A Spatial Perspective on Burstiness Measure for Spatiotemporal Event Analysis

Exploring space-time structures is essential to understanding dynamic geographic/environmental events that are localized in space and time. A bursty pattern, where extremely long inter-event times alternate with many short inter-event times, is one of the characteristic temporal patterns of events that has been paid attention to in the field of complex systems [1]. The concept and measures of bursts have been developed and applied in many theoretical and applied studies [2-3,6]. However, spatial perspectives have been seldom integrated with analyzing the bursty phenomena, although spatial dimension of events including a spatial scale and the environmental context of an event can affect temporal dynamics of phenomena [4]. In measuring patterns of geographic/environmental events, there is the modifiable areal unit problem (MAUP) that a spatial aggregation unit choice may affect measurement outcomes, well-known in the field of geography [5]. Particularly, the shape and
scale of spatial aggregation units or spatial sampling methods of events can potentially influence the detected temporal patterns. The same applies to the burstiness measure (e.g., [2,6]); it is crucial to understand how different settings of a spatial dimension of measurements changes the burstiness measurement results even with the same input data. In this study, I explore the impact of various spatial parameters including the shape and size of spatial aggregation units and geographical sampling methods on burstiness measures of Kim & Jo
(2016) [2] and Goh & Barabasi (2008) [6], and provide guidelines for constructing a spatial aggregation unit as a spatial container of events and selecting spatial sampling methods for those burstiness indicators. As a proof-of-concept, I apply the proposed guideline to the burstiness analysis of traffic accident events in Switzerland.