betaFeedback

WorksComputer Science

Bayesian Learning of Coupled Biogeochemical-Physical Models

Abstract

Predictive dynamical models for marine ecosystems are used for a variety of needs. Due to sparse measurements and limited understanding of the myriad of ocean processes, there is however significant uncertainty. There is model uncertainty in the parameter values, functional forms with diverse parameterizations, level of complexity needed, and thus in the state fields. We develop a Bayesian model learning methodology that allows interpolation in the space of candidate models and discovery of new models from noisy, sparse, and indirect observations, all while estimating state fields and parameter values, as well as the joint PDFs of all learned quantities. We address the challenges of high-dimensional and multidisciplinary dynamics governed by PDEs by using state augmentation and the computationally efficient GMM-DO filter. Our innovations include stochastic formulation and complexity parameters to unify candidate models into a single general model as well as stochastic expansion parameters within piecewise function approximations to generate dense candidate model spaces. These innovations allow handling many compatible and embedded candidate models, possibly none of which are accurate, and learning elusive unknown functional forms. Our new methodology is generalizable, interpretable, and extrapolates out of the space of models to discover new ones. We perform a series of twin experiments based on flows past a ridge coupled with three-to-five component ecosystem models, including flows with chaotic advection. The probabilities of known, uncertain, and unknown model formulations, and of state fields and parameters, are updated jointly using Bayes' law. Non-Gaussian statistics, ambiguity, and biases are captured. The parameter values and model formulations that best explain the data are identified. When observations are sufficiently informative, model complexity and functions are discovered.

Introduction

The ability to predict and understand marine ecosystems is essential for addressing many of the grand challenges faced by humanity, such as climate change, food security, and sustainability. In broad terms, marine ecosystems can be seen as food webs, or flows of food/energy from nutrients to phytoplankton, to zooplankton, to fish, and finally recycling back to the nutrients [1, 2]. However, there does not yet exist a single generic model that accurately represents all the components in marine food webs due to the presence of highly complex biological processes with many unknown interactions. Therefore, many approximations are made in such ecosystem models. In addition, only parts of a food web are commonly modeled. The interactions of what is modeled with the portions of the food web that are not modeled are then either neglected or parameterized in terms of the modeled variables. Biology is also forced by complex nonlinear physics. Most biogeochemical-physical modeling systems thus broadly categorize the nutrients and individual species, representing them as continuous state variable fields, defined as concentrations of nutrients, biomass, or number of organisms per unit volume of water. The dynamics of these fields consists of reaction terms representing biogeochemical processes such as nutrient uptake, grazing, death, etc., and of forcing by physical processes such as advection, diffusion, and sunlight. Each reaction term or physical process is commonly modeled mathematically, using functional forms or terms that contain multiple parameters and have different levels of accuracy.

A plethora of biogeochemical modeling systems have been proposed, each of which with many model formulations [3, 4]. The models differ in their complexity, or ability to resolve different biological processes. Models of higher complexity have more biological components, functional terms, and parameters. However, process terms and parameters are often poorly known, which hampers the utility of highly complex models [5, 6, 7]. The simplest models are 3-component nutrient-phytoplankton-zooplankton (NPZ) models [8, 9]. NPZ models are easily understood and serve an important role in ocean research. Including the intermediate state of detritus leads to four component NPZ-Detritus biological models [10]. Intermediate complexity models involve around 7 to 10 components, adding bacteria, nitrate, ammonium, and dissolved organic nitrogen [11], or related state variables [12]. One of the most complex lower-trophic-level marine ecosystem models is the European Regional Seas Ecosystem Model (ERSEM, [13, 14, 15]), originally developed for the North Sea. Many choices of functional forms exist for each of the biological processes [5], leading to application-specific variants of the above models.

Biogeochemical models are commonly developed semi-empirically, leading to uncertainty in their parameters, functional forms, and level of complexity. What is adequate for a particular ocean region may not work elsewhere or may need to be updated due to seasonal or other variabilities [16]. Such model uncertainties transfer to the state variables predicted, complicating model learning by direct comparisons of state variables with in situ data. As a result, when observations are employed to develop models, it is often in an offline mode, fitting parameter values or functional forms to data in controlled experiments. With data assimilation, we could however use observations in a direct Bayesian sense, to jointly learn state variables, parameter values, and discriminate/discover functional forms with quantifiable uncertainty [17]. Most biogeochemical data assimilation [18, 19] can be categorized broadly into two categories. The first is parameter estimation, where model parameters are calibrated by minimizing misfits between model output fields and independent observations [20, 21, 6, 22, 23]. The second is sequential estimation, where observations collected are used to update model states during the forward model integration [12, 24, 25, 26, 27]. However, very few studies deal with the simultaneous estimation of parameters, state variables, and model equations. Doron et al. [28] used a Monte Carlo ensemble of 200 simulations lasting 30-days in the North Atlantic and conducted idealized twin experiments with surface observations of phytoplankton to estimate parameters and states with a Kalman filter-based scheme and state augmentation. Jones et al. [29] performed state and parameter estimation in a nonlinear phytoplankton-zooplankton model using two Markov Chain Monte Carlo (MCMC) algorithms in an identical-twin setting. Mattern et al. [30] used a nonlinear particle filter scheme to assimilate satellite sea surface color and jointly estimate the state and parameters of a three-dimensional biological ocean model. Lately, along with state and parameter estimation, the selection of optimal complexity of biogeochemical models has become a new area of research [17, 6, 31, 32]. Because of the multiscale and intermittent variability of marine ecosystems, there is also a need for generalized and adaptive modeling, where models can learn and adapt during run-time [16, 33, 34, 35].

Several machine learning methods have been developed for the discovery of model equations. The sparse regression-based methods (SINDy; [36, 37]) are promising as they do not require prior knowledge, however, they often require large data sets. Variations of SINDy include weak SINDy to learn PDEs [38], adaptive generation of features to increase the library of models [39], and extensions to Bayesian identification [40]. Deep learning methods have been derived to obtain marine ecosystem closure models [41, 42]. Genetic algorithms [43] and reinforcement learning [44, 45, 46] have been used to search the space of candidate models. However, most of these approaches do not provide uncertainty estimates for the discovered models. Methods have also combined prior knowledge about underlying governing equations for model recovery and refinement. For example, Raissi and Karniadakis [47] used Gaussian processes to learn the values of the parametric response of partially-known nonlinear differential equations. Unfortunately, data and knowledge of governing laws are luxuries in the case of realistic biogeochemical models.

