Revealing the Impact of COVID-19 on Urban Residential Travel Structure Based on Floating Car Trajectory Data: A Case Study of Nantong, China

Fei Tao, Junjie Wu, Shuang Lin, Yaqiao Lv, Yu Wang, Tong ZhouISPRS Int. J. Geo Inf 2023 · DOI

TL;DR

This paper uses taxi trajectory data from Nantong, China, to analyze how COVID-19 lockdowns impacted urban residential travel patterns. By developing a novel method to infer travel purposes using urban functional areas (UFAs) and machine learning, the study reveals significant shifts in travel hotspots, trip types, and residents' underlying psychological needs, offering valuable insights for urban planning during public health crises.

What Problem Does This Paper Address?

The COVID-19 pandemic drastically altered urban mobility, but understanding the nuanced impact on residential travel, considering various trip purposes and the underlying psychological drivers, remains a challenge. Existing studies often focus on single travel modes or purposes, or rely on less granular data like mobile phone records. This paper aims to fill this gap by:

  • Quantifying changes in the structure of residential travel: This includes origin-destination (OD) points, travel flows, and the proportion of different travel purposes.
  • Inferring travel purpose accurately: Traditional methods using Points of Interest (POI) data struggle when multiple POIs are near a destination. This paper proposes a more robust approach using Urban Functional Areas (UFAs).
  • Connecting travel behavior to psychological needs: The study explores how shifts in travel patterns reflect changes in residents' needs, framed by Maslow's hierarchy.

Methods

The research employs a multi-faceted approach:

  • Inferring Travel Purpose with UFAs and Machine Learning: The core innovation lies in using Urban Functional Areas (UFAs) to classify the purpose of residential travel. Taxi pick-up and drop-off points are mapped to UFAs (e.g., residential areas, shopping malls, hospitals, schools, traffic centers, scenic spots). Machine learning models (k-means++, kNN, MLC-ResNets, YOLO v3) are used to identify UFA attributes from taxi trajectory data and remote sensing imagery. This UFA-based classification aims to improve accuracy over traditional POI-based methods, especially in complex urban environments.
  • Establishing Residential Travel Structure: This involves creating an Origin-Destination (OD) matrix representing travel flows between different UFAs. The study then analyzes the distribution of OD points, the volume of travel flows, and the proportion of different travel purposes.
  • Analyzing Spatiotemporal Characteristics: Kernel density estimation is used to identify travel hotspots, revealing how their location and density changed over different phases of the pandemic. The Net Flow Rate (NFR) indicator is employed to analyze the overall travel patterns of different UFAs.
  • Categorizing Travel Purposes: Travel purposes are categorized into commuting, shopping, business, medical, tourism, and random elastic travel. These are then mapped onto Maslow's hierarchy of needs (physiological, safety, love/belonging, esteem, self-actualization) to understand the psychological drivers behind travel behavior.

Datasets

  • Taxi Trajectory Data: Provided by the Nantong Taxi Management System, covering December 2019 to March 2020. This dataset includes license plate, phone number, time, longitude, latitude, speed, direction, and passenger status.
  • Remote Sensing Data: Google Maps imagery of Chongchuan District, Nantong, China, from December 2019 to March 2020, with a resolution of 0.5m. This data was used for UFA identification.
  • Points of Interest (POI) Data: Collected via web crawler and Baidu Maps API, totaling 88,331 pieces of POI data in Chongchuan District. This data was used for validation and comparison.

Key Insights and Ideas

  • Travel Hotspots Contracted: During the pandemic, especially the outbreak period, residential travel hotspots significantly decreased in number and spatial distribution, concentrating more in the main urban areas. This suggests a narrowing of residents' travel scope.
  • Commuting and Medical Travel Became Dominant: The proportion of commuting and medical travel increased, reflecting a prioritization of essential needs and work-related travel. Conversely, leisure and shopping travel saw sharp declines.
  • Psychological Needs Shifted: The study provides a novel link between travel patterns and Maslow's hierarchy. During the outbreak, physiological and safety needs became paramount, while love/belonging and esteem needs increased as society recovered, indicating a return to more complex social interactions.
  • UFA-based Purpose Identification is Promising: The paper demonstrates that using UFAs for travel purpose identification can improve accuracy, especially in complex urban settings where multiple POI types coexist near a single destination.

Key Results

  • Travel Volume Decline and Rebound: Residential travel volume significantly decreased from December 2019 to February 2020 and began to rebound in March 2020.
  • Shift in Travel Purpose Proportions:
    • Commuting and medical travel ratios increased from 43.8% to 45.7% and 7.1% to 8.1%, respectively, during the outbreak period.
  • Shopping and tourism travel proportions decreased significantly.
  • Hotspot Reduction: Figures 7a-c clearly show a substantial reduction in urban hotspots of residential travel from the normal to the outbreak period.
  • NFR Trends: The absolute values of NFR for most UFAs decreased during the pandemic, indicating reduced overall residential travel, and gradually increased as the situation improved.

Limitations and Future Work

  • Limited to Taxi Data: The study primarily relies on taxi trajectory data. Future work should incorporate other modes of transport (buses, subways, shared bikes, walking) for a more comprehensive understanding of urban mobility.
  • Single City Study: The analysis was conducted on a single city. Future research should explore the synergistic effects of travel structure changes among cities at an urban agglomeration or regional scale.
  • Smartphone Data for Purpose Identification: Incorporating other social sensing data, such as smartphone data, could further enhance the accuracy of residential travel purpose identification.

Summary updated

BibTeX
@article{tao2023revealing,
  author = {Fei Tao and Junjie Wu and Shuang Lin and Yaqiao Lv and Yu Wang and Tong Zhou},
  title = {{Revealing the Impact of COVID-19 on Urban Residential Travel Structure Based on Floating Car Trajectory Data: A Case Study of Nantong, China}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2023},
  doi = {10.3390/ijgi12020055},
}
Figure