Smelly Maps: The Digital Life of Urban Smellscapes
University of Cambridge
University of Torino
Luca Maria Aiello
Royal College of Art & CCCU
Smell has a huge inﬂuence over how we perceive places.
Despite its importance, smell has been crucially over-
looked 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 ur-
ban 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 governmen-
tal air quality indicators, adding validity to our study.
Smells impact our behavior, attitudes and health. Street food
markets, for example, have dramatically changed the way
we perceive entire streets of global cities.
Despite its importance (which we will detail in Sec-
tion 2), smell has been crucially overlooked (Section 3). City
planners are mostly concerned with managing and control-
ling bad odors. Scientists have focused on the negative re-
search aspects of smell as well: they have studied air pollu-
tion characteristics (often called ‘environmental stressors’)
rather than the more general concept of smell. As a result,
the methodological tools at the disposal of researchers and
practitioners are quite limited. Smell is simply hard to cap-
To enrich the urban smell toolkit, we here explore the pos-
sibility of using social media data to reliably map the smells
of entire cities. In so doing, we make the following contri-
• One of the authors of this paper ventured out in the urban
world and conducted “smellwalks” around seven cities in
2015, Association for the Advancement of Artiﬁcial
Intelligence (www.aaai.org). All rights reserved.
UK, Europe, and USA (Section 4.1). Locals were asked
to walk around their city, identify distinct odors, and take
notes. Smell descriptors are taken verbatim from the smell
walkers’ original hand-written notes. As a result of those
sensory walks, smell-related words were recorded and
classiﬁed, resulting in the ﬁrst urban smell dictionary,
which we will make publicly available to the research
• For the cities of Barcelona and London, we collected geo-
referenced tags from about 530K Flickr pictures and 35K
Instagram photos, and 113K geo-referenced tweets from
Twitter (Section 4.2). We matched those tags and tweets
with the words in the smell dictionary.
• We found that smell-related words are best classiﬁed in
ten categories (Section 4.3). Our classiﬁcation, generated
automatically from social media, is very similar to clas-
siﬁcation systems obtained manually as a result of ﬁeld
• We also found that speciﬁc categories (e.g., industry,
transport, cleaning) correlate with governmental air qual-
ity indicators (Section 5), and that speaks to the validity of
our study. Finally, we show that, using social media data,
we are able to capture not only a city’s dominant smells
(base smell notes) but also localized ones (mid-level smell
These results open up new opportunities (Section 6). Our
ultimate goal is to open up a new stream of research that
celebrates the positive role that smell has to play in city life.
2 Why Smell
Our daily urban experiences are the product of our per-
ceptions and senses (Quercia, Schifanella, and Aiello 2014;
Quercia et al. 2015), yet the complete sensorial range is
strikingly absent from urban studies. Sight has been his-
torically privileged over the other senses. In the early six-
ties, Jane Jacobs stressed the importance of visual order in
the city (Jacobs 1961), and Kevin Lynch focused on the vi-
sual dimensions of urban design (Lynch 1960). When odor
is mentioned in the built environment literature, it is gener-
ally in negative terms. We do not have to go that far back
in time to ﬁnd the ﬁrst positive reference to smell by a cel-
ebrated architect: in 2005, Juhani Uolevi Pallasmaa brieﬂy
arXiv:1505.06851v1 [cs.SI] 26 May 2015
highlighted smell in the second part of his well-known book
‘The Eyes of the Skin’ (Pallasmaa 2012).
The problem is that not knowing what smells exist in
cities may result in:
Partial views of the collective image of the city. Good cities
are those that have been built and maintained in a way that
they are imageable, i.e., that their mental maps are clear
and economical in terms of mental effort (Lynch 1960).
Urban studies have explored how imageability is affected
by the ease with which people memorize visual images of
the city. Yet, memory is affected not only by what we see
but also by what we smell. Smell and long-term memory
are closely related and, more importantly, odor associa-
tions are retained for much longer time periods than vi-
sual images (Engen 1991). Research into how we build
the collective image of the city has overlooked the signif-
icant role that smell plays in our urban well being.
