pith. sign in

arxiv: 2510.19745 · v3 · submitted 2025-10-22 · 💻 cs.SI

Substitution or Complement? Uncovering the Interplay between Ride-hailing Services and Public Transit

Zhicheng Jin , Xiaotong Sun , Li Zhen , Weihua Gu , Huizhao Tu This is my paper

Pith reviewed 2026-05-18 04:33 UTC · model grok-4.3

classification 💻 cs.SI
keywords ride-hailingpublic transitcomplementaritysubstitutionTNCShanghaidata classificationSHAP analysis
0
0 comments X

The pith

In a mature ride-hailing market, the share of trips that complement public transit has risen while the substitutive share has fallen.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper asks whether ride-hailing services are shifting from substitutes to complements for public transit once the market becomes saturated. It draws on trip records from 96,716 vehicles operating in Shanghai in September 2022 and applies an enhanced classification method that sorts each trip into one of four categories: first-mile complement, last-mile complement, substitute, or independent. The resulting counts show a 9.22 percent rise in the complementary ratio and a 9.06 percent drop in the substitutive ratio relative to earlier work. A separate machine-learning step identifies nonlinear influences, most notably from distance to the nearest metro station and local bus-stop density. Readers would care because the findings suggest that, in developed markets, ride-hailing can support rather than erode existing transit networks and therefore alter how cities plan integrated mobility.

Core claim

The central claim is that the interplay between transportation network companies and public transit in Shanghai shows a substantial increase in the complementary ratio of 9.22 percent and a relative decline in the substitutive ratio of 9.06 percent compared to previous studies, derived from classifying nearly 97,000 ride-hailing vehicles' data into four relationship categories and examining nonlinear influences with CatBoost and SHAP methods.

What carries the argument

An enhanced data-driven framework that classifies individual ride-hailing trips into first-mile complementary, last-mile complementary, substitutive, and independent categories based on vehicle travel data.

If this is right

  • Urban planners can treat ride-hailing as a first- and last-mile extender rather than a competitor when designing transit integration policies.
  • The observed nonlinear effects imply that small changes in metro-station distance or bus-stop density can shift a trip from substitutive to complementary.
  • Regulations in mature TNC markets may need updating to encourage rather than restrict complementary ride-hailing services.
  • Targeted location-specific interventions become feasible once the key factors driving each relationship type are known.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same shift toward complementarity could appear in other large cities once ride-hailing reaches high penetration, producing more hybrid mobility systems.
  • Transit agencies might gain from partnerships that let ride-hailing apps display real-time connections to nearby bus or metro departures.
  • If the pattern holds, some cities could slow certain transit expansions and instead subsidize complementary ride-hailing in low-density corridors.

Load-bearing premise

The enhanced data-driven framework accurately classifies individual trips into the four relationship types using the available vehicle data without substantial mislabeling, and the September 2022 Shanghai dataset represents typical behavior in a saturated TNC market.

What would settle it

Re-running the classification on the same Shanghai trip records with an alternative rule set that yields no meaningful rise in the complementary ratio, or repeating the study in another saturated city and obtaining a dominant substitutive pattern instead.

Figures

Figures reproduced from arXiv: 2510.19745 by Huizhao Tu, Li Zhen, Weihua Gu, Xiaotong Sun, Zhicheng Jin.

