Urban Currents

Agent-based modelling of urban non-exhaust emissions reveals the limits of fleet electrification

An agent-based model of the Phibsborough junction in Dublin, calibrated with street-level Google Air View data, was developed to simulate non-exhaust traffic emissions. The model evaluated passenger-preserving policies, finding that a 50% electric vehicle fleet only reduces ultrafine particles (UFP) by 19% if regenerative braking is used, while increasing PM2.5 by 1% to 5% due to heavier vehicle mass. In contrast, shifting 50% of trips to buses yielded the largest reductions of 30% for UFP and 34% for PM2.5, while a 30 km/h speed limit cut UFP by 16.5% regardless of fleet composition.

Why it matters — The study demonstrates that vehicle electrification is insufficient on its own to solve traffic-related particulate pollution, proving that traffic dynamics and modal shifts to public transit are more effective levers for reducing non-exhaust emissions than changing fuel types.

Caveat: The model's calibration assumes that frequent rainfall suppresses road-surface wear and resuspended dust, meaning the findings may not directly apply to drier climates where these emission channels remain active.

Transportation Research Part D Transport and Environment · pdf · doi

A precise urban vehicle navigation method based on 3D velocity Mamba constraint

This paper introduces a GNSS/INS/Mamba-ODO navigation framework that maps raw IMU outputs directly to 3D velocity using asymmetric sequence segmentation, multi-scale temporal modeling, and a state-space sequence learning backbone. Tested in real-world ground vehicle experiments, the method achieves a mean 3D velocity estimation error of 0.22 m/s and maintains a 3D positioning error under 0.20 m during a 120-second GNSS outage.

Why it matters — It eliminates the reliance on manual installation angle calibration between the IMU and the vehicle frame, allowing vehicles to maintain sub-meter positioning accuracy in dense urban canyons where GNSS signals are blocked.

Urban Informatics · pdf · doi

Opportunities and risks of integrating urban allotments as niche innovations into transformative, sustainable urban development

This study analyzes how innovative urban allotment gardens can be integrated into municipal planning to drive sustainable urban development. Using a qualitative, multi-scalar research design in Germany, the researchers combined document analysis of municipal planning instruments, expert interviews with local administrations and allotment organizations, and site visits of selected innovative garden initiatives.

Why it matters — It demonstrates that the transformative capacity of urban allotments relies on embedding niche experiments into formal planning instruments and stable governance structures rather than relying on isolated pilot projects. The study provides a transferable evaluation grid that municipalities can use to identify planning gaps and leverage these gardens as multifunctional, climate-resilient infrastructure.

Caveat: The findings and the resulting evaluation framework are based on qualitative data and best-practice cases specifically within the German planning and governance context.

Discover Cities · pdf · doi

Informal urban development in Herat Afghanistan as a product of governance practices and institutional neglect

This study analyzes the persistence of informal urban development in Herat, Afghanistan, using data from one focus group discussion, 24 semi-structured interviews with urban officials and actors in Herat and Kabul, and a review of policy documents. The thematic analysis reveals that informality is driven by delayed planning, selective regulatory enforcement, institutional ambiguity, and reliance on extra-legal actors, spanning both low-income and elite-driven commercial and residential developments.

Why it matters — It demonstrates that informal urban development is not merely a consequence of state incapacity, but is actively reproduced through governance practices and legal ambiguity, functioning as an integral component of post-conflict urban governance rather than an exception to it.

Caveat: The empirical findings are primarily based on qualitative interviews and policy reviews focused on a single Afghan city, which may limit direct generalizability to other post-conflict contexts.

Discover Cities · pdf · doi

Asymmetric and spatially heterogeneous urban sensitivity to inter-day temperature variability: Evidence from geo-tagged social media

This study analyzed 1.9 million geotagged temperature-related social media posts across 367 Chinese cities in 2017, 2020, and 2022 to construct a city-scale thermal sensitivity index. The analysis revealed that public perception sensitivity to cooling is 2.3 times higher (0.114%/°C) than to warming (0.049%/°C) on average, though the gap narrows under extreme temperatures. High-sensitivity hotspots are clustered in southwestern China, while low-sensitivity coldspots are concentrated in the northwest and northeast.

