Weekly城市规律 | 高温与城市出行：基于机器学习的韧性分析

本文字数：2244字

阅读时间：约8分钟

2025

LIFE

点击上方蓝字关注我们

高温与城市出行 —— 基于机器学习的

上海共享单车韧性分析

Heat and mobility: Machine

learning perspectives on

bike-sharing resilience in Shanghai

研究背景

在全球气候变化的背景下，极端高温事件日益频繁，严重影响城市出行系统，尤其是共享单车等非机动出行方式。现有研究多关注温度对骑行量的整体影响，却缺乏对极端高温条件下“城市出行韧性”（UMR）的系统量化与机制解析。特别是建成环境与高温的交互作用尚未被深入揭示，因此有必要构建新的数据驱动框架来识别共享单车在极端气候下的韧性模式。

上下滑动阅读英文

Background:

Against the backdrop of global climate change, extreme heat events are becoming more frequent, severely affecting urban mobility systems, especially non-motorized modes such as bike-sharing. Existing studies mostly focus on the overall relationship between temperature and cycling volume, but lack systematic quantification and mechanistic analysis of “Urban Mobility Resilience” (UMR) under extreme heat. In particular, the interactions between the built environment and heat have not been thoroughly explored, highlighting the need for a new data-driven framework to identify resilience patterns of bike-sharing under extreme weather.

文献综述

已有研究证明气温与骑行行为存在非线性关系，中等气温促进出行，而极端高温则显著抑制。同时，建成环境的密度、功能混合度、公共交通可达性等因素都会影响共享单车的使用。然而，以往的研究多停留在定性或静态空间尺度上，对极端高温情境下的出行韧性缺乏量化指标与细粒度分析。因此，提出在“气候—环境—出行”三者交互机制中引入韧性视角具有重要理论意义。

上下滑动阅读英文

Literature Review:

Previous studies demonstrate a nonlinear relationship between temperature and cycling, where moderate temperatures encourage ridership while extreme heat significantly suppresses it. Built environment factors such as density, functional mix, and public transport accessibility also influence bike-sharing use. However, most existing work remains at qualitative or static spatial scales, lacking quantitative indicators and fine-grained analysis of resilience under extreme heat. Thus, introducing a resilience perspective into the climate–environment–mobility nexus has important theoretical significance.

数据与方法

研究以上海市中心七个核心城区为对象，选取 2016 年连续两周极端高温期作为案例，采集了共享单车出行大数据（480 万条）、建成环境指标（建筑密度、高度、容积率、交通设施）、自然环境指标（NDVI、水体）、社会经济数据（人口密度、夜间灯光）以及高温气象数据。通过 500×500m 网格划分构建分析单元，利用 LightGBM 模型并结合 SHAP 可解释性方法，对工作日与周末的共享单车韧性进行了建模与机制解读。

上下滑动阅读英文

Data and Methodology:

The study focuses on seven core districts in central Shanghai, using two consecutive weeks of extreme heat in August 2016 as the case period. Datasets include over 4.8 million bike-sharing trips, built environment indicators (building density, height, floor area ratio, transit facilities), natural environment measures (NDVI, water bodies), socio-economic variables (population density, nighttime lights), and meteorological data. The city was divided into 500×500 m grid cells as analysis units. A LightGBM model, combined with SHAP interpretability analysis, was applied to explore bike-sharing resilience on weekdays and weekends.

图1：研究区域

Fig1: Study area

The findings reveal distinct temporal patterns of resilience: weekday commuting trips follow a “camelback-line” with limited heat impact, while weekend leisure trips form a “horseback-line” with much greater reductions under heat. Spatially, resilience is higher in central areas and lower at the edges, with weekend low-resilience zones more widespread. Mechanism analysis shows that building density within a reasonable range significantly enhances resilience, bike-sharing serves as an effective feeder mode within 1,000–1,500 m of metro stations, while road density and greening sometimes negatively affect resilience. Population and economic vitality exhibit threshold effects: moderate agglomeration improves resilience, but excessive concentration or dispersion reduces it.