The proliferation of clone towns. Odors contribute to a
place’s identity. Place identity odors are often overlooked
by city planning professionals who either do not notice
them, or do not attribute any value to them. This re-
sults into the proliferation of homogenized, sterile and
controlled areas that have “an alienating sense of place-
ness.” (Drobnick 2006). Because of globalization, the typ-
ical UK town or city has become a ‘clone town’: “a place
that has had the individuality of its high street shops re-
placed by a monochrome strip of global and national
chains”. Clone towns can only offer homogenized olfac-
tory environments (Reynolds 2009).
Reinforcing socio-economic boundaries. Odors contribute
to the construction of a place’s socio-economic identity.
The greasy odors coming from fast food restaurants are
often associated with rundown areas and with the evening
economy. (Macdonald, Cummins, and Macintyre 2007)
found a signiﬁcant positive relationship between the loca-
tion of the four most popular fast food chains in UK and
neighborhood socio-economic deprivation. Smells that
provide insights into the social life of cities are used as
an invisible marker in reinforcing socio-economic bound-
aries. If we do not know what smells exist, we are likely
to reinforce those boundaries without even knowing it.
3 Related work
People are able to detect up to 1 trillion smells (Bushdid et
al. 2014). Despite that, there are limited maps of this poten-
tially vast urban smellscape. One reason is that smell is prob-
lematic to record, to analyse and to depict visually. Here we
review a variety of methodological approaches for recording
Recording odors with devices. Olfactometers have been
used to collect information about distinct odor molecules.
They look like ‘nose trumpets’ and capture four main as-
pects: odor character, odor intensity, duration, and fre-
quency. Public agencies usually use them to verify com-
plaints about odor nuisances. Other smell recording tech-
nologies include a head-space ‘smell camera’. This device
traps volatile odor molecules in a vacuum and is able to cap-
ture permanent (i.e., non ﬂeeting) smells.
Recording odors with the Web. Online participatory map-
ping allows web users to annotate pre-designed base maps
with odor markers (Henshaw 2013). This method promises
to be scalable, but engaging enough people to participate is
Recording odors with sensory walks. Social science re-
search has increasingly used the methodology of sensory
walks. The earliest example of a sensewalk was undertaken
in 1967 by Southworth with a focus on the sonic environ-
ments of cities (Southworth 1967). During the soundwalk,
participants were involved in increasingly focused tasks of
attentive listening. Urban smellwalks are similarly designed.
They consist of incrementally broadening experiences to
cover the wider olfactory environment. A few multi-sensory
researchers have engaged in such walks.
In Vienna in 2011 the philosopher Madalina Diaconu
ran a project exploring the meanings and associations of
the tactile and olfactory qualities of the city researching
through the noses of a group of students (Diaconu 2011).
Meanwhile, Victoria Henshaw conducted her smellwalks
in Doncaster, England (Henshaw 2013). Most of our work
is based upon her research ﬁndings. The problem is that
the sensory walk methodology collects ﬁne-grained data
but is not scalable. To see why, consider that an individual
walk typically took Henshaw three non-continuous months,
involved six participants, and covered approximately 160
To sum up, previous odor collection methodologies are
not scalable. Web-based methodologies would be only under
the unrealistic assumption of massive public engagement.
By contrast, sensory walks successfully engage one individ-
ual at the time but, when carried out over several years, they
only result in data about limited geographic areas. We thus
need a new way of collecting odor information at scale with-
out requiring a massive public engagement.
This work proposes a new way of doing so from data implic-
itly generated by social media users. The idea is to search for
smell-related words on geo-referenced social media content.
To do so, we need those words and the content itself, both of
which are described in the following section.
4.1 Urban Smell Dictionary
When attempting to control and enforce odor law and
policies, city authorities face the well-known difﬁculty of
recording, measuring, describing, and classifying odors.
Smells do not lend themselves easily to quantitative mea-
surement. However, scientists have long attempted to pro-
pose a uniﬁed odor categorization system.
Aristotelean Classiﬁcation. Aristotle, for example, divided
odours into six separate classes, later amended by Linnaeus
City #participants #smells
Amsterdam 44 650
Pamplona 58 374
Glasgow 20 55
Edinburgh 10 40
Newport 30 80
Paris 10 25
New York 20 43
Table 1: The set of smellwalks whose data has been used to
integrate previous odor classiﬁcations. For example, in Am-
sterdam, the smellwalks involved 44 local residents over 4
days in April 2013 and resulted in the collection of 650 smell
perceptions, which include background smells, episodic and
in 1756 to seven; aromatic, fragrant, alliaceous (garlic), am-
brosial (musky), hircinous (goaty), repulsive and nauseous.
