Landscape Mapping using Remote Sensing and Neural Networks

Date Published
A convolutional neural network (CNN) approach produced highly accurate vegetation classifications. Hyper-spectral datasets (e.g., AVIRIS) were most useful for our machine learning approaches. Accurate and high-resolution datasets generated using our approach are needed for Arctic models.
Objective
  • Develop high-resolution maps of Arctic vegetation using machine learning and satellite imagery (e.g., NASA).
New Science
  • A convolutional neural network (CNN) approach produced highly accurate vegetation classifications. Hyper-spectral datasets (e.g., AVIRIS) were most useful for our machine learning approaches. Accurate and high-resolution datasets generated using our approach are needed for Arctic models.
Impact
  • Remote sensing products were combined and maps of vegetation distribution evaluated against field data for the Seward Peninsula
Image with caption
Image

A clustering-based stratification method previously used to map vegetation at NGEE Arctic field sites near Utqia─ívik, AK.

Citation(s)
Funding

This research was supported by the Director, Office of Science, Office of Biological and Environmental Research of the US Department of Energy under Contract No. DE-AC02-05CH11231 as part of the Next-Generation Ecosystem Experiments (NGEE Arctic) project.

For more information, please contact:

Forrest Hoffman

hoffmanfm@ornl.gov