Task-oriented data classification and design of choropleth maps (aChor)
Final Report Abstract
Choropleth maps are probably the most frequently used type of thematic map. In order to achieve a better overview or faster readability, the attribute values to be displayed are very often classified in advance. The classification methods commonly used and implemented in software packages (such as equidistance, quantiles, Jenks, etc.) are data-driven, i.e. the intervals are determined exclusively on the basis of the existing frequency distribution of the original values. However, the spatial context of the underlying data, which is important for many questions, is completely neglected in such a division along the number line. This means that information about spatial relationships or patterns (or desired statements on a map) can be lost. Instead of the data-driven approach the project aChor pursues the reverse way – a task-oriented approach. Firstly, spatial “tasks” like the preservation of local outliers, large value differences, hot/cold spots or clusters are specified. Based on this, task-specific specific data classification algorithms that also take the spatial relationships into account have been developed, implemented and successfully tested. For all patterns under consideration, the first step of these algorithms is a definition of neighboring polygons that fulfill a certain condition (e.g., showing a large value difference or being a hot spot). Based on this, setting class breaks is performed by using adapted line sweep algorithms. These methods have been embedded as a plug-in tool into the open source QGIS software. Several open source python modules such as GDAL (Geospatial Data Abstraction Library), PySAL (Python Spatial Analysis Library), Fiona, Shapely and RTree were used. The plugin tool together with a detailed documentation was published in the GitLab repository. But not only the classification as such but also their visualization is of importance for the purpose of visually detecting and interpreting spatial patterns. Here, a specific and so far not investigated aspect of visualization has been treated. It could be shown theoretically and empirically that the use of nonlinear color schemes for non-liner data classes is not feasible – other aspects like the number of classes or the complexity of the map content are more dominant factors in the process of visual pattern detection.
Publications
- (2017) Data Classification for Highlighting Polygons with Local Extreme Values in Choropleth Maps. In: Peterson M.P. (ed.) Advances in Cartography and GIScience, Lecture Notes in Geoinformation and Cartography, Springer: 449- 459
Schiewe, J.
(See online at https://doi.org/10.1007/978-3-319-57336-6_31) - (2018) An open source tool for preserving local extreme values and hot/coldspots in choropleth maps. Kartographische Nachrichten, 68(6): 307-309
Chang, J. & Schiewe, J.
(See online at https://doi.org/10.1007/bf03544625) - (2018) Preserving Local Extreme Values in Choropleth Maps. Proceedings der Tagung FOSSGIS 2018, Bonn (ISBN 978-3-00-059170-9), 148-152
Chang, J. & Schiewe, J.
- (2018) Task-oriented Data Classification of Choropleth Maps for Preserving Local Extreme Values. In: Kersten, T. et al. (Ed.): Photogrammetrie - Fernerkundung - Geoinformatik - Kartographie-2018, 38: 213-218
Chang, J. & Schiewe, J.
- (2019): Erhaltung und Betonung räumlicher Muster in Choroplethenkarten. GIS.Science, 1/2019: 1-8
Schiewe, J. & Chang, J.
