August 19, 2021 Dataset Restricted Access

Blickensdörfer, Lukas; Schwieder, Marcel; Pflugmacher, Dirk; Nendel, Claas; Erasmi, Stefan; Hostert, Patrick

Dublin Core Export

<?xml version='1.0' encoding='utf-8'?>
<oai_dc:dc xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ http://www.openarchives.org/OAI/2.0/oai_dc.xsd">
  <dc:creator>Blickensdörfer, Lukas</dc:creator>
  <dc:creator>Schwieder, Marcel</dc:creator>
  <dc:creator>Pflugmacher, Dirk</dc:creator>
  <dc:creator>Nendel, Claas</dc:creator>
  <dc:creator>Erasmi, Stefan</dc:creator>
  <dc:creator>Hostert, Patrick</dc:creator>
  <dc:date>2021-08-19</dc:date>
  <dc:description>Detailed maps of agricultural landscapes are a valuable data source for manifold applications, such as environmental modelling, biodiversity monitoring or the support of agricultural statistics. Satellites from the European Copernicus program, especially, Sentinel-1 and Sentinel-2, as well as the Landsat missions operated by NASA/USGS, acquire data with a spatial resolution (10 m to 30 m) that is sufficient to identify field structures in complex agricultural landscapes. Time series of combined Sentinel-2 and Landsat data facilitate to differentiate crop types with a high thematic detail based on differences in land surface phenology. However, large data gaps due to frequent cloud cover may hamper such classification approaches.  

We thus combined dense interpolated times series of Sentinel-2A/B and Landsat data with monthly composites of Sentinel-1 backscatter data to overcome periods with high cloud contamination. To further account for regional variations along the agroecological gradient within Germany, we additionally included a broad set of spatially explicit environmental data in a random forest classification model.  

All optical satellite data were downloaded, pre-processed and structured in an analysis-ready data (ARD) cube using the open-source software FORCE - Framework for Operational Radiometric Correction for Environmental monitoring (Frantz, D., 2019; https://force-eo.readthedocs.io/en/latest/ last accessed: 19. August 2021), before environmental and SAR data were included in the ARD cube.  

For each year (2017, 2018 and 2019) we trained an individual random forest model with 24 agricultural classes. Each model was independently validated with area adjusted overall accuracies of 80% (2017), 79% (2018), and 78% (2019). Further details regarding the data and methods used as well as class wise accuracies can be found in Blickensdörfer et al. (2022). 

The final models were applied to areas in Germany that were defined as agricultural land in ATKIS DLM 2018 (Geobasisdaten: © GeoBasis-DE / BKG (2018)). Post-processing of the final maps included applying a sieve filter, the exclusion of classes other than grasslands and small woody features above 900 m (based on the Digital Elevation Model for Germany BKG (2015)) and the exclusion of grapevine/hops areas that were not labelled as the respective permanent crop in ATKIS DLM (labelled as other agricultural areas in the final map). 

The maps are provided as GeoTiff files together with a QGIS legend file for visualization. 

Class catalogue:

10     Grassland
31     Winter wheat
32     Winter rye
33     Winter barley
34     Other winter cereal
41     Spring barley
42     Spring oat
43     Other spring cereal
50     Winter rapeseed
60     Legume
70     Sunflower
80     Sugar beet
91     Maize
92     Maize (grain)
100    Potato
110    Grapevine
120    Strawberry
130    Asparagus
140    Onion
150    Hops
160    Orchard
181    Carrot
182    Other vegetables
555    Small woody features
999    Other agricultural areas

 

Blickensdörfer, L., Schwieder, M., Pflugmacher, D., Nendel, C., Erasmi, S., &amp; Hostert, P. (2022). Mapping of crop types and crop sequences with combined time series of Sentinel-1, Sentinel-2 and Landsat 8 data for Germany. Remote Sensing of Environment, 269, 112831

BKG, Bundesamt für Kartographie und Geodäsie (2015). Digitales Geländemodell Gitterweite 10 m. DGM10. https://sg.geodatenzentrum.de/web_public/gdz/dokumentation/deu/dgm10.pdf (last accessed: 19. August 2021). 

BKG, Bundesamt für Kartographie und Geodäsie (2018). Digitales Basis-Landschaftsmodell. 
https://sg.geodatenzentrum.de/web_public/gdz/dokumentation/deu/basis-dlm.pdf (last accessed: 19. August 2021).

Frantz, D. (2019). FORCE—Landsat + Sentinel-2 Analysis Ready Data and Beyond. Remote Sensing, 11, 1124.

 

National-scale crop type maps for Germany © 2021 by Blickensdörfer, Lukas; Schwieder, Marcel; Pflugmacher, Dirk; Nendel, Claas; Erasmi, Stefan; Hostert, Patrick is licensed under CC BY 4.0. </dc:description>
  <dc:identifier>https://zenodo.org/record/5153047</dc:identifier>
  <dc:identifier>10.5281/zenodo.5153047</dc:identifier>
  <dc:identifier>oai:zenodo.org:5153047</dc:identifier>
  <dc:language>eng</dc:language>
  <dc:relation>doi:10.5281/zenodo.5153046</dc:relation>
  <dc:rights>info:eu-repo/semantics/restrictedAccess</dc:rights>
  <dc:subject>Agriculture</dc:subject>
  <dc:subject>Remote sensing</dc:subject>
  <dc:subject>Land cover</dc:subject>
  <dc:subject>Machine learning</dc:subject>
  <dc:subject>Maps</dc:subject>
  <dc:subject>crop type</dc:subject>
  <dc:subject>Germany</dc:subject>
  <dc:title>National-scale crop type maps for Germany from combined time series of Sentinel-1, Sentinel-2 and Landsat 8 data (2017, 2018 and 2019)</dc:title>
  <dc:type>info:eu-repo/semantics/other</dc:type>
  <dc:type>dataset</dc:type>
</oai_dc:dc>