Mapping Landslide Activity in SAR Imagery

Purpose of this project is to evaluate the effectiveness of Synthetic Aperture Radar in mapping surface damage caused by landslides in reported locations. We also intend for this project to help enhance the understanding of detecting, measuring, and visualizing landslide locations on a large scale.

General Info
Data Sources
Example Data
Workflow
Installation
Running the Workflow
Project Status
Acknowledgements
Contact
Citations
License

General Information

This project is built upon a previously developed low-accuracy large spatial scale landslide database. Space-based SAR imagery will be utilized in order to improve the accuracy of landslide detection. This is a capstone project in the University of Colorado-Boulder Earth Analytics certifcation program.
Predicting landslides is challenging due to the many variables that should be considered when trying to identify what triggered a landslide. There is a need to better identify landslide locations across a large spatial scale. Can combining a low-accuracy large spatial scale landslide database with SAR imagery improve the accuracy of landslide detection across a global scale?
Utilize human-assisted, programmatic, and possibly machine learning methods to identify landslides from satellite imagery. We hope that this will bridge the gap with landslide event detection and future event prediction.

Data Sources

[NASA Global Landslide Catalog (GLC)] Link: https://data.nasa.gov/Earth-Science/Global-Landslide-Catalog/h9d8-neg4 Info: The GLC considers all types of mass movements triggered by rainfall, which have been reported in the media, disaster databases, scientific reports, or other sources. Type: Geospatial data can be downloaded as: KLM, KMZ, Shapefile or GeoJSON
[Sentinel-1 SAR] Link: https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar Info: SENTINEL-1 is an imaging radar mission providing continuous all-weather, day-and-night imagery at C-band. Type: The SENTINEL-SAFE format wraps a folder containing image data in a binary data format and product metadata in XML.

Example Data

Workflow

Installation

Installation and usage of this repository requires an account on Google Earth Engine in order to access the Sentinel-1 data. As of 25 April, area of interests (AOIs) are ingested into the workflow through a json.load() function. The json files are located within the ./inputs path. The virtual environment is best constructed within an Anaconda Powershell prompt. To initiate the envrionment after forking the repository, open an Anaconda Powershell prompt, change the local directory path to the ./env path, then execute the command: "conda env create -f environment.yml"

Project requirements/dependencies:
Google Earth Engine account access

Running the Workflow

Begin by opening an Anaconda Powershell or similiar python command prompt on your computer. Activate the by typing conda activate earth-analytics. Once activated, open a Jupyter notebook in the IDE of your choice (e.g., VSCode, Spyder, or PyCharm). Conversely, you can type jupyter notebook at the Powershell prompt. This will open jupyter notebook in a browser tab. Navigate to the file located at the root directory of this repository. Run the notebook by selecting earth-analytics Kernel from the menu bar and then Restart & Run All. The notebook is formatted to inform the reader of processes within each code block as well as the overall current status of the project.

Future Work

Continue to investigate Google Earth Engine more and make sure we are leveraging all of their tools the best way that we can.
Iterate over a provided database or spreadsheet file.
Look into cross referencing with higher resolution SAR imagery offerred by commercial imaging companies such as ICEYE, Airbus, or Capella Space.

Acknowledgements

This project was inspired by Dr. Elsa Culler, CU Boulder Earth Lab
This project was based on:

Project Contacts

Citations

Our Github is currently private and we intend to keep it private until the end of the summer course. We will be linking Zenodo with our Git Hub repository to ensure that our code is visible and citable!

License

This project is open source and available under the Apache License 2.0.

mthomp89/landslide-detect