Prediction of radiation transmittance and reflectance in ocean-sea fog coupled environment driven by wind speed forecasts

Frequent fog occurrences over the ocean surface pose significant challenges to remote sensing and target detection in marine environments. Considering the complexity and dynamism of the ocean surface, we propose a predictive transmittance and reflectance model (PTR), which is driven by wind speed and utilized for predicting the transmittance and reflectance of radiative transfer in coupled ocean-sea fog environments. The PTR combines a hierarchical two-dimensional convolutional time series prediction network named Conv2Times with the decomposable Monte Carlo (DMC) algorithm for calculating radiative transport. Conv2Times is designed for short-term wind speed forecasting tasks and features a unique time series reconstruction method termed "close to the correct answer (CCA)". Sea surface wind speed data were sourced from two buoy stations managed by the National Oceanic and Atmospheric Administration (NOAA). Six wind speed prediction experiments were conducted with forecast lengths of 40, 60, and 80 minutes. Conv2Times achieved a mean absolute error (MAE) in these tasks of 0.621 m/s, 0.678 m/s, 0.721 m/s, 0.459 m/s, 0.514 m/s, and 0.562 m/s, respectively, demonstrating lower prediction errors compared to seven other benchmark algorithms. Incorporating forecasted wind speeds into the DMC model for predicting radiative transmittance and reflectance, a pronounced trend consistency is evident between the curves resulting from predicted and actual wind speeds. These results indicate that PTR effectively predicts radiation transmission characteristics, which is significant for remote sensing and target detection in marine environments.