UCLA Odor Classiﬁcation. One issue is that, to describe
smells, people often name potential sources of smell rather
than actual odors. They, for example, use terms such as
sulphurous, eggy, ﬂoral, earthy, or nutty. To partly ﬁx this
problem, (Curren 2012) developed an urban odor descriptor
wheel that includes the words people use to describe speciﬁc
odors (e.g., grease) along with their chemical names (e.g, 2-
Henshaw’s classiﬁcation. For her PhD, Victoria Henshaw
set out to document odors present in the city of Doncaster.
She did so by conducting a number of smellwalks. These
smellwalks followed a pre-planned set of routes. Each route
was designed in a way that provided exposure to a range of
different smellscapes. The route included a set of stopping
points (e.g., mixed-used developments, busy bus routes, eth-
nically diverse residential areas, business areas, markets). At
each stopping point area, a range of questions regarding the
smells that the participants detected were asked. Those ques-
tions invited insights not only about annoyance and distur-
bance (as it is usually done) but also about positive percep-
tions of smell. After each smellwalk, the weather, temper-
ature, time, and activities taking place were recorded. The
different walks were carried out in periods of cold weather
(e.g., January-March) and of warmer weather (e.g., April-
July), on weekdays and weekends, and at different times of
the day from 7am to 8pm. Henshaw also combined her in-
sights with those offered by the Vivacity2020 Project that
investigates urban environmental quality. This combination
resulted into a classiﬁcation of urban odors along 11 types:
trafﬁc emissions, industrial odors, food and beverages, to-
bacco smoke, cleaning materials, synthetic odors, waste,
people and animals, odors of nature, building materials, and
Additional Smellwalks. One of the authors of this pa-
per complemented Henshaw’s classiﬁcation by conducting
smellwalks in other cities across the UK, Europe, and USA
(Table 1). These walks mainly involved local people. Partic-
ipants identiﬁed distinct odors and recorded their location,
description, expectation, intensity, personal association and
hedonic scale. Smell descriptors are taken verbatim from the
original hand-written notes. Figure 1 overlays some notes
gathered at the Amsterdam’s smellwalk on the city map.
Comprehensive smell dictionary. To build a smell dictio-
nary, we hand-code the previously discussed literature and
the hand-written notes from the smellwalks. Speciﬁcally, we
use line-by-line coding to generate a set of words conceptu-
ally associated with smell. Three annotators independently
generated a list of words that relate to olfactory perceptions.
We then combine the three lists using the most conserva-
tive approach; we take their intersection (rather than union).
We double-checked the resulting list removing potentially
ambiguous tags (e.g., the word ‘orange’ can refer to a fruit,
a color, or a smell). The result of the processes explained
above is the ﬁrst urban smell dictionary containing some
285 English terms
. Since our analysis considers not only
London but also Barcelona, we also manually translated the
terms into Spanish. By visual inspection, one sees that all the
words in the dictionary are related to the domain of smell.
However, by no means, do they represent an exhaustive list.
Therefore, it is not clear whether we will observe any rela-
tionship between the presence of speciﬁc smell words in a
place and the actual smell of the place.
4.2 Social Media Data
Having deﬁned the smell words, our next step is to gather
social media data against which those words are matched.
Flickr. Out of the set of all the public geo-referenced Flickr
pictures, we selected a random sample of 17M public photos
taken within the bounding boxes of London and Barcelona.
For each picture, we collected the anonymized owner identi-
ﬁer and the free-text tags attached to the photo by the owner.
Instagram. To obtain a sizable sample of Instagram pic-
tures, we collected data for a random set of 5.1M users
whose accounts were public. We collected all of their
“feeds” for a three-year period between December 2011 to
December 2014. The collection resulted in about 154M im-
ages and videos along with their meta data including hash-
tags, captions, and geo-references. Using the picture geo-
location, we selected photos taken in London and Barcelona,
for a total of 436K images.
