Visual abstraction of large-scale geographical point data with credible spatial interpolation

With the increasing size of geographical point data, scatterplot often suffers from serious overdraw problems, which greatly hinders the visual exploration and analysis of data. At present, a variety of sampling methods considering data features have been proposed to simplify the large-scale geographical point data to alleviate this problem. However, there is still no attempt to simplify data from the perspective of geostatistics in the sampling methods, which will be greatly beneficial to explore the spatial information of unknown points and restore the original data features. In this paper, a sampling model is proposed to generate a representative subset from the large-scale geographical point data to improve the interpolation quality of the sampled points and preserve attribute features of original data, in which a semivariable function is applied to capture geostatistical characteristics of data attributes. A set of visual interfaces are further implemented enabling users to visually evaluate the sampled results of different methods and effectively conduct geospatial analysis. Case studies and quantitative comparisons based on the real-world geographical datasets further demonstrate the validity of our interpolation-driven sampling model in the abstraction and analysis of large-scale geographical point data. Graphic abstract