Urban Currents

Unveiling stochasticity: Universal multi-modal probabilistic modeling for traffic forecasting

This study introduces a method to convert deterministic traffic forecasting models into probabilistic ones by replacing the final output layer with a Gaussian Mixture Model (GMM) layer. Evaluated across multiple traffic datasets and architectures, the modified models are trained using only Negative Log-Likelihood loss and consistently improve the Continuous Ranked Probability Score (CRPS) and predictive interval calibration compared to unimodal baselines.

Why it matters — It provides a universal, plug-and-play adaptation that equips existing spatio-temporal traffic models with uncertainty quantification capabilities without requiring changes to data preparation, optimizers, or hyperparameters.

Caveat: While the GMM adaptation improves L2 error metrics, it introduces trade-offs that can degrade L1 error performance.

Transportation Research Part C Emerging Technologies · pdf · doi · code

Perceived traffic safety’s impact on cycling route choice with street-level images

Using a stated choice experiment and a street-level image dataset, this study extracted safety perception scores from cycling-perspective images using a computer vision model. Mixed logit models revealed that cyclists are willing to accept 64 additional seconds of travel time for a one-unit increase in perceived safety on a scale from -2 to +2. The analysis shows that older cyclists, recreational riders, and those with positive cycling attitudes prioritize safety, whereas commuters prioritize speed.

Why it matters — It quantifies the trade-off between travel time and perceived safety in monetary or time-equivalent terms, allowing planners to incorporate subjective safety improvements directly into cost-benefit analyses and targeted infrastructure upgrades.

Caveat: The findings are based on a stated choice experiment using street-level images rather than observed GPS tracks of actual route choices.

Transportation · pdf · doi · code

FlexBSS: A flexible multi-objective framework for bike-sharing station optimization

The researchers developed FlexBSS, a multi-objective optimization framework for bike-sharing station placement that integrates configurable spatial factors, network-level accessibility metrics on directed cycling networks, and slope-adjusted distances. The framework uses a genetic algorithm to generate station layouts under minimum-distance constraints, and was tested in Barcelona using open data across scenarios targeting demand, multimodal integration, and social equity, including an expansion scenario in the l'Eixample district.

Why it matters — It replaces rigid, single-objective planning models with a flexible system where planners can weight and combine any number of custom objectives, while accounting for physical topography and network directionality that directly affect cycling effort and station accessibility.

Journal of Urban Mobility · doi

Forced or desired residential relocation? Moving voluntariness and its effects on well-being and travel behaviour

This study tracks individuals before (n=1,234) and after (n=411) a residential relocation using longitudinal panel survey data to analyze how the voluntariness of a move affects well-being and travel behavior. Using a new subjective measure of voluntariness, the researchers found that voluntary movers experienced significant improvements in mental health, life satisfaction, residential satisfaction, and place attachment, whereas involuntary movers showed non-significant changes. While subjective voluntariness did not directly alter travel attitude-behavior consistency, involuntary movers exhibited weaker car orientations and lower fulfillment of mobility needs.

Why it matters — It demonstrates that subjective perceptions of moving voluntariness, rather than traditional objective classifications, reveal distinct differences in post-move well-being and highlight how involuntary movers face persistent disadvantages in mobility satisfaction and perceived control.

Caveat: The findings are based on a relatively small post-move sample size of 411 respondents, and the study relies on self-reported subjective measures of voluntariness and well-being.

Travel Behaviour and Society · doi

Group representation of vulnerable travelers in household travel surveys: an assessment of 43 U.S. surveys

This study introduces the Group Representation Ratio (GR-ratio), a metric derived from sampling weights to measure how accurately survey data reflects population subgroups. The researchers applied this metric to 43 U.S. household travel surveys, including the 2017 National Household Travel Survey and 42 regional and state datasets, to evaluate the representation of low-income, transit-dependent, unemployed, and Black-led households. The analysis revealed that 85% of the surveys overrepresented white households, and only 11 achieved reasonable representation of vulnerable subgroups.

Why it matters — It provides transportation planners with a standardized, equity-focused metric to replace traditional response rates, which often mask systematic nonresponse biases. It also identifies specific survey design choices—such as income-based stratified sampling, mixed-mode data collection, and targeted incentives—that successfully improve the representation of marginalized travelers.

Caveat: The utility of the GR-ratio is constrained by the availability of specific survey items and requires careful, consistent definitions of population subgroups.

Transportation · pdf · doi

Revealing intercity connectivity of less-than-truckload logistics and associated determinants in Pearl River Delta

This study analyzes the spatial structure and drivers of the less-than-truckload (LTL) logistics network across the Pearl River Delta, China, using inter-city route data from an Alibaba-affiliated digital freight platform. Applying a reversed gravity model, the analysis reveals that manufacturing value added is the primary driver of inter-city connectivity, especially at destination cities, while major gateway ports like Shenzhen and Guangzhou channel flows toward inland manufacturing hubs.

Why it matters — It provides empirical evidence on the spatial organization of LTL logistics, which support small and medium-sized enterprises, demonstrating how industrial clustering, international port access, and physical distance co-shape regional supply chains.