Twitter. We gather geo-referenced tweets. Using the Twitter
API, we collected 5.3M tweets during year 2010 and from
October 2013 to February 2014. Out of those, we selected
the 1.7M geo-referenced tweets in London and Barcelona
after ﬁltering out retweets and direct replies.
4.3 Urban Smell Classiﬁcation
We then textually parsed our geo-referenced items (which
are tags in Flickr, hashtags and captions in Instagram, and
To support future work in the area, the link to the urban dic-
tionary will be provided in the camera-ready version.
Figure 1: Visualization of the hand-written annotations taken by participants of the smellwalk in Amsterdam.
Users Items Smell words Street segments Users Items Smell words Street segments
Flickr 28.381 454.484 593.602 27.232 8.366 74.381 102.876 14.952
Instagram 5.509 30.432 58.522 11.654 1.513 5.637 11.314 5.380
Twitter 16.214 109.269 125.137 9.373 816 3.915 4.670 2.245
Table 2: Dataset statistics for Flickr and Instagram photos and for tweets.
tweets in Twitter) and searched for exact matches with the
dictionary words. Table 2 summarizes the size our datasets
together with the total number of matched smell words. To
verify whether those words matched pictures that actually
related to smell, we manually checked 100 random Flickr
pictures and found that 85% of the pictures did so.
The next stage was to create a structure for a large and
apparently unrelated dataset of smell words through a sys-
tem of classiﬁcation. We ﬁrst built a co-occurrence net-
work where nodes are smell words and undirected edges
are weighted with the number of times the two words co-
occur in the same Flickr pictures as tags (Flickr is the dataset
containing the highest number of smell words). We built
this co-occurrence network because the semantic relatedness
among words naturally emerges from the network’s commu-
nity structure: semantically related nodes are those that are
highly clustered together and weakly connected to the rest
of the network. To determine the community structure, we
could use any of the literally thousands of different com-
munity detection algorithms that have been developed in the
last decade (Fortunato 2010). None of them always returns
the “best” clustering. However, since Infomap has shown
very good performance across several benchmarks (Fortu-
nato 2010), we opt for using it to obtain the initial partition
of our network (Rosvall and Bergstrom 2008). This parti-
tion results in many clusters containing semantically-related
words, but it also results in some clusters that are simply
too big to possibly be semantically homogeneous. To fur-
ther split those clusters, we iteratively apply the community
detection algorithm by Blondel et al. (2008), which has been
found to be the second best performing algorithm (Fortunato
2010). This algorithm stops when no node switch between
communities increases the overall modularity
. The result of
those two steps is the grouping of smell words in hierarchical
categories. Since a few partitions of words might be too ﬁne-
grained, we manually double-check whether this is case and,
if so, we merge all those sub-communities that are under the
same hierarchical partition and that contain strongly-related
If one were to apply Blondel’s right from the start, the resulting
clusters are less coherent than those produced by our approach.
Figure 2: Urban smellscape taxonomy. Top-level categories
are in the inner circle; second-level categories, when avail-
able, are in the outer ring; and examples of words are in the
Figure 2 sketches the resulting classiﬁcation. It has ten
main categories, each of which has a hierarchical structure
with variable depth from 0 to 3. For brevity, the ﬁgure re-
ports only the ﬁrst level.
This classiﬁcation is of good quality because of two main
reasons. First, despite spontaneously emerging from word
co-occurrences, our classiﬁcation strikingly resembles Hen-
shaw’s. The only difference between the two is that ours has
the category metro
Second, our smell categories are ecological valid and are
mostly orthogonal to each other. They are ecologically valid
because, later on, will study their distribution across Lon-
don streets and see that they behave as expected (Section 5).
For now, just consider the pairwise correlations between the
presence of a category and that of another category at street
level in London (Figure 3). By looking, for example, at the
last row (that of the category emissions), we see that the
most complementary category to it is nature: this means that
gas emissions are rarely found where greenery is found (and
vice-versa). More importantly, the whole correlation matrix
suggests that the vast majority of category pairs show no
correlation, and that is good news because it implies that our
categories are orthogonal and, as such, the clustering algo-
rithms have done a good job.
