Real-time Probabilistic Coupled Ocean Physics-AcousticsForecasting and Data Assimilation for Underwater GPS

C. Mirabitoa, P. F. J. Lermusiauxa∗, P. J. Haley, Jr.a, W. H. Alia,E. Dorfmanb, A. Laferriereb, A. Koffordb, G. Shepardb, M. Goldsmithb,

K. Heaneyc, E. Coelhoc, J. Boylec, J. Murrayc, L. Freitagd, A. Morozovd

a Massachusetts Institute of Technology; b Raytheon BBN Technologies; c OASIS, Inc.;d Woods Hole Oceanographic Institute; ∗ Corresponding Author: pierrel@mit.edu

Introduction. The POSYDON program aims to develop a Global Positioning System (GPS) for under-water assets. The primary goals of our MIT effort are to: (1) Employ and develop our regional oceanmodeling, data assimilation, and uncertainty quantification for the estimation of sound speed variability,coupled oceanographic-acoustic forecasting and inversion relevant to the POINT effort; (2) Apply our the-ory and schemes for optimal placement, path planning, and persistent ocean sampling with varied assets andacoustic source platforms; and (3) Further quantify the ocean dynamics and variability of the regional areasof interest, utilizing our multi-resolution data-assimilative ocean modeling and process studies.

As part of this program, we completed regional ocean modeling and forecasting for the Middle AtlanticBight. Our realistic data-assimilative modeling involved real-time forecasting and data-driven simulationsand analyses of the sound speed variability. To do so, we built on our experiences, especially on large andcollaborative research initiatives. Our methods and software were used and further developed for POINT.

For this project, we also characterized and forecasted the oceanographic variability and uncertainty.Our MIT-MSEAS PE model of the temporal and spatial evolution of physical features and circulationshas been validated through extensive measurements, and analysis in many regions. However, due to theuncertain initial and boundary conditions, and sub-grid-scale parameters, the variability of the environmentalpropagating medium is uncertain. Just as we now utilize probabilities for rain or bad weather on a dailybasis, the proposed underwater communication and global positioning system for deep ocean navigation canalso utilize and benefit from such information. Real-time integrated oceanographic-acoustic predictions mustaccount for and forecast these uncertainties and their effects on sound propagation and communications.

For this “Precision Ocean Interrogation, Navigation, and Timing (POINT)” effort, we utilized our MITMultidisciplinary Simulation, Estimation, and Assimilation System [MSEAS; 5, 6, 4]. The MSEAS softwareis used for fundamental research and for realistic simulations and predictions in varied regions of the world’socean, including monitoring, ecosystem prediction and environmental managementand, importantly for thepresent project, real-time oceanographic-acoustic predictions and coupled ocean-acoustic data assimilation.For this exercise, we mainly employed our MIT-MSEAS hydrostatic PE code with a nonlinear free surface,based on second-order structured finite volumes.

Real-time exercise. The POSYDON Sea Exercise 2018 occurred in the Middle Atlantic–New York BightRegion for August 2018. In collaboration with the POINT team, our objectives were to utilize the MITMSEAS system to: (i) forecast the probability of high-resolution ocean fields using our multiscale ErrorSubspace Statistical Estimation (ESSE) methodology; (ii) transfer the corresponding distribution of thesound speed field to three-dimensional underwater sound propagation uncertainties; (iii) collect sufficientdata to evaluate the accuracy of the Bayesian tomographic inversion and of its posterior estimates of rangebetween transducers and sound velocity profiles (SVPs).

The MSEAS modeling system was set up in an implicit two-way nesting configuration. The oceanforecasts were initialized from 1/12◦ HYCOM analyses [1], downscaled to higher resolution and updatedwith ocean data from varied open sources of opportunity (CTDs, ARGO floats, gliders, SST, etc.) and withour MSEAS feature models for additional corrections. These ocean simulations are forced by atmosphericflux fields forecast by the GFS 1/4◦ model [7] from the National Centers for Environmental Prediction(NCEP) and tidal forcing from TPXO8 [2, 3], but updated for the high-resolution bathymetry, coastlinesand for our quadratic bottom drag. For uncertainty predictions, ensemble forecasts were initialized using ourESSE procedures. For these ensemble forecasts 48 closely related tidal estimates were used for tidal forcingsand small perturbations were applied to the atmospheric forcings.

1

Mirabito, C., P.F.J. Lermusiaux, P.J. Haley, Jr., W.H. Ali, E. Dorfman, A. Laferriere, A. Kofford, G. Shepard, M. Goldsmith, K. Heaney, E. Coelho, J. Boyle, J. Murray, L. Freitag, and A. Morozov, 2020. Real-time Probabilistic Coupled Ocean Physics-Acoustics Forecasting and Data Assimilation for Underwater GPS. In: OCEANS '20 IEEE/MTS, 5-30 October 2020, sub-judice.

Results. Daily deterministic and ensemble forecasts of 3-day duration were issued for the period 9–24 Au-gust 2018. Three regions of distinct water masses (on the continental shelf, over the continental slope andin the Sargasso Sea) were identified and the updates to HYCOM were made separately for each region, thenmelded across the shelfbreak front and the Gulf Stream, respectively. These corrected fields provided thesubtidal initial and boundary conditions for the deterministic forecasts. Similarly, the ensemble perturba-tions were applied to the deterministic ICs in each region and melded across the respective fronts. The dailyensemble forecasts contained between 100–300 members. Both deterministic and ensemble forecasts wereverified by independent data of opportunity, and both showed skill out to 3 days, as shown in fig. 1.

(a) 2 m vorticity (b) Deterministic verification (c) 100 m sound speed uncer-tainty

(d) Ensemble verification

Figure 1: MSEAS-PE fields and verification: (a) 2 m vorticity, (b) deterministic verification against CTDdata, (c) 100 m standard deviation in sound speed, and (d) verification of ensemble against CTD data (notethat realizations encompass data and standard deviation of the ensemble is a reasonable proxy for RMSE).

References

[1] James A. Cummings and Ole Martin Smedstad. Variational data assimilation for the global ocean. In Seon KiPark and Liang Xu, editors, Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications, volume II,pages 303–343. Springer Berlin Heidelberg, Berlin, Heidelberg, 2013.

[2] Gary D. Egbert and Svetlana Y. Erofeeva. Efficient inverse modeling of barotropic ocean tides. Journal ofAtmospheric and Oceanic Technology, 19(2):183–204, 2002.

[3] Gary D. Egbert and Svetlana Y. Erofeeva. TPXO8-ATLAS. OSU, 2013.

[4] P. J. Haley, Jr., A. Agarwal, and P. F. J. Lermusiaux. Optimizing velocities and transports for complex coastalregions and archipelagos. Ocean Modeling, 89:1–28, 2015.

[5] Patrick J. Haley, Jr. and Pierre F. J. Lermusiaux. Multiscale two-way embedding schemes for free-surface prim-itive equations in the “Multidisciplinary Simulation, Estimation and Assimilation System”. Ocean Dynamics,60(6):1497–1537, December 2010.

[6] MSEAS Group. MSEAS Software, 2013.

[7] National Centers for Environmental Prediction (NCEP). Global Forecast System, 2018.

2