Caveat: The study relies on data from a single digital freight platform, which may not capture the entire spectrum of LTL logistics in the region.

Computational Urban Science · pdf · doi

Barriers and enablers to implementing active Urbanism for health in the UK

This study conducted a qualitative expert consultation process using purposive and snowball sampling, alongside document analysis, to evaluate how UK built environment professionals receive the 'Invigorating Landscape' concept. The research identifies key actors, regulatory barriers, and insurance-related liabilities that influence the implementation of physically challenging cityscapes designed to promote bone, muscle, and cardiovascular health.

Why it matters — It identifies specific institutional barriers—namely UK safety regulations, insurer recommendations, and liability concerns regarding fall risks—that prevent theoretical active urban design concepts from being implemented by developers and policymakers.

Caveat: The findings are based on qualitative interviews and document analysis within the specific regulatory and insurance context of the United Kingdom, which may not generalize to other legal jurisdictions.

Discover Cities · pdf · doi

Out of business: design constraints and financial burdens in repurposing dead shopping malls

This study evaluates the technical and financial feasibility of adaptive reuse for abandoned shopping malls using a four-stage methodology that combines multi-criteria evaluation, concept design, cost appraisal, and Monte Carlo simulations. Tested on a case study of a dying mall in Italy, the analysis compares nine repurposing options and finds that converting the mall into a data center yields the highest base-case net present value of ‐12.8 million (7.7% IRR), followed by senior housing at ‐3.4 million (5.3% IRR) and a medical facility at ‐2.2 million (4.4% IRR). The data center conversion requires minimal structural intervention, whereas residential and healthcare options demand extensive spatial reconfiguration.

Why it matters — It provides developers and planners with quantified financial trade-offs and risk probabilities for mall conversions, demonstrating that while data centers are the most financially viable and least structurally invasive option, their success remains highly sensitive to revenue fundamentals and existing property book values.

Caveat: The financial and structural findings are based on a single case study of an Italian mall, meaning the specific cost structures and viability rankings may not directly translate to other regulatory or real estate markets.

Cities · pdf · doi

Uneven resilience: Constructing a climate environmental vulnerability index for South Africa's urban core

The study developed a Climate Environmental Vulnerability Index (CEVI) for municipalities in South Africa's Gauteng Province using 17 standardized indicators spanning exposure, sensitivity, and adaptive capacity from 2010 to 2024. Using Principal Component Analysis, the framework measures and maps localized climate vulnerability across geographically proximate urban areas.

Why it matters — It demonstrates that local differences in adaptive capacity and socio-economic conditions, rather than climate exposure itself, are the primary drivers of vulnerability variation, proving that uniform provincial climate policies are insufficient compared to targeted municipal-level interventions.

Caveat: The index is applied only to the municipalities within a single province, Gauteng, which may limit its immediate generalizability to rural or less urbanized regions of South Africa.

Cities · pdf · doi

For whom the road tolls: perceived benefits, commuter segmentation, and congestion pricing acceptability in Hyderabad

Based on a face-to-face survey of 435 car users in Hyderabad, India, this study uses benefit-based commuter segmentation and Meta Ranking to analyze how different groups perceive the advantages of congestion pricing. While travel-time minimization is the top perceived benefit overall, the analysis identifies four distinct commuter segments that prioritize different combinations of benefits.

Why it matters — It demonstrates that car users are not a homogeneous group in their expectations of congestion pricing, providing a segmentation methodology that allows planners to tailor policy design and communication strategies to specific user concerns before implementation.

Caveat: The findings are based on a relatively small sample of 435 respondents in a single city, which may limit the direct generalizability of the specific segment profiles to other urban contexts.

Case Studies on Transport Policy · pdf · doi

Yield decisions and perceived right-of-way at junctions: road traffic rule knowledge, simple road crossings, warning sign and pedestrian behaviour effects

Two online studies evaluated UK drivers' yield decisions and perceived right-of-way at junctions following recent pedestrian priority rule changes. The first study tested a simple, non-prescribed zebra crossing design alongside pedestrian stance and vehicle turn direction, while the second study tested an alternative footway crossing design, warning signs, and pedestrian gaze. Both studies found that driver knowledge of the priority rule, the presence of simple crossings, and warning signs significantly increased the likelihood of deciding to yield and the perception of pedestrian priority.

Why it matters — The research demonstrates that low-cost engineering countermeasures, particularly simple zebra crossings and warning signs, can actively reinforce pedestrian priority rules at junctions where driver compliance is otherwise uncertain.

Caveat: The findings are based on self-reported driver decisions and perceptions in online image-based studies rather than observed real-world behavior at physical junctions.

Journal of Transport & Health · doi

Ten years of Brazilian airport concessions, learned lessons

This study analyzes a decade of Brazilian airport concession contracts to identify how specific clauses and external shocks influenced contract duration and asset devolution. The analysis correlates contractual variables, regulatory rounds, and competition levels with the outcomes of these long-term agreements.