It is clear that fundamental methods for identifying dynamical models that best explain sparse data in accord with prior governing laws and uncertainties would be most useful. The Bayesian theory and schemes of Lu and Lermusiaux [48, 49] address several of the above needs and drawbacks, using noisy, sparse, and indirect observations for joint Bayesian inference of states and parameters along with probabilistic discrimination among candidate models. However, several questions remain: Could we avoid assimilating observations independently in each candidate model when there are so many models to choose from? And if none of these models are that accurate, could the Bayesian machine find the elusive true formulations? Could it interpolate within and extrapolate out of known model spaces, while providing accurate joint probability distributions for model states, parameters, and formulations? Could such Bayesian learning be efficient and accurate with high-dimensional and multidisciplinary physical-biogeochemical stochastic PDEs? The overall goal of the present paper is thus to extend and generalize the discrimination-based model learning developed in [48, 49] to allow for interpolation in the space of candidate models and discovery of new models, in an efficient fashion. Our novel learning and discovery of differential models are achieved by introducing stochastic formulation parameters, stochastic complexity parameters, and piecewise function approximations assembled with stochastic expansion parameters. We address the challenges of multidisciplinary dynamics and develop a rigorous PDE Bayesian learning framework using state augmentation and the Gaussian Mixture Model - Dynamically Orthogonal (GMM-DO) filter [50, 51]. The final estimates are notably joint probability distributions for all learned quantities. To our knowledge, it is the first time that sequential Bayesian data assimilation is developed to predict and update the joint probability distributions of state variables, parameters, and known, uncertain, and unknown model formulations, enabling the Bayesian discovery of model functional forms and model complexities, with applications to high-dimensional ocean physical-biogeochemical dynamical systems.

In Sect. 2, we present the problem statement. In Sect. 3, we develop the general Bayesian learning methodology with novel parameters for model learning and discovery. In Sect. 4, we describe the stochastic biogeochemical-physical equations and simulated experiments. In Sect. 5, we apply our methodology to four sets of experiments of varying complexities and learning objectives. Conclusions are provided in Sect. 6.

Problem Statement

A single mathematical model that exactly captures all the physical and biological processes occurring in the real world does not yet exist. Hence, there is inherent model uncertainty that manifests in many forms, including: initial and boundary condition uncertainties; unreliable parameter values; multiple competing candidate model functions; unknown functional forms; missing model terms; and, debatable complexity of the model. In this work, we consider discriminating among candidate models, learning among compatible models, and discovering new model formulations. Compatible models are models that can be related to a single dynamical system theoretically and that can also be combined numerically. Compatible models can nonetheless represent different dynamics, e.g., our goals include learning which dynamics are or are not present based on observations.

In general, we consider a stochastic dynamical modeling system defined on a domain D\mathcal{D}, governing the uncertain spatiotemporal dynamics of ϕ(x,t;ω):RN×[0,T]→RNv\phi(x,t;\omega): \mathbb{R}^{N} \times[0,T] \to\mathbb{R}^{N_v}, the stochastic state vector comprising NvN_v dynamical state variable fields (e.g., physical fields, biogeochemical concentration fields, etc.). The realization index ω\omega belongs to a measurable sample space Ω\Omega and the model depends on a vector θ(ω)\theta(\omega) of NθN_{\theta} uncertain parameters. The main notation used is defined in Table D.2. To encompass the majority of scenarios, we write the general form of the uncertain dynamical modeling system as follows,

∂ϕ(x,t;ω)∂t=L[ϕ(x,t;ω),θ(ω),x,t]+L^[ϕ(x,t;ω);ω]+L~[ϕ(x,t;ω);ω],\frac{\partial\phi(x,t;\omega)}{\partial t} = \mathcal{L}[\phi(x,t;\omega),\theta(\omega),x,t] + \widehat{\mathcal{L}}[\phi(x,t;\omega);\omega] + \widetilde{\mathcal{L}}[\phi(x,t;\omega);\omega],
x∈D, t∈[0,T], ω∈Ω,(1)x \in\mathcal{D},\ t \in[0,T],\ \omega\in\Omega, \tag*{(1)}

with ϕ(x,0;ω)=ϕo(x;ω)\phi(x,0;\omega)=\phi_o(x;\omega),

and B[ϕ(x,t;ω)]=b(x,t;ω), x∈∂D, t∈[0,T], ω∈Ω\mathcal{B}[\phi(x,t;\omega)]=b(x,t;\omega),\ x \in\partial\mathcal{D},\ t \in[0,T],\ \omega\in\Omega,

where ϕo(x;ω)\phi_o(x;\omega), B\mathcal{B}, and b(x,t;ω)b(x,t;\omega) are the stochastic initial conditions, boundary condition operators, and boundary values respectively. The functional form of the first dynamics term L[ϕ(x,t;ω),θ(ω),x,t]\mathcal{L}[\phi(x,t;\omega),\theta(\omega),x,t] is assumed to be known, but with uncertain parameters θ(ω)\theta(\omega). The second term L^[ϕ(x,t;ω);ω]∈{L^1[ϕ(x,t;ω);ω],…,L^Nm[ϕ(x,t;ω);ω]}\widehat{\mathcal{L}}[\phi(x,t;\omega);\omega] \in\{\widehat{\mathcal{L}}_1[\phi(x,t;\omega);\omega],\ldots,\widehat{\mathcal{L}}_{N_m}[\phi(x,t;\omega);\omega]\} represents a set of compatible candidate functional forms, where NmN_m is the number of candidates. For example, for reaction terms, model functions are often from the polynomial, exponential, and/or sinusoidal families, and can be rational or irrational functions. The third term L~[ϕ(x,t;ω);ω]\widetilde{\mathcal{L}}[\phi(x,t;\omega);\omega] has a functional form completely unknown. Each of these three functional terms has uncertainties, hence the ω\omega dependence. Their summation encompasses common scenarios, e.g., their multiplication simply absorbs the more known types into the most unknown type. The stochastic initial and boundary condition formulations can also have uncertain function forms, similar to the dynamical modeling system itself, i.e., they can be known, belonging to a family, or unknown.

