Analysis of User Behaviour While Interpreting Spatial Patterns in Point Data Sets.

Volunteered geographic information is often generated as voluminous point data, leading to geometric and thematic clutter when presented on maps. To solve these clutter problems, cartography provides various point generalization operations such as aggregation, simplification or selection. While these operations reduce the total number of points and therefore improve the readability, information preservation could be harmed when specific spatial patterns disappear through the generalization process, possibly leading to false interpretations. However, sets of map generalization constraints that maintain spatial pattern characteristics of point data are still missing. To define constraints that support synoptic interpretation tasks, user behaviour while solving these tasks has to be analysed first. We conduct a study where participants have to perform such interpretation tasks, using a new method that combines think-aloud interviews and techniques from visual analytics. We reveal that the point density of a dataset has the biggest impact on the user behaviour and the respective task-solving strategy, independently from the actual task type executed. Furthermore, our results show that the graphical map complexity only has a minor impact on the user behaviour, and there is no evidence that point data cardinality influences task execution and the solution-finding strategies.