News
Big Earth Observation Analytics: Rapid detection of illegal tropical deforestation
Earth observation is a rapidly developing technology field. New satellites provide increasingly more and better data, which makes close monitoring of areas possible. These advancements provide multiple uses, one of these is the detection of illegal deforestation.
Big data challenge
New satellites monitor the earth surface with a higher resolution than before. The new Sentinel-1 radar (global coverage since 2014) even overcomes the biggest monitoring limitation: clouds. This method of earth observation provides more options for close monitoring, however it also increases the amount of data produced. The amount of satellite data is doubling every two years, which requires an increasing amount of storage space. A visual representation of these large amounts of data can be seen in Figure 1. Sentinel-1 satellite data is available for free from ESA.
Satellite data is becoming too large and diverse to store and analyse locally, therefore bringing the users and their software to the data rather than vice versa becomes inevitable. Cloud services provide solutions to these storage and analysis issues faced by users, but choosing the right one remains a challenging task. Different cloud services are available and they can be divided into three types:
- Commercial data centers: Google Earth Engine, Amazon and others.
- Publicly-funded data centers: Copernicus Data and Information Access Service.
- Cooperative data centers: EODC, Science Centers.
Cloud services differ in aspects such as performance, costs and legal conditions such as protection of data, software and intellectual property. Some commercial parties offer free services, but these may not provide the same ownership protection as others. Furthermore, public investments are made to provide cloud services and several cooperatives are developing services.
Detecting illegal deforestation
Regular earth monitoring with a high spatial detail of 10 metres is possible using Sentinel-1 Radar based sensors, but also historical – less accurate – data proves to be very valuable in change detection. The images captured using these satellites are used in machine learning methods in order to detect changes in the landscape. In this case, it involves detecting forest clearings appearing over time. This method has been tested on an area with known forest harvesting activities, whereby new forest clearings can be detected. The method can subsequently be used to detect (illegal) deforestation.