Urban Currents

Mapping and comparing climate equity policy practices using RAG LLM-based semantic analysis and recommendation systems

The study introduces the Retrieval-Augmented Policy Analysis Framework (RAPAF), a system built with LangChain and ChatGPT to extract and analyze policy, strategy, and action items from climate equity and action plans. It also evaluates planning-related job postings to assess the demand for AI skills versus traditional planning roles, and implements a content-based recommendation system to identify semantic similarities in climate policies across different cities.

Why it matters — It demonstrates how large language models can automate the comparison of complex, unstructured policy documents across jurisdictions, allowing planners to quickly find cities with similar policy themes and strategies without manual coding.

Computational Urban Science · pdf · doi

Measuring willingness to pay for travel time savings, CO2 emissions reduction and calories burned when multiple modes are available: an application of the partial choice set approach

This study estimated willingness to pay (WTP) for travel time savings, carbon reductions, and physical effort using a stated choice experiment with a partial choice set design. Based on survey data from 1,255 residents in the car-dependent Solent region of the UK, the researchers modeled preferences across 15 long-journey and 8 short-journey transport alternatives. The analysis revealed a mean WTP of £4.31 per kg of CO2 reduced, showed that walking has the highest value of travel time savings among all modes, and found that willingness to burn calories during travel decreases with age.

Why it matters — It integrates environmental and health attributes directly into a multimodal transport choice model, demonstrating how environmental attitudes and age-related physical preferences segment the demand for active and sustainable travel.

Caveat: The findings are based on stated choice experiments from a single, highly car-dependent region in the UK, which may limit their generalizability to areas with different transport cultures or infrastructure.

Transportation · pdf · doi

More trees, more restorativeness? Exploring the influence of campus pedestrian space on stress recovery using street view imagery

The study combined 1,060 street-view images across four university campuses in Nanning with on-site Restoration Outcome Scale (ROS) scores to model perceived stress recovery. Using multiple regression and random forest models, the researchers achieved an R-squared of 0.602 (RMSE = 0.078) in predicting restorativeness, identifying vegetation cover and sky visibility as the most critical visual predictors.

Why it matters — It demonstrates that the relationship between campus greening and stress recovery is nonlinear, showing that vegetation cover actually reduces restorative quality after exceeding a certain threshold. This challenges the common planning assumption that maximizing green volume always yields better mental health outcomes.

Caveat: The predictive models and threshold findings are based on data from only four university campuses in a single city, which may limit their generalizability to other urban or regional contexts.

Computational Urban Science · pdf · doi

Spatio-temporal patterns of short-term rental clustering and land-use distribution in the greater Banjul Area, The Gambia

This study mapped the spatial and temporal distribution of short-term rentals (STRs) in the Greater Banjul Area, Gambia, comparing DBSCAN, HDBSCAN, and K-Means clustering methods. The analysis overlaid these clusters onto planned land-use zones and Points of Interest (POIs), finding that STR listings concentrate heavily during the peak tourism season of November to April. Residential zones contained the highest volume of listings with the lowest average price of USD 51.80, whereas designated tourism zones commanded the highest nightly rates at USD 100.67.

Why it matters — It provides some of the first empirical evidence on how short-term rentals interface with formal zoning and tourism seasonality in a rapidly urbanizing West African city, establishing a replicable spatial analysis framework for regional planners.

Caveat: The study relies on spatial clustering and proximity correlations to POIs, which may not fully capture the underlying socioeconomic drivers of STR development.

Discover Cities · pdf · doi

Teleconnection driven drought to flood sequences and spatially compound events in urban water systems for resilient cities

This review synthesizes evidence on how large-scale climate teleconnections, such as ENSO, the Indian Ocean Dipole, and the Madden-Julian Oscillation, drive sequential and compound drought-to-flood events in urban areas. It evaluates operational strategies for anticipatory urban water management, including teleconnection-aware seasonal forecasting, diversified storage portfolios, and modular stormwater controls.

Why it matters — It reframes urban water resilience from a reactive crisis response into a predictive, multi-hazard framework by linking global climate diagnostics directly to local infrastructure planning and governance.

Discover Cities · pdf · doi

Data-driven urban planning: Real population estimation using water consumption records and tourism data

The study developed a computational framework to estimate monthly de facto urban populations by combining smart-meter water consumption records, census data, climate normalization, and tourism registers. Applied to Alicante, Spain, between 2022 and 2024, the model uses a thermal-correction algorithm to isolate temperature-driven water demand and K-means clustering with Dynamic Time Warping to classify neighborhood seasonal patterns. Out-of-sample validation for 2024 achieved a Mean Absolute Percentage Error of 6.34% and an R-squared of 0.9874, capturing a summer population surge of 8.78% (+32,749 people) citywide, with coastal tourist zones growing by over 50% and wealthier central districts shrinking by up to 22.2%.

