Data-constrained modeling is now routine (Edwards et al 2015). However, most models barely resolve, incorrectly model, or exclude many features that have been shown to strongly influence sound propagation (see next section).…”

“…Sound behavior in 4D-variable mesoscale ocean structures can be modeled with datadriven methods given adequate data (Lermusiaux et al 2010;Edwards et al 2015), but gravity-wave features that affect sound are more difficult to capture accurately with dynamical models. It is safe to say that all sound is influenced by internal waves, just as all undersea sound is influenced by local mesoscale conditions.…”

“…Hence, 4D-variable environmental input fields, if available, could be exploited in acoustic models. Because the contemporary ocean modeling and prediction capability is data-starved, as are most oceanographic endeavors, leveraging data with sparse sampling in space and time by infilling with data-constrained (data-assimilating (DA)) dynamical modeling (or state-space estimation) is a reasonable approach for efficiently building reliable 4D fields (Robinson et al 1998;Edwards et al 2015). Although this method is progressing for features that are resolved by the models, it is not yet possible for small-scale ocean features of first-order acoustic importance.…”

Validation of genetic algorithm-based optimal sampling for ocean data assimilation

“…In order to maintain forecast capability by accounting for errors in the initial and boundary conditions, as well as errors within the scales not resolved by the computational model, a continuous set of measurements must be taken. The technique of measuring the ocean and integrating these measurements is termed dataassimilation and has been shown to be successful (e.g., Robinson et al, 1998;Lermusiaux et al, 2006b;Edwards, al. 2015).…”