Fractional Vegetation Cover Mapping - UAS RGB and Multispectral Imagery, CNN algorithms, Semi-Arid Australian Ecosystems Coverage
Authors/Creators
-
1.
University of Tasmania, School of Geography, Planning and Spatial Sciences
-
2.
The University of Adelaide, School of Biological Sciences
-
3.
Deutsches Zentrum für Luft- und Raumfahrt e. V. (DLR)
-
4.
Consejo Nacional de Investigaciones Científicas y Técnicas
- 5. CONICET Mendoza
-
6.
University of Tasmania
Description
This dataset is produced using the RGB and multispectral UAS/drone data at 1 centimetre-scale.
Data description
Format: GeoTIFF and ESRI Shapefile
Projection: EPSG:7854 - GDA2020 / MGA zone 54
Resolution: 1 cm
Size: 10.3 GB
Contact: Laura Sotomayor
Email: laura.sotomayor@utas.edu.au
Products
This dataset delivers the inputs to build the reference/labelling data in order to run the U-net CNN-based segmentation models. There are three main folders related to three stratum vegetation covers: low, medium and dense vegetation. Each contain the mask and predictors 5 bands at 1 cm scale use as rasters for the U-net CNN model. The predictors also contains the RGB data, multispectral data (5 bands) and composite colour.
1 cm FVC multiclass segmentation
Float32 GeoTIFF product showing FVC at 1 cm resolution
U-net CNN-based segmentation models using a structured nomenclature with five classes and identifiers (IDs):
- bare ground (BE = 0)
- non-photosynthetic vegetation (NPV = 1)
- photosynthetic vegetation (PV = 2)
- shadow (SI = 3)
- water (WI = 4).
1 cm FVC predictors
Float32 GeoTIFF product showing FVC at 1 cm resolution with 5 bands provided byt the MicaSense RedEdge-MX sensor (’Blue-475’, ’Green-560’, ’Red-668, ’Red Edge-717’, and ’NIR-842’).
Method
The full CNN workflow for data derivation and modelling is described in Sotomayor, L.N., et al. (2025). Mapping fractional vegetation cover in UAS RGB and multispectral imagery in semi-arid Australian ecosystems using CNN-based semantic segmentation.
Landscape Ecology, 40, 169. https://doi.org/10.1007/s10980-025-02193-y
Authors:
Laura N. Sotomayor (1*) | School of Geography, Planning, and Spatial Sciences, University of Tasmania, Australia | [laura.sotomayor@utas.edu.au]
Arko Lucieer (1) | School of Geography, Planning, and Spatial Sciences, University of Tasmania, Australia | [arko.lucieer@utas.edu.au]
Darren Turner (1) | School of Geography, Planning, and Spatial Sciences, University of Tasmania, Australia | [darren.turner@utas.edu.au]
Megan Lewis (2) | School of Biological Sciences, University of Adelaide, Australia | [megan.lewis@adelaide.edu.au]
Teja Kattenborn (3) | Sensor-based Geoinformatics (geosense), University of Freiburg, Germany | [teja.kattenborn@geosense.uni-freiburg.de]
Acknowlegments:
RGB (1 cm) and Multispectral (5 cm) orthomosaics processing: TERN Landscapes, TERN Surveillance Monitoring, Stenson, M., Sparrow, B., & Lucieer, A. (2022). Drone RGB and Multispectral Imagery from TERN plots across Australia. Version 1. Terrestrial Ecosystem Research Network. Dataset.
Contribution for reference/labelling dataset process:
Prof. Megan Lewis (School of Biological Sciences, University of Adelaide), Dr Krishna Lamsal (School of Geography, Planning, and Spatial Sciences, UTAS), Sophia Hoyer (School of Geography, Planning, and Spatial Sciences, UTAS) and Molly Marshall (School of Geography, Planning, and Spatial Sciences, UTAS).