It is estimated that 50 percent of the injuries caused when a high magnitude earthquake affects an area are because of falls or falling hazards. This means that most of these injuries could have been prevented if the population had a few seconds of warning to take cover. Grillo, a social impact enterprise focused on seismology, created a low-cost solution using AWS that senses earthquakes and alerts the population in real time about the dangers in the area.
Earthquakes can happen at any time, and there are two actions cities can take to mitigate the damages. First is structural refitting, that is, building structures that can resist earthquakes. This solution doesn’t apply to many areas because they require big investments. The second solution is to send an alert to the affected population before the shaking reaches them. Ten to sixty seconds can be enough time for people to take action by getting out of a building, taking cover, or turning off a dangerous machine.
Earthquake Early Warning (EEW) systems provide rapid detection of earthquakes and alert people at risk. However, because of the hardware, infrastructure, and technology involved, traditional EEW systems can cost hundreds of millions of US dollars to deploy—a cost too high for most countries.
Andrés Meira was living in Haiti during the 2010 earthquake that claimed over 100,000 human lives and left many people homeless and injured. It is estimated that the earthquake affected three million people. He later moved to Mexico, where in 2017, he experienced another high-magnitude earthquake. As a result, Andrés founded Grillo to develop an accessible EEW system, and its solution has been operating successfully in Mexico since 2017.
Grillo developed a low-cost EEW system using sensors and cloud computing. This system uses off-the-shelf sensors that are placed in buildings near seismically active zones. Grillo sensors cost approximately $300 USD, compared to the traditional seismometers that cost around $10,000 USD. Because of these inexpensive sensors, Grillo can offer a higher density of sensors, which reduces the time needed to issue an alert and gives people more time for action. This benefits the population because higher density increases the accuracy of the location detection, reduces false positives, and reduces times to alert.
Grillo’s sensors transmit data to the cloud as the shaking is happening. The cloud platform Grillo built on AWS uses machine learning models that can determine and alert in almost real time, with an average latency of 2 to 3 seconds if an earthquake is happening, depending on the data sent by the different sensors. When the cloud platform detects earthquake risk, it sends alerts to nearby populations via a native phone application, IoT loudspeakers placed in populated areas, or by SMS.
In addition, Grillo founded the OpenEEW initiative to enable EEW systems for millions of people who live in areas with earthquake risks. This features the sensor hardware schematics, firmware, dashboard, and other elements of the system as open source, with a permissive license for anyone to use freely.
In this initiative, they also share on the Registry of Open Data on AWS all the data produced from the sensors deployed in Mexico, Chile, Puerto Rico, and Costa Rica for different organizations to learn from it and also to train machine learning models.
Grillo in Haiti
Haiti ranks among the countries with the highest seismic risk in the world. Large magnitude earthquakes hit Haiti in 2020 and 2021. Currently, Grillo is working to establish their low-cost EEW system in southern Haiti, where most of the large seismic events in the past decade have occurred. This area is home to over three million people.
Over the course of 2021, Grillo installed over 100 sensors in Puerto Rico. And during 2022, they have focused on deploying sensors in the nationwide cell tower network of Haiti. Also during this year, they will calibrate the machine learning models with data from the new sensors in order to correctly predict when there is earthquake risk. Finally, they will develop an SMS alert system with Digicel, a local telecommunication company. Grillo plans to complete the deployment of the south Haiti EEW system by the end of 2022.
Grillo partnered with the AWS Disaster Response team to achieve their goals. AWS helped Grillo to migrate their initial system to AWS and provided expert technical assistance on how to use Amazon SageMaker and AWS IoT services. AWS also provided credits to run the system and financial help to build the sensors.
Источник: AWS News BlogAmazon Machine Learning Amazon SageMaker Internet of Things