Why it matters — It demonstrates that rigid contract terms prevent operators from adapting to external crises like the pandemic, and shows that auction duration is driven by competition levels rather than just asset value. This evidence supports specific drafting reforms, such as standardized learning clauses and competition filters, to prevent aggressive overbidding and premature asset returns.

Caveat: The findings and proposed contractual adjustments are derived entirely from the specific regulatory and economic context of Brazil's airport privatization rounds.

Research in Transportation Economics · doi

Pedestrian mobility and innovative data sources: a literature review focused on app-based GPS data

This literature review synthesizes 95 studies that utilize smartphone application GPS data to analyze pedestrian mobility. It categorizes these data sources into three distinct types—fitness/activity trackers, application aggregators, and travel-survey apps—and details how their specific collection, sharing, and processing methods shape their utility for pedestrian research.

Why it matters — It establishes a systematic framework for understanding the trade-offs of different app-based data sources, allowing researchers and planners to match specific pedestrian planning questions—such as assessing spatial inequalities or walking environments—with the most appropriate GPS data type.

Travel Behaviour and Society · pdf · doi

Hyperlocal monitoring of urban activity reveals responses to heat exposure

The study analyzed urban activity patterns across 10 German cities over a two-month period in 2024 using hourly telecommunication data at a spatial resolution of 100m by 155m, totaling over 100 million data points. By linking these location counts with hourly weather records and point-of-interest data, it found that sustained hot periods of three or more days above 25°C reduce overall city-wide presence by 1.5 percentage points. During these hot spells, activity shifts toward leisure and cultural amenities like swimming pools and cafés by up to 10 percentage points, while presence at public services like schools and healthcare facilities declines.

Why it matters — It provides empirical, high-resolution evidence of how populations dynamically redistribute themselves across specific urban amenities during heatwaves, moving beyond static vulnerability indexes to show where heat adaptation interventions are most urgently needed.

Caveat: The findings are based on a relatively short observation window of two months during a single year.

arXiv · pdf

Link-Based Multimodal Traffic Dynamics Model in Continuous-Time Framework

The study develops a multimodal link transmission model (M-LTM) that simulates continuous road traffic and discrete tramway traffic in a continuous-time framework. The model incorporates moving bottleneck theory to capture inter-modal interactions, alongside node models that simulate tram dwell times, queue spillbacks, and tram priority at intersections. Validation on synthetic networks and the real-world network of Dresden, Germany, demonstrated road traffic GEH statistics below 4 and average tram arrival time absolute errors under 10 seconds.

Why it matters — It provides a computationally efficient network-level simulation tool capable of capturing the mutual congestion impacts between public transit and private vehicles, which was previously difficult to model dynamically at scale.

arXiv · pdf

Cascading Effects of the COVID-19 Pandemic on Barangays in the Philippines

This study analyzed the cascading socio-economic impacts of the COVID-19 pandemic on Philippine barangays using survey data from 2,122 households collected between May and June 2021. The researchers constructed a Bayesian network model and an influence diagram to map the conditional dependencies among mobility, public service access, financial health, food security, education, and physical health. The model identified that interventions targeting food production, market price stabilization, and income generation yielded the highest improvements in overall community outcomes.

Why it matters — It provides local government units and policymakers with a quantified, systemic map of how household vulnerabilities interact, demonstrating that economic and food security interventions are more effective at driving systemic recovery than isolated health or mobility policies.

Caveat: The analysis relies on self-reported household survey data collected during a specific two-month window in 2021, which may not capture long-term structural changes or post-pandemic recovery dynamics.

arXiv · pdf

CityReal: Human-Aligned Urban Behavior and City Dynamics Simulation with Large-Scale LLM Agents

The researchers developed CityReal, a modular urban simulation framework that models agents as intention-driven decision makers pursuing coherent daily plans rather than isolated choices. The framework incorporates textual adapters for behavior modules to align agent decisions with observed population statistics, and it scales to simulate tens of thousands of agents.

Why it matters — It overcomes the limitation of previous LLM-based simulations that rely on few-shot prompting, which often causes agents to reflect the model's pre-existing behavioral biases rather than the actual target population.

arXiv · pdf

Context-Aware Slum Mapping in Sub-Saharan Africa Using Sentinel-1 Texture and Local Climate Zones

This study developed an Optical-SAR framework that combines Sentinel-2 spectral data with Sentinel-1 structural information to map informal settlements (LCZ 7) and distinguish them from compact low-rise areas (LCZ 3). Tested in Nairobi and Eldoret, Kenya, the model incorporates calibrated backscatter, GLCM textures, and a custom physics-guided feature to capture structural disorder. The framework achieved overall accuracies of 0.816 in the dry season and 0.807 in the wet season, reducing the misclassification between LCZ 3 and LCZ 7 to 7% compared to a 0.704 overall accuracy for the WUDAPT baseline.

Why it matters — The approach overcomes the seasonal vulnerability of optical-only mapping by using SAR-derived textures to stabilize informal settlement detection across wet and dry seasons. This provides a reliable, reproducible method for mapping complex urban morphologies in data-scarce Sub-Saharan African environments.

Caveat: The model's performance in cross-city transfer is limited without local adaptation strategies.

arXiv · pdf