Why it matters — It provides a low-cost, privacy-preserving method to measure highly dynamic, seasonal population shifts at a neighborhood scale without relying on restrictive mobile phone data or static census records.

Caveat: The methodology is dependent on the presence of domestic smart-metering infrastructure, limiting its immediate application to cities already equipped with this technology.

Cities · pdf · doi

Why smart mobility alone does not ensure accessibility? Social, spatial, and digital dimensions of mobility as a service in island contexts

This study evaluates the social, spatial, and digital dimensions of Mobility as a Service (MaaS) for airport-to-city travel in Sardinia, Italy. It combines a survey of 391 users analyzing the relationships between perceived accessibility, satisfaction, and safety with a Space Syntax topological analysis of the physical transit corridors. The findings indicate that while public transit safety and affordability are rated highly, fragmented digital tools, poor pedestrian infrastructure, and peak-hour unreliability degrade the actual travel experience.

Why it matters — It demonstrates that technological integration alone cannot guarantee successful MaaS deployment, showing that digital mobility platforms fail unless they are supported by physical pedestrian connectivity and reliable transit operations.

Caveat: The empirical findings are limited to airport-to-city transit corridors within a single island region.

Cities · doi

Bridging circularity and inclusion in urban metabolic transitions: Through the lens of waste system evolution in Almere, the Netherlands

This study tracks the evolution of the municipal solid waste infrastructure system in Almere, Netherlands, from the 1970s to the 2020s. Using a combination of qualitative and quantitative indicators, it maps three distinct phases of sociotechnical change to evaluate how circular economy strategies interact with social inclusion.

Why it matters — It demonstrates that regulatory reforms and infrastructure upgrades alone are insufficient for equitable transitions, showing instead that the most transformative circular outcomes occur when grassroots innovations and inclusive governance are integrated into the formal waste system.

Caveat: The findings are based on a single longitudinal case study of a planned Dutch city, which may limit direct applicability to older cities or different national regulatory contexts.

Cities · pdf · doi

Advancing sustainable construction: A trust-enhanced model of AI adoption and realised use

This study developed and validated a trust-enhanced Technology Acceptance Model (TAM) using structural equation modelling on survey data from 1,233 construction professionals across Australia, the UK, and the US. The model measures how trust in AI, perceived ease of use, and perceived usefulness influence behavioural intention and self-reported actual usage of AI tools in construction workflows. The analysis reveals that trust significantly shapes both perceived ease of use and usefulness, while behavioural intention serves as the strongest direct predictor of actual AI use.

Why it matters — It shifts the focus of AI adoption research in the built environment from mere behavioural intention to self-reported actual usage, demonstrating that trust is a critical prerequisite for professionals to accept AI tools that affect project costs, safety, and professional liability.

Caveat: The measurement of actual AI deployment relies entirely on self-reported usage rather than objective system logs or administrative data.

Sustainable Cities and Society · doi

A framework for enriching survey datasets using passive data and machine learning, with an application to transferring attitudinal variables across transport surveys

This study develops and tests a predictive transfer learning framework that uses machine learning, regression algorithms, and passive data augmentation to enrich shortened survey datasets with variables from external sources. The framework was validated by transferring psychometric attitudinal variables to the U.S. National Household Travel Survey (NHTS), explaining up to 25% of the variance in observed attitudes and achieving correlations of up to 0.5 between predicted and actual values.

Why it matters — It establishes a systematic method to recover missing behavioral and psychological dimensions in large-scale surveys, allowing researchers to maintain high response rates through shorter questionnaires without sacrificing the explanatory power of attitudinal variables.

Caveat: The framework's predictive performance is moderate, capturing at most a quarter of the variance in the transferred attitudinal variables.

Transportation Research Part A Policy and Practice · doi

Induced e-bike adoption from income-conditioned rebates in British Columbia: who are the marginal purchasers?

This study analyzed survey data from 1,110 participants in two income-conditioned e-bike rebate programs in British Columbia, Canada, where rebate values ranged from CA$350 to CA$1,600. Using a latent class model, the researchers identified five distinct purchaser types and found that the average likelihood of an induced (marginal) purchase was 56%, scaling from 21% to 75% as the rebate value increased. The likelihood of a marginal purchase was significantly higher among lower-income households, regular commuters, younger buyers, and those who did not cycle prior to receiving the rebate.

Why it matters — It provides empirical evidence that income-conditioned rebate structures successfully target marginal buyers who would not have otherwise purchased an e-bike, offering a concrete strategy to maximize both the cost-effectiveness and social equity of active transportation subsidies.

