Fair-Weather Nearshore Surface Winds: Observational Insights and Gaussian Process Regression Analysis

Abstract

The fair-weather (wind speeds 10 m/s), surface ($z=$ 4 m), nearshore wind field modification is examined with multiple cross-shore arrays spanning from the coastline to 40 km offshore, deployed within four-month-long field experiments, measuring winds and air-water temperatures. The over-water to coastline winds ratio, $R_U$, was previously explored for offshore winds with limited observations and minimally for onshore and alongshore winds. Array observations provide a complete picture of the nearshore wind field for all wind orientations. The over-water wind and temperature are linearly related to coastline values near the coastline, with $r^2$ decreasing with distance from shore, $X^W$, with a decorrelation scale of 10 km. Mean $R_U$ as a function of $X^W$ differs per wind orientation, consistent with prior work. An ${\hat{R}}_U$ model is developed from exponential Gaussian Process Regression (GPR), which accurately predicts the wind field with 20% data set holdback to elucidate the cross-shore patterns and variable co-dependence. The modeled Partial Dependence Plots provide $R_U$ dependency as a function of $X^W$ on coastline winds, and temperature differences without preconceptions. A consistent nearshore wind slowing occurs that varies in amplitude and distance, and changes with variable co-dependence for wind orientation. The onshore wind slowing is counterintuitive, though consistent with sophisticated numerical models. Wind gustiness exhibits $X^W$ dependence, with linear normalization akin to the open ocean but larger. The $R_U$ observations and GPR highlight nearshore winds' cross-shore extent and complexity, which are important for atmospheric and oceanic studies.