CNN-based reconstruction of near-surface atmospheric turbulence using surface wave measurements

The small-scale turbulence resulting from wind-wave interactions profoundly affects the interfacial air-sea flux exchanges and, consequently, the long-term climate trends and short-term weather events. However, the correlation between this turbulence and surface wave characteristics has yet remained challenging due to the complexity of near-surface dynamics. It is, in fact, extremely challenging to resolve the near-surface turbulence using either high-resolution experimental/field measurements or numerical techniques in most wind-wave conditions.

Over the past few years, deep learning methods have been increasingly used to estimate turbulence from indirect and limited measurements. In this study, we developed a CNN model to reconstruct the turbulent flow above surface waves based on surface observations, such as surface elevation and surface velocity. The model is trained to minimize the induced reconstruction errors using the existing dataset of high-resolution measurements above wind-generated surface waves obtained using particle image velocimetry (PIV) and laser-induced fluorescence (LIF) techniques in the wind-wave tunnel facility. This novel approach seeks to enhance the understanding of near-surface turbulence and improve the predictive capabilities that are crucial for wind-wave interaction estimates.