Figure 1
Figure 1. Figure 1: TNC trip classifications. that ride-hailing services primarily substitute for PT, with relatively low complementary ratios (Henao, 2017; Tirachini and Del Río, 2019; Meredith-Karam et al., 2021). However, these conclu￾sions may be outdated for today’s China, given the recent developments in the world’s largest TNC market. Since 2022, the ride-hailing market has been approaching saturation in cities like Sh… view at source ↗
Figure 2
Figure 2. Figure 2: Spatial distribution of ride-hailing trips and public transit stations in Shanghai. 3.2. TNC-PT relationship recognition To facilitate a better identification of TNC-PT relationships, the framework proposed by Meredith￾Karam et al. (2021) is adopted with three key extensions. First, the first- and last-mile complemen￾tary TNC trips are respectively identified by whether the destination or origin of the TNC… view at source ↗
Figure 3
Figure 3. Figure 3: Framework of recognizing TNC-PT relationships. values are derived from the integrated transportation network information in the Amap geographic information system, offering realistic approximations 3 . Part 2 is a condition-based recognition mechanism involving six sequentially checked condi￾tions. First, for each TNC trip, Condition 1 checks whether its PT alternative (bus or metro) is in service during t… view at source ↗
Figure 4
Figure 4. Figure 4: Statistics of complementary and substitutive trips. age ride-hailing fare is less than half of common fares (roughly 10RMB/km) in developed countries (Rangel et al., 2022), indirectly reflecting the intense competition within Shanghai’s ride-hailing market. Compared against complementary trips, substitutive trips exhibit a higher median price (4.99 RMB/km) and a greater proportion (7.62%) in higher fare ra… view at source ↗
Figure 5
Figure 5. Figure 5: Temporal distribution of complementary and substitutive trips and ratios. approximately 10% of total trip volume. Following this initial surge, both ratios maintain a rela￾tively stable level during mid-day hours but rise again to a second peak in the evening. The peak in the evening may be driven by both return commutes and recreational (e.g., dine-out) trips, for which the convenience and directness of T… view at source ↗
Figure 6
Figure 6. Figure 6: Spatial distribution of complementary trips and ratios. relationships between TNC and PT. On the other hand, in subcenters such as Nanhui and Jiading New Cities, higher first- and last-mile complementary ratios are observed. A potential reason is that the densely populated residential zones in these subcenters may lead residents to rely on TNCs to access nearby public transit nodes [PITH_FULL_IMAGE:figure… view at source ↗
Figure 7
Figure 7. Figure 7: Spatial distribution of substitutive trips and ratios. City and the Nanhui New City. Although these subcenters are typically equipped with bus depots or subway lines, due to the longer travel distances to the city center, some people prefer ride-hailing services that save both travel times and inconvenient transfers. In more remote areas of Shanghai, both departure and arrival substitutive trips and ratios… view at source ↗
Figure 8
Figure 8. Figure 8: SHAP beeswarm diagrams (left). The color represents the value of the feature from low to high. The abscissa denotes each feature’s contribution to the output. Charts of the relative importance (right) represent the relative contributions of the features to the model output. Jin et al.: Preprint submitted to Elsevier Page 35 of 48 [PITH_FULL_IMAGE:figures/full_fig_p036_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: SHAP value scatter and partial dependence plots. Jin et al.: Preprint submitted to Elsevier Page 36 of 48 [PITH_FULL_IMAGE:figures/full_fig_p037_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Daily OD flows of complementary trips. flows further reveal that Pudong District (home to the Pudong International Airport) and Min￾hang District (home to the Hongqiao Transport Hub) register the highest substitutive trip volumes, particularly in flows to and from the city center. This pattern reveals the fact that passengers trav￾eling to or from these major regional transportation centers are high-speed… view at source ↗
Figure 11
Figure 11. Figure 11: Daily OD flows of substitutive trips. 7.2. Complementary trips further classified by connected PT stations Fig. 12a further divides each of the first- and last-mile complementary trips into two trip types based on whether they connect to bus or metro services. For first-mile complementary trips, 84.6% connect to metro versus 15.4% to buses, whereas for last-mile trips the split is 55.1% to metro and 44.9%… view at source ↗
Figure 12
Figure 12. Figure 12: Types of complementary trips. for last-mile—is a significant finding. It reveals that a majority of travelers bypass their nearest metro station, instead choosing to connect to a more distant station (likely on a different line) or a multi-line hub to reduce transfers in the PT network and improve overall travel convenience, which echoes the PDP of the distance to the nearest multi-line metro hub in Secti… view at source ↗
read the original abstract

