Modelling human perception of urban landscape and its spatial association with shared e-scooter usage in Lund

Abstract

Shared micromobility has been recognized as a promising contributor to sustainable urban transportation systems. Owing to their convenience and efficiency, shared e-scooters have gained widespread adoption globally and emerged as a rapidly growing mobility option. Understanding the relationship between subjective human perception of urban landscape and e-scooter usage intensity is crucial, as it could help planners identify street segments where perceptual improvements could promote e-scooter usage. Previous studies have focused on the impacts of socio-demographic and built-environment factors on e-scooter usage, while the role of subjective perceptual factors remains underexplored. This study investigates how urban perception from six dimensions (i.e.: aesthetic, boringness, depression, safety, vitality, and wealth) influences e-scooter usage intensity in Lund. Utilizing the MIT Place Pulse 2.0 voting dataset and Google Street View imagery from Lund, a deep CNN model (i.e.: ResNet50) was trained to predict urban perception scores along the cycling road. The resulting spatial distribution map reveals that the city center is perceived more positively, especially for boringness and vitality, whereas negative perceptions dominate in suburban areas. Strong positive global autocorrelation and varying degrees of local spatial dependence are detected across Lund, which justifies the spatial heterogeneity analysis. While OLS provides a global linear baseline and GWR captures local linear relationships, GWRF extends the analysis by integrating additional predictors, including POI-related built-environment factors. The OLS model detects multicollinearity between the six perceptual factors, and spatial non-stationarity justifies the usage of further analyses. The GWR results suggest that areas with higher e-scooter ridership tend to be perceived more positively, though the instability of local models warrants cautious interpretation. GWRF together with SHAP analysis further reveals that e-scooter usage is primarily associated with accessibility to urban facilities and potential travel demand, while perceived boringness and vitality play a secondary yet notable role. In light of the findings, operators should position shared e-scooters in areas that not only exhibit high demand but also feature livelier and more engaging urban environments. Lund municipality could invest in streetscape improvements, particularly in areas with low perceptual scores but high accessibility and population density, potentially unlocking shared e-scooter demand.

Popular Abstract

Electric scooters have grown rapidly in popularity in recent years, which offers an environmentally friendly way to travel short distances around the city. But what factors affect people’s choice of whether to ride one? Most research has focused on objective factors such as street connectivity, population density, and weather conditions. This study takes a different angle by examining whether the subjective perception, that is to say the appearance and vibe of streetscapes, affects how people use shared e-scooters in Lund. This question was explored using the MIT Place Pulse 2.0 dataset, Google Street View imagery across the cycling road in Lund, and artificial intelligence. The deep learning model, trained on over one million image comparisons made by volunteers around the world, was used to predict perception scores along the street from six dimensions: how beautiful, boring, depressing, safe, lively, and wealthy it looks. These predicted scores were then compared with data on where e-scooter trips actually took place in Lund. The results show that the city center of Lund is generally perceived as livelier and less boring, whereas rural areas tend to feel more negative. Also, areas where more e-scooter trips occur tend to be perceived more positively. However, when examining what is strongly associated with e-scooter usage, practical factors emerge as the most influential. For example, proximity to restaurants, hospitals, and shops, as well as the overall level of travel demand in the area. That said, how lively or boring a street feels plays a secondary but still meaningful role. In practice, the findings suggest that e-scooter operators should position the fleet in areas that are not only characterized by high demand but also perceived as lively and engaging, given their association with higher ridership. For Lund municipality, a worthwhile investment would be to make cycling streets more visually attractive. Such improvements may be associated with greater uptake of sustainable transport options, potentially supporting a shift toward greener urban mobility.

• author: Zong, Yue
• supervisor: Pengxiang Zhao; Rachid Oucheikh
• organization: Department of Earth and Environmental Sciences (MGeo)
• course: NGEM21 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: shared e-scooter, urban perception, perceived built environment, deep convolutional neural network, geographically weighted random forest, spatial heterogeneity, Google Street View, objective built environment
• publication/series: Student thesis series INES
• report number: 787
• language: English
• id: 9240606
• date added to LUP: 2026-06-24 09:30:36
• date last changed: 2026-08-03 11:56:17