Urban Currents

City landscape in sight: A crowdsourced framework for unlocking urban-scale window view perceptions from real estate imagery

The study developed a framework to map urban-scale window view perceptions using 12,334 real estate images from residential listings in Wuhan, China. Researchers collected 27,477 pairwise comparisons across six perceptual dimensions from 304 participants viewing 499 images, using these data to train a hybrid neural network that predicted perceptions across the entire dataset. The model revealed that higher floors correlate with preferred and extensive views, lower floors correlate with quietness and vividness, and high ratios of sky, trees, and low-rise buildings non-linearly enhance view preferences.

Why it matters — It establishes a scalable method to measure actual, rather than simulated, residential window views across an entire metropolitan area, proving that real estate listings can serve as a viable data source for urban visual quality assessments.

Caveat: The findings and predictive model are based on a single city, and the image dataset is limited to properties actively listed on real estate platforms, which may bias the sample toward newer or higher-value housing stock.

arXiv · pdf · doi

Reimagining urban planning for disaster resilience and post-pandemic sustainability in South Africa

This study synthesized scholarly literature, policy documents, and municipal planning frameworks published between 2000 and 2023 using a PRISMA-guided systematic qualitative review and thematic comparative synthesis. The analysis evaluated how South Africa's historical spatial inequality, fragmented governance, and infrastructure fragility interact to compound urban disaster vulnerability, using the 2020–2022 COVID-19 pandemic as a systemic stress test.

Why it matters — It reframes post-pandemic urban resilience in South Africa as a challenge of compounded risk governance rather than isolated hazard management, identifying specific integration points for resilience within municipal Integrated Development Plans (IDPs) and Spatial Development Frameworks (SDFs).

Caveat: The findings are based on a desktop-based qualitative review of existing literature and policy documents rather than new empirical field data.

Discover Cities · pdf · doi

Out-of-home toilet access, transport disadvantage, and dignified ageing in Africa

This conceptual paper establishes a framework linking out-of-home toilet access, transport disadvantage, and dignified ageing in Africa. It conceptualises transport disadvantage in later life as an embodied experience shaped by structural transport barriers, age-related physiological changes, and bodily needs during travel.

Why it matters — It shifts the research focus on toilet infrastructure and later-life mobility from Global North contexts to the distinct social, spatial, and infrastructural realities of Africa, establishing a new interdisciplinary agenda between transport geography, sanitation, and population health.

Caveat: The paper is a conceptual framework and agenda-setting piece rather than an empirical study presenting new primary data.

Journal of Transport & Health · doi

Relaxing or imposing constraints? Exploring the associations between digitalization and space-time flexibilities in Qinghe, Beijing

This study analyzes how foreground and background information and communications technology (ICT) use relates to perceived spatial and temporal flexibility. Using a 2-day combined daily activity and internet activity diary survey from 2021 in Qinghe, Beijing, the researchers applied multilevel ordered logistic regression at the activity level and ordinary least squares regression at the daily level to compare light and heavy ICT users.

Why it matters — It demonstrates that the impact of digitalization on daily schedules depends on how technology is used; while foreground ICT use generally correlates with higher perceived flexibility, frequent background ICT use among heavy users is actually associated with lower perceived temporal flexibility.

Caveat: The findings are based on self-reported perceived flexibility and diary data from a single subdistrict in Beijing, which may limit generalizability to other urban contexts.

Journal of Transport Geography · doi

The patronising city: adult-centrism and youth exclusion in urban environments

This paper introduces the concept of the 'patronizing city' and establishes an analytical framework that connects urban governance, institutional practices, and the design of built environments to adult-centered planning. The framework is illustrated through qualitative case studies to demonstrate how adult-centrism systematically excludes young people from political, social, and spatial urban domains.

Why it matters — It provides a new conceptual vocabulary and structural framework to analyze how urban planning practices actively marginalize youth, shifting the conversation on urban sustainability to include age-based spatial justice.