In some cases, candidate models have different complexities,

Mi:{∂ϕ1i(x,t;ω)∂t=L1i[ϕ1i(x,t;ω),…,ϕNv(i)i(x,t;ω),θi(ω),x,t;ω]⋮∂ϕNv(i)i(x,t;ω)∂t=LNv(i)i[ϕ1i(x,t;ω),…,ϕNv(i)i(x,t;ω),θi(ω),x,t;ω)],i=1,…,Nm(2)\mathcal{M}_i:\quad \begin{cases} \frac{\partial\phi^i_1(x,t;\omega)}{\partial t} = \mathcal{L}^i_1[\phi^i_1(x,t;\omega),\ldots,\phi^i_{N_v(i)}(x,t;\omega),\theta^i(\omega),x,t;\omega] \\ \vdots\\ \frac{\partial\phi^i_{N_v(i)}(x,t;\omega)}{\partial t} = \mathcal{L}^i_{N_v(i)}[\phi^i_1(x,t;\omega),\ldots,\phi^i_{N_v(i)}(x,t;\omega),\theta^i(\omega),x,t;\omega)] \end{cases} ,\quad i=1,\ldots,N_m \tag*{(2)}

where each model, Mi\mathcal{M}_i, has Nv(i)N_v(i) number of state variable fields ({ϕ1i,…,ϕNv(i)i}\{\phi^i_1,\ldots,\phi^i_{N_v(i)}\}) from a pool of candidates, and their aggregates. In such situations, the candidate models can often remain compatible with each other, for example, low-complexity models are embedded in higher-complexity ones. We refer to such classes of candidate models as, compatible-embedded models. The number NvN_v of dynamical state variable fields then denotes the number of state variables needed to encompass all models Mi\mathcal{M}_i’s. Of course, in general, uncertainty in parameter values, functional forms, and complexities occur simultaneously, thus, each term {L1i,…,LNv(i)i}\{\mathcal{L}^i_1,\ldots,\mathcal{L}^i_{N_v(i)}\} in equation (2) can encompass the L^\widehat{\mathcal{L}} and L~\widetilde{\mathcal{L}} terms introduced in equation (1). Such scenarios are exemplified in our series of experiments in Sect. 5. For example, in Experiments-2, three- and four-component biogeochemical models are considered to be candidate Mi\mathcal{M}_i’s; in Experiments-1& 4, the zooplankton mortality function is considered to be either linear or quadratic, corresponding to L^\widehat{\mathcal{L}}; and in Experiments-3, the zooplankton mortality function is assumed completely unknown, corresponding to L~\widetilde{\mathcal{L}}.

Let Φ(t;ω)∈RNvNx\Phi(t;\omega) \in\mathbb{R}^{N_vN_x} denote the spatially discretized state vector of the continuous field ϕ(x,t;ω)\phi(x,t;\omega), where NxN_x denotes the dimension of the discretized state space. In the experiments, we assume that all observations Y(t;ω)\mathcal{Y}(t;\omega) are noisy, sparse, and indirectly related to Φ(t;ω)\Phi(t;\omega) by a stochastic linear measurement model from the state to the data space,

Y(t;ω)=HΦ(t;ω)+V(t;ω),V(t;ω)∼N(0,R)(3)\mathcal{Y}(t;\omega) = H\Phi(t;\omega) + V(t;\omega),\qquad V(t;\omega) \sim\mathcal{N}(0,R) \tag*{(3)}

where NyN_y is the number of available observations; H∈RNy×NvNxH \in\mathbb{R}^{N_y \times N_vN_x} the observation matrix; and V∈RNyV \in\mathbb{R}^{N_y} a zero-mean, uncorrelated Gaussian observation noise with covariance matrix R∈RNy×NyR \in\mathbb{R}^{N_y \times N_y}. The latter noise in the observations is larger than the sensor noise as it also contains representation errors [52]: in our experiments, it will be a fraction of the variability in the state variables [53, 54]. The noisy observations are also sparse (Ny≪NvNxN_y \ll N_vN_x) and indirect: they are available only at discrete time-instants, tkt_k for k=1,2,…,Kk=1,2,\ldots,K, at a very limited number of spatial locations, and only for one state variable. As in reality, the specific variable being measured will vary from experiment to experiment.

In summary, our specific objectives are thus two-fold, first to solve the stochastic forward-modeling system (Eqs. 1 & 2), taking into account all the associated uncertainties including compatible, compatible-embedded, and unknown model terms; and second to simultaneously learn, in the Bayesian sense, the state fields, parameters, and model equations based on the stochastic, sparse, and indirect observation model (Eq. 3). Our Bayesian learning thus needs to evolve the prior and posterior joint probabilities of state fields, parameters, and model formulations, given the noisy observations available and all other uncertainties. The overall goal is to accurately represent these probability density functions (pdfs), including the marginal probabilities of known, uncertain, and unknown model formulations. It is only if the noisy observations are sufficiently informative about either the state fields, parameters, and model formulations, that the Bayesian machine can identify the true state variables, true parameters, and true model. If the noisy observations are not sufficiently informative, the perfect Bayesian machine will not lead to a unique identification, but provide the exact posterior probabilities of the models, parameter values, and state variable fields.

General Bayesian Learning Methodology

Prior to developing our general methodology, we briefly review the Bayesian learning for rigorous discrimination among candidate differential dynamical models [48, 49]. In this Bayesian discrimination, each candidate model then evolves the joint pdf of its state variables and parameters, independently from other models, and provides probability distributions that are conditional on the candidate model. In other words, each model runs its own probabilistic forecast. When noisy observations are made, the model-conditional state variables and parameters (all contained in Φ\boldsymbol{\Phi}), and also the model pdfs themselves, are updated using Bayes’ rules [55, 56],

pΦ∣Y,M(Φ∣y,Mi)=pY∣Φ,M(y∣Φ,Mi)pY∣M(y∣Mi)pΦ∣M(Φ∣Mi),∀ Φ∈RNvNx, ∀ i∈{1,…,Nm},pM∣Y(Mn∣y)=pY∣M(y∣Mi)pY(y)pM(Mi),∀ i∈{1,…,Nm},(4)p_{\boldsymbol{\Phi} \mid\mathcal{Y},\mathcal{M}}(\boldsymbol{\Phi}\mid y,\mathcal{M}_i)=\frac{p_{\mathcal{Y}\mid\boldsymbol{\Phi},\mathcal{M}}(y\mid\boldsymbol{\Phi},\mathcal{M}_i)}{p_{\mathcal{Y}\mid\mathcal{M}}(y\mid\mathcal{M}_i)}p_{\boldsymbol{\Phi}\mid\mathcal{M}}(\boldsymbol{\Phi}\mid\mathcal{M}_i),\quad\forall\ \boldsymbol{\Phi}\in\mathbb{R}^{N_vN_x},\ \forall\ i\in\{1,\ldots,N_m\}, \\ p_{\mathcal{M}\mid\mathcal{Y}}(\mathcal{M}_n\mid y)=\frac{p_{\mathcal{Y}\mid\mathcal{M}}(y\mid\mathcal{M}_i)}{p_{\mathcal{Y}}(y)}p_{\mathcal{M}}(\mathcal{M}_i),\quad\forall\ i\in\{1,\ldots,N_m\}, \tag*{(4)}

