Automated change detection with annual update for mapping urban dynamics

The new EvoLand candidate prototype, named “Automated Land Use Mapping of Urban Dynamics,” aims to simplify some of the processes used to study urban changes by enabling the automatic detection and characterisation of modifications in urban landscapes annually. This approach aims to not only enhance the frequency and accuracy of urban monitoring, but also, it will facilitate timely urban planning and management.

In the EO business, there are numerous products available for use. So, the question arises: why would this specific product play a different role in the urban planning monitoring? Let us briefly list the four key advancements of this prototype over current urban dynamics mapping products:

• Reduced Time Step: Current products map changes over periods of 3 to 6 years. The prototype will aim to reduce this period to just one year, allowing for more immediate detection of changes.
• Automated Change Detection: Unlike existing methods that may require manual intervention, the prototype will automatically detect changes.
• Improved Delineation: The prototype will bring automated improvements of delineations of the detected change polygons.
• Automated Class Characterisation: It will also automatically characterise the classes of detected changes, providing detailed insights into the nature of these changes.

The above figure illustrates the same building shown as a comparison 2020 and 2021, Camargue, France

Certainly, achieving the accurate development of an EO product requires establishing a clear methodology that incorporates data from diverse sources. This is precisely the case for this prototype, where one of the primary data sources is Sentinel-2 Level 3A (S2 L3A), which provides monthly synthesis without cloud cover, ensuring high-quality and consistent data. This allows for uninterrupted monitoring and analysis of urban changes. Additionally, Super-Resolved Sentinel-2 Level 2A (S2 L2A) was tested, but it was found not to detect more objects compared to S2 L3A, reinforcing the latter’s suitability for the prototype’s objectives. The added value of using these data sources lies in their ability to provide frequent, clear, and detailed imagery. This is crucial for accurate and timely change detection, enabling urban planners and researchers to make informed decisions. Moreover, the integration of these data sources supports automated processes in detecting and characterising changes, thereby improving the efficiency and effectiveness of urban dynamics monitoring.

In addition to using Earth Observation (EO) data, the prototype will also integrate various in-situ data sources to ensure the product reflects reality as accurately as possible. Among these, three main ones are listed below:

• Cadastral data, which will be employed to improve geometric accuracy of limits detected automatically, although its coverage may be limited depending on the country.
• The CLC+ 2018 dataset, crucial for training the characterisation model, ensuring precise and accurate change characterization.
• Finally, WorldCover 2021, which focuses on urban areas and is available globally, will provide a valuable reference for the prototype.

By combining the data sources explained in the two previous paragraphs, the EvoLand prototype aims to deliver a robust tool for urban planners and researchers. This integration will facilitate better decision-making for sustainable urban development, helping to monitor and manage urban landscapes with unprecedented accuracy and efficiency. As the prototype continues to evolve, it holds great promise for shaping the future of urban monitoring and planning.

In-line image credit: CLS

Cover image credit: European Union, Copernicus Sentinel-2 imagery

Would you be interested to follow the results of EvoLand and the development of this candidate prototype? Sign up for our newsletter (here below or via this form), follow us on LinkedIn, Mastodon or X (Twitter). 

Questions about the article or the project? E-mail us at contact@evo-land.eu.  

== 

This article is part of a series providing more details on all 11 EvoLand candidate prototypes.

Previous article in the series: 

Innovating land cover monitoring by complementing annual land cover maps with near real-time observations