Caveat: The paper relies on illustrative cases rather than a systematic empirical dataset or comparative urban analysis.

npj Urban Sustainability · pdf · doi

Applying Natural Language Processing to Identify Critical Textual Phrases for Classification of Wrong-Way Driving Crashes

The study applied deep learning-based natural language processing models to classify wrong-way driving crashes from police-written crash narratives. Testing various n-gram tokenization approaches, the classifiers achieved prediction accuracies between 84% and 92%, with unigrams and bigrams yielding higher accuracy but more generic keywords, while trigrams and tetragrams identified more descriptive, critical phrases.

Why it matters — It demonstrates that automated text classification can reliably identify wrong-way driving incidents within large crash databases, offering traffic safety analysts a way to extract specific crash types from unstructured narratives without manual review.

Caveat: The performance of the classification model is dependent on the specific phrasing and reporting quality of law enforcement officers' written narratives.

Journal of Transportation Engineering Part A Systems · doi

Joint Lane Selection and Coordinated Signal Control for Mixed Traffic Arterials with CAV-Dedicated Lanes

This study develops a hierarchical control framework, called LS-MAPPO, that dynamically reallocates connected and automated vehicles (CAVs) between dedicated and mixed lanes alongside coordinated signal control. Simulation testing shows that under high traffic demand, the controller reduces average vehicle delay by 18.6% to 40.3% compared to benchmark methods. The analysis also reveals that a single CAV-dedicated lane reduces delay by 13.8% to 21.1% under heavy demand with a 40% to 60% CAV penetration rate, but increases delays by 31.7% to 85.9% at low penetration levels.

Why it matters — It quantifies the corridor-level impacts of CAV-dedicated lanes, demonstrating that restricting CAVs to dedicated lanes can be counterproductive at low penetration rates, and provides a dynamic lane-selection mechanism to prevent saturation imbalance.

Caveat: The performance and safety benefits of the control framework are evaluated solely through traffic simulations rather than real-world deployment.

Journal of Transportation Engineering Part A Systems · doi

Reclaiming the city: more-than-human solidarities between women and street dogs in Eskişehir

Through qualitative interviews in Eskiřehir, Turkey, this study examines how everyday care practices between women and street dogs shape the social production of urban space. It documents a reciprocal relationship where women advocate for the legitimacy of street dogs as urban co-habitants, while the presence of the dogs alters women's spatial experiences, allowing them to access and reclaim urban areas and times of day they would otherwise avoid due to safety concerns.

Why it matters — The study demonstrates that urban solidarity and space-making are not exclusively human activities, showing how multispecies relationships can actively disrupt and reshape patriarchal and anthropocentric urban structures.

Caveat: The findings are based on qualitative interviews in a single city, focusing on specific gendered and cultural dynamics that may not directly translate to other urban contexts.

Social & Cultural Geography · doi

Dyads and silences: critical plant studies and craft choices in writing Celtic rainforest

This paper outlines a creative writing methodology applied to the threatened ancient temperate rainforests of Coed Felenrhyd and Ceunant Llennyrch in North Wales. Drawing on critical plant studies and the concept of 'phytographia', the project produced a prose piece titled 'Litany of Hesitations' alongside a series of flash-fiction works focusing on micro-interactions between species.

Why it matters — It introduces two transferable writing techniques—paradoxical incantation and the 'lyric now' present tense—that help environmental writers and researchers move away from human-centered framings and capture precise, localized ecological relationships.

Caveat: The paper is a reflective, creative practice-based study focused on a specific literary project rather than an empirical or generalizable ecological assessment.

Cultural Geographies · doi

Sustainable Urban Greening of Tropical Asia: A Lightweight Vegetative Tile for Conventional Sloped Roofs of Sri Lanka

Researchers engineered and tested a lightweight, modular vegetative roof tile measuring 900 mm × 1200 mm with a wave-shaped corrugated profile to replace conventional sloped roofing materials in Sri Lanka. The fiber-reinforced plastic tile has a depth of 2.5 cm and a total saturated weight of 52.5 kg/m², meeting Eurocode structural standards. Biological testing demonstrated a 100% survival rate and over 80% canopy coverage for the grass species Axonopus compressus and Zoysia matrella.

Why it matters — This design provides a direct structural replacement for conventional sloped roofs, bypassing the high structural conversion costs and slope limitations that previously prevented the adoption of green roofs in tropical Asian cities.

Caveat: While the design methodology is intended for tropical Asia, the physical and biological testing was limited to Sri Lankan roofing standards and specific local grass species.

