Source-linked AI summary
Smelly Maps: The Digital Life of Urban Smellscapes
Daniele Quercia, Rossano Schifanella, Luca Maria Aiello, Kate McLean
TL;DR
Urban smell strongly shapes how people perceive places, yet it has been overlooked partly because it is difficult to record and analyze at scale. The paper builds an urban smell dictionary from smellwalks and matches its terms to geo-referenced social-media data from Barcelona and London. It finds ten smell categories and reports that selected categories correlate with governmental air-quality indicators, supporting the study’s validity.
Problem
Urban smell influences place perception but is overlooked by planners and scientists because it is difficult to capture and existing methods are limited.
Method
The paper builds an urban smell dictionary from smellwalks and matches its words with geo-referenced Flickr, Instagram, and Twitter data from Barcelona and London.
Results
Smell-related words were best classified into ten categories, while categories including industry, transport, and cleaning correlated with governmental air-quality indicators.
Takeaways & Limitations
Social-media data can map both dominant and localized urban smells and provide methodological tools for studying urban smell environments.
Takeaways & Limitations
Most smell groups would occur in the vast majority of cities, especially in densely populated areas containing food, waste, and materials.
Abstract
from arXiv · showhide
Smell has a huge influence over how we perceive places. Despite its importance, smell has been crucially overlooked by urban planners and scientists alike, not least because it is difficult to record and analyze at scale. One of the authors of this paper has ventured out in the urban world and conducted smellwalks in a variety of cities: participants were exposed to a range of different smellscapes and asked to record their experiences. As a result, smell-related words have been collected and classified, creating the first dictionary for urban smell. Here we explore the possibility of using social media data to reliably map the smells of entire cities. To this end, for both Barcelona and London, we collect geo-referenced picture tags from Flickr and Instagram, and geo-referenced tweets from Twitter. We match those tags and tweets with the words in the smell dictionary. We find that smell-related words are best classified in ten categories. We also find that specific categories (e.g., industry, transport, cleaning) correlate with governmental air quality indicators, adding validity to our study.
1 Introduction
Urban smell matters but remains difficult to capture at scale, limiting the tools available to planners and researchers. This paper explores social-media-based citywide smell mapping using a smell dictionary built from smellwalks.
- Smell affects behavior, attitudes, health, and perceptions of urban places.Street food markets are given as an example of smells changing how entire streets are perceived.
- Urban planners and scientists have largely overlooked general smell, while existing tools remain limited because smell is difficult to capture.Research has often focused on bad odors or air-pollution characteristics rather than broader urban smell.
- The paper explores using social-media data to reliably map the smells of entire cities.This approach is presented as a way to enrich the methodological toolkit for urban smell research.
- Smellwalks across seven cities produced verbatim smell descriptors that were classified into the first urban smell dictionary.Locals identified distinct odors and recorded them in handwritten notes; the dictionary is intended for public release.
- Social-media data from Barcelona and London combined geo-referenced Flickr, Instagram, and Twitter content with dictionary words.The datasets included about 530K Flickr pictures, 35K Instagram photos, and 113K geo-referenced tweets.
- Smell-related words were best classified in ten categories, with the automated classification closely resembling manually produced field-research classifications.The paper also reports correlations between categories such as industry, transport, and cleaning and governmental air-quality indicators.
2 Why Smell
Smell contributes to how people remember, identify, and socially experience cities, yet urban studies and planning have often prioritized visual information. Ignoring urban odors can leave collective city images incomplete and contribute to homogenized places and reinforced socio-economic boundaries.
- Urban experience is shaped by multiple senses, but urban studies have historically privileged sight over smell.The paper frames smell as part of the broader sensory range missing from conventional urban analysis.
- Not knowing city smells can produce partial views of the collective image of a city.Odors affect memory and are retained for longer periods, linking smell to how places are mentally represented.
- Odors contribute to place identity, so overlooking them can encourage homogenized, sterile, and controlled urban areas.The paper connects neglected place-identity odors with the proliferation of “clone towns.”
- Smells can reveal and reinforce socio-economic boundaries within cities.Fast-food odors are associated with rundown areas, and the paper links fast-food-chain locations with neighborhood deprivation.
3 Related work
Existing methods for recording urban odors are difficult to scale: devices capture selected odor properties, web mapping depends on substantial participation, and sensory walks provide fine-grained but geographically limited data. The paper therefore identifies a need for scalable odor collection without massive public engagement.
- People can detect up to 1 trillion smells, but urban smellscapes remain poorly mapped because smell is difficult to record, analyze, and visualize.The paper reviews several methodological approaches for recording urban smells.
- Odor-recording devices such as olfactometers capture odor character, intensity, duration, and frequency, while smell cameras trap volatile molecules.Public agencies commonly use olfactometers to verify odor-nuisance complaints.
- Web-based participatory odor mapping can scale in principle, but it depends on engaging enough people to contribute.Users annotate pre-designed maps with odor markers.
- Sensory walks collect fine-grained olfactory data through incrementally broadening experiences of the urban environment.Urban smellwalks build on the sensory-walk tradition used in research on cities’ sonic and tactile environments.
- Previous odor-collection methods are not scalable without unrealistic public engagement or coverage limited to small geographic areas.The paper motivates a new way to collect odor information at scale without requiring massive participation.