This category’s words refer to public transportation facilities,
and might well be a Flickr-speciﬁc artifact: London subway sta-
tions have long been of photographic interest and, as a result, might
be overrepresented on Flickr.
Figure 3: Pairwise correlations between presence of smell
categories at street level in London
4.4 Air quality of streets
The olfactory experience of a city is inevitably inﬂuenced
also by the quality of the air, measured by the amount of
pollutants that are emitted in the atmosphere by several hu-
man activities. It is useful to clarify the differences between
air pollutants and odors. Air pollutants are chemicals that,
when released into the air, pose potential harm to human and
environmental health. These chemicals may or may not be
detected through the human senses (McGinley, Mahin, and
Pope 2000). Some air pollutants have odors (e.g., benzene
has a sickly sweet odor) while others, such as carbon monox-
ide, cannot be detected through the senses of smell. Air pol-
lution is the world’s largest single environmental health risk,
being the cause of one in eight of the total premature global
deaths, according to the World’s Health Organization
few pollutants are systematically measured in cities:
CO. Carbon Monoxide is a colorless, odorless poisonous
gas produced by the incomplete or inefﬁcient combustion
of fuel. The gas affects the blood’s transport of oxygen
around the body and to the heart.
. Nitrogen oxides are formed during high-temperature
combustion processes from the oxidation of nitrogen in
the air. It is a noxious gas with serious health implica-
tions: eye irritation, irritation of the respiratory system,
and shortness of breath.
. Ozone is not directly emitted, but is formed by a com-
plex set of chemical reactions. Like N O
, high levels of
can irritate and inﬂame the lungs, possibly causing mi-
graine and coughing.
P M10, P M 2.5. These are coarse particles (P M10) and
ﬁne particles (P M 2.5) that are linked to lung cancer
and asthma. They are named for the size, in microns, of
the particles. Particulate matter smaller than about 10µm
(P M10) can settle in the bronchi and lungs and cause
health problems. PM2.5 is the smallest and most danger-
ous sort of particulate matter (particles less than 2.5µm in
diameter) and can enter deep into the respiratory system.
. Sulphur dioxide results from burning coal or oil, and
it makes buildings crumble and lungs sting.
It is not easy to assess health risks by comparing pollu-
tants, not least because pollutants come at different concen-
tration levels at any point in time. To ease comparison, the
air-quality index (AQI) has been introduced. This rescales
the concentrations of a given pollutant in the range from 1
(low risk) to 10+ (“hazardous” for all).
In London, we collect air quality indicators as the pub-
lic API provided by the environmental research group at
King’s College London
allows us to do. More speciﬁcally,
we are able to collect AQI values directly from air qual-
ity tracking stations: AQI values for NO
from 90 sta-
tions, for P M10 from 77, for O
from 25, for P M 2.5
from 20, and for SO
from 13. From those numbers, it
is clear that not all pollutants are measured by all track-
ing stations. We are also able to collect the predicted pol-
lutant concentration values of NO
, P M 10, and P M 2.5
for every single street. These values are accurately esti-
mated by advanced models of dispersion assessments (Beev-
ers et al. 2013). In a similar way, in Barcelona, we gather
the predicted N O
pollution concentration for every street.
The values are estimated by the regression models devel-
oped within the ESCAPE project (Eeftens, M et al. 2012;
Beelen et al. 2013).
4.5 Mapping data onto streets
All this social media and air quality data now needs to be
mapped. A street segment is the unit with the ﬁnest-grained
spatial resolution that is common to all our sets of data. A
segment is a street’s portion between two road intersections.
We gathered street segment data for Central London (36.755
segments) and Barcelona (44.044 segments) from Open-
StreetMap (OSM) (a global group of volunteers who main-
tain free crowdsourced online maps). After mapping our so-
cial media data onto street segments, each segment ends up
being characterized by the presence or absence of words
(i.e., of smell categories) within it. Since geo-referencing
comes with positing errors, we buffer each street’s poly-
line with an area of 22.5 meters on each side. This means
that data mistakenly positioned within that buffer area is still
considered part of the segment.
By having our social media and air quality data mapped onto
street segments, we are now able to study how those two
datasets are statistically related. However, before doing so,
we need to introduce the concept of odor notes.