研究结果

结果表明共享单车出行韧性在时间上呈现差异化：工作日通勤出行构成“驼峰线”，受高温影响相对有限，而周末休闲出行呈“马背线”，在高温下减少更为显著。在空间上，市中心韧性普遍高于边缘区，且周末低韧性区域更为集中。

机制分析揭示，建筑密度在合理区间内显著提升韧性，地铁 1000–1500m 范围内的共享单车接驳作用明显，而道路密度和绿化在某些条件下反而削弱韧性。同时，人口与经济活力存在阈值效应，适度集聚能增强韧性，而过度集中或分散则效果不佳。

Results:

The findings reveal distinct temporal patterns of resilience: weekday commuting trips follow a “camelback-line” with limited heat impact, while weekend leisure trips form a “horseback-line” with much greater reductions under heat. Spatially, resilience is higher in central areas and lower at the edges, with weekend low-resilience zones more widespread. Mechanism analysis shows that building density within a reasonable range significantly enhances resilience, bike-sharing serves as an effective feeder mode within 1,000–1,500 m of metro stations, while road density and greening sometimes negatively affect resilience. Population and economic vitality exhibit threshold effects: moderate agglomeration improves resilience, but excessive concentration or dispersion reduces it.

图 4. 极端高温日与正常日共享单车使用的日内时段模式。(a) 工作日的平均使用模式；(b) 周末的平均使用模式

Fig. 4. Patterns of time-of-day bike-sharing usage on extreme heat days and normal days. (a) Average usage pattern on weekdays; (b) Average usage pattern on weekends

图5. 共享单车 UMR 的空间分布。(a）工作日的平均分布；（b）周末的平均分布

Fig.5.Spatial distribution of UMR of bike-sharing. (a) Average distribution on weekdays; (b) Average distribution on weekends

图 5. 所有因素的 SHAP 汇总图。(a）工作日的结果；（b）周末的结果

Fig. 5. SHAP summary plot for all factors. (a) Results on weekdays; (b) Results on weekends

讨论

研究进一步揭示了建成环境、公共交通、自然环境与社会经济因素对共享单车韧性的非线性作用机制。高密度与功能混合区在通勤需求驱动下更能维持出行韧性，而公共交通系统的合理布局可在高温条件下发挥缓冲作用。绿化在工作日对韧性有积极贡献，但在周末因缺乏可达性反而限制出行。人口与经济活力在适度水平下能够提升韧性，但一旦超过阈值则会导致资源压力与服务效率下降。

Discussion:

The study reveals the nonlinear influence of built environment, public transit, natural environment, and socio-economic factors on bike-sharing resilience. High-density, mixed-use areas maintain stronger resilience under commuting needs, while well-integrated transit systems buffer against heat impacts. Greening contributes positively during weekdays but may reduce resilience on weekends due to accessibility issues. Moderate levels of population and economic vitality enhance resilience, but once thresholds are surpassed, resource stress and efficiency losses emerge.

政策建议

从规划实践角度，应推动中等强度的高密度开发，保持功能多样性与就业居住平衡，提升共享单车与公共交通的衔接效率，并合理规划慢行网络。绿化建设需兼顾布局与可达性，避免孤立大绿地的使用障碍。在人口与经济活动的空间引导上，应保持合理集聚，避免过度集中或分散化格局，从而在极端气候下提升城市出行系统的整体韧性。

Policy Recommendations:

From a planning perspective, moderate high-density development should be encouraged, with balanced functional diversity and job-housing mix. Integration of bike-sharing with public transport must be strengthened, alongside the rational design of slow-mobility networks. Green space planning should emphasize layout and accessibility, avoiding isolated large parks that hinder usage. Spatial guidance for population and economic activity should promote moderate clustering, avoiding both over-concentration and excessive decentralization, thereby enhancing urban mobility resilience under extreme heat.

局限与展望

本研究的局限在于仅以上海为样本、时间跨度较短，且数据来自 2016 年，因而存在代表性和时效性的不足。同时方法上主要停留在相关性分析，因果推断有限。未来研究可拓展到多城市、多气候区的对比分析，结合长时序与面板数据，甚至引入生成式建模与情境预测，以进一步推动共享单车及慢行系统在气候适应性和城市韧性研究中的理论深化与实践应用。

Limitations and Future Work:

This study is limited to Shanghai, with a short time span and data from 2016, which affects representativeness and timeliness. Methodologically, the analysis remains correlational, lacking strict causal inference. Future research should expand to multiple cities and climate zones, incorporate long-term or panel data, and explore generative modeling and scenario simulations to deepen theoretical and practical understanding of bike-sharing and slow mobility systems in climate adaptation and urban resilience.