• Search
  • LinkedIn
  • Instagram

Australia’s shifting sands

Words by Hannah Bird
4 September 2023

Despite Australia’s relative tectonic stability, earthquakes do still occur together with their associated hazards, such as liquefaction (when loosely consolidated, water-logged sediment loses strength due to intense ground shaking). Four such events have been documented in recorded history in Australia, the most recent being in Perth in 1979.

Ratiranjan Jena at the University of Technology Sydney, Australia, and colleagues combine Probabilistic Seismic Hazard Assessments with artificial intelligence (specifically, deep learning, which includes statistical and predictive modelling) to produce national-scale liquefaction hazard maps for Australia. The researchers use three different approaches to create maps that are based on numerical modelling of a magnitude 6 earthquake and account for a variety of conditional factors, such as regional variations in the soils’ clay and water content, thickness and bulk density, as well as shear-wave velocity and ground slope. The maps are used to estimate a liquefaction potential index (LPI), which is subsequently used to estimate the liquefaction hazard index (LHI) – a guide to the potential damage anticipated from each progressive level of the LPI.

Depending on the approach used, the results show that about 45% of Australia (largely concentrated in the north and east of the country) is considered non-hazardous in the event of an earthquake of magnitude 6. Western and central Australia are most susceptible to liquefaction and the associated hazards, though there are pockets to the south and north of the country that are also predicted to have a liquefaction risk. Up to 19% of the country – notably those regions with silty and sandy alluvial sediments and a shallow groundwater table – falls into a category defined by the authors as highly hazardous. The team stress that a country-scale investigation has limits – high-resolution geophysical and geotechnical data are critical for local-scale hazard assessments.

The project aims to encourage the use of deep learning software to inform infrastructure investment for the future, as well as planning for post-event reconstruction priorities.

Hannah Bird


Geosci. Front. 14 (1), 101460; doi.org/10.1016/j.gsf.2022.101460

Related articles