by Markus Neteler, https://www.mundialis.de/
presented at:
Open Data Science Europe workshop 2022, 2022-06-14, 11:00-12:30, Workshop room 2 - C223
https://opendatascience.eu/workshop-2022/
Abstract
GRASS GIS supports time-series processing for vector, raster, and volume data. This workshop offers a micro-introduction to Sentinel satellite data archives, and the various ways to access them. It also explores the i.sentinel toolset which allows querying Sentinel data coverage for a region of interest, downloading from multiple data sources, performing atmospheric and topographic corrections, and cloud/shadow masking. This workshop also gives a preparation of data for multitemporal analyses through automatic creation of a space-time raster dataset (strds), It explores the computation of NDVI time series. Eventually we run a simple RandomForest landuse classification on Sentinel-2 data.
This course is based on workshop material by Martin Landa, CVUT.
We will run GRASS GIS 8.2 through a Jupyter Notebook.
Content
- Why Jupyter Notebooks and how to use them?
- GRASS GIS & Python
- Setup of the Google Colab instance with GRASS GIS 8 (only Google Colab version of Jupyter Notebook)
- Paths and variables, connecting to GRASS GIS backend
- Data upload to notebook session
- Initialization of GRASS GIS in the Jupyter notebook session
- Creating an area of interest map
- Importing geodata into GRASS GIS
- Sentinel-2 processing overview
- Computing NDVI
- Time series data processing
- Creating an image stack (imagery group)
- Object recognition with image segmentation
- Supervised Classification: RandomForest
- What's next?
Jupyter Notebooks
Google Colab (https://colab.research.google.com/) notebook version:
Standard Jupyter Notebook for local usage or in Jupyterhub:
There are several options available for running Jupyter notebooks, each offering different functions. Here are some ways to run Jupyter notebooks:
A) Your local computer:
- Standard Jupyter Notebook Server: Install the
python3-jupyter-core
server locally to run Jupyter notebooks
B) Cloud providers to use your own Jupyter notebooks without having to install anything on your local machine:
- Binder: run Notebooks stored in GitHub, GitLab, Google Drive, etc. (https://mybinder.org/)
- Whole Tale: mainly for research purposes (https://wholetale.org/)
- Google Colab: run Notebooks stored in Google Drive and GitHub; offers good hardware options (CPU, GPU, TPU) (https://colab.research.google.com/)
- and other providers
C) Own JupyterLab server:
- JupyterLab: Deployment of a notebook server in your own server infrastructure, for multiple users (https://jupyter.org/hub)