Change detection technique comparison in long-term wetland monitoring: datasets and maps of the Poitevin Marsh (France)

For a full description of the methodology and results, please see the following article:

# Datasets

Points datasets are projected in WGS84 (EPSG:4326), and are provided in the open source GeoPackage format.

The first dataset (Dataset_1.gpkg) contains training and validation points for random forest classification of EUNIS habitats in the Poitevin Marsh. This dataset consists of 3360 training and 840 validation points (total:  4200).
Fields description:

• "ID": unique identifier
• "CLASS": EUNIS first level habitat type, classified as following:
  
  • 1: EUNIS habitat A
  • 2: EUNIS habitat B
  • 3: EUNIS habitat C1J5
  • 4: EUNIS habitat C3
  • 5: EUNIS habitat E
  • 6: EUNIS habitat G
  • 7: EUNIS habitat I
  • 8: EUNIS habitat J
• "DATE": Date associated with EUNIS habitat sample
• "LON": Point longitude in decimal degrees
• "LAT": Point latitude in decimal degrees
• "TYPE": Either training ("train") or validation ("test") sample

The second dataset (Dataset_2.gpkg) contains points for the Olofsson correction method. This dataset consists of 326 points where the change classes are classified as following: -10 (wetland loss), 10 (wetland gain), 100 (stable existing wetland), and 200 (stable damaged wetland).
Fields description:

• "ID": unique identifier
• "LON": Point longitude in decimal degrees
• "LAT": Point latitude in decimal degrees
• "REFERENCE": Change class reference
• "CCDC": Change class obtained from the Continuous Change Detection and Classification approach
• "PCCD": Change class obtained from the Post-Classification Change Detection approach

Supplementary layout files (Dataset_1.qml and Dataset_2.qml) support formatting of the points in QGIS software.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

# EUNIS habitat

Maps are projected in WGS84 (EPSG:4326), and are provided in the GeoTiff format at 30m of spatial resolution. 

Habitat maps are given for the two approaches in years 1984 and 2022:

• CCDC: Continuous Change Detection and Classification (CCDC_HABITAT_1984.tif and CCDC_HABITAT_2022.tif)
• PCCD: Traditional post-classification approach (PCCD_HABITAT_1984.tif and PCCD_HABITAT_2022.tif)

Supplementary layout files (CCDC_HABITAT_1984.qml, CCDC_HABITAT_2022.qml, PCCD_HABITAT_1984.qml, PCCD_HABITAT_2022.qml) support formatting of raster layers in QGIS software.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

# Change detection during the 1984-2022 period

Maps are projected in WGS84 (EPSG:4326), and are provided in the GeoTiff format at 30m of spatial resolution. Raster values follow the classification scheme used in Dataset_2.

Change detection maps are given for the two approaches:

• CCDC: Continuous Change Detection and Classification (CCDC_CHANGE_1984_2022.tif)
• PCCD: Traditional post-classification approach (PCCD_CHANGE_1984_2022.tif)

Supplementary layer files (CCDC_CHANGE_1984_2022.qml and PCCD_CHANGE_1984_2022.qml) support formatting of the raster layers in QGIS software.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

# GEE repository

To get direct access to GEE scripts and assets, please follow those two links:

https://code.earthengine.google.com/?accept_repo=users/demarquetquentin/CCDC_Poitevin

https://code.earthengine.google.com/?asset=projects/ee-quen-dem/assets/CCDC_Poitevin