4 Methodology
The paper combines smellwalk observations, a 285-term urban smell dictionary, and geo-referenced social-media data to classify and map urban smells. It constructs ten hierarchical categories from word co-occurrences and validates their structure against prior research and street-level patterns.
- Urban smell dictionary: Smellwalks across cities recorded participants’ odor descriptions, locations, expectations, intensity, associations, and hedonic responses for building the dictionary.Descriptors were transcribed from handwritten notes and integrated with previous odor classifications.
- Urban smell dictionary: The dictionary contains about 285 English smell terms, translated into Spanish, after conservative three-annotator coding and removal of ambiguous words.The authors caution that the list is not exhaustive and may not fully represent actual place smells.
- Social-media data: Geo-referenced Flickr, Instagram, and Twitter content from London and Barcelona was matched against the dictionary’s smell words.The datasets included public photos, captions, hashtags, tags, and tweets; a manual check found that 85% of 100 sampled Flickr pictures related to smell.
- Urban smell classification: A Flickr co-occurrence network connected smell words by their frequency of appearing together in picture tags.Community detection used Infomap followed by iterative modularity-based splitting, with manual merging of overly fine-grained subcommunities.
- Urban smell classification: The resulting classification has ten main categories arranged in hierarchical structures with depths ranging from 0 to 3.Figure 2 displays only the first hierarchical level for brevity.
- Validation: The automatically generated classification closely resembles Henshaw’s manual system, while adding a metro category and producing mostly uncorrelated category pairs.Street-level correlations suggest that emissions and nature are complementary, with emissions rarely found where greenery is present.
5 Results
The analysis maps urban smell at citywide and localized scales, comparing smell categories with air-quality indicators and examining how spatial choices affect correlations. It finds distinct city smell profiles, positive emissions–pollutant and negative nature–pollutant relationships, and localized patterns for several categories.
- Base Notes of Urban Smell: Base notes are citywide smell patterns: Barcelona is dominated by food and nature, whereas London is characterized by traffic emissions and waste.The paper computes each category’s fraction of Flickr tags as a high-level olfactory footprint.
- Base Notes of Urban Smell: Across Flickr, Instagram, and Twitter, pollutant concentrations correlate positively with emissions and negatively with nature.Correlations are lower for Instagram and Twitter, especially in Barcelona, with dataset size, geographic salience, and location errors offered as explanations.
- Base Notes of Urban Smell: 25 meters best balances relevant street-level data against sparsity, while larger buffers slightly degrade correlations and buffers below 20 meters excessively restrict the study unit.The authors associate oversized buffers with loose smell–street links and undersized buffers with data sparsity.
- Base Notes of Urban Smell: Heatmaps use z-scores, with zero indicating a city-average presence of a smell category or pollutant and positive or negative values indicating above- or below-average conditions.This transformation supports visual comparison of smell and air-quality patterns across street segments.
- Mid-level Notes of Urban Smell: Mid-level smell notes are localized rather than dominant: food clusters around markets and restaurants, while waste and smoking cluster in evening-economy areas.Examples include Boqueria and Borough markets, restaurant districts in Barcelona, and evening-economy areas in both cities.
- Mid-level Notes of Urban Smell: Cleaning and chemical smells appear around Shoreditch, industrial facilities, hospitals, and major railway stations.The reported locations include Sant Adria, Hospital San Pau, and King’s Cross.
6 Discussion
The discussion frames urban smell mapping as a promising but inherently limited way to interpret cities, because odor perception varies across people, cultures, environments, and time. It identifies applications for planning, computing, arts, and public engagement while emphasizing that social-media traces only partly address smellwalk biases.
- Limitations: Urban odor perception is individually, socially, and contextually situated, making the resulting dataset nuanced rather than exhaustive.Personal characteristics, cultural setting, city layout, weather, and time of day can all shape detected odors.
- Limitations: The classification may omit odors from cities with extreme climates because it is based primarily on northern European contexts.The authors nevertheless state that many groups should recur in densely populated cities, including food, waste, and materials.
- Urban Planning: Urban smell maps could help planners manipulate airflow, traffic, vegetation, and stopping points while preserving or celebrating pleasant smells.The authors connect these interventions to reducing emissions, depicting green-space olfactory perception, and creating multisensory places.
- Computer Science, Arts & Humanities, and Public Engagement: The methodology could support olfactorily informed way-finding tools, creative mapping practices, and public participation in evaluating urban smell.Proposed applications include pleasant-route suggestions, artistic visualization, and a public voice on smell’s positive and negative roles.
- Opportunities: Social media can reduce smellwalk sample and response biases by enabling unobtrusive data collection from potentially more representative user samples.The authors report that digital traces can track urban odors when combined with smell-related words.
7 Conclusion
The conclusion presents the work as an unobtrusive contribution to research on how people sense cities, addressing the relative lack of work on urban smell. It also identifies future work on multisensory comparison and fleeting odors that may require new crowdsourcing tools.
- Conclusion: This work is presented as the first examination of using social media to map urban smell environments unobtrusively.The authors aim to provide methodological tools and practical insights for designers, researchers, and city managers.
- Future Work: Future research will compare sound, visual, and olfactory aesthetics within the same city.The authors describe this as a more comprehensive multisensory evaluation.
- Future Work: Fleeting odors remain difficult to capture through social media because they are localized in space and time.The authors suggest that new mobile applications may be needed to crowdsource these high notes of the urban smellscape.