Caveat: The findings rely on self-reported survey data regarding purchase motivations and hypothetical behavior, which may introduce recall or social desirability biases.

Transportation Research Part A Policy and Practice · pdf · doi

Optimizing cycling network design in the city of Parma

The study develops a Mixed-Integer Linear Programming (MILP) formulation to optimize bicycle network design under a constrained municipal budget, testing it on real-world data from Parma, Italy. The optimization strategy determines which infrastructure interventions to implement—such as building new lanes or upgrading existing ones—and recommends routes for specific origin-destination pairs by minimizing perceived travel costs based on cyclist preferences for safety, length, and practicability. To solve the model, the researchers introduce a branch-and-bound algorithm alongside two dynamic programming heuristic procedures based on knapsack problems.

Why it matters — It provides municipal planners with a mathematical framework to maximize the utility of limited infrastructure budgets by directly linking physical network investments to the subjective route preferences of cyclists.

Caveat: The optimization model's real-world performance is demonstrated only on the network of a single mid-sized Italian city.

Transportation Research Part A Policy and Practice · doi

Does public debt hinder transport decarbonization? Theoretical analysis and empirical evidence

This study analyzes the relationship between public debt and transport-sector carbon emissions using a panel CS-ARDL model on data from 31 economies spanning 2000 to 2021. The empirical analysis confirms a non-linear, U-shaped long-run relationship where public debt initially reduces transport emissions but increases them after surpassing a specific threshold. Additionally, the model evaluates the emission-reduction impacts of environmental taxes and climate change mitigation technologies.

Why it matters — It establishes a direct empirical link between national fiscal health and transport decarbonization, demonstrating that high public debt levels eventually constrain a government's capacity to sustain climate-aligned transport strategies, especially in low-growth and politically unstable economies.

Caveat: The study relies on macro-level national panel data, which may mask subnational variations in how local debt and municipal transport policies interact.

Transportation Research Part D Transport and Environment · pdf · doi

Developing and testing the effectiveness of a truck driver hazard anticipation training module: a simulator study

The study developed and evaluated a video-based hazard anticipation training module for novice truck drivers using a truck driving simulator and eye-tracking technology. Researchers measured driving metrics and glance behaviors across eight hazardous scenarios involving vulnerable road users (VRUs) before and after training, comparing a treatment group to a placebo control group. The trained drivers spent a significantly higher percentage of time looking at VRU-relevant areas of interest, though changes in driving speed and deceleration metrics were not statistically significant.

Why it matters — It demonstrates that a targeted, video-based training intervention can measurably improve how novice truck drivers allocate their attention toward vulnerable road users, offering a concrete method to enhance entry-level commercial driver safety programs.

Caveat: The study relied on a simulator environment rather than real-world driving conditions, and the abstract does not disclose the sample size used to establish statistical significance.

Transportation Research Interdisciplinary Perspectives · doi

An Embarrassingly Simple Rule-based Visiting Circulation Approach to Trip Destination Prediction

The Rule-based Visiting Circulation (RVC) model predicts trip destinations in a target metropolitan area without using any destination training data from that area. It relies on rule-based heuristics derived from individual revisiting behaviors and origin-destination relationships, bypassing supervised learning entirely. The model achieved second place in the IEEE Big Data Cup 2022 competition, outperforming supervised learning baselines in offline and leaderboard evaluations.

Why it matters — It demonstrates that simple, rule-based heuristics leveraging behavioral patterns can outperform complex supervised learning models in zero-shot spatial transfer tasks where destination labels are completely missing.

Caveat: The model's performance was evaluated within the specific constraints and dataset structure of a single data science competition.

arXiv · pdf · doi

Mobility and Contact Networks Shape Epidemic Outcomes: A Large-Scale Agent-Based Modeling Study

Using a synthetic population of one million agents representing an urban environment, this study evaluated how five different mobility models alter contact networks and epidemic trajectories. The models varied across activity patterns (empirical versus randomized) and destination choices (empirical popularity, distance-based, or random) while keeping disease parameters constant. The simulations revealed that mobility assumptions alone generate vastly different peak incidences and outbreak timings, driven by the spatial and social structure of contact opportunities rather than overall movement volume.

Why it matters — It demonstrates that epidemic simulations and policy interventions modeled in agent-based frameworks can be highly sensitive to uncalibrated mobility assumptions, meaning model outputs may reflect arbitrary movement rules rather than actual disease dynamics.

Caveat: The findings are based on a synthetic urban population and simulated mobility models rather than observed real-world epidemic outcomes.

arXiv · pdf