Why it matters — It establishes that human sensitivity to temperature fluctuations is highly asymmetric and spatially heterogeneous, providing a method to use social media as a real-time sensor for subjective thermal stress to inform targeted urban early warning systems.

Caveat: The study relies on geotagged social media posts, which may introduce demographic biases as social media users do not perfectly represent the entire urban population.

Applied Geography · doi

What rights and whose rights count? Housing activism in urban redevelopment in Shenzhen

This study analyzes housing activism and urban redevelopment in Shenzhen, examining the actions of both property owners and non-owners, including migrant tenants, non-governmental organizations, and intellectuals. It details how municipal redevelopment initiatives, regulatory frameworks, market compensation, consent rules, police mobilization, and digital censorship collectively marginalize non-owning residents.

Why it matters — It demonstrates that Chinese urban redevelopment operates on a rationality of 'privileged recognition' that systematically prioritizes property owners and aesthetic space, showing that the 'right to the city' framework must be expanded to include non-owners who are typically cast as passive victims.

Caveat: The findings are based on a qualitative analysis of a single Chinese megacity, which may limit direct generalizability to cities with different regulatory and property rights regimes.

Cities · pdf · doi

Safe speeds in rear-end collisions: Injury-risk thresholds from population-weighted event data recorder analysis

This study analyzed 216 two-vehicle rear-end crashes from the US Crash Investigation Sampling System (2016–2024) using Event Data Recorder measurements to model injury risk. Binary logistic regression models calculated the probability of minor, moderate, and serious-or-fatal injuries based on both striking-vehicle speed and relative impact speed. The findings show that moderate-injury risk reaches a 10% probability at an impact speed of 43 km/h, while serious-injury risk hits the 10% threshold at an impact speed of 88 km/h.

Why it matters — It establishes empirical, population-weighted speed thresholds for rear-end crash injuries, correcting previous unweighted estimates that overrepresented crash severity. These specific velocity limits provide concrete benchmarks for calibrating autonomous emergency braking systems and setting urban speed limits under Safe System frameworks.

Caveat: The analysis is based on a relatively small sample size of 216 crashes, which may limit the precision of the risk curves at extreme speeds.

Transport Policy · doi

An AI- and IoT-Enabled intelligent transportation architecture for multimodal corridor Integration: A Port-Centric case study

This study designs a layered digital architecture for Iraq's Grand Faw Port and the broader Development Road Corridor, integrating IoT systems, Port Community Systems, Terminal Operating Systems, AI predictive planning, and digital twins. The framework is built using a semi-quantitative benchmarking of leading global smart ports, a Digital Maturity Index to evaluate transferable practices, and conceptual systems modeling.

Why it matters — It provides a blueprint for developing countries to transition from isolated port technologies to integrated, corridor-scale digital systems, addressing the institutional and governance barriers that typically stall multimodal transport efficiency.

Caveat: The proposed digital architecture and maturity index are conceptual models and have not yet been deployed or empirically tested in the Grand Faw Port.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Strategizing sustainable freight transport: structural insights and prioritization among economic, social, and environmental pillars

This study analyzes the structural dynamics of the Sustainable Freight Transport (SFT) index using a four-stage analytical pipeline that combines expectation-maximization clustering, Bayesian network tree-augmented naive Bayes classification, partial least squares path modeling, and importance-performance map analysis. The model evaluates the interrelationships between economic, social, and environmental pillars, finding that the environmental dimension is the strongest driver of overall performance, while the social pillar acts as a key mediator transmitting the benefits of economic and environmental improvements.

Why it matters — It establishes the structural architecture and hierarchical dependencies between different sustainability dimensions in freight systems, moving beyond isolated pillar analyses to show how social outcomes depend on environmental and economic progress.

