TL;DR
This paper uses taxi GPS data from Shenzhen to analyze the spatial and temporal patterns of urban tourism travel. By extracting tourist trips and employing advanced spatial analysis techniques, the study reveals how urban morphology, tourism resources, and passenger preferences shape tourist movement, offering valuable insights for urban tourism planning and development.
What Problem Does This Paper Address?
Understanding urban tourism travel is crucial for effective tourism management and urban planning. While traditional methods like surveys provide insights, they are often costly, time-consuming, and lack real-time dynamism. Existing studies using digital footprints often focus on general urban travel or lack the granularity to capture specific tourism behaviors. This paper addresses the gap by specifically analyzing tourism travel using taxi trajectory data, a rich source of information that reflects "point-to-point" mobility and is closely tied to human activities and attraction visitation. The authors aim to fill the void in understanding the spatial and temporal dynamics of intra-city tourism flows, particularly during peak periods, and how these flows interact with the urban environment.
Methods
The study employs a multi-stage methodological framework:
- Tourist Trip Extraction: Taxi trajectory data is pre-processed to remove anomalies. A map-matching algorithm (ST-Matching) is used to align trajectories with the road network. Tourist trips are then identified by analyzing taxi drop-off locations near the entrances of 26 selected tourist attractions.
- Spatial Aggregation and Dependence Analysis:
- Kernel Density Estimation (KDE): Used to visualize and quantify the spatial aggregation of tourist trip origins, identifying "hotspots" of tourist activity.
- Getis-Ord Gi*: Applied to identify statistically significant hot and cold spots of tourist trip origins, revealing local spatial dependencies.
- Correlation with Geographic Contextual Factors:
- Geographically Weighted Regression (GWR): This is a key technique. POI (Point of Interest) data (categorized into 11 types like catering, shopping, transportation facilities) is used. Buffers of varying sizes (20-300m) are created around trip origins, and the number of POIs within these buffers is extracted. GWR models the relationship between trip origin density and these POI types, accounting for spatial heterogeneity.
- Travel Network Analysis:
- Complex Network Metrics: Travel networks are constructed for each tourist attraction, with nodes representing origins, attractions, and road intersections, and edges representing road segments. Metrics like average degree, network diameter, average path length, and clustering coefficient are used to characterize the structure and connectivity of these networks.
- Spatial Interaction Intensity Map: Visualizes the flow of tourist trips between attractions, categorized into three classes based on trip volume.
Datasets
- Taxi Trajectory Data: Collected from 16,828 GNSS-equipped taxis in Shenzhen during the "May Day" holiday period (May 1-3, 2015). Each record includes taxi ID, coordinates, speed, time, and occupancy state. This resulted in 69.16 million records, from which 37,878 tourist trips were extracted.
- POI Data: Crawled from Gaode Maps, containing over 1.7 million records as of September 2018. POI data includes name, address, type, longitude, and latitude, categorized into 11 types.
- Road Network Data: Used for map-matching and constructing travel networks.
- Tourist Attraction Data: The top 26 attractions in Shenzhen, ranked by Ctrip.com, were selected.
Key Insights and Ideas
- Tourism Travel is Distinct: Tourist taxi trips exhibit different temporal patterns than residential trips, with two distinct peaks (11:00 and 15:00) and a downward trend on the last day of the holiday, unlike the cyclical pattern of residential travel.
- Short Trips Dominate: The majority of tourist taxi trips are short in duration (less than 15 min) and distance (within 8 km), suggesting a focus on exploring nearby attractions or immediate surroundings.
- Spatial Heterogeneity is Key: Tourist trip origins are not evenly distributed but are concentrated in areas with high tourist attractions and urban entry/exit points. This spatial distribution is significantly influenced by urban morphology and the availability of amenities.
- POI Influence is Spatially Variable: The impact of different POI types (e.g., transportation facilities, accommodation) on tourist trip origins varies significantly across different areas of the city, highlighting the need for spatially explicit analysis (like GWR).
- Network Structure Matters: Tourist travel networks exhibit grouping and hierarchical characteristics, with some attractions forming tight clusters and others being more peripherally connected, reflecting the overall tourism system's structure.
Key Results
- Temporal Patterns: Tourist trips peak at 11:00 and 15:00, with a decline on the final day of the holiday.
- Travel Characteristics: 67.89% of tourist trips are under 15 minutes, and 71.25% are under 8 km. The Generalized Extreme Value (GEV) distribution best models these travel time and distance patterns.
- Spatial Aggregation: Tourist trip origins are concentrated in areas like Baishizhou, Hongshuwan, Huaqiaochen, Dameisha, Haiyabaihuo, and Guomao.
- GWR Findings: Transportation facilities (TF), accommodation services (AS), residential housing (RH), sports and leisure services (SLS), and catering services (CS) show positive correlations with tourist travel. Scientific, educational, and cultural services (SECS) and shopping services (SS) show negative correlations. The GWR model achieved a high goodness-of-fit (R²-Adjusted = 0.948554) with a 100m buffer.
- Network Coverage: Attractions like Lianhuashan Park, Window of World, Happy Coast, and Talent Park have the greatest network coverage, while Guanlan Printmaking Village and Gankeng Hakka People Small Town have the least.
- Spatial Interaction: Tourist flows show grouping and hierarchy, with attractions like Window of World, Jinxiu China Folk-Custom Village, Happy Coast, and Happy Valley forming tight clusters.
Limitations and Future Work
- Limited Transport Modes: The study focuses solely on taxi travel during a specific holiday period ("May Day"). It acknowledges that other modes (metros, buses, car sharing) are also important for urban tourism and that a more comprehensive analysis would include these.
- Tourist Trip Extraction Challenges: Defining precise tourist drop-off areas at attractions can be challenging due to the random nature of taxi pick-ups and potential GPS inaccuracies. The authors attempted to mitigate this through careful correction using street view imagery.
- Data Availability: The raw data for this study is part of an ongoing project and cannot be shared at this time.
Future Work Suggestions:
- Obtain data for the same period over multiple years to conduct comparative analyses across different tourist seasons.
- Expand the analysis to include other tourist seasons in China (e.g., National Day).
- Incorporate environmental semantic features of traffic around tourist attractions to aid in tourism product development and tourist moderation.
- Investigate the impact of emerging transport modes like car sharing.
Summary updated