automated-glacier-terminus

This code automatically delineates marine glacier terminus positions in Landsat 7 and Landsat 8 imagery using the adapted 2D Wavelet Transform Modulus Maxima (WTMM) segmentation method (Liu et al., 2021). The 2D WTMM method calculates gradients in brightness throughout each image at 50 unique spatial scales and produces maxima chains correpsonding to the maximum brightness gradients. These maxima chains delineate regions of high contrast in brightness, such as the glacier terminus margin. The terminus line is selected through an optimized thresholding approach. Relative terminus positions are calculated along three glacier flowlines, as such:

For more details, please see:

    Liu,J., Enderlin, E. M., Marshall, H.P., and A. Khalil (2021). Automated Detection of Marine Glacier Calving Fronts Using the 2-D Wavelet Transform Modulus Maxima Segmentation Method. IEEE Transactions on Geoscience and Remote Sensing, vol. 59, no. 11, pp. 9047-9056. doi: 10.1109/TGRS.2021.3053235

We politely request that you provide attribution to this work if you intend to use the code for published research by citing the paper above.

Set up your working directory structure as follows:

• glacier_files/
  • Box001/
  • Box002/
  • BoxNNN/
• LSaws/
  • Box001/
  • Box002/
  • BoxNNN/

Begin by obtaining or generating rectangular boxes over each glacier's terminus area as ESRI shapefiles. Name them BoxNAMEID.shp. These will be stored in your glacier_files folder (basepath variable). When the BoxID subfolders are generated automatically in the first two scripts, move the shapefiles into their respective folders. OPTIONAL: Repeat for each glacier's Randolph Glacier Inventory outline if available. Name them RGI_BoxNAMEID.shp.

The LSaws folder (downloadpath variable) will contain the Landsat images automatically downloaded in the first script.

Run the scripts in this order:

1. LS_image_download_AWS.ipynb*
   
2. preprocess.ipynb
   
3. wtmm2d_terminuspick.ipynb (python only) OR wtmm2d_terminuspick_Xsmurf.ipynb (requires Xsmurf**)
   
4. analyze_wtmm_results.ipynb
   

*The first script in this workflow automatically downloads subsets of Landsat 7 and 8 imagery available at low cost through the Amazon Web Services (AWS) s3 bucket. This requires a requester pays account through AWS. If you would like to input your own set of Landsat 7 and 8 images, skip the first script and start with preprocess.ipynb.

**Please contact jukesliu@u.boisestate.edu and andre.khalil@maine.edu if you would like to install Xsmurf.

Each .ipynb file contains text describing the analysis in that step.

System requirements:

Requires GDAL version 3.2 or newer with command terminal functionality (e.g., gdalwarp, gdal_translate). Several steps require the ImageMagick command line software to be installed. If downloading images through AWS, the AWS command line interface must also be installed. Other packages required are listed in the autoterm.yml file included. Set up a local micromamba environment with the .yml file. From the repository root, run in a terminal:

Installing the autoterm environment with micromamba

(0) Install micromamba

"${SHELL}" <(curl -L micro.mamba.pm/install.sh)

Option A: from autoterm.yml

micromamba env create --file autoterm.yaml

Option B: manual installation

(1) Create a new environment named "autoterm" with python 3.9.15

micromamba create -n autoterm python=3.9.15

(2) Activate the environment. Then, download the following packages into the environment in the following order:

micromamba activate autoterm

micromamba install notebook numpy pandas matplotlib pyfftw geojson -c conda-forge

micromamba install boto3=1.24.60 scipy=1.9.0 pyproj=3.3.1 opencv=4.5.5 -c conda-forge

micromamba install rasterio=1.2.10 fiona=1.8.21 geopandas=0.11.0 -c conda-forge!

make sure autoterm is active prior to opening Jupyter notebook and running the scripts:

micromamba activate autoterm