Caveat: The analysis relies on cross-sectional data, which limits the ability to observe how these structural relationships and priorities evolve over time.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Resilient Cities and Inclusion: Urban Design as a Response to the Ukrainian Crisis in Poland

This study evaluated the urban integration of collective accommodation centers (CACs) for Ukrainian refugees in Poland using an experimental, three-phase research framework that combined artificial intelligence with GIS environments. Focusing on the five largest CACs, the analysis assessed their integration into the surrounding neighborhoods against the principles of the 15-minute city and documented local integration initiatives.

Why it matters — It demonstrates that while local integration events occurred, they lacked permanence, meaning these accommodation centers missed the opportunity to leave a lasting, positive physical or social imprint on the surrounding urban landscape.

Caveat: The study focuses exclusively on the five largest collective accommodation centers, which may not represent the dynamics of the wider, more dispersed network of smaller centers across Poland.

Journal of Urban Planning and Development · doi

<i>Printing Nueva York: Spanish-language print culture, media change, and democracy in the late nineteenth century</i> , by Kelley Kreitz

This study analyzes the history of Spanish-language print culture, newspapers, and periodicals published in New York City during the late nineteenth century, examining how these media channels interacted with democratic processes and media change.

Why it matters — It documents the historical role of Spanish-language media in shaping democratic discourse and public spheres within a major US metropolis during a period of rapid media evolution.

Caveat: The provided abstract contains very little detail about the specific empirical findings, sample sizes, or archival scope of the book.

Journal of Urban Affairs · doi

<i>Unruly comparison: Queerness, Hong Kong, and the Sinophone</i> , by Alvin Wong

This book develops a comparative framework at the intersection of queer theory, Sinophone studies, and Hong Kong cultural studies to analyze alternative ways of seeing Hong Kong's cultural landscape.

Why it matters — It provides a new theoretical lens for understanding marginalized sexualities and cultural identities within the specific geopolitical context of Hong Kong and the broader Sinophone world.

Journal of Urban Affairs · doi

SnapScope: A Platform for City-Scale Collection and Exploration of Public Snap Map Data

The researchers developed SnapScope, an open-source platform for harvesting and analyzing public, geotagged Snapchat stories at a city scale. They tested the system in Riyadh, Saudi Arabia, collecting 515,364 unique snaps over 23 days using a 1 km grid of 2,740 query points, with a saturation probe showing that 94.8% of returned records in consecutive runs were duplicates.

Why it matters — It provides a reproducible, city-agnostic pipeline and interactive interface to collect and explore ephemeral social media data that was previously inaccessible due to the lack of a documented API.

arXiv · pdf

Mapping Armenian Paris: Extracting and Geocoding Commercial Advertisements from the 20th-Century Diaspora Press

The researchers built an end-to-end pipeline using vision-language models to locate, read, parse, and geocode commercial advertisements from 20th-century digitised Armenian newspapers in France. The pipeline was evaluated on a newly created corpus of 500 pages of the Western Armenian press containing 3,270 annotated advertisements, successfully mapping the historical Parisian Armenian commercial community despite the curved scans and lack of standard OCR support for Western Armenian.

Why it matters — It demonstrates that vision-language models can successfully bootstrap structured spatial data from highly under-resourced historical languages and distorted physical scans where conventional optical character recognition workflows fail.

arXiv · pdf

Fuzzy network jump models for soft dynamic clustering of graph-structured data

The study introduces a fuzzy network jump model designed for soft dynamic clustering of time-varying data on weighted graphs, incorporating spatial and temporal regularization to ensure smooth cluster transitions. The model's performance was validated through simulations across varying spatial dependencies and applied to a traffic-network dataset from San Francisco to identify evolving traffic regimes across connected road segments.

Why it matters — It provides a new capability to model and cluster time-varying network data with soft, probabilistic memberships rather than hard boundaries, allowing researchers to track how spatial-temporal phenomena like traffic congestion gradually transition between states.

