🚨🌍 NEW ARTICLE 🌍🚨
We geocoded the #mobility of over 2 million #Erasmus students across #Europe from 2014 to 2022 with @miladmzdh Oula Inkeröinen & Olle Järv. The data descriptor article is published in #ScientificData, and is an output from the #MobiTwin project.
https://doi.org/10.1038/s41597-025-04789-0
#GIScience #Geospatial #GIS #MobiTwin #OpenData #OpenScience
We used #Photon and #Nominatim #geocoding services to convert the textual place names into geographical locations and aggregated them on the #NUTS and #LAU levels to support usability of our data.
Geocoding the data was not simple. Many place names were entered into the Erasmus+ database manually resulting in diverse place names containing national spellings, typographical mistakes, character encoding errors, and national scripts, which required manual corrections.
To support using this data, we have also developed an #EdgeBundling tool to aid in visualising complex origin-destination flow data:
https://doi.org/10.5281/zenodo.14532547
We also provide a ready-made edge-bundled network on NUTS 2 level lead by Oula Inkeröinen:
Edge-bundling tool for regional mobility flow data This repository hosts the scripts to perform edge-path bundling (Wallinger et al. 2022) for flow data. It's primary use case is to support visualization of complex mobility data, and has been used to bundle human mobility flows across NUTS regions in Europe. The tool's inputs are two CSV files, one for point feature data and associated coordinates, and another for flows (edges) to be bundled. After bundling, the tool outputs a GeoPackage file. The script expects the data to be in WGS84 coordinate reference system. The scripts in this repo are repurposed versions of the original scripts written by Peterka (2023). The updates to the original code aim to make the code more usable for analytical purposes. This tool is an additional output of the Mobi-Twin research project. Requirements The scripts within the repo require Python 3.10 or newer version with the following packages: pandas geopandas tqdm shapely On top of these Python requirements, the script expects the input CSV data (centroids and edges) to have a certain structure. Data structure for input files Centroid file | ID_COLUMN | X | Y | | ---- | :----- | :---------- | | Unique identifier for centroid (e.g., NUTS code) | X coordinate (WGS84) of the centroid | Y coordinate (WGS84) of the centroid | N.B.: The ID_COLUMN in the above is an example name, use the column name you have in your data. Edge file | ORIGIN | DESTINATION | OD_ID | COUNT | | ---- | :----- | :---------- | :---------- | | ID code of origin | ID code of destination | ID made of origin and destination codes joined by an underscore (_) | Integer/floating point number of flow strength | N.B.: The ID codes of origins and destinations have to match the IDs of your centroid file. Usage Clone this repository, and run the tool by typing in the following command: python bundle_edges.py -c /path/to/centroids.csv -id ID_COLUMN -ew /path/to/edges.csv -o /path/to/output.gpkg If you want to adjust some parameters of the bundling, such as weights or bundling threshold use the flags -ew for edge weights (default is 2), and -t for bundling threshold (default is 2). The edge weights dictate how powerful the "gravity" of long edges are. The bundling threshold sets the distance limit for how many times longer the bundled edges can be compared to straight line distances, flows that are longer than the threshold are not bundled but remain as straight line geometries in the output. Please note, the script expects the coordinates to be in WGS84 (EPSG:4326) Test files We have provided two test CSV files that demonstrate the data structure of the required CSV files. These files can be found under the example_data directory. References Wallinger, M., Archambault, D., Auber, D., Nöllenburg, M., & Peltonen, J. (2022). Edge-Path Bundling: A Less Ambiguous Edge Bundling Approach. IEEE Transactions on Visualization and Computer Graphics, 28(1), 313–323. https://doi.org/10.1109/TVCG.2021.3114795 Peterka, O. (2024). Xpeterk1/edge-path-bundling. https://github.com/xpeterk1/edge-path-bundling (Original work published 2023). Related links Mobi-Twin project official webpage Digital Geography Lab webpage
The full student mobility data and Python scripts used to produce the data are openly available in our #Zenodo:
https://zenodo.org/records/14332354
Read more in our blog post: https://blogs.helsinki.fi/digital-geography/2025/03/26/over-2-million-erasmus-student-mobility-flows-across-europe-geocoded-open-data-description-article-out/
General information This repository holds the spatially enriched Erasmus+ student mobility data from 2014 to 2022. It is based on the official Erasmus+ mobility raw data available from http://data.europa.eu/88u/dataset/erasmus-mobility-raw-data. The spatially enriched Erasmus+ student mobility data covers the period from 2014 to 2022, and is presented as four tabular files. For a detailed description on how these data were produced, please see the Scientific Data article of the same name. The files included are Erasmus_2014-2022_individual.csv The full individual-level student mobility dataset. Erasmus_2014-2022_individual.parquet.gzip The full individual-level student mobility dataset as a compressed parquet file. Erasmus_2014-2022_aggregate_LAU.csv The aggregated student mobility flows between LAU units. Erasmus_2014-2022_aggregate_NUTS.csv The aggregated student mobility flows between NUTS units. scripts.zip The analysis scripts used to produce this data, for most up-to-date scripts, remember to check out our GitHub repository README_DATA.md The README file describing each variable in the data. If you are using the data or scripts herein, please cite the Scientific Data article. Links related to the data publication: Mobi-Twin project official webpage Digital Geography Lab webpage
Tuomas Väisänen 📼🧟♂️
adingbatponder