Understanding earthquake hazard in Malta ‒ new scientific insights

The Maltese islands are not known for devastating earthquakes, but that doesn’t mean they are immune to them. Over the years, smaller tremors have occasionally been felt, and historical records remind us that stronger earthquakes have struck in the past.

The Maltese archipelago is located in a seismically active region, close to several geological features that can generate earthquakes. These include the Hyblean-Malta Plateau, responsible for the devastating 1693 earthquake that affected both Malta and Sicily, the Malta Escarpment, and the Sicily Channel Rift Zone.

Historically, sizeable events originating from the Hellenic Arc, a highly active earthquake zone in Greece, have also affected the Maltese islands, despite their considerable distance.

How seismic waves spread

When an earthquake occurs, the energy it releases doesn’t stay in one place − it spreads outward, similar to how ripples move across a pond when a stone is dropped. However, just as ripples fade, seismic waves also weaken as they travel. This process is known as attenuation.

The first part of this study, undertaken as part of my doctoral thesis, focused on understanding how seismic waves travel through the Central Mediterranean and how they affect the Maltese islands.

Over 21,000 earthquake records from Italy, Sicily, Greece and the Central Mediterranean (Figure 1) were analysed in order to develop Ground Motion Prediction Equations (GMPEs). GMPEs are mathematical models used to estimate how strongly the ground would shake at different distances from an earthquake and  thus help predict the level of shaking Malta would experience from hypothetical earthquakes.

Understanding this is essential for designing earthquake-resistant buildings, updating safety regulations and ensuring Malta is as prepared as possible for future seismic activity.

The role of AI in earthquake studies

To develop GMPEs, two approaches were used: traditional statistical methods and artificial intelligence (AI), specifically through the application of artificial neural networks (ANNs).

The use of ANNs was innovative, as their application in similar studies is still relatively new, with only a handful of published research studies available.

Unlike traditional statistical models, ANNs mimic the way the human brain processes information. They can learn patterns from data and make better predictions, even when dealing with complex and uncertain situations.

When applied to earthquake data, the AI-based models were able to capture subtle trends that traditional methods sometimes miss. They were especially effective at predicting shaking for larger earthquakes and at greater distances.

Seismic hazard analysis for the Maltese islands

The attenuation relationships developed were then used as inputs for a seismic hazard analysis undertaken for the Maltese archipelago.

Seismic hazard refers to the potential for earthquake-related ground shaking in a given region over a certain period. Since earthquakes are unpredictable, seismic hazard analysis estimates the likelihood of different intensities of shaking occurring at a particular location. The results are crucial for designing safer buildings and infrastructure.

The strength of ground shaking during an earthquake is commonly expressed in terms of Peak Ground Acceleration (PGA) and Peak Ground Velocity (PGV). These values help engineers and policymakers set safety standards for buildings, minimising damage and protecting lives.

The Probabilistic Seismic Hazard Analysis (PSHA) approach was used in this study because it provides a more comprehensive understanding of seismic risks by considering uncertainties in earthquake occurrence and ground-shaking intensity since it takes into consideration all possible earthquake sources that may affect the study site. The process involved four main steps:

• Identifying all potential seismic sources that could affect Malta (Figure 2);

• Using statistical models to estimate the frequency of earthquakes of different magnitudes in each source zone;

• Implementing the GMPEs developed earlier to estimate how ground-shaking changes with distance from an earthquake’s epicentre;

• Using a logic tree approach, a statistical method that considers different scenarios and uncertainties, to combine probabilities and uncertainties, improving the reliability of the final results.

Figure 2: Map showing study site and all seismic source zones and faults considered.

Once the analysis was completed, results were presented in the form of hazard curves and hazard maps. Hazard curves describe the probability of ground shaking exceeding a certain level at a given location, while hazard maps show the expected level of ground shaking at different locations across Malta. Both help engineers design earthquake-resistant buildings by identifying areas more prone to stronger shaking.

Hazard maps are shown for different time frames, known as return periods, which help us understand the likelihood of different levels of ground shaking over time. A short return period (e.g., 50 years) represents more frequent, lower-intensity earthquakes that people might experience in their lifetime. A longer return period (e.g., 475 or 975 years) represents rarer but stronger earthquakes that could cause significant damage.

By using different return periods, engineers and policymakers can design buildings and infrastructure to withstand both common and extreme earthquake scenarios, ensuring safety and preparedness for the future.

Main results and observations

One of the main findings of this study was that including distant seismic sources within the Hellenic Arc provided a more accurate assessment compared to past studies, which underestimated hazard by ignoring these sources.

Additionally, the study considered local site effects, which were previously overlooked. It found that areas with soft soils and underlying blue clay (BC) would potentially experience greater shaking than areas with outcropping Globigerina Limestone. This is due to the fact that soft soils and clay layers have natural frequencies that can amplify seismic waves, especially if they match the dominant frequencies of the earthquake. This can lead to prolonged and intensified shaking.

The study confirms that seismic hazard must be considered in urban planning and construction. Understanding seismic hazard is crucial for earthquake preparedness. The study highlights the need for:

• Updated building codes and regulations that consider Malta’s seismic risk;

• Public education on earthquake preparedness to reduce potential damage and injuries. It is also vital to enhance public awareness and overcome complacency regarding earthquake risks;

• Investments in seismic monitoring to improve earthquake detection and hazard assessment.

As more data becomes available and AI models improve, earthquake prediction techniques will become even more accurate. This will ensure that Malta stays prepared for future seismic activity, reducing risks and increasing safety for all.

Acknowledgements

This research work was funded by the Tertiary Education Scholarships Scheme (TESS). The author also thanks Prof. Sebastiano D’Amico, her Ph.D supervisor, for his guidance throughout her studies, Prof. Francesco Panzera for his advice on seismic hazard analysis, as well as Dr Gianluca Valentino and Dr Reuben Farrugia for their expertise and assistance with programming and the application of artificial neural networks (ANNs).