Dynamical Attribution of Non-Equilibrium Directional Wave States Using Autonomous Observations and Coupled Modeling
Scientific motivation
Ocean surface waves do not adjust instantaneously to changes in atmospheric forcing. Previous studies have demonstrated that directional adjustment depends on frequency, wind history, and the development state of the wave field (e.g., van Vledder and Holthuijsen, 1993; Jackson and Jensen, 1999; Chen et al., 2016). Consequently, an observed wave spectrum can retain signatures of earlier or remotely generated forcing that are not explained by the instantaneous local wind. Here, I refer to this information as sea-state memory. Existing forcing-response relationships, however, have largely been constrained using prescribed forcing transitions, relatively homogeneous conditions, or bulk wave properties. How frequency- and direction-dependent adjustment evolves when atmospheric forcing, wave propagation, and ocean currents vary continuously in space and time remains poorly constrained.
Tropical cyclones (TCs) provide a natural laboratory for investigating these processes. Rapidly evolving winds and strong spatial gradients produce complex wave fields containing locally forced and remotely generated energy, organized into a continuous, directionally skewed spectrum rather than distinct wind-sea and swell systems (e.g., Young, 2006; Hu and Chen, 2011). Nonlinear spectral transfer, wave propagation, and wave-current interactions further modify this structure (e.g., Fan et al., 2009; Tamizi et al., 2021), while recent observations demonstrate that wind-wave alignment influences surface roughness under hurricane conditions (Davis et al., 2025). Thus, the instantaneous wave state reflects multiple interacting processes operating across different spatial and temporal scales. My research will determine how forcing history is encoded in directional wave state, identify the physical processes that create and maintain this memory, and establish how much of this dynamical information can be obtained from autonomous observations.
Preliminary research and hypotheses
My preliminary analysis of autonomous directional-wave observations under multiple TCs suggests that this frequency and direction-dependent response is structured rather than arbitrary. Compared with instantaneous wind forcing alone, incorporating historical forcing improves predictions of spectral properties (Fig. 1b). Historical crosswind forcing improves predictions of the directional offset between peak and high-frequency components by ~15%, while historical wind curl reduces HF directional-spread prediction error by ~14%. These improvements occur across all four held-out storms, suggesting that different aspects of directional wave structure retain complementary information about atmospheric forcing history.
Multivariate analyses (PCA, Isomap, and clustering) further indicate that observed wave states occupy a continuous, relatively low-dimensional space rather than separating into distinct regimes. Although constructed only from instantaneous wave properties, observations from translating storms organize coherently in chronological order without time being included as an input (Fig. 1a). High-frequency directional spreading and historical wind curl, neither used to construct the embedding, also exhibit systematic gradients across this space (Fig. 1c). Different storms traverse distinct portions of this shared continuum, and differences in trajectory direction and curvature may reflect transitions between physical processes or changes in their relative influence. Establishing these relationships could help explain how common physical processes produce different wave-field evolutions across storms. However, this organization alone does not establish dynamical significance. Demonstrating that connection is a central objective of the proposed research.
I will test three hypotheses:
H1 — Frequency-dependent memory: Different spectral components retain distinct signatures of atmospheric forcing history, reflected in their directional structure and spreading. H2 — Dynamical organization of wave states: The observed low-dimensional continuum reflects changes in the balance of physical processes governing wave evolution, with state-space position and trajectory geometry providing information about these dynamics. H3 — Dynamical information from autonomous observations: Distinct physical processes produce identifiable spectral signatures that can be recovered from sparse directional-wave observations, with quantifiable limits imposed by sampling trajectories, spatial variability, and the wave properties measured.
Research approach
I will first extend my ongoing analysis of multi-storm directional-wave observations collected by drifting Spotter buoys through ongoing ONR-funded projects, including SASCWATCH MURI. Directional spectra reconstructed from measured Fourier coefficients will be used to characterize spectral shape, directional spreading, wind-wave alignment, mean square slope, and directional offsets between spectral components. Instantaneous forcing will be compared with fixed-lag, weighted, and wave-following wind histories, including wave-relative directional forcing and spatial wind gradients, to identify reproducible signatures of preceding forcing and their frequency dependence. I will then evaluate the robustness of the preliminary state-space structure across storms, reconstruction methods, and feature subsets, using dimensionality reduction, feature ablation, and cross-storm validation. Temporal information will be introduced only after constructing the state space to examine the coherence and geometry of individual storm trajectories. These analyses will determine whether differences in observed evolution are robust and identify the minimum observable properties needed to represent them. Existing datasets and analysis tools developed through my preliminary work, including computational tools for spectral reconstruction, Doppler correction, forcing-history analysis, and multivariate evaluation, allow the proposed research to begin immediately.