5.1 Odor Notes
To best interpret our results, we should think about the dif-
ferent levels at which place odors can be considered. To
ease illustration, we resort to Victoria Henshaw’s analogy
taken from the perfume industry. When a new perfume is
created, different top, middle, and base note ingredients are
combined to make the new fragrance. Those notes differ in
terms of their tenacity. Top notes are those perceived im-
mediately (e.g., citrus fruits, aromatic herbs) and, since they
are intense, they are also volatile and evaporate quickly. By
contrast, base notes are those adding depth and stay on the
skin for hours (e.g., wood, moss, amber, and vanilla). Mid-
dle notes sit somewhere in between (e.g., ﬂowers, spices,
berries). The urban smellscape is similarly composed:
Base notes. The macro-level base notes for the urban
smellscape are those that are likely smelled by a city’s
ﬁrst-timer visitors. That is because known odors are un-
consciously processed by people, while only unfamiliar or
strong odors are brought to people’s attention (as potential
threats or sources of pleasure). As a result, residents are
not likely to pay attention to their city’s base notes, while
visitors would be able to consciously process them.
Mid-level notes. As one moves through the city, the base
notes blend with dominant smells that are localized in spe-
ciﬁc areas (e.g., factories, ﬁsh markets).
High notes. Finally, the micro-level high notes are short-
lived odors (e.g., goods from a leather shop). These are
emitted in points that are very localized in space and time.
With our analysis based upon social media, we are after
the detection of base notes (uniformly distributed across the
city) and mid-level notes (localized in speciﬁc areas of the
city). High notes are likely to go undetected because of data
sparsity and because of our spatial unit of analysis being a
5.2 Base Notes of Urban Smell
To capture the base notes of the urban smellscape, we con-
sider our 10-category classiﬁcation of urban smells (Fig-
ure 2) and compute the fraction of Flickr tags that match
each of them. This gives us a high-level olfactory foot-
print of the city. Barcelona is predominantly characterized
by smells related to food and nature, while London is char-
acterized by smells related to, trafﬁc emissions and waste
(Figure 4). As The Economist puts it: “In 2013 the an-
nual mean concentration of N O
on Oxford street [in Lon-
don] was one of the highest levels found anywhere in Eu-
rope.” (TheEconomist 2015). The predominance of trafﬁc
over nature comes as no surprise since smells of trafﬁc pol-
lution have been found to overlay and mask more subtle
smells. Air pollutants also have been found to reduce the
ability of ﬂoral scent trails to travel through air.
However, critics might rightly say that the predominance
of certain smell words over others might well come from
data artifacts and might not reﬂect the actual street odors ex-
perienced on the streets. To partly counter that argument, we
test weather air quality conditions are related to the presence
of speciﬁc smell words by answering this question:
Figure 4: Distribution of smell categories in London and
Q1. Do air quality indicators correlate with speciﬁc smell
categories as expected? To answer that, we compute the
fraction of each segment’s tags that belong to a given smell
category. We do that for segments having at least 30 geo-
referenced smell tags to avoid data sparsity. For each seg-
ment, we thus have a 10-element smell vector (as there are
10 smell categories) and a set of air quality vectors reﬂect-
ing the pollutant concentrations on the segment. We need
to compute the correlation between the smell vectors and
the air quality vectors. However, when high spatial auto-
correlation occurs, traditional metrics of correlation such as
Pearson require independent observations and cannot thus
be directly applied. To overcome this problem, we used a
statistical method introduced by Clifford et al. (Clifford,
Richardson, and Hemon 1989). This approach addresses the
“redundant, or duplicated, information contained in georef-
erenced data” (Grifﬁth and Paelinck 2011) – the effect of
spatial auto-correlation – through the calculation of a re-
duced sample size. After doing so, we ﬁnd our hypothe-
sized relationships to hold: pollutant concentrations are pos-
itively correlated with the category of (trafﬁc) emissions in
the smell vectors (r = 0.47 in London and r = 0.29 in
Barcelona for N O
, p < 0.001) and are negatively corre-
lated with the category of nature (r = −0.33 in London,
r = −0.35 in Barcelona for N O
, p < 0.001). One might
wonder whether those results change across social media
Q2. To which extent the smell categories of emissions and
nature correlate with the air quality indicators, if the data
comes from different social media sites? Across the three
sites of Flickr, Instagram and Twitter (Figure 5), the corre-
lations of pollutant concentrations are consistently positive
with emissions (red bars) and negative with nature (green
bars). The correlations are lower for Instagram and Twitter,
slightly in London and moderately in Barcelona. That might
be explained by three main reasons. First, the smaller the
dataset, the lower the correlations (and Twitter is the small-
est dataset among the three, as Table 2 showed). Second,
Twitter is less geographically salient than Flickr, as it has
Figure 5: Correlations between smell and pollutants, across
Figure 6: Correlation with NO
vs. buffer size
been previously shown (Quercia et al. 2013). Third, the lo-
cation errors might differ across users of different services.
