Abstract

Seismic wavefields are exceptional carriers of information about their sources and propagation paths, providing long-range, multiscale, and spatially explicit data through non-invasive, cost-efficient, and scalable acquisition. Their well-understood and often linear physics enables applications across a wide range of natural processes. Yet many systems—such as earthquakes, soils, and wildlife behaviour—are inherently complex: multi-scalar, agent-driven networks with emergent dynamics and abrupt transitions. In the absence of general theoretical frameworks for such systems, comprehensive datasets are essential but often difficult to obtain.

This talk presents collaborative advances in seismic data acquisition and modelling that address three challenges across seven orders of magnitude—each constrained by intrinsic data poverty—and demonstrates how seismic methods can expand our understanding of complex systems, or deduce valuable inference for science and society alike.

• Earthquake hazards: We developed a machine-learning framework that generates synthetic ground shaking in 3D Earth models quasi-instantaneously for arbitrary earthquake scenarios. This bridges empirical data catalogues and physics-based modelling towards probabilistic hazard assessment.

• Wildlife behaviour stands at the frontline of climate events, biodiversity loss and land degradation. Our seismic recordings from the Kenyan savanna capture the movements and communications of large mammals across landscapes. These data can help understand interspecies interaction and support early-warning systems that reduce human–wildlife conflict.

• Soil health: Degraded soils threaten food security and ecosystem stability. Through the non-profit Earth Rover Program, we demonstrate how 1000-Hz seismology can infer key soil properties and scale towards providing precision interventions needed for the transition to sustainable agriculture.