To identify the physical mechanisms responsible for this organization, I will use high-resolution coupled atmosphere-ocean-wave simulations through the established modeling framework and computational resources of the Seo Coupled Ocean–Atmosphere Research (SCOAR) Lab. The coupled WRF–ROMS–WAVEWATCH III system will provide dynamically consistent wind, current, and directional-wave fields, including frequency- and direction-resolved wave-action budgets, wind input, dissipation, nonlinear transfer, propagation, and current-induced effects. Following evaluation against observed wave properties, the unmodified simulations will be used to diagnose the processes actively maintaining and modifying the realized wave field. Offline wave-model replay experiments with controlled forcing changes, supplemented by selected fully coupled sensitivity simulations, will distinguish direct process effects from subsequent coupled feedbacks. The model-resolved tendencies terms will then be related to independently constructed observational state-space coordinates and the geometry of individual storm trajectories to test is whether the changes in trajectory direction, curvature, or separation correspond to identifiable shifts in the modeled dynamical balance.
Finally, I will use the modeled four-dimensional wave field to establish how much of this dynamical information can be recovered from sparse autonomous observations. Modeled properties and process wave-action budgets will be sampled along observed and synthetic drifting trajectories, fixed locations, and wave-following paths. These virtual-observing experiments will distinguish true local evolution from apparent temporal changes caused by motion through storm-relative spatial gradients and test whether state-space position and trajectory geometry reliably diagnose underlying dynamics. Relationships will be evaluated on independent storms, and where data permit, other rapidly-evolving high-wind environments. Where distinct physical balances produce indistinguishable observations, the analysis will also establish limits on dynamical inference. The ultimate objective is a reduced-order, physically interpretable framework that extracts information about sea-state history and evolving dynamics from autonomous directional-wave measurements.
Research milestones and expected publications
Year 1 —Observational synthesis and wave-state organization (H1–H2). I will complete the multi-storm forcing-history and spatial-gradient analysis, establish the robustness and dimensionality of the observed continuum, and characterize the geometry of individual storm trajectories. A first manuscript will describe frequency-dependent wave adjustment and the emergent organization of directional wave states.
Year 2 — Coupled dynamical attribution (H2). I will evaluate the coupled simulations, diagnose frequency- and direction-resolved wave-action budgets, and conduct controlled sensitivity experiments. A second manuscript will present the dynamical interpretation of the observed continuum and the processes governing its evolution.
Year 3 — Dynamical observability and autonomous sampling (H3). I will conduct virtual-buoy experiments to quantify spatial-sampling effects and limits of the identified relationships across independent storms and other high-wind environments. A third manuscript will address the dynamical information recoverable from autonomous wave observations, including uncertainties and limitations. Reproducible analysis tools and a compact, history-aware observational framework will provide outcomes applicable to future autonomous observing systems and coupled model evaluation.
Relevance to the Office of Naval Research
This research directly supports ONR priorities in non-equilibrium air-sea interaction, autonomous environmental sensing, high-resolution coupled prediction, and physically based model development. Autonomous platforms enable critical observations in severe marine environments, but their motion through strong spatial gradients and transient variability makes it difficult to distinguish local environmental evolution from spatial sampling. By identifying how local wind forcing, nonlocal wave propagation, nonlinear transfer, and ocean currents shape directional wave fields, this research will determine which underlying dynamics can be inferred from sparse autonomous observations and quantify the limits of that inference. These physically based constraints will improve the interpretation of autonomous measurements and provide new benchmarks for evaluating coupled ocean–atmosphere–wave prediction models. The project directly leverages and complements ONR investments in SASCWATCH, SAFARI, and related high-wind air–sea interaction research.