Development of automatic global land use/land cover classification methods for selected thematic classes

Project title: Development of automatic global land use/land cover classification methods for selected thematic classes
Project Manager:  dr inż. Artur Nowakowski
Implementing institution: The Faculty of Geodesy and Cartography of the Warsaw University of Technology (WUT) in collaboration with two other faculties of WUT: the Faculty of Building Services, Hydro and Environmental Engineering (project executors: prof. dr hab. inż. Jarosław Zawadzki i dr inż. Karol Przeździecki), and the Faculty of Electronics and Information Technology
Source of funding: the BEYOND POB II funding within the Excellence Initiative – Research University Programme

Project summary:

Thanks to the development of satellite monitoring programs and the availability of new classification techniques it is possible to provide more and more accurate and reliable information on land cover. This information is crucial in many applications, e.g. , in the field of environmental protection, sustainable development, management of natural resources or analyzes of trends in urbanization.

Since 2016, thanks to the European Space Agency (ESA), image data from the Sentinel-2 mission with an unprecedented terrain spatial resolution of 10 m has been available to the public and free of charge. This data consists of 13 spectral bands and covers all land areas with an average revisit time of 5 days. Thanks to this data, it was possible to develop the first-ever global land cover map with a resolution of 10 m, called WorldCover, which was published this month (10/2021). This map covers 11 basic land cover classes, providing unique information allowing the analysis of structures not visible on earlier maps with a lower spatial resolution. Taking into account the current state of research, it is not possible to further increase the spatial resolution of the land cover classification without losing the reliability of the results, but it is still possible and desirable to refine the thematic map.

In this research, it is proposed to develop methods for producing two thematic classifications with a spatial resolution of 10 m from the Sentinel-2 images. Firstly, the class ‘built-up’ available in the WorldCover system will be divided into buildings and road infrastructure, secondly, the class of ’Tree cover’, also from the WorldCover system will be divided into deciduous and coniferous tree cover. The uncertainty of the classification of individual pixels will also be calculated.