Monitoring Offshore Ocean Wave Heights from Land Seismic Networks: Multi-Location ANN Estimations and Insights from Microseism Wavefield Simulations

This study investigates the potential of using land based seismic observations of secondary microseisms to estimate oceanic significant wave height (SWH) in the Northeast Atlantic. A feedforward artificial neural network (ANN) was trained using physical buoy measurements, hindcast wave data and seismic amplitudes recorded at multiple broadband stations across Ireland. The model successfully reproduces temporal variability in SWH at different locations and demonstrates strong predictive capability across both real and virtual offshore sites. Spatial analysis reveals that model performance is not uniform, with higher accuracy and stronger signal coherence observed at northern and northwestern locations compared to sites further south. To understand this variability, numerical simulations of microseism propagation were performed for different source regions. The simulations show that microseism wavefield propagation from northern sources is less distorted than for southern sources, where wavefields interact strongly with the shelf. These results suggest that the spatial variability in ANN performance is consistent with and may be partly influenced by the physical characteristics of microseism wavefield propagation, particularly the degree of scattering along source to receiver paths. The agreement between ANN predictions and simulated wavefields demonstrates that the model captures physically meaningful relationships between ocean wave activity and seismic response. This work highlights the potential of terrestrial seismic networks as a complementary tool for offshore wave monitoring and provides new insight into the limitations imposed by spatial variations in microseism wavefield complexity.