Caveat: The empirical application of the model is demonstrated on a single city's road network, meaning its performance on different network topologies remains to be fully documented.

arXiv · pdf

Accurate Localization of Road Traffic Objects on the Road Plane Using Surveillance Camera Imagery

A two-stage geometry-aware pipeline was developed to localize vehicles by projecting their footprints onto the road plane using monocular roadside surveillance cameras. The system combines a YOLO-based detector with a ResNet34 regression network to predict the four corner points of a vehicle's base, which was trained on CARLA synthetic data and fine-tuned on the DAIR-V2X dataset. On real-world imagery, the method reduced the mean image-space localization error by 51.8% from 31.77 to 15.30 pixels, and cut median ground-plane errors for medium-range vehicles from 5.52 m to 0.90 m.

Why it matters — It overcomes the severe localization inaccuracies caused by perspective distortion and parallax in standard bounding-box-center methods, particularly for large or distant vehicles captured by elevated cameras.

Caveat: The pipeline's performance gains rely heavily on contextual information surrounding the detector bounding box, which may vary under different environmental or occlusion conditions.

arXiv · pdf

Organizational and Socio-Technical Challenges in UAV Incidents: Evidence from a Practitioner Focus Group

This study conducted an in-person focus group with UAV and counter-UAV practitioners from US government and industry organizations to analyze how drone incidents are managed in real-world settings. Qualitative analysis of the session identified five primary operational hurdles: poor airspace visibility, fragmented interorganizational reporting, technical forensic limitations, legal gaps, and insufficient training capacity.

Why it matters — It shifts the focus of UAV forensics from purely technical data extraction to the socio-technical and organizational realities of incident response, mapping out systemic coordination and policy barriers that practitioners face.

Caveat: The findings are based on a single qualitative focus group with US-based practitioners, which may not capture the regulatory and operational realities of other jurisdictions.

arXiv · pdf

Investigating Artificial Intelligence Digital Sovereignty in Mobile Shopping Apps: A Case Study of Nigeria

This study analyzes the presence of AI features and the transparency of their disclosure in Nigerian Android mobile shopping applications using a forensic analysis of selected apps combined with contextual document analysis. The findings reveal widespread implementation of AI technologies across these e-commerce platforms, contrasted by very limited disclosure and transparency regarding how the AI is used.

Why it matters — It provides empirical evidence on how the lack of platform transparency in a rapidly growing e-commerce market undermines individual digital sovereignty, establishing a baseline of user awareness and control in a developing digital economy.

Caveat: The study relies on a qualitative, interpretive approach and forensic analysis of selected apps rather than a comprehensive, market-wide audit of all mobile shopping platforms.

arXiv · pdf

To See a World in a Living Context: Unified Indoor-Outdoor Urban World Generation

The researchers developed HoloWorld, a text-driven 3D generation framework that synthesizes unified indoor and outdoor urban environments using a continuously updated cross-scale world context. The system autoregressively generates urban exteriors across blocks and grounds them in 3D building footprints to generate geometry-constrained, matching indoor layouts. In evaluations, the framework improved the average Aesthetic Quality Score (AQS) by 7.68% over current state-of-the-art models and achieved the highest average Realistic Detail Ratio (RDR).

Why it matters — It establishes a method to generate coherent 3D virtual cities where individual building interiors maintain explicit spatial, geometric, and visual correspondence with their external structures, a capability previously limited by the independent synthesis of indoor and outdoor environments.

arXiv · pdf

The Impact of Cut-ins on Mixed-Autonomy Traffic Flow: Bridging Microscopic Game-Theoretic Model and Macroscopic Flow Analysis

This study integrates a microscopic Stackelberg game-theoretic model of human-CAV cut-in maneuvers into a macroscopic kinematic wave framework to analyze mixed-autonomy traffic. The model identifies an exploitation window where human drivers execute aggressive merges in front of defensive automated vehicles, translating these interactions into a continuous friction parameter that modifies the traffic conservation law. Numerical simulations reveal a convex relationship between CAV penetration and system efficiency, with a critical instability regime peaking at approximately 45 percent penetration.