The literature on transportation network companies (TNCs), also known as ride-hailing services, has often characterized these service providers as predominantly substitutive to public transit (PT). However, as TNC markets expand and mature, the complementary and substitutive relationships with PT may shift. To explore whether such a transformation is occurring, this study collected travel data from 96,716 ride-hailing vehicles during September 2022 in Shanghai, a city characterized by an increasingly saturated TNC market. An enhanced data-driven framework is proposed to classify TNC-PT relationships into four types: first-mile complementary, last-mile complementary, substitutive, and independent. Our findings reveal a substantial increase in the complementary ratio (9.22%) and a relative decline in the substitutive ratio (9.06%) compared to previous studies. Furthermore, to examine the nonlinear impact of various influential factors on these ratios, a machine learning method integrating categorical boosting (CatBoost) and Shapley additive explanations (SHAP) is proposed. The results show significant nonlinear effects in some variables, including the distance to the nearest metro station and the density of bus stops.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper analyzes the relationship between ride-hailing (TNC) services and public transit using trip data from 96,716 vehicles in Shanghai during September 2022. It proposes an enhanced data-driven framework to classify individual trips into four categories—first-mile complementary, last-mile complementary, substitutive, and independent—and reports a 9.22% increase in the complementary ratio and a 9.06% decline in the substitutive ratio relative to prior studies. The study additionally applies CatBoost combined with SHAP to identify nonlinear effects of factors including distance to the nearest metro station and bus stop density.

Significance. If the trip classification framework is shown to be accurate and robust, the results would indicate a meaningful shift toward complementarity in saturated TNC markets, with direct relevance for urban transportation policy and modal integration strategies. The integration of gradient boosting with SHAP for nonlinear factor analysis provides a useful methodological contribution for examining threshold effects in transportation data.

major comments (2)
  1. [Framework and Methods] The description of the enhanced data-driven framework (abstract and methods) provides no explicit definition of the classification thresholds or rules for first-mile complementary, last-mile complementary, or substitutive trips, nor any validation against ground-truth labels, sensitivity tests on alternative thresholds, or inter-rater reliability metrics. Because the reported 9.22% and 9.06% ratio shifts are computed directly from these classifications of the 96,716 trips, the absence of these details renders the central empirical claims difficult to assess.
  2. [Results and Discussion] The comparison of ratios to previous studies does not specify the exact baseline ratios from those studies or adjust for differences in data granularity, market maturity, or classification methodologies, which directly affects the interpretation of the claimed 'substantial increase' and 'relative decline'.
minor comments (2)
  1. [Abstract] The abstract states that 'significant nonlinear effects' are observed in some variables but lists only two examples; a short enumeration of all variables with notable nonlinear patterns would improve clarity.
  2. [Data] Clarify the sampling frame and potential selection biases in the 96,716-vehicle dataset to strengthen claims about representativeness in a saturated TNC market.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript. We address each of the major comments below and outline the revisions we will make to enhance the transparency and interpretability of our findings.

read point-by-point responses

  1. Referee: [Framework and Methods] The description of the enhanced data-driven framework (abstract and methods) provides no explicit definition of the classification thresholds or rules for first-mile complementary, last-mile complementary, or substitutive trips, nor any validation against ground-truth labels, sensitivity tests on alternative thresholds, or inter-rater reliability metrics. Because the reported 9.22% and 9.06% ratio shifts are computed directly from these classifications of the 96,716 trips, the absence of these details renders the central empirical claims difficult to assess.

    Authors: We agree that additional details on the classification framework are necessary to allow readers to fully evaluate our results. In the revised manuscript, we will expand the Methods section to explicitly define the thresholds and rules used for classifying trips (e.g., distance thresholds to transit stops for first- and last-mile complementarity, and origin-destination matching criteria for substitution). We will also include sensitivity analyses by varying key thresholds and reporting the resulting changes in ratios. While ground-truth validation via labeled data is not available in our dataset, we will discuss this as a limitation and justify our approach based on established methods in the literature. These additions will make the empirical claims more assessable. revision: yes

  2. Referee: [Results and Discussion] The comparison of ratios to previous studies does not specify the exact baseline ratios from those studies or adjust for differences in data granularity, market maturity, or classification methodologies, which directly affects the interpretation of the claimed 'substantial increase' and 'relative decline'.

