Abstract
There exists a kind of trajectories of dynamic geographic phenomena, which have splitting, merging, or merging-splitting branches. Clustering these complex trajectories may help to more deeply explore and analyze the evolution mechanism of geographic phenomena. However, few methods explore the clustering patterns of such trajectories. Thus, we propose a Process-oriented Spatiotemporal Clustering Method (PoSCM) for clustering complex trajectories with multiple branches. The PoSCM includes the following three parts: the first represents the trajectories with a “process-sequence-node” structure inspired by a process-oriented semantic model; the second designs a hierarchical similarity measurement method to calculate the similarity of space, time, thematic attributes and evolution structure between any two trajectories; the last uses a density-based clustering algorithm to mine the trajectories' clustering patterns. Simulation experiments are used to evaluate PoSCM and to demonstrate the advantages by comparing against that of the VF2 algorithm. A case study of sea surface temperature abnormal variation (SSTAV) trajectories in the Pacific Ocean is addressed. The clustering results not only validate well-known knowledge but also provide some new insights about the evolution characteristics of SSTAVs during El Nino Southern Oscillation (ENSO); these insights may provide new references for further study on global climate change.
Highlights
Geographical phenomena are dynamic and complex [1], [2]
The trajectory clustering patterns of such dynamic phenomena may reveal the evolution mechanism and their relationship with global climate change signals, such as El Niño Southern Oscillation (ENSO) events, which are crucial for monitoring global climate change and regional disasters [11], [12]
As we do not focus on the discussion of advantages and disadvantages between different clustering algorithms, Process-oriented Spatiotemporal Clustering Method (PoSCM) uses Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to group the trajectories into clusters based on the similarity matrix An×n (n is the total number of process trajectory (PT)) obtained from Hierarchical Similarity Measurement Method (HSMM)
Summary
Geographical phenomena are dynamic and complex [1], [2]. There exists a type of dynamic geographic phenomena that has spatiotemporal continuity [3], [4] and may have complex evolutionary behaviors along with dynamic changes, e.g., ocean eddies, storm events, and marine anomaly variations. The main contributions of our study include the following: 1) we proposed a process-oriented representation model based on geographical process semantics for representing complex trajectories This representation model retains the evolutionary structure and spatiotemporal characteristics of dynamic phenomena and further reduces the complexity of the original trajectory, which makes the similarity measurement and cluster identification easier; 2) a hierarchical similarity measurement method is designed for obtaining the similarity of space, time, thematic attributes and evolutionary structure between complex trajectories. This method provides a comprehensive measurement strategy for measuring the multi-attribute similarity of the spatiotemporal evolution process; 3) we have found some interesting trajectory clustering patterns of sea surface temperature anomaly variations in the Pacific Ocean.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