References

  1. [1]C. Lalli, T. R. Parsons, Biological Oceanography: An Introduction, Elsevier Butterworth-Heinemann, 1997.
  2. [2]W. Fennel, T. Neumann, Introduction to the Modelling of Marine Ecosystems:(with MATLAB programs on accompanying CD-ROM), Vol. 72 of Oceanography, Elsevier, 2014.
  3. [3]E. Hofmann, M. A. Friedrichs, Predictive modeling for marine ecosystems, The Sea 12 (2002) 537–565.
  4. [4]K. Fennel, J. P. Mattern, S. C. Doney, L. Bopp, A. M. Moore, B. Wang, L. Yu, Ocean biogeochemical modelling, Nature Reviews Methods Primers 2 (1) (2022) 76.
  5. [5]P. J. S. Franks, NPZ models of plankton dynamics: their construction, coupling to physics, and application, Journal of Oceanography 58 (2) (2002) 379–387.DOI
  6. [6]B. A. Ward, M. A. M. Friedrichs, T. R. Anderson, A. Oschlies, Parameter optimisation techniques and the problem of underdetermination in marine biogeochemical models, Journal of Marine Systems 81 (1) (2010) 34–43.DOI
  7. [7]K. L. Denman, Modelling planktonic ecosystems: parameterizing complexity, Progress in Oceanography 57 (3-4) (2003) 429–452.DOI
  8. [8]P. J. S. Franks, J. S. Wroblewski, G. R. Flierl, Behavior of a simple plankton model with food-level acclimation by herbivores, Marine Biology 91 (1) (1986) 121–129.DOI
  9. [9]E. H. D. J. McGillicuddy, Mesoscale and submesoscale physical-biological interactions, The sea 12 (2002) 113–185.
  10. [10]C. S. Davis, J. H. Steele, Biological/physical modeling of upper ocean processes, Tech. rep., Woods Hole Oc. Inst. (1994).DOI
  11. [11]M. J. R. Fasham, H. W. Ducklow, S. M. McKelvie, A nitrogen-based model of plankton dynamics in the oceanic mixed layer, Journal of Marine Research 48 (3) (1990) 591–639.DOI
  12. [12]Ş. T. Beşiktepe, P. F. J. Lemursiaux, A. R. Robinson, Coupled physical and biogeochemical data-driven simulations of Massachusetts Bay in late summer: Real-time and post-cruise data assimilation, Journal of Marine Systems 40–41 (2003) 171–212. doi:10.1016/S0924-7963(03)00018-6.DOI
  13. [13]J. W. Baretta, W. Ebenhöhl, P. Ruardij, The European Regional Seas Ecosystem Model, a complex marine ecosystem model, Netherlands Journal of Sea Research 33 (3) (1995) 233–246.DOI
  14. [14]J. W. Baretta, Preface to the European Regional Seas Ecosystem Model II, Journal of Sea Research 38 (3) (1997) 169–171.
  15. [15]J. C. Blackford, J. J. Allen, F. J. Gilbert, Ecosystem dynamics at six contrasting sites: a generic modelling study, Journal of Marine Systems 52 (1) (2004) 191–215.DOI
  16. [16]P. F. J. Lermusiaux, C. Evangelinos, R. Tian, P. J. Haley, Jr, J. J. McCarthy, N. M. Patrikalakis, A. R. Robinson, H. Schmidt, Adaptive coupled physical and biogeochemical ocean predictions: A conceptual basis, in: Computational Science - ICCS 2004, Vol. 3038 of Lecture Notes in Computer Science, Springer Berlin Heidelberg, 2004, pp. 685–692. doi:10.1007/978-3-540-24688-6_89.DOI
  17. [17]P. F. J. Lermusiaux, Adaptive modeling, adaptive data assimilation and adaptive sampling, Physica D: Nonlinear Phenomena 230 (1) (2007) 172–196. doi:10.1016/j.physd.2007.02.014.DOI
  18. [18]A. R. Robinson, P. F. J. Lermusiaux, Data assimilation for modeling and predicting coupled physical–biological interactions in the sea, in: A. R. Robinson, J. J. McCarthy, B. J. Rothschild (Eds.), Biological-Physical Interactions in the Sea, Vol. 12 of The Sea, John Wiley and Sons, New York, 2002, chapter 12, pp. 475–536.
  19. [19]M. Dowd, E. Jones, J. Parslow, A statistical overview and perspectives on data assimilation for marine biogeochemical models, Environmetrics 25 (4) (2014) 203–213.DOI
  20. [20]M. A. M. Friedrichs, et al., Assessment of skill and portability in regional marine biogeochemical models: Role of multiple planktonic groups, Journal of Geophysical Research: Oceans 112 (C8) (2007).DOI
  21. [21]S. N. Losa, G. A. Kivman, V. A. Ryabchenko, Weak constraint parameter estimation for a simple ocean ecosystem model: what can we learn about the model and data?, Journal of Marine Systems 45 (1) (2004) 1–20.DOI
  22. [22]J. P. Mattern, K. Fennel, M. Dowd, Estimating time-dependent parameters for a biological ocean model using an emulator approach, Journal of Marine Systems 96 (2012) 32–47.DOI
  23. [23]T. Toyoda, et al., Improved state estimations of lower trophic ecosystems in the global ocean based on a green’s function approach, Progress in Oceanography 119 (2013) 90–107.
  24. [24]J. P. Mattern, M. Dowd, K. Fennel, Sequential data assimilation applied to a physical–biological model for the bermuda atlantic time series station, Journal of Marine Systems 79 (1) (2010) 144–156.
  25. [25]J. I. Allen, M. Eknes, G. Evensen, An Ensemble Kalman Filter with a complex marine ecosystem model: hindcasting phytoplankton in the Cretan Sea, in: Annales Geophysicae, Vol. 21, 2003, pp. 399–411.
  26. [26]L. J. Natvik, G. Evensen, Assimilation of ocean colour data into a biogeochemical model of the North Atlantic: Part 1. Data assimilation experiments, Journal of Marine Systems 40 (2003) 127–153.
  27. [27]J. Hu, K. Fennel, J. P. Mattern, J. Wilkin, Data assimilation with a local ensemble kalman filter applied to a three-dimensional biological model of the middle atlantic bight, Journal of Marine Systems 94 (2012) 145–156.DOI
  28. [28]M. Doron, P. Brasseur, J.-M. Brankart, Stochastic estimation of biogeochemical parameters of a 3D ocean coupled physical–biogeochemical model: Twin experiments, Journal of Marine Systems 87 (3) (2011) 194–207.DOI