    Authors: We appreciate this point regarding the comparability of our results. In the revised paper, we will explicitly report the baseline complementary and substitutive ratios from the prior studies we reference, along with their sources. Additionally, we will add a discussion acknowledging potential differences arising from variations in data sources, city-specific market conditions, and methodological choices in trip classification. This will provide a more nuanced interpretation of the observed shifts in Shanghai's context. revision: yes

Circularity Check

0 steps flagged

No significant circularity: ratios computed directly from new trip classifications on fresh Shanghai data

full rationale

The paper collects primary data from 96,716 ride-hailing trips in September 2022 Shanghai and applies a proposed data-driven classification framework that assigns each trip to one of four relationship types using observable attributes (e.g., distance to nearest metro station, bus-stop density). The complementary ratio (9.22%) and substitutive ratio (9.06%) are direct aggregates of these classifications rather than outputs of any model fitted to the target ratios or prior parameter estimates. The CatBoost-SHAP component analyzes nonlinear effects of covariates on the already-computed ratios but does not redefine or predict them. Comparisons to earlier studies function as external benchmarks, not self-citation chains that define the result. No equation or step reduces the central empirical claims to tautological inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The work rests on an empirical classification framework whose exact decision rules are not specified in the abstract and on the assumption that the collected vehicle data faithfully captures trip purposes and transit connections.

free parameters (1)
  • Classification thresholds for first-mile, last-mile, and substitutive trips
    The framework must use distance, time, or spatial buffers to assign trips to categories; these cutoffs are not reported and function as free parameters.
axioms (1)
  • domain assumption The September 2022 ride-hailing trip records accurately reflect real user behavior and proximity to transit infrastructure.
    The classification into complementary or substitutive types depends on this assumption about data quality and representativeness.

pith-pipeline@v0.9.0 · 5745 in / 1537 out tokens · 49629 ms · 2026-05-18T04:33:35.119031+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    Accident Analysis & Prevention 211, 107896

    How does distraction affect cyclists’ severe crashes? A hybrid CatBoost-SHAP and random parameters binary logit approach. Accident Analysis & Prevention 211, 107896. doi:10.1016/j.aap.2024.107896. Akiba, T., Sano, S., Yanase, T., Ohta, T., Koyama, M.,

    work page doi:10.1016/j.aap.2024.107896 2024

  2. [2]

    Transport Policy 114, 280–297

    Why they don’t choose bus service? Understanding special online car-hailing behavior near bus stops. Transport Policy 114, 280–297. doi:10.1016/j.tranpol.2021.09.013. Cats, O., Kucharski, R., Danda, S.R., Yap, M.,

    work page doi:10.1016/j.tranpol.2021.09.013 2021

  3. [3]

    PLOS ONE 17, e0262496

    Beyond the dichotomy: How ride-hailing competes with and complements public transport. PLOS ONE 17, e0262496. doi:10.1371/journal.pone.0262496. Cui, X., Zhuang, C., Jiao, Z., Tan, Z., Li, S.,

    work page doi:10.1371/journal.pone.0262496

  4. [4]

    Journal of Transport Geography 111, 103669

    How can urban built environment (BE) influence on-road (OR) carbon emissions? A road segment scale quantification based on massive vehicle trajectory big data. Journal of Transport Geography 111, 103669. URL:https:// linkinghub.elsevier.com/retrieve/pii/S0966692323001412, doi:10.1016/j.jtrangeo.2023.103669. Erhardt,G.D.,Mucci,R.A.,Cooper,D.,Sana,B.,Chen,M...

    work page doi:10.1016/j.jtrangeo.2023.103669 2023

  5. [5]

    Transportation Research Part C: Emerging Technologies 178, 105181

    Novel operational algorithms for ride-pooling as on-demand feeder services. Transportation Research Part C: Emerging Technologies 178, 105181. URL:https://www.sciencedirect.com/science/article/pii/S0968090X25001858, doi:10.1016/j.trc.2025.105181. Fan, Z., Gou, J., Weng, S., 2024b. A Feature Importance-Based Multi-Layer CatBoost for Student Performance Pre...

    work page doi:10.1016/j.trc.2025.105181 2025

  6. [6]

    Annals of statistics , 1189–1232

    Greedy function approximation: a gradient boosting machine. Annals of statistics , 1189–1232. Gan,Z.,Yang,M.,Feng,T.,Timmermans,H.J.,2020.Examiningtherelationshipbetweenbuiltenvironmentandmetroridershipatstation-to-station level. Transportation Research Part D: Transport and Environment 82, 102332. Gao,F.,He,S.Y.,Han,C.,Liang,J.,2024. Theimpactofsharedmob...

