This repository contains the code of the author's Master's thesis.
The repository also contains additional data from the appendix of the thesis. This data
can be found in the directory Appendix.
The thesis attempts to assess the temporal quality of OpenStreetMap (OSM) data in relation to land use/land cover (LULC) data, using the example of the expansion of built-up areas. For this purpose, LULC maps are created for the Areas of Interest (AoI) based on OSM data for two different points in time. The maps follow the class definitions of the WorldCover (WC) data. The WC data covering the AoIs are downloaded for the same two points in time to be used as reference data. The OSM-based maps are evaluated for completeness and accuracy in comparison to the WC data. It is also determined to what percentage an expansion of the built-up class corresponds in the respective datasets. In addition, the calculation of the mapping saturation indicator of the Ohsome Quality API is used to calculate how complete the respective LULC classes are mapped in the OSM maps and a correlation analysis is carried out between the characteristics of the AoI and the correspondence of the development expansion.
In order to run the code, the required environment must be set up. This can be done by installing it from the pyproject.toml file with Poetry. Furthermore, the user must set the configurations in the config.py file. The code can be run by executing the run_full_analysis.py file.
- Python >=3.9, <=3.11
- Poetry installed as Python Package Manager
- User Account for Terrascope Platform
- Adapted Version of the Ohsome Quality API (see section 'Mapping Saturation Indicator Calculation' for further information)
Up to 999 Inputfiles in the GeoJSON Format can be specified. Each file can hold multiple Polygons. Those must have an attribute "id" which must be an integer between 0 and 999. Furthermore, in order to be able to run the correlation analysis, each polygon must have the following attributes assigned: "Population" (int), "X-Coordinate" (float), "Y-Coordinate"(float), "University" (boolean), and "Disposable Income" (int).
As the filters defined in this repository are not default topics in the ohsome quality API (OQAPI), the script calculate_oqapi.py does not work with public OQAPI. Therefore, it is recommended to set up a local OQAPI instance, define the filters used in this analysis as topics in OQAPI, and use it for the calculation. For instructions on how to set up a local OQAPI instance, see the Ohsome Quality API Repository. Each filter must be defined as a topic in OQAPI. Fur further information on OQAPI topics, check out the topic documentation of OQAPI.
In order to define which OSM features should be considered per LULC class, a filter was defined for each class. These filters can be adapted. However, the filter format must be kept. The filters are stored in the filter directory.
The first line of the file contains the OSM tags which should be used to query Polygons. The second line contains the OSM tags which should be used to query line features. Each file must not have more than 2 lines. If only line features are queried, they must be written in the second line leaving the first line empty.
The file buffersizes.txt contains the width with which line
features should be buffered. Each line must contain a Key-Value-Pair and the buffer
width. The tag and width must be divided by ,, not =.
The width of lines must be stated as total width, not per lane. The required unit is
meters.
The file keyconfidences.txt contains the confidence value
which will be assigned to each queried OSM feature based on the key used to query the
respective feature. When adapting the filter definitions, this file must be updated as
well. The file must contain one OSM key per line, paired with the to be associated level
of confidence. The key and confidence level must be divided by ,, not =.
The confidence level must be an integer between 1 and 4. It defines the confidence that
features tagged with the respective key are associated with the correct LULC class. It
is used to define which features will be overwritten by others in the final map if
features overlay each other. The higher the confidence level, the less likely the
feature will be overwritten by another feature and consequently the more likely the
feature will be represented in the final map.
If Poetry fails to install the dependencies with the following error message:
Retrieved digest for link
terracatalogueclient-0.1.14-py3-none-any.whl(md5:xxxxx)
not in poetry.lock metadata
['sha256:xxxxx']
run
poetry config experimental.new-installer false
to use the old (warning: not recommended) installer and try again