Animal Alerts: Dogs Will Aid AI in Predicting Earthquakes

Let's take the catastrophe in Haiti, where in 2010, over 220,000 people perished, around 300,000 were injured, and about 3 million (one-third of the population) were left without shelter. The earthquake disrupted water and power supplies, and all hospitals were destroyed. A 7.0 magnitude quake nearly dissolved the very governance of the Republic of Haiti, briefly shifting control to street gangs. It took the incredible efforts of the international community to restore order.   

Sadly, no methods have yet been developed to counteract the movement of tectonic plates, and it seems unlikely that they ever will be. Seismologists have come to understand the phenomenon well and have pinpointed regions at elevated risk. Yet, the current early warning systems that detect subterranean shifts provide authorities and rescue teams with only a few hours—or even minutes—of advance notice. 

Moreover, humanity has never before possessed such a potent tool for prediction as artificial intelligence. For instance, AI is already used for creating long-term weather forecasts, such as on the Earth-2 platform, which models weather conditions in Taiwan using Nvidia's AI model CorrDiff. However, gathering and interpreting data for weather forecasts—like temperature, wind, solar activity, and ocean currents—is much simpler than accessing information from deep beneath the surface where earthquakes begin.

Historically, humans have been adept at observing unusual animal and insect behavior well before seismic tremors. The earliest documented observations of this phenomenon date back to Ancient Greece in 373 BC, where our ancestors noted the anxious behaviors of domestic animals and the mass migration of wild animals days before an earthquake.   

In modern times, especially in countries like China and Japan, sudden changes in animal behaviors have garnered significant scientific interest, with this data now being incorporated into earthquake early warning systems. A notable instance occurred in 1975 in Haicheng, China, a city in a seismically active region. The local authorities evacuated 90,000 residents following widespread unusual behaviors in dogs, cats, and other animals. Just a few hours later, a 7.3 magnitude earthquake devastated Haicheng.  

This event led to intensive research into animal behavior. The theory that animals perceive high-frequency sound vibrations from tectonic shifts was inadequate because such vibrations are common in nature, yet animals seem to uniquely identify those that signify danger. Thus, installing simple high-frequency sensors was not a sufficient solution.  

Artificial intelligence has emerged as the most effective tool for analyzing large datasets of such phenomena.   

The American company PetPace, which specializes in AI-powered gadgets for pets, has launched the Animal Alerts project. The centerpiece of this initiative is a biometric dog collar that helps AI predict earthquakes. This device uses non-invasive, electrode-free sensors to monitor various biological indicators in animals, such as stress levels, pulse, heart rate, temperature, and breathing.

Biometric Dog Collar Source: Animal Alerts

All this data is automatically sent to an AI platform that gauges the stress levels the dogs are experiencing. The AI then correlates these readings with seismic activity data and looks for patterns linking dog behavior to changes in weather, wind strength, occurrences of snow avalanches, loads on regional infrastructure, and numerous other factors. The aim is to pinpoint peaks in the animals’ physiological responses that are specifically linked to subterranean tremors.  

The choice of dogs for this experiment was deliberate. They are ubiquitous companions to humans across all continents and living situations. Given that around 360,000 earthquakes are recorded globally each year, ranging from minor tremors to catastrophic disasters, the stream of data from these AI-enabled collars is expected to be highly relevant.  

The initiative is currently being piloted in Lima, the capital of Peru, a country prone to seismic activities. Since 1900, Lima has endured three earthquakes with magnitudes exceeding 8.0. Just since the beginning of 2024, there have already been 53 earthquakes in Lima with magnitudes of up to 5.2. Therefore, if significant numbers of animals in different city areas exhibit similar stress patterns during such events, it could be an indicator of an impending natural disaster.   

To maximize the visibility of forecasts based on pet behavior, Animal Alerts' media partners have agreed to disseminate geo-targeted alerts through devices owned by residents, online platforms, digital billboards, and local radio broadcasts. To involve as many pets as possible in this monitoring, the project team is also conducting various educational activities.   

There is an urgent need for innovative approaches to warn the public before an earthquake. Especially since Lima's infrastructure is extremely vulnerable to seismic activity. And so, a timely alert from Animal Alerts could help save countless lives

says Carlos G. Brown, Disaster Risk Management Expert, TECHO Peru.

While it's uncertain if the experiment will ultimately be successful, this straightforward approach to equipping humanity with a reliable early earthquake warning system certainly warranted exploration.