UC Santa Barbara

In recent years, the easy access to geospatial data has increased the opportunities and demand for geospatial analysis across disciplines. Yet, it is often not obvious for non-geographers how to use Geographic Information Systems (GIS) or other geospatial analysis tools. A popular place for users to seek answers to their questions is online forums, such as GIS Stack Exchange. Studying questions on these forums may help uncover what users usually ask about GIS in order to answer their questions about the world. In this paper, a Latent Dirichlet Allocation (LDA) topic model is used to explore the topics users ask about on GIS Stack Exchange. The 15 topics identified from over 40,000 posts on GIS Stack Exchange cover a broad spectrum, ranging from geospatial information representation (e.g., coordinate systems), geospatial information retrieval (e.g., geospatial querying) to geospatial computing (e.g., raster and vector computations) and geospatial data visualization (e.g., mapping). The number of posts, answer rates and answer times vary by topic clusters, implying their relative popularity and difficulty. To complement the analysis, I also compared the top 20 tags used in GIS and Cross Validated (Statistics) Stack Exchange. The domination of software-related tags in GIS contrasts with the prevalence of tags standing for software-independent fundamental concepts (e.g., hypothesis test, probability and distributions) for statistics questions. This supports my observation, from the LDA topic model, that the questions users ask about GIS are mainly about data models and software procedures. It indicates that the conceptual basis for GIS has not yet reached the clarity and consensus found in statistics. This study contributes empirical evidence on gaps in the conceptual basis underlying GIS design, education, and training. It also offers insights to GIS educators and software developers who aim at making GIS more accessible and easier to use across disciplines, inspired by the success of statistics along these lines.