    work page doi:10.1016/j.tbs.2024.100842 2020

  7. [7]

    Trans- portation Research Part D: Transport and Environment 108, 103325

    Analysis of air quality spatial spillover effect caused by transportation infrastructure. Trans- portation Research Part D: Transport and Environment 108, 103325. URL:https://linkinghub.elsevier.com/retrieve/pii/ S1361920922001535, doi:10.1016/j.trd.2022.103325. Henao,A.,2017.Impactsofridesourcing-lyftanduber-ontransportationincludingVMT,modereplacement,p...

    work page doi:10.1016/j.trd.2022.103325 2022

  8. [8]

    Travel Behaviour and Society 40, 100994

    What promotes the integration of metro and ridesourcing? Analysis of first/last mile ridesourcing origin–destination trips. Travel Behaviour and Society 40, 100994. doi:10.1016/j.tbs.2025.100994. Jin, Z., Mao, H., Chen, D., Li, H., Tu, H., Yang, Y., Attard, M.,

    work page doi:10.1016/j.tbs.2025.100994 2025

  9. [9]

    Journal of Transport Geography 109, 103579

    A data-driven approach to estimating dockless electric scooter service areas. Journal of Transport Geography 109, 103579. URL:https://www.sciencedirect.com/science/article/pii/S0966692323000510, doi:10. 1016/j.jtrangeo.2023.103579. Kong, H., Zhang, X., Zhao, J.,

    work page arXiv 2023

  10. [10]

    Journal of Transport Geography 86, 102769

    How does ridesourcing substitute for public transit? A geospatial perspective in Chengdu, China. Journal of Transport Geography 86, 102769. doi:10.1016/j.jtrangeo.2020.102769. Li, W., Shalaby, A., Habib, K.N.,

    work page doi:10.1016/j.jtrangeo.2020.102769 2020

  11. [11]

    Transportation 49, 765–789

    Exploring the correlation between ride-hailing and multimodal transit ridership in toronto. Transportation 49, 765–789. doi:10.1007/s11116-021-10193-5. Liu, X., Ye, Q., Li, Y., Yang, K., Shao, X.,

    work page doi:10.1007/s11116-021-10193-5

  12. [12]

    TransportationResearchPartA:PolicyandPractice192,104339

    Investigating relationships between ridesourcing and public transit using big data analysis and nonlinearmachinelearning: AcasestudyofShanghai,China. TransportationResearchPartA:PolicyandPractice192,104339. doi:10.1016/ j.tra.2024.104339. Lundberg, S.M., Lee, S.I.,

    work page arXiv 2024

  13. [13]

    Transportation Research Part C: Emerging Technologies 153, 104192

    Integrated design of a bus-bike system considering realistic route options and bike availability. Transportation Research Part C: Emerging Technologies 153, 104192. URL:https://www.sciencedirect.com/science/article/pii/S0968090X2300181X, doi:10.1016/j.trc.2023.104192. Luo,X.,Gu,W.,Fan,W.,2021. Jointdesignofshared-bikeandtransitservicesincorridors. Transpo...

    work page doi:10.1016/j.trc.2023.104192 2023

  14. [14]

    Journal of Transport Geography 97, 103219

    The relationship between ridehailing and public transit in Chicago: A comparison before and after COVID-19. Journal of Transport Geography 97, 103219. doi:10.1016/j.jtrangeo.2021.103219. Murray,A.T.,Davis,R.,Stimson,R.J.,Ferreira,L.,1998.PublicTransportationAccess.TransportationResearchPartD:TransportandEnvironment 3, 319–328. doi:10.1016/S1361-9209(98)00...

    work page doi:10.1016/j.jtrangeo.2021.103219 2021

  15. [15]

    https://www.marketplace.org/2025/01/14/chinas-rideshare-taxi- drivers-longer-hours-less-pay/

    China’s ride-hailing and taxi drivers face longer hours and less pay. https://www.marketplace.org/2025/01/14/chinas-rideshare-taxi- drivers-longer-hours-less-pay/. Pan, Y., Qiu, L.,

    work page 2025

  16. [16]