Why it matters — It mathematically links individual strategic driving behaviors to network-level traffic instability, proving that the defensive programming of early-stage automated vehicles paradoxically triggers phantom traffic jams and reduces overall road capacity.

Caveat: The findings are based on theoretical analysis and numerical simulations rather than empirical, real-world mixed-autonomy traffic data.

arXiv · pdf

Thermodynamic statistics of given names in USA and France

This study analyzes the frequency and popularity distributions of given names in the USA and France over a period of more than 100 years using official government datasets. By applying Lorenz and Pareto curves, the analysis reveals that name distributions remain stable over time with a Gini coefficient consistently between 0.85 and 0.95. The researchers model these distributions using the Rayleigh-Jeans thermalization and condensation framework from physics, finding that name popularity levels behave like energy states with stable time correlations until the mid-twentieth century, after which they changed significantly.

Why it matters — It demonstrates that the dynamics of cultural choices, specifically naming conventions, follow the same thermodynamic statistical laws and inequality distributions as wealth accumulation and physical systems.

arXiv · pdf

Longitudinal wearable monitoring and polygenic risk for incident major depressive disorder in the All of Us Research Program

The study integrated genomic data, electronic health records, and longitudinal Fitbit wearable data from 3,030 adults of genetically inferred European ancestry to model the risk of major depressive disorder (MDD). Using time-varying Cox models, it tracked 284 incident MDD cases over a baseline period of 180 days, measuring the interaction between MDD polygenic risk scores (PRS) and monthly physical activity and sleep metrics. Combining PRS, baseline and monthly wearable features, and their interactions improved model discrimination, raising the C-index from 0.637 to 0.705.

Why it matters — It demonstrates that objective, long-term wearable data and genetic risk scores provide complementary predictive value for depression, showing that individuals with high genetic liability require more favorable behavioral patterns, such as higher step counts and stable sleep, to maintain a comparable risk profile to those with lower genetic risk.

Caveat: The study population was limited to participants of genetically inferred European ancestry, which restricts the generalizability of the genetic risk findings to more diverse populations.

arXiv · pdf

Studying People to Study AI: Expert Perspectives on the Epistemic Fit and Barriers of Human Research in AI Safety & Ethics

This study surveyed 93 experts and interviewed 17 researchers across technical, sociotechnical, governance, and normative backgrounds to analyze the barriers to adopting human-subject research in AI Safety and Ethics. The findings reveal that while researchers generally value human-centered evidence, its integration is hindered by perceived validity issues, resource constraints, and methodological preferences, with technical researchers showing the lowest valuation of human methods and the lowest rates of interdisciplinary collaboration.

Why it matters — It identifies and documents the specific epistemic and structural barriers that prevent human-centered empirical research from being integrated into AI safety evaluations, offering a framework to prevent 'human-washing' in technical benchmarks.

Caveat: The findings rely on self-reported survey and interview data from a self-selected sample of AI safety and ethics researchers.

arXiv · pdf

Hybrid Machine Learning Framework for Herd-Level Cattle Growth Pattern and Weight Gain Forecasting in Grazing-Based Production Systems

A hybrid machine learning framework was developed to forecast herd-level cattle weight using automated sensing data collected between 2022 and 2024 in southeastern Australia. The system aggregates animal-level predictions into herd trajectories using weekly live weights, demographics, and lagged environmental variables. A cascade architecture combining Gradient Boosting, Random Forest, and Neural Networks outperformed recurrent sequential baselines, achieving a test R^2 of 0.889, an RMSE of 21.319 kg, and an MAE of 15.462 kg.

Why it matters — The framework demonstrates that hybrid machine learning architectures maintain predictive robustness even when livestock observations are sparse and irregular, a common limitation in commercial grazing operations.

Caveat: Forecasting errors scale up progressively as the prediction horizon is extended.

arXiv · pdf