Direkt zum InhaltDirekt zur SucheDirekt zur Navigation
▼ Zielgruppen ▼

Faculty of Mathematics and Natural Sciences - Earth Observation Lab


student group on a boat in BrazilYou plan to focus on Earth Observation?

Here you find all information needed to structure your studies with a focus on Earth Observation: course descriptions, and good practice examples of theses or MAPs from previous courses.



Cropland and forest change mapsTopographic correction matters: Land-cover mapping in the Caucasus

In our study we found that topographic correction matters for land-cover mapping, especially for discriminating forest types in steep terrain, and we examined three decades of Landsat imagery revealing that cropland loss was the most prevalent land-cover change in the Caucasus. Find out more here.

simulated EnMap mosaics "Spring Summer Fall"Three-season simulated EnMAP mosaics for the San Francisco Bay Area, USA

Interested in using hyperspectral EnMAP imagery? Check out our simulated EnMAP mosaics derived from airborne AVIRIS imagery acquired over three dates in 2013. AVIRIS images were simulated to match expected EnMAP characteristics, and secondary geometric and brightness gradient corrections make this dataset analysis ready. Data can be downloaded from GFZ Dataservices.

Vegetation cover trends in Berlin, Germany30-years Landsat time series for relating population, land use and vegetation in a city

Green growth in a city? Our new publication develops a methodology for contrasting vegetation- and population density developments over long time periods using a Landsat time series and spectral unmixing in a densifying city and different land use classes. Read the article here.

figure - different Satellite slices: Google, EnMap, LandsatUnmixing vegetation fractions with simulated EnMAP imagery

We compared hyperspectral and multispectral image analysis for mapping hierarchical vegetation classes in the San Francisco Bay Area, USA using regression based unmxing. We found that EnMAP-based analysis consistently performed better than Landsat across ecoregions and classes, and particularly for more complex vegetation classes. Find out more here.

satellite image classified - figure 11 of publicationLand Cover Fraction Mapping with Synthetically Mixed Training Data from Sentinel-2 STMs

This study maps sub-pixel fractions of built-up surfaces and two vegetation types in Germany and Austria at 10m spatial resolution. We used all available Sentinel-2 imagery from 2017 and 2018 and find that spectral-temporal metrics (STM) can be used to create synthetic training data as an input for regression-based unmixing across large areas. Read and share the article and explore our results here.