As this might be yet another factor contributing to those dif-
ferences in the correlation results, we test it next.
Q3. Do our correlations depend on the size of the street
buffer? They do but not to a great extent. As the size in-
creases, the correlations slightly degrade (Figure 6). There-
fore, if the buffer is too large, then matched tags are only
loosely related to what is actually happening on the street.
By contrast, if the buffer is too small (say below 20 meters),
the spatial unit of study is excessively restricted, resulting
into data sparsity problems. For both cities, a buffer size of
25 meters best strike a good balance between having relevant
data and avoiding sparsity.
After having analyzed the relationship between pollutants
and smell categories, we test whether, by mapping those cat-
egories, we are able to conﬁrm what we have found so far.
For an easier representation on heatmaps, we transform our
smell and air quality vectors into the corresponding z-scores.
Street segments with zero values are those experiencing the
city’s average presence of a smell category or of a pollutant.
Those segments with values below (above) zero are experi-
encing conditions below (above) the city’s average. As one
expects, the nature category is present where the emissions
category is absent, and vice-versa (Figure 7). In London,
Hyde Park experiences high levels in the nature category,
and, conversely, the trafﬁcked streets at its boundaries expe-
rience high levels in the emissions category. In Barcelona,
(a) London, nature (b) London, emissions (c) London, animals
(d) Barcelona, nature (e) Barcelona, emissions (f) Barcelona, animals
Figure 7: Heatmaps of smell-related tag intensity
the same goes for Montjuic Park and Park Guell, and for the
nearby streets of Ronda Litoral and the Travessera de Dalt.
Finally, the last column of Figure 7 shows the heatmaps for
animal smells, that is registered around both Barcelona’s and
London’s zoos. In London, other smaller animal small hold-
ings (e.g., Vauxhall City Farm) also emerge.
5.3 Mid-level Notes of Urban Smell
With social media, we have seen that we are able to cap-
ture background smells (base-level smell notes). Now we
will show that we are also able to capture smells localized
in speciﬁc areas (mid-level smell notes). Consider the maps
showing the presence of some of the remaining smell cate-
gories (other than the three we have discussed) in London
(Figure 8) and Barcelona (Figure 9). Those maps suggest
that the remaining smell categories are not dominant but are
localized in speciﬁc areas:
Smells of food are localized around food markets (Bo-
queria market, Barcelona; Borough Market, London) and
in areas where restaurants tend to cluster (Born and
Smells of waste and smoking are found in areas enjoying
the evening economy (Barceloneta, and Bogatell beach,
Barcelona; Blackfriars, and Elephant and Castle, Lon-
don). That is partly because, the encouragement of the
evening economy has increased the levels of waste in
city streets (e.g., odors of urine, and cigarette) (Hen-
shaw 2013), in particular following the criminalization of
smoking in enclosed public places.
Strong presence of cleaning (product) smells are detected
in Shoreditch, London. More expectedly, those smells
plus chemical ones are detected around big industrial fa-
cilities (Sant Adria, Barcelona), hospitals (Hospital San
Pau, Barcelona), and big railways stations (King’s Cross,
This work aims to engage with academics and built environ-
ment professionals who are passionate about the multisen-
sory experience of cities. It highlights the positive role that
‘smell’ as opposed to ‘air pollution’ can play in the environ-
mental experience. Next, we discuss some of the limitations
of our work and some of the opportunities it opens up.
The way people perceive odors is individually