Documentation for mapping of AK glaciers using DL

This project aims to create composite raster images and vector datasets for the period 1984-2020 for all of the glaciated area in Alasaka. We use images from the Landsat archive and we use the Landtrendr algorithm to create a spectrally homogeous time series of optical data. We combine spectral indices from that time series with topographic indices calculated from the Alaska IfSAR 5m DEM in a convolutional neural network (CNN). We use the CNN to make predictions about glacier and debris covered glacier areas across the state of Alaska for bi or tri annual composites.

Dependencies:

• numpy

• pandas

• geopandas

• pytorch

• multiprocessing

• subprocess

Script examples:

• Create training dataset
  • Code description
• GlacierCoverNet model
• Convert raster to vector
• Postprocess vectors
• Error analysis
• Temperature data
• Create figures
  • Code description

Indices and tables

• Index

• Module Index

• Search Page