Tutorial: Geospatial Data Formats

Geospatial data allows the communication of information in a wide range of fields. Over time, many data formats have emerged to support this diversity. This tutorial introduces these various formats, their evolution, their advantages and disadvantages, and provides guidance on how and in which contexts to use them most effectively. It will be divided into several jupyter notebooks:

1. Raster formats
2. Vector formats
3. Data cube formats
4. Point clouds
5. Data optimization

For each tutorial, sample data will be provided in the sample_data directory or automatically downloaded (in the sample_data directory by default). This means that an internet connection is required to run these notebooks. Alternatively, you can manually download data from other providers and use them to run the notebooks. In some examples, benchmarks are ran and performances are compared between different formats, but the data used for these tests is generally quite small, mostly used as example. Using real, larger datasets highlights better the performance discrepancies between formats. For example, in the raster tutorial you can use real Sentinel-2 data, downloadable from geodes or copernicus. For the point clouds or datacube tutorials, you can use SWOT data, available on hydroweb next (with netcdf files for example). For SWOT data, you can also check out this tutorial (TBD).

Some notebooks showcase how to visualize the data from the different file types, but you can find more detailed tutorials here.

Installation

To run a notebook, first create a virtual environment using pip. Then, install the required packages (if not already done), using the requirements.txt file.

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Then you can simply use jupyter notebook xxx.ipynb to run the corresponding notebook.