  29. [29]E. Jones, J. Parslow, L. Murray, A bayesian approach to state and parameter estimation in a phytoplankton-zooplankton model, Australian Meteorological and Oceanographic Journal 59 (SP) (2010) 7–16.
  30. [30]J. P. Mattern, M. Dowd, K. Fennel, Particle filter-based data assimilation for a three-dimensional biological ocean model and satellite observations, Journal of Geophysical Research: Oceans 118 (5) (2013) 2746–2760.
  31. [31]E. Giricheva, Aggregation in ecosystem models and model stability, Progress in Oceanography 134 (2015) 190–196.
  32. [32]B. A. Ward, et al., When is a biogeochemical model too complex? objective model reduction and selection for north atlantic time-series sites, Progress in Oceanography 116 (2013) 49–65.
  33. [33]C. Evangelinos, R. Chang, P. F. J. Lermusiaux, N. M. Patrikalakis, Rapid real-time interdisciplinary ocean forecasting using adaptive sampling and adaptive modeling and legacy codes: Component encapsulation using XML, in: Computational Science–ICCS 2003, Springer, 2003, pp. 375–384. doi:10.1007/3-540-44864-0_39.DOI
  34. [34]R. C. Tian, P. F. J. Lermusiaux, J. J. McCarthy, A. R. Robinson, A generalized prognostic model of marine biogeochemical-ecosystem dynamics: Structure, parameterization and adaptive modeling, Harvard Reports in Physical/Interdisciplinary Ocean Science 67, Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA (May 2004).
  35. [35]P. F. J. Lermusiaux, P. J. Haley, W. G. Leslie, A. Agarwal, O. Logutov, L. J. Burton, Multiscale physical and biological dynamics in the Philippine Archipelago: Predictions and processes, Oceanography 24 (1) (2011) 70–89, Special Issue on the Philippine Straits Dynamics Experiment. doi:10.5670/oceanog.2011.05.DOI
  36. [36]S. L. Brunton, J. L. Proctor, J. N. Kutz, Discovering governing equations from data by sparse identification of nonlinear dynamical systems, PNAS 113 (15) (2016) 3932–3937.DOI
  37. [37]S. Rudy, A. Alla, S. L. Brunton, J. N. Kutz, Data-driven identification of parametric partial differential equations, SIAM Journal on Applied Dynamical Systems 18 (2) (2019) 643–660.
  38. [38]D. A. Messenger, D. M. Bortz, Weak sindy for partial differential equations, Journal of Computational Physics (2021) 110525.
  39. [39]C. S. Kulkarni, A. Gupta, P. F. J. Lermusiaux, Sparse regression and adaptive feature generation for the discovery of dynamical systems, in: F. Darma, E. Blasch, S. Ravela, A. Aved (Eds.), Dynamic Data Driven Application Systems. DDDAS 2020., Vol. 12312 of Lecture Notes in Computer Science, Springer, Cham, 2020, pp. 208–216. doi:10.1007/978–3–030–61725–7_25.
  40. [40]R. K. Niven, A. Mohammad-Djafari, L. Cordier, M. Abel, M. Quade, Bayesian identification of dynamical systems, in: Multidisciplinary Digital Publishing Institute Proceedings, Vol. 33, 2020, p. 33.
  41. [41]A. Gupta, P. F. J. Lermusiaux, Neural closure models for dynamical systems, Proceedings of The Royal Society A 477 (2252) (2021) 1–29. doi:10.1098/rspa.2020.1004.DOI
  42. [42]A. Gupta, P. F. J. Lermusiaux, Generalized neural closure models with interpretability, Scientific ReportsSub-judice (2023). doi:10.48550/arXiv.2301.06198.
  43. [43]M. Maslyaev, A. Hvatov, A. Kalyuzhnaya, Data-driven pde discovery with evolutionary approach.(2019), arXiv preprint arXiv:1903.08011.
  44. [44]M. Bassenne, A. Lozano-Durán, Computational model discovery with reinforcement learning, arXiv preprint arXiv:2001.00008 (2019).arxiv.org/abs/2001.00008
  45. [45]G. Novati, H. L. de Laroussilhe, P. Koumoutsakos, Automating turbulence modelling by multi-agent reinforcement learning, Nature Machine Intelligence 3 (1) (2021) 87–96.
  46. [46]Y. Wang, Z. Shen, Z. Long, B. Dong, Deep Learning to discretize: solving 1d scalar conservation laws via deep reinforcement learning, arXiv preprint arXiv:1905.11079 (2019).
  47. [47]M. Raissi, G. E. Karniadakis, Hidden physics models: Machine learning of nonlinear partial differential equations, Journal of Computational Physics 357 (2018) 125–141.
  48. [48]P. G. Y. Lu, P. F. J. Lermusiaux, Pde-based bayesian inference of high-dimensional dynamical models, MSEAS Report 19, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA (2014).
  49. [49]P. Lu, P. F. J. Lermusiaux, Bayesian learning of stochastic dynamical models, Physica D: Nonlinear Phenomena 427 (2021) 133003. doi:10.1016/j.physd.2021.133003.
  50. [50]T. Sondergaard, P. F. J. Lermusiaux, Data assimilation with Gaussian Mixture Models using the Dynamically Orthogonal field equations. Part I: Theory and scheme., Monthly Weather Review 141 (6) (2013) 1737–1760. doi: 10.1175/MWR-D-11-00295.1.
  51. [51]T. Sondergaard, P. F. J. Lermusiaux, Data assimilation with Gaussian Mixture Models using the Dynamically Orthogonal field equations. Part II: Applications., Monthly Weather Review 141 (6) (2013) 1761–1785. doi:10.1175/MWR-D-11-00296.1.
  52. [52]T. Janjić, N. Bormann, M. Bocquet, J. Carton, S. Cohn, S. L. Dance, S. Losa, N. K. Nichols, R. Potthast, J. A. Waller, et al., On the representation error in data assimilation, Quarterly Journal of the Royal Meteorological Society 144 (713) (2018) 1257–1278.
  53. [53]P. F. J. Lermusiaux, D. G. M. Anderson, C. J. Lozano, On the mapping of multivariate geophysical fields: Error and variability subspace estimates, Quarterly Journal of the Royal Meteorological Society 126 (565) (2000) 1387–1429. doi:10.1256/smsqj.56509.DOI
