Spatiotemporal patterns of urban residents'travel activities under the impact of emergencies
Throughout the course of human societal development,emergencies often trigger profound changes in individuals'lives and behaviors,leaving a lasting impact.While existing research primarily focuses on overarching trends and general characteristics of residents'travel activities,there remains a paucity studies delving into the finer spatiotemporal differences of such activities.Moreover,these studies often grapple with issues concerning disconnected spatiotemporal dimensions and lack comprehensiveness.This study employs the COVID-19 pandemic as a case study and conducts an extensive analysis spanning a significant period.Utilizing bike-sharing data and point of interest(POI)data,the study represents bike-sharing destination sites as feature vectors composing the proportions of various categories of POI points in their proximity.Three traditional clustering algorithms are compared,and the k-means algorithm,demonstrating superior performance,is utilized to cluster the sites.Subsequently,the study explores the land use characteristics of the bike-sharing sites obtained from each cluster.The clustering label of the destination site is used as the spatial feature of the trip,and when combined with the travel time feature,converts the trip record of the shared bicycle into the trip label containing spatial-temporal information.The entire dataset of trip label corpora is then used to train the latent Dirichlet allocation(LDA)model.This process yields the topic distribution for each document,i.e.,the topic distribution of the set of resident travel tags divided by month,and the word distribution for each topic,i.e.,the distribution of travel tags for each topic.The study observes the trend of topic distribution over time,with particular focus on the dominant topics prevalent during different periods.Based on the distribution of dominant topics'travel tags,the study explores the characteristics and changes in spatiotemporal patterns of resident travel before and after the emergency event.The results indicate significant changes in the spatiotemporal patterns of residents'travel activities for various purposes following emergencies.Notably,the proportions of commuting and non-essential entertainment trips have notably decreased,with residents opting for outdoor recreational activities in smaller gatherings as a substitute for group entertainment.There is also an increase in the proportions of travel activities related to purchasing necessities and attending to personal affairs.The demand for medical services has substantially increased,accompanied by an earlier start time of the morning peak hour for such activities.Over time,the impact of the emergency gradually diminished,and people's travel activity patterns gradually revert to a state akin to the pre-emergency period.This study enhances our understanding of risks and uncertainties,establishing a more comprehensive spatiotemporal knowledge service system.Furthermore,the methods and approaches employed in this study exhibit strong flexibility and wide applicability,providing a transferrable framework for studying residents'spatiotemporal travel patterns under the influence of emergencies.This offers valuable insights for urban management and planning departments in formulating reasonable emergency management strategies.Future studies could benefit from exploring alternative data sources and intergrating diverse datasets to further investigate various aspects of residents'travel patterns.
emergencytravel activitiespoint of interesttrip purposecluster analysisspatiotemporal patternslatent Dirichlet allocation
NETL
NSTL
万方数据
