YIP Physics-Data Driven Surface Flux Parameterization for Air-Sea Interaction
Abstract
Covering approximately two-thirds of Earth s surface, seas and oceans serve as pivotal determinants for weather patterns, climate, and various facets of human life. The expansive interface between air and sea facilitates the exchange of mass, momentum, and energy, which is significantly influenced by surface wave dynamics within the air-sea boundary layer. Acting as roughness elements, these waves alter wind stress through skin friction and form drag across diverse lengths and time scales. Despite the importance of this phenomenon, contemporary research and parameterizations of wind stress over the ocean have inadequately addressed the sea state s influence on drag. Key factors such as wave height, age, slope, and wind-wave alignment are often neglected. For instance, many operational models utilize a simplistic bulk parameterization following the Monin#Obukhov similarity theory (MOST). While convenient, this approach is only valid under equilibrium conditions and fails to consider sea-state variability. As a result, forecasting the dynamics of wave fields is a common challenge faced by oceanic and atmospheric models. This issue extends to predicting weather in intricate high-resolution environments like coastal zones and cyclones, and inhibits the effectiveness of global and general circulation models for seasonal and climate-scale forecasts. Addressing this problem necessitates comprehensive research into the relationships between sea-surface drag and sea-state parameters, and the development of parameterizations that incorporate sea-state considerations.This aligns directly with the research priorities of the Department of Defense (DoD). An emerging field of potential is data-drivendiscovery, regarded as the fourth pillar of scientific discovery, providing an opportunity to confront this issue head-on. This project introduces an innovative, sea-state-aware, physics-informed, and data-driven parameterization for surface stresses. It posits that a physics-informed machine learning model#trained with high-resolution process-resolving Large Eddy Simulation (LES) data and bound by physical constraints#will yield more accurate estimates of area-aggregate surface fluxes, outperforming the MOST model acrossvaried air-sea interaction regimes. To validate this hypothesis, an extensive database of process-resolving LES of air-sea interaction will be compiled, incorporating programmatically varied properties of the ocean wave field, wind forcing, and wind-wave angle. The data will be employed to develop and verify a Physics-Informed Neural Network for surface momentum Fluxes (PINN-FLUX). A standoutfeature of the proposed model is its capacity to map input parameters to spatially localized flow statistics, exceeding the conventional area-aggregate approach for evaluating surface fluxes and facilitating the assimilation of available measurements for parameter calibration. The performance of PINN-FLUX will be validated under a broad range of sea state conditions using observations from the Air-Sea Interaction Tower (ASIT). The model parameters will be calibrated within a Bayesian framework, and predictions will be extensively compared against those from MOST. PINN-FLUX s capacity to capture a broader scope of physical processes and seamlessly assimilate available airside statistics promises a more accurate and nuanced depiction of surface fluxes under complex sea-state conditions. Furthermore, compared to strictly physics-based approaches, the proposed model can be easily extended, via transfer learning techniques, to encapsulate a wider range of flow phenomena impacting surface fluxes. Thus, PINN-FLUX is expected to significantly enhance the DoD#s capabilities by reducing uncertainty in numerical weather and climate model predictions.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 09, 2024
- Source ID
- N000142412604
Entities
People
- Marco G Giometto
Organizations
- Office of Naval Research
- Trustees of Columbia University in the City of New York
- United States Navy