Science / Regional research

Research rooted
in the Azores.

How Earth observation, environmental data and machine learning came together in SOILRISK’s regional development.

Discover the research
Research setting
The Azores
Approach
Machine learning
Original model inputs
20+ conditioning factors

01 / The question behind the research

Where might
the next landslide
happen?

In the Azores, prolonged and heavy rainfall can trigger landslides, putting people, homes and infrastructure at risk.

The original SOILRISK case study set out to address a practical challenge: identifying where a landslide might occur in the coming days, with enough spatial detail to support preparedness and emergency planning.

The research brought together information about the terrain and changing environmental conditions. The aim was to move from understanding which slopes are susceptible to understanding how risk may develop over time.

02 / The approach

Many factors.
One connected view.

A proprietary early-warning algorithm, built around a machine-learning approach.

The original SOILRISK approach describes more than 20 conditioning factors. Land cover, geology, vegetation, soil moisture, wind, precipitation and ground movement contribute different pieces of the picture.

Combining these factors allows the model to consider both the characteristics of a slope and the conditions that change around it.

03 / The information behind it

From satellites
to local conditions.

Four data sources described in the original technical overview.

01 / Ground deformation

Following the
movement of slopes.

Satellite radar interferometry (InSAR) helps identify ground-surface movement. These observations add information about how slope activity changes across space and time and can help identify isolated instabilities.

02 / Vegetation

Understanding
what covers the ground.

Vegetation, soil characteristics and slope help identify susceptible areas. The research used Sentinel-2 imagery to derive NDVI, an index of vegetation condition.

03 / Soil moisture

Tracking changes
in moisture.

The original method used Sentinel-1 radar data and change detection to estimate soil moisture. Land-cover maps were used to mask areas where those estimates were less reliable.

04 / Weather

Adding the conditions
that keep changing.

Precipitation and wind feed the algorithm as dynamic variables. Where available, hourly local weather-station observations provide context, alongside weather forecast services.

04 / From research to the app

A regional foundation.
A broader perspective.

The Azores work is part of the scientific foundation behind Eyecon’s own SOILRISK models.

The app brings regional SOILRISK models together with global LHASA information, weather and compatible local measurements. Each source has its own coverage, resolution and forecast horizon.

Check the model and information available for your location in the app. The regional research described here should not be read as a worldwide specification.

Explore the science behind SOILRISK

Continue the conversation

Want to go deeper?

Talk to the team about the Azores work, methodology or regional research.

Talk about the research

Based on the Azores case study and technical overview published on the original SOILRISK website ↗.

Photography, video & illustration

Images &
video.

The imagery illustrates environmental settings. It does not show events predicted by SOILRISK.