  54. [54]P. F. J. Lermusiaux, On the mapping of multivariate geophysical fields: Sensitivities to size, scales, and dynamics, Journal of Atmospheric and Oceanic Technology 19 (10) (2002) 1602–1637. doi:10.1175/1520-0426(2002)019<1602:OTMOMG>2.0.CO;2.doi.org/10.1175/1520-0426(2002)019<1602:otmomg>2.0.co;2
  55. [55]T. Bayes, M. Price, An Essay towards Solving a Problem in the Doctrine of Chances. By the Late Rev. Mr. Bayes, F. R. S. Communicated by Mr. Price, in a Letter to John Canton, A. M. F. R. S., Philosophical Transactions (1683-1775) 53 (1763) 370–418. doi:10.1098/rstl.1763.0053.
  56. [56]D. Bertsekas, J. N. Tsitsiklis, Introduction to probability, Vol. 1, Athena Scientific, 2008.
  57. [57]L. N. Trefethen, Approximation Theory and Approximation Practice, Extended Edition, SIAM, 2019.
  58. [58]J. Lambers, MAT 772 fall semester 2010-11 lecture 17 notes, Available at https://www.math.usm.edu/lambers/mat772/fall10/lecture17.pdf (2023/01/20).
  59. [59]T. P. Sapsis, P. F. J. Lermusiaux, Dynamically orthogonal field equations for continuous stochastic dynamical systems, Physica D: Nonlinear Phenomena 238 (23–24) (2009) 2347–2360. doi:10.1016/j.physd.2009.09.017.DOI
  60. [60]T. P. Sapsis, P. F. J. Lermusiaux, Dynamical criteria for the evolution of the stochastic dimensionality in flows with uncertainty, Physica D: Nonlinear Phenomena 241 (1) (2012) 60–76. doi:10.1016/j.physd.2011.10.001.DOI
  61. [61]F. Feppon, F. J. Lermusiaux, A geometric approach to dynamical model-order reduction, SIAM Journal on Matrix Analysis and Applications 39 (1) (2018) 510–538. doi:10.1137/16M1095202.DOI
  62. [62]A. Gelb, Applied optimal estimation, MIT Press, 1974.
  63. [63]G. Casella, R. L. Berger, Statistical inference, Cengage Learning, 2021.
  64. [64]P. E. Hart, D. G. Stork, R. O. Duda, Pattern classification, Wiley Hoboken, 2000.
  65. [65]P. A. Newberger, J. S. Allen, Y. H. Spitz, Analysis and comparison of three ecosystem models, Journal of Geophysical Research: Oceans (1978–2012) 108 (C3) (2003).
  66. [66]P. J. Haley, Jr., P. F. J. Lermusiaux, Multiscale two-way embedding schemes for free-surface primitive equations in the “Multidisciplinary Simulation, Estimation and Assimilation System”, Ocean Dynamics 60 (6) (2010) 1497–1537. doi:10.1007/s10236-010-0349-4.DOI
  67. [67]P. J. Haley, Jr., A. Agarwal, P. F. J. Lermusiaux, Optimizing velocities and transports for complex coastal regions and archipelagos, Ocean Modeling 89 (2015) 1–28. doi:10.1016/j.ocemod.2015.02.005.DOI
  68. [68]J. C. McWilliams, The nature and consequences of oceanic eddies, Ocean modeling in an eddying regime 177 (2008) 5–15.DOI
  69. [69]M. W. Hecht, H. Hasumi, Ocean modeling in an eddying regime, Vol. 177, John Wiley & Sons, 2013.
  70. [70]J. H. Ferziger, M. Perić, R. L. Street, Computational methods for fluid dynamics, Vol. 3, Springer, 2002.
  71. [71]C. S. Kulkarni, P. F. J. Lermusiaux, Advection without compounding errors through flow map composition, Journal of Computational Physics 398 (2019) 108859. doi:10.1016/j.jcp.2019.108859.DOI
  72. [72]D. N. Subramani, P. F. J. Lermusiaux, Energy-optimal path planning by stochastic dynamically orthogonal level-set optimization, Ocean Modeling 100 (2016) 57–77. doi:10.1016/j.ocemod.2016.01.006.DOI
  73. [73]D. N. Subramani, Q. J. Wei, P. F. J. Lermusiaux, Stochastic time-optimal path-planning in uncertain, strong, and dynamic flows, Computer Methods in Applied Mechanics and Engineering 333 (2018) 218–237. doi:10.1016/j.cma.2018.01.004.DOI
  74. [74]M. Branicki, A. J. Majda, Fundamental limitations of polynomial chaos for uncertainty quantification in systems with intermittent instabilities, Communications in mathematical sciences 11 (1) (2013) 55–103.DOI
  75. [75]D. McGillicuddy, D. Lynch, A. Moore, W. Gentleman, C. Davis, C. Meise, An adjoint data assimilation approach to diagnosis of physical and biological controls on pseudocalanus spp. in the gulf of maine–georges bank region, Fisheries Oceanography 7 (3-4) (1998) 205–218.
  76. [76]P. F. J. Lermusiaux, Evolving the subspace of the three-dimensional multiscale ocean variability: Massachusetts Bay, Journal of Marine Systems 29 (1) (2001) 385–422. doi:10.1016/S0924-7963(01)00025–2.DOI
  77. [77]J. Pineda, V. Starczak, J. C. da Silva, K. Helfrich, M. Thompson, D. Wiley, Whales and waves: Humpback whale foraging response and the shoaling of internal waves at stellwagen bank, J. of Geophysical Res.: Oceans 120 (4) (2015) 2555–2570.
  78. [78]R. Tian, C. Chen, J. Qi, R. Ji, R. C. Beardsley, C. Davis, Model study of nutrient and phytoplankton dynamics in the gulf of maine: patterns and drivers for seasonal and interannual variability, ICES J. of Marine Science 72 (2) (2015) 388–402.
  79. [79]A. Pershing, et al., Temperature and circulation conditions in the gulf of maine in 2050 and their expected impacts, Scientific scenario paper, Gulf of Maine 2050 International Symposium, 2019.
  80. [80]T. L. Silva, State of the science report: An addendum to the stellwagen bank national marine sanctuary 2020 condition report (2021).
  81. [81]M. P. Ueckermann, P. F. J. Lermusiaux, 2.29 Finite Volume MATLAB Framework Documentation, MSEAS Report 14, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA (2012). URL http://mseas.mit.edu/?p=2567
  82. [82]B. Van Leer, Towards the ultimate conservative difference scheme. iv. a new approach to numerical convection, Journal of computational physics 23 (3) (1977) 276–299.
