Eyes in the Sky: Geohazard Monitoring with Terradue and Elasticsearch

Landslides, earthquakes, ground settlement, and floods. Earth is an unquiet planet, and sometimes that’s trouble for the life that inhabits it, including humans. But it’s hard to predict natural disasters. The best way to protect life and infrastructure from geohazards is to prepare in advance. Post-disaster studies show that mitigation efforts pay off, but scientists need the best data they can get to analyze risk. One place they get this good data is from Terradue.

Terradue is a European Space Agency (ESA) spin-off company that specializes in massive processing of Earth science data using Elastic. Emmanuel Mathot, a technical leader at Terradue, wrote a blog post last May about their Elasticsearch-powered query engine that helps monitor geohazards from space. Since last year, Terradue has done even more exciting work that Mathot discussed in his March 8, 2017 presentation at Elastic{ON}.

Earth observation satellites orbit the planet 24/7 taking fantastically precise radar images of its surface. “This [satellite] instrument is able to deliver data we use to track movements down to a few millimeters,” said Emmanuel Mathot. It can even or measure how fast individual buildings settle into the ground, like San Francisco’s famously sinking Millennium Tower.

As the satellites orbit, they take pictures of the planet’s entire surface once every six days. All the images and metadata about satellite position, velocity, and orbital track get indexed into Terradue’s Elasticsearch query engine, Mathot explained. This engine provides the best data to monitor ground movement along a fault or near a volcano, or to map flood plains, among other uses. It does this by selecting the most appropriate images to compare to each other by minimizing some parameters within the metadata.

One good illustration of this involves earthquakes, specifically the 6.2 magnitude earthquake that struck central Italy in August of 2016. To create the most accurate picture of earthquake-related ground movement, researchers compared the satellite’s first post-event image of the disaster area with the best image taken in the months before the earthquake. There are typically around 50 of these pre-event images to pick from and which one is “best” depends on many factors.

To do so, Terradue uses Elasticsearch to compare the satellite’s position when it took the post-event picture to its position when it took each the pre-event image. It also analyzes the time span between when the images were taken, the satellite’s orbital track and velocity, and other metadata. The query engine compares all of the possible images and selects the best one, and it can do it in about ten seconds.

“Without this information in Elasticsearch, in our catalog,” said Mathot, “It’s not possible. The user has to [make] assumptions.” With these accurate pictures, scientists can locate the fault lines that previously went unnoticed. With the earthquake in Italy, the pictures confirmed Italian scientists’ field observations that the ground had moved a whopping 70 centimeters.

The query engine images offer more knowledge than just ground displacement — it can even delineate water movement if you ask the data the right questions. Nigerian scientists have partnered with ESA and Terradue to map flooding along the Niger River in Africa.

“The impact of the floods on the people and the culture is really critical, and understanding and forecasting the intensity and extent is key to reduce the impact of the flooding,” said Mathot. To find patterns, they need an accurate picture of what the river looks like during the rainy and dry seasons over several years.

The trouble is, liquid water doesn’t show up in images the way the ground and buildings do. But scientists figured out that when the satellite images a patch of ground, rougher surfaces bounce back higher intensity microwave backscatter profiles, while smooth surfaces — like water — bounce back low intensity backscatter data, all of which is indexed into Elasticsearch. Using this knowledge they query Elasticsearch to find the best reference images of an area prior to flooding and also track the greatest extent of the flooding. As with the earthquake in Italy, the engine can pick the best image in seconds.

Terradue compiles these very accurate Earth observation images into one collaborative, virtual workspace that also contains over 20 years’ worth of Earth observation data. This workspace is part of a major European Space Agency initiative to provide scientists with a single interface to use this data and interact with an online community of researchers. Terradue uses Elastic software to power three of its seven platforms. This ESA initiative allows scientists across the globe to search and process this wealth of Earth observation data to its full research potential.

To learn more about the details of how Terradue is using the Elastic Stack watch their full presentation.