Abstract: Significant wave height (H_s) is a key parameter in ocean engineering and marine operations, directly influencing the design and safety assessment of offshore structures and coastal facilities. Current in-situ H_s observations primarily rely on buoys and satellite remote sensing, which are costly to maintain and often limited by extreme weather conditions. The technique for estimating H_s from ocean surface gravity waves (OSGWs) recorded by seafloor distributed acoustic sensing (DAS) offers a potential solution to these challenges. However, existing studies are mostly limited to short-range H_s inversion, and approaches using single-channel seafloor DAS for long-range estimation suffer from low accuracy. The previous approach for H_s inversion does not fully utilize the advantages offered by the multi-channel sensing capability of the DAS. To address this issue, we propose a deep learning-based modeling approach to extract spatiotemporal features from the effective DAS signals recorded by multiple seafloor DAS channels, enabling the inversion of H_s observed by an Acoustic Wave and Current Profiler (AWAC) in the distant ocean area (~17.8km). Our method significantly improves the accuracy of long-range H_s estimation and extends the effective range of seafloor DAS-based H_s retrieval, highlighting its potential for cost-effective and reliable OSGW parameter retrieval in ocean engineering applications.

Key words: Significant wave height, Distributed acoustic sensing, Deep learning, Long-Distance Estimation