Visual Abstraction of Geographical Point Data with Spatial Autocorrelations

Zhiguang Zhou, Xinlong Zhang, Zhendong Yang, Yuhua Liu, Yuanyuan Chen, Ying Zhao, Wei Chen

View presentation: 2020-10-30T14:45:00Z GMT-0600 Change your timezone on the schedule page
2020-10-30T14:45:00Z
Exemplar figure
A case study for the visual abstraction of a UK census dataset. (a) and (b) present the sampled geographical points by means of Z-order and our method (sampling rate is specified as 10%). Both the geographical views present similar spatial densities according to the heatmaps located at the bottom right corners. (e) presents the Moran Scatterplots, in which spatial autocorrelations are classified into four categories. The retention of the spatial autocorrelations with two sampling methods is listed in (c) and (d). It can be found that our method outperforms the other in preserving the spatial autocorrelations, which are of great significance for geospatial modeling and analysis.
Keywords

Visual Abstraction, Spatial Autocorrelation, Sampling, Geospatial Analysis

Abstract

Scatterplots are always employed to visualize geographical point datasets, which often suffer from an overdraw problem due to the increase of data sizes. A variety of sampling strategies have been proposed to reduce overdraw and visual clutter with the spatial densities of points taken into account. However, informative attributes associated with the points also play significant roles in the exploration of geographical datasets. In this paper, we propose an attribute-based abstraction method to simplify the cluttered visualization of large-scale geographical points. Spatial autocorrelations are utilized to measure the attribute relationships of points in local areas, and a novel attribute-based sampling model is designed to generate a subset of points to preserve both density and attribute characteristics of original geographical points. A set of visual designs and user-friendly interactions are implemented, enabling users to capture the spatial distribution of geographical points and get deeper insights into the attribute features across local areas. Case studies and quantitative comparisons based on the real-world datasets further demonstrate the effectiveness of our method in the abstraction and exploration of large-scale geographical point datasets.