The third and final part of the Winter Challenge!

New here? Go to part 1 and part 2 to find out more!

Winter challenge prognostica: The final

In order to determine biodiversity in agriculture, it helps to know the exact field boundaries of agricultural land. With this knowledge, satellite data can be analyzed so that the use of different areas (pasture, meadow, grain, …) can be precisely monitored.

Data on field boundaries is available from various sources. For this task, we have provided you with three different data sets for the same region as GeoJSON files:

• reference.geojson: This file contains the references. Use this information to determine the quality of the other two files.
• boundaries_1.geojson and boundaries_2.geojson: These datasets contain data from other providers. You can compare the individual fields via their ID with those of the reference (in reality, this is usually not the case - a preliminary step here would be to find a matching approach via the spatial information).

Now you need to find out which of the two sources provides better information. Find a way to assess the quality of the data and justify why you would choose one over the other. Write a function in Python find_best_fitting_version() that reads in the data and determines which of the two versions 1 or 2 is closer to the reference. The result should be output in a JSON with the following format:

{"boundaries_1": {"best_version": "True"},"boundaries_2": {"best_version": "False"} }

You are welcome to store further results as additional fields in the JSON. Show us in a Jupyter notebook how best to call your function. You are welcome to use short texts and other information to illustrate your approach.

You can find the data here; it comes from the Schlaginfo Portal (© ML/SLA Niedersachsen (2024), dl-de/by-2-0 (www.govdata.de/dl-de/by-2-0), data edited). We have modified them slightly for this task. If you have never worked with GeoJSON files before, don’t worry, GeoPandas has a very detailed introduction.

Send us your submission to winter@challenge.prognostica.de. We are already looking forward to your approach!