  83. [83]A. Gupta, Scientific machine learning for dynamical systems: Theory and applications to fluid flow and ocean ecosystem modeling, Ph.D. thesis, Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts (Sep. 2022).
  84. [84]M. P. Ueckermann, P. F. J. Lermusiaux, T. P. Sapsis, Numerical schemes for dynamically orthogonal equations of stochastic fluid and ocean flows, Journal of Computational Physics 233 (2013) 272–294. doi:10.1016/j.jcp.2012.08.041.DOI
  85. [85]F. Feppon, P. F. J. Lermusiaux, Dynamically orthogonal numerical schemes for efficient stochastic advection and Lagrangian transport, SIAM Review 60 (3) (2018) 595–625. doi:10.1137/16M109394.
  86. [86]Solve system of nonlinear equations - MATLAB, Accessed on January 21, 2023. URL https://www.mathworks.com/help/optim/ug/fsolve.html
  87. [87]L. Bengtsson, M. Ghil, E. Källén, Dynamic meteorology: data assimilation methods, Springer, 1981.
  88. [88]K. Ide, M. Ghil, Extended kalman filtering for vortex systems. part 1: Methodology and point vortices, Dynamics of Atmospheres and Oceans 27 (1–4) (1998) 301–332.
  89. [89]K. Ide, M. Ghil, Extended kalman filtering for vortex systems. part ii: Rankine vortices and observing-system design, Dynamics of Atmospheres and Oceans 27 (1–4) (1998) 333–350.
  90. [90]P. F. J. Lermusiaux, Data assimilation via Error Subspace Statistical Estimation, part II: Mid-Atlantic Bight shelf-break front simulations, and ESSE validation, Monthly Weather Review 127 (7) (1999) 1408–1432. doi:10.1175/1520–0493(1999)127<1408:DAVESS>2.0.CO;2.
  91. [91]T. Lolla, P. J. Haley, Jr., P. F. J. Lermusiaux, Time-optimal path planning in dynamic flows using level set equations: Realistic applications, Ocean Dynamics 64 (10) (2014) 1399–1417. doi:10.1007/s10236–014–0760–3.
  92. [92]A. Gupta, P. J. Haley, D. N. Subramani, P. F. J. Lermusiaux, Fish modeling and Bayesian learning for the Lakshadweep Islands, in: OCEANS 2019 MTS/IEEE SEATTLE, IEEE, Seattle, 2019, pp. 1–10. doi:10.23919/OCEANS40490.2019.8962892.
  93. [93]A. Bilmes, et al., A gentle tutorial of the EM algorithm and its application to parameter estimation for Gaussian mixture and hidden Markov models, International Computer Science Institute 4 (510) (1998) 126.
  94. [94]P. Stoica, Y. Selen, Model-order selection: a review of information criterion rules, Signal Processing Magazine, IEEE 21 (4) (2004) 36–47.
  95. [95]G. Wornell, Inference and information. lecture notes for mit course 6.437 in spring 2013 (May 2016).
  96. [96]J. Jin, Bayesian learning for high-dimensional nonlinear systems: Methodologies, numerics and applications to fluid flows, Ph.D. thesis, Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts (Sep. 2020).
  97. [97]R. O. Duda, P. E. Hart, et al., Pattern classification, John Wiley & Sons, 2006.
  98. [98]A. Gupta, Bayesian inference of obstacle systems and coupled biogeochemical-physical models, Master’s thesis, Indian Institute of Technology Kanpur, Kanpur, India (2016).
  99. [99]P. F. J. Lermusiaux, D. N. Subramani, J. Lin, C. S. Kulkarni, A. Gupta, A. Dutt, T. Lolla, P. J. Haley, Jr., W. H. Ali, C. Mirabito, S. Jana, A future for intelligent autonomous ocean observing systems, Journal of Marine Research 75 (6) (2017) 765–813, the Sea. Volume 17, The Science of Ocean Prediction, Part 2. doi:10.1357/002224017823524035.DOI
  100. [100]P. F. J. Lermusiaux, et al., A future for intelligent autonomous ocean observing systems, Journal of Marine Research 75 (6) (2017) 765–813, the Sea. Volume 17, The Science of Ocean Prediction, Part 2. doi:10.1357/002224017823524035.DOI
  101. [101]P. F. J. Lermusiaux, et al., Optimal planning and sampling predictions for autonomous and Lagrangian platforms and sensors in the northern Arabian Sea, Oceanography 30 (2) (2017) 172–185, special issue on Autonomous and Lagrangian Platforms and Sensors (ALPS). doi:10.5670/oceanog.2017.242.
  102. [102]D. N. Subramani, Probabilistic regional ocean predictions: Stochastic fields and optimal planning, Ph.D. thesis, Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts (Feb. 2018).
  103. [103]D. Subramani, P. F. J. Lermusiaux, Probabilistic ocean predictions with dynamically-orthogonal primitive equations, in preparation (2023).
  104. [104]K. A. Gkirkis, Stochastic ocean forecasting with the dynamically orthogonal primitive equations, Master’s thesis, Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts (Jun. 2021).
  105. [105]K. A. Gkirkis, P. F. J. Lermusiaux, Massive probabilistic forecasts for the Gulf of Mexico: Dynamically-orthogonal primitive equations, in preparation (2023).
  106. [106]P. F. J. Lermusiaux, A. R. Robinson, Data assimilation via Error Subspace Statistical Estimation, part I: Theory and schemes, Monthly Weather Review 127 (7) (1999) 1385–1407. doi:10.1175/1520-0493(1999)127<1385:DAVESS>2.0.CO;2.
  107. [107]P. F. J. Lermusiaux, Estimation and study of mesoscale variability in the Strait of Sicily, Dynamics of Atmospheres and Oceans 29 (2) (1999) 255–303. doi:10.1016/S0378-3774(99)00008-1.
  108. [108]P. F. J. Lermusiaux, C.-S. Chiu, A. R. Robinson, Modeling uncertainties in the prediction of the acoustic wavefield in a shelfbreak environment, in: E.-C. Shang, Q. Li, T. F. Gao (Eds.), Proceedings of the 5th International Conference on Theoretical and Computational Acoustics, World Scientific Publishing Co., 2002, pp. 191–200, refereed invited manuscript. doi:10.1142/9789812777362_0020.DOI