    Production and Operations Management 31, 906–927

    How Ride-Sharing Is Shaping Public Transit System: A Counterfactual Estimator Approach. Production and Operations Management 31, 906–927. doi:10.1111/poms.13582. Pereira,R.H.,Herszenhut,D.,Saraiva,M.,Farber,S.,2024. Ride-hailingandtransitaccessibilityconsideringthetrade-offbetweentimeandmoney. Cities 144, 104663. doi:10.1016/j.cities.2023.104663. Prokhore...

    work page doi:10.1111/poms.13582 2024

  17. [17]

    Transportation Research Part D: Transport and Environment 114, 103533

    Is ride-hailing competing or complementing public transport? A perspective from affordability. Transportation Research Part D: Transport and Environment 114, 103533. doi:10.1016/j.trd.2022.103533. Rangel, T., Gonzalez, J.N., Gomez, J., Romero, F., Vassallo, J.M.,

    work page doi:10.1016/j.trd.2022.103533 2022

  18. [18]

    Transportation 49, 373–393

    Exploring ride-hailing fares: an empirical analysis of the case of madrid. Transportation 49, 373–393. Rayle,L.,Dai,D.,Chan,N.,Cervero,R.,Shaheen,S.,2016. Justabettertaxi? asurvey-basedcomparisonoftaxis,transit,andridesourcingservices in san francisco. Transport Policy 45, 168–178. Sadowsky, N., Nelson, E.,

    work page 2016

  19. [19]

    URL:https://mp.weixin.qq.com/ s/kIhrfDSzR4ankJW0kmbIqw

    Shanghai transport industry development report. URL:https://mp.weixin.qq.com/ s/kIhrfDSzR4ankJW0kmbIqw. ShanghaiMunicipalTransportationCommission,2024. Shanghaibuslineinformationenquiry. URL:https://zwdt.sh.gov.cn/govPortals/ bsfw/item/cf1ce7ac-bb4b-40ff-a066-1abd93169b4a. Tirachini, A., Del Río, M.,

    work page 2024

  20. [20]

    Transport Policy 82, 46–57

    Ride-hailing in Santiago de Chile: Users’ characterisation and effects on travel behaviour. Transport Policy 82, 46–57. doi:10.1016/j.tranpol.2019.07.008. Tuffour, J.P.,

    work page doi:10.1016/j.tranpol.2019.07.008 2019

  21. [21]

    Case Studies on Transport Policy 10, 2229–2243

    Contested mobility interactions: Characterizing the influence of ride-sharing services on the adoption and use of public transit system. Case Studies on Transport Policy 10, 2229–2243. doi:10.1016/j.cstp.2022.10.004. Tukey, J.W., et al.,

    work page doi:10.1016/j.cstp.2022.10.004 2022

  22. [22]

    URL:https://www.worldpop.org/doi/10.5258/SOTON/WP00660, doi:10.5258/SOTON/WP00660

    The spatial distribution of population in 2020 with country total adjusted to match the corresponding UNPD estimate, China. URL:https://www.worldpop.org/doi/10.5258/SOTON/WP00660, doi:10.5258/SOTON/WP00660. Wu,L.,Gu,W.,Fan,W.,Cassidy,M.J.,2020.Optimaldesignoftransitnetworksfedbysharedbikes.TransportationresearchpartB:methodological 131, 63–83. Young, M., ...

    work page doi:10.5258/soton/wp00660 2020

  23. [23]

    Journal of Transport Geography 82, 102629

    Measuring when Uber behaves as a substitute or supplement to transit: An examination of travel-time differences in Toronto. Journal of Transport Geography 82, 102629. doi:10.1016/j.jtrangeo.2019.102629. Yu, C., Deng, Y., Qin, Z., Yang, C., Yuan, Q.,

    work page doi:10.1016/j.jtrangeo.2019.102629 2019

  24. [24]

    Transportation Research Part D: Transport and Environment 124, 103935

    Traffic volume and road network structure: Revealing transportation-related factors on PM2.5 concentrations. Transportation Research Part D: Transport and Environment 124, 103935. doi:10.1016/j.trd.2023.103935. Jin et al.:Preprint submitted to ElsevierPage 47 of 48 Zhen,L.,Gu,W.,2024. Feederbusservicedesignunderspatiallyheterogeneousdemand. Transportation...

    work page doi:10.1016/j.trd.2023.103935 2023