Scholarlink

<p>Rental-return imbalances commonly arise in bike share systems due to uneven travel demand across urban areas. However, existing bike-sharing research has mainly focused on large metropolitan areas, leaving usage patterns in smaller cities relatively underexplored. Meanwhile, earlier built environment measures tend to ignore the impact of magnitude and proximity, potentially leading to biased inferences of travel behavior. To address these, this study aims to compare different built environment measures and examine bike-sharing usage patterns by taking magnitude and proximity into account and using Tartu (Estonia) as a case study. For doing so, this study firstly compared commonly used point of interest (POI)-based built environment measures with magnitude-based counterparts calculated from building, land use, and population register data; the Gaussian distance decay function was further incorporated into calculations for evaluating the impact of proximity. The results reveal that POI-based measures substantially misrepresented educational and residential characteristics around bike stations. Built environment measures were widely overestimated, if no proximity weighting applied. Retail characteristics, however, were relatively robust to both magnitude and proximity. Furthermore, using the decayed magnitude-based built environment measures, this study acquired five types of bike-sharing usage patterns in Tartu (Estonia) by the k-means + + clustering algorithm. The results reveal that commercial areas saw the highest bike-sharing usage, far exceeding that of all other areas. By contrast, sparsely populated residential areas experienced the lowest bike-sharing usage. Peak hours of bike rentals and returns varied substantially across different urban areas, especially on weekdays. For instance, commercial areas and school surroundings had sharper peaks of bike returns in the morning but sharper peaks of bike rentals in the afternoon, whereas opposite temporal demands were observed in the densely populated residential areas. This suggests that shared bikes were widely used for people’s commuting purposes. These findings offer valuable insight into both biases and usability associated with POI data and highlight the importance of magnitude and proximity for representing built environments when investigating travel behavior. They also provide actionable guidance for addressing rental–return imbalances in small-scale bike share systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Magnitude and proximity matter: examining bike-sharing usage patterns by k-means + + clustering of station-surrounding built environments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月13日

查看全文
  • Xiao Cai,
  • Siiri Silm,
  • Amnir Hadachi

摘要

Rental-return imbalances commonly arise in bike share systems due to uneven travel demand across urban areas. However, existing bike-sharing research has mainly focused on large metropolitan areas, leaving usage patterns in smaller cities relatively underexplored. Meanwhile, earlier built environment measures tend to ignore the impact of magnitude and proximity, potentially leading to biased inferences of travel behavior. To address these, this study aims to compare different built environment measures and examine bike-sharing usage patterns by taking magnitude and proximity into account and using Tartu (Estonia) as a case study. For doing so, this study firstly compared commonly used point of interest (POI)-based built environment measures with magnitude-based counterparts calculated from building, land use, and population register data; the Gaussian distance decay function was further incorporated into calculations for evaluating the impact of proximity. The results reveal that POI-based measures substantially misrepresented educational and residential characteristics around bike stations. Built environment measures were widely overestimated, if no proximity weighting applied. Retail characteristics, however, were relatively robust to both magnitude and proximity. Furthermore, using the decayed magnitude-based built environment measures, this study acquired five types of bike-sharing usage patterns in Tartu (Estonia) by the k-means + + clustering algorithm. The results reveal that commercial areas saw the highest bike-sharing usage, far exceeding that of all other areas. By contrast, sparsely populated residential areas experienced the lowest bike-sharing usage. Peak hours of bike rentals and returns varied substantially across different urban areas, especially on weekdays. For instance, commercial areas and school surroundings had sharper peaks of bike returns in the morning but sharper peaks of bike rentals in the afternoon, whereas opposite temporal demands were observed in the densely populated residential areas. This suggests that shared bikes were widely used for people’s commuting purposes. These findings offer valuable insight into both biases and usability associated with POI data and highlight the importance of magnitude and proximity for representing built environments when investigating travel behavior. They also provide actionable guidance for addressing rental–return imbalances in small-scale bike share systems.