Urban Science · doi

ChatPlanner: A Large Language Model Framework for Personalized Public Transit Routing

The ChatPlanner framework integrates Large Language Models with Retrieval-Augmented Generation to extract routing parameters and user preferences from natural language queries, embedding them directly into a public transit routing algorithm's objective function. The system was trained and evaluated using custom preference-aware datasets featuring eight distinct personas and five contexts to calibrate routing scores. Testing across three experiments demonstrated that combining fine-tuning with retrieval-augmented generation successfully enforces structured outputs and resolves ambiguous conversational expressions.

Why it matters — It establishes a method for translating unstructured, subjective passenger preferences into quantitative routing constraints, allowing transit engines to generate personalized route alternatives that standard shortest-path planners overlook.

Caveat: The framework's performance and routing quality were validated using selected case studies rather than a large-scale, real-world deployment with diverse transit networks.

arXiv · pdf

Bridging Geographic Bias in Urban Streetscape Inference via Lifelong Learning with Visual-Semantic Pivoting

The researchers developed HVSP-LL, a lifelong learning framework that uses a three-tier visual-semantic ontology (macro, meso, and micro) and an equity-aware rehearsal mechanism to reduce geographic bias in streetscape perception models. Tested on a panoramic streetscape benchmark spanning 12 cities across four continents and seven perceptual dimensions, the model achieved a 0.834 Spearman correlation on held-out cities. This represents a 6.1 point improvement and a 38% reduction in the inter-city perception gap compared to the strongest continual learning baseline.

Why it matters — It provides a method for adapting streetscape inference models to new geographic regions sequentially without suffering from catastrophic forgetting or propagating systematic biases against underrepresented districts.

arXiv · pdf

A UAV-Based Multi-Modal Vision System for Automated Sideslope Deformation Monitoring and Hazard Detection

This study develops an automated workflow for monitoring expressway sideslope hazards using UAV-borne LiDAR. The system integrates a ground-surface extraction stage, a single-observation hazard-screening branch powered by RandLA-Net, and a multi-epoch deformation-monitoring branch that calculates grid-wise elevation differences. Field tests and controlled experiments demonstrate that the workflow successfully extracts ground-surface point clouds beneath vegetation cover and detects centimeter-level elevation changes.

Why it matters — It replaces manual, high-risk physical inspections of deteriorated slopes with an automated, aerial monitoring system capable of detecting subtle, centimeter-scale structural shifts over large areas.

arXiv · pdf

AI for Social Good: An Investigation of the Causal Relationship Between Environmental Regulations and Their Effects on Air Pollution in London, UK

An uncertainty-aware Bayesian deep learning framework was developed to estimate the cumulative impact of 32 air pollution policies on daily PM2.5 concentrations in Inner London from 2010 to 2020. The model integrates daily monitoring station data, meteorological variables, annual socioeconomic indicators, and policy status using a Bayesian LSTM and propensity score-based adjustments. The analysis found that London's regulations reduced PM2.5 levels by an average of 1.88 μg/m³ (a 12.35% relative reduction), with the strongest improvements occurring in 2018 and 2019.

Why it matters — It provides a method to isolate the causal impact of overlapping, non-randomly implemented urban environmental policies from confounding meteorological and socioeconomic trends, establishing a clearer timeline of when cumulative regulatory efforts actually began to yield measurable air quality improvements.

Caveat: The study relies on PM2.5 data specifically from Inner London monitoring stations, which may not fully capture the regulatory impacts on Outer London or areas lacking monitoring infrastructure.

arXiv · pdf

Reducing health and climate impacts of the global food system

A new, spatially-explicit model of the global food system was developed to analyze the intersection of dietary health and greenhouse gas emissions. The model demonstrates that while reducing red meat intake lowers emissions, its health benefits are modest, revealing that public health and climate mitigation goals in the food system are surprisingly independent.

Why it matters — It challenges the common assumption that dietary shifts automatically yield joint, proportional benefits for both human health and climate mitigation, proving instead that these two policy goals can be pursued and achieved independently.

Caveat: The summary of findings relies on a global modeling framework, which may smooth over localized nutritional needs or specific regional agricultural contexts.

arXiv · pdf