  109. [109]G. Cossarini, P. F. J. Lermusiaux, C. Solidoro, Lagoon of Venice ecosystem: Seasonal dynamics and environmental guidance with uncertainty analyses and error subspace data assimilation, Journal of Geophysical Research: Oceans 114 (C6) (Jun. 2009). doi:10.1029/2008JC005080.DOI
  110. [110]P. F. J. Lermusiaux, C. Mirabito, P. J. Haley, Jr., W. H. Ali, A. Gupta, S. Jana, E. Dorfman, A. Laferriere, A. Kofford, G. Shepard, M. Goldsmith, K. Heaney, E. Coelho, J. Boyle, J. Murray, L. Freitag, A. Morozov, Real-time probabilistic coupled ocean physics-acoustics forecasting and data assimilation for underwater GPS, in: OCEANS 2020 IEEE/MTS, IEEE, 2020, pp. 1–9. doi:10.1109/IEEECONF38699.2020.9389003.DOI
  111. [111]P. J. Haley, Jr., A. Gupta, C. Mirabito, P. F. J. Lermusiaux, Towards Bayesian ocean physical-biogeochemical-acidification prediction and learning systems for Massachusetts Bay, in: OCEANS 2020 IEEE/MTS, IEEE, 2020, pp. 1–9. doi:10.1109/IEEECONF38699.2020.9389210.DOI
  112. [112]T. Lolla, P. F. J. Lermusiaux, A Gaussian mixture model smoother for continuous nonlinear stochastic dynamical systems: Theory and scheme, Monthly Weather Review 145 (2017) 2743–2761. doi:10.1175/MWR-D-16-0064.1.DOI
  113. [113]T. Lolla, P. F. J. Lermusiaux, A Gaussian mixture model smoother for continuous nonlinear stochastic dynamical systems: Applications, Monthly Weather Review 145 (2017) 2763–2790. doi:10.1175/MWR-D-16-0065.1.DOI
  114. [114]K. D. Heaney, P. F. J. Lermusiaux, T. F. Duda, P. J. Haley, Jr., Validation of genetic algorithm based optimal sampling for ocean data assimilation, Ocean Dynamics 66 (2016) 1209–1229. doi:10.1007/s10236-016-0976-5.DOI
  115. [115]K. D. Heaney, G. Gawarkiewicz, T. F. Duda, P. F. J. Lermusiaux, Nonlinear optimization of autonomous undersea vehicle sampling strategies for oceanographic data-assimilation, Journal of Field Robotics 24 (6) (2007) 437–448. doi:10.1002/rob.20183.DOI
  116. [116]P. F. J. Lermusiaux, P. J. Haley, Jr., N. K. Yilmaz, Environmental prediction, path planning and adaptive sampling: sensing and modeling for efficient ocean monitoring, management and pollution control, Sea Technology 48 (9) (2007) 35–38.
  117. [117]S. R. Ramp, R. E. Davis, N. E. Leonard, I. Shulman, Y. Chao, A. R. Robinson, J. Marsden, P. F. J. Lermusiaux, D. M. Fratantoni, J. D. Paduan, F. P. Chavez, F. L. Bhah, S. Jiang, W. Leslie, Z. Li, Preparing to predict: The second Autonomous Ocean Sampling Network (AOSN-II) experiment in the Monterey Bay, Deep Sea Research Part II: Topical Studies in Oceanography 56 (3–5) (2009) 68–86. doi:10.1016/j.dsr2.2008.08.013.DOI
  118. [118]D. Wang, P. F. J. Lermusiaux, P. J. Haley, Jr., D. Eickstedt, W. G. Leslie, H. Schmidt, Acoustically focused adaptive sampling and on-board routing for marine rapid environmental assessment, Journal of Marine Systems 78 (Supplement) (2009) S393–S407. doi:10.1016/j.jmarsys.2009.01.037.DOI
  119. [119]S. Petillo, D. Y. H. Schmidt, P.F.J. Lermusiaux, A. Balasuriya, Autonomous & adaptive oceanographic front tracking on board autonomous underwater vehicles, in: Proceedings of IEEE OCEANS’15 Conference, IEEE, Genoa, 2015. doi:10.1109/oceans-genova.2015.7271616.DOI
  120. [120]M. Cococcioni, B. Lazzerini, P. Lermusiaux, Adaptive sampling using fleets of underwater gliders in the presence of fixed buoys using a constrained clustering algorithm, in: Proceedings of IEEE OCEANS’15 Conference, IEEE, Genoa, 2015. doi:10.1109/oceans-genova.2015.7271446.DOI
  121. [121]K. Rajan, F. Aguado, P. Lermusiaux, J. B. de Sousa, A. Subramaniam, J. Tintore, METEOR: A Mobile (portable) ocEan roboTic obsErvatOry, Marine Technology Society Journal 55 (3) (2021) 74–75. doi:10.4031/MTSJ.55.3.42.DOI
  122. [122]S. V. T. Lolla, Path planning and adaptive sampling in the coastal ocean, Ph.D. thesis, Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts (Feb. 2016).
  123. [123]P. F. J. Lermusiaux, P. J. Haley, Jr., S. Jana, A. Gupta, C. S. Kulkarni, C. Mirabito, W. H. Ali, D. N. Subramani, A. Dutt, J. Lin, A. Shcherbina, C. Lee, A. Gangopadhyay, Optimal planning and sampling predictions for autonomous and Lagrangian platforms and sensors in the northern Arabian Sea, Oceanography 30 (2) (2017) 172–185, special issue on Autonomous and Lagrangian Platforms and Sensors (ALPS). doi:10.5670/oceanog.2017.242.DOI
  124. [124]A. Gupta, W. H. Ali, P. F. J. Lermusiaux, Boundary conditions for stochastic DO equations, MSEAS Report, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA (2016).
  125. [125]M. J. Humara, Stochastic acoustic ray tracing with dynamically orthogonal equations, Master’s thesis, Massachusetts Institute of Technology, Joint Program in Applied Ocean Science and Engineering, Cambridge, Massachusetts (May 2020).
  126. [126]M. J. Humara, W. H. Ali, A. Charous, M. Bhabra, P. F. J. Lermusiaux, Stochastic acoustic ray tracing with dynamically orthogonal equations, in: OCEANS 2022 IEEE/MTS, IEEE, Hampton Roads, VA, 2022, pp. 1–10. doi:10.1109/OCEANS47191.2022.9977252.DOI
  127. [127]A. Charous, M. J. Humara, W. H. Ali, M. S. Bhabra, A. Gupta, P. F. J. Lermusiaux, Dynamically orthogonal ray equations with adaptive reclustering, The Journal of the Acoustical Society of America 150 (4) (2021) A209–A209. doi:10.1121/10.0008139.DOI

Paper details

vlnc:W_4m8ywvw6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents