Exploring cycling flow dynamics and their interaction with built environment features: evidence from bike-sharing data

Understanding urban travel patterns through the lens of flow space is essential for analyzing urban structures and their dynamics. Cycling flows, in particular, offer valuable insights into these patterns. However, previous studies have largely overlooked the factors driving cycling flow patterns and their spatial dynamics within urban contexts. This study addresses these gaps by introducing a systematic framework to investigate cycling flow patterns and their underlying formation mechanisms. Using bike-sharing data in Shenzhen as a case study, cycling flows are categorized into three types: clustering, divergent, and convergent patterns. To quantify the nonlinear effects of various built environment features on each pattern type, interpretable machine learning models are employed. The findings reveal that cycling flow patterns reflect urban polycentric structures and functionally diverse nature of large urban environments. Additionally, the impact of built environment features depends on the context, rather than being simply positive or negative. These features also have broad and varied effects across different areas. These results provide insights for optimizing urban built environments, promoting sustainable urban mobility, and fostering balanced spatial interactions.

借助流动空间视角解析城市出行模式，对于剖析城市结构及其动态演化规律至关重要。其中，骑行流尤为关键，能够为这类出行模式研究提供极具价值的洞察。然而，既往研究大多忽略了城市场景下驱动骑行流模式及其空间动态变化的影响因素。本研究通过构建系统化分析框架，对骑行流模式及其内在形成机制展开探究，以此填补上述研究空白。本研究以深圳共享单车（bike-sharing）数据为案例样本，将骑行流划分为集聚型、发散型与收敛型三类模式。为量化各类建成环境（built environment）特征对不同骑行流模式的非线性影响，本研究采用可解释机器学习模型。研究结果表明，骑行流模式能够反映大城市的多中心城市结构与功能多元性特征。此外，建成环境特征的影响效应因场景而异，并非简单的正向或负向作用。这类特征在不同城市区域还会产生广泛且差异化的影响。本研究结果可为优化城市建成环境、推动可持续城市出行以及促进均衡空间互动提供决策参考。