SHA - Sentinels for Habitats

You are here

Managing and monitoring nature reserves currently involves many hours of field work and manual labour. Additionally, the results can be inconsistent between different ecologists performing the surveys, and budgets are a big constraint on the frequency and comprehensiveness of which areas can be monitored. uses Artificial Intelligence (AI) combined with remote sensing data (e.g. aerial photography) to address these issues. Our solution automatizes the process of boundary delimitation and flora classification. This helps ecologists to improve their analyses in the field, and work more efficiently. Identifying different types of nature with remote sensing required data across multiple seasons in order to distinguish different growth onsets and other cyclical characteristics of different valuable types of forest, grasslands, peatlands, wetlands and swamps. In Sentinels for Habitats, Sentinel-2 data is leveraged to improve the extent to which our service can distinguish between different types of nature. This results in significantly more accurate mapping and classification of nature reserves, because similarly to humans, the more relevant information provided, the better the performance.

Users and their needs is a collaboration between Sobolt and the Dutch region Overijssel. Its main target users are land owners  and ecologists tasked with assessing valuable natural habitats. Provinces are responsible for reporting the quality and quantity of different types of nature in various areas. These tasks are then delegated to ecologists. The challenges and needs of ecologists are summarized as follows:


  • Draw borders in nature areas
  • Classify the type of nature in accordance with (inter)national standards
  • Improve initial map with field measurements


  • Physically demanding fieldwork
  • Boundary delimitations can be subjective
  • Drawing maps is time consuming
  • Hard to survey low access areas


  • Repeatable and consistent reporting
  • Efficient (time and physical involvement)
  • Higher monitoring frequency

The project’s current country target is the Netherlands; however, the intention is to eventually scale services to the rest of Europe.

Service/ system concept

The service is an online tool that facilitates the mapping of nature areas. It facilitates producing accurate maps of different vegetation types. This allows the user to compute many types of analyses, assessing habitats in natural reserves, and informing nature management decisions. The architecture of the system from raw data to usable maps on the web application are described in the figure below.

Users  engage with by logging into the web application, which gives them access to various functionalities to facilitate the workflow of nature area mapping. Ecologists are assisted with predictions derived from the AI algorithm using remote sensing data, including Sentinel information.

This allows users to then view the collected data through maps, interactively edit these maps, add notes, assign actions to colleagues, and easily export results for use in other visualization applications.

Space Added Value uses Sentinel data to improve automated nature mapping. The spectral bands of Sentinel-2 are designed for applications in agriculture and implicitly contain a lot of information about vegetation type and its health. Without this information, Sentinels for Habitat would not be able to generate inferences about flora type and condition. Additionally, compared to aerial images, Sentinel-2 has a higher temporal frequency for these areas. This provides seasonal information that is relevant to habitat types. This information allows us to better understand crop and leaf development across the seasons. This was previously not possible when using only aerial imaging. The added value of using these constellations can be observed in the figure below. The algorithm is able to automatically take in a raw image and classify at the pixel level the vegetation type. This processing allows ecologists to cover large areas in a more efficient, consistent and affordable manner.

Current Status



In the project, the web application Naturaai has been tailored to assist ecologists with monitoring tasks. Through, they have access to both aerial and satellite data when performing nature assessment tasks.

The Artificial Intelligence (AI) algorithms in have provided them with suggestions on the type of nature they are inspecting.


As a part of this project, they provide the regional government of Overijssel with feedback on their nature management plans in and current assessment of nature types. This feedback is used to further improve the AI algorithms.

About to start

By not only using aerial data but also Sentinel-2 data, predictions of will take a leap forward in assessing nature areas during the remainder of Sentinels for Habitats. Field measurements will be used to quantify improvements over previous results.

Prime Contractor(s)

Status Date

Updated: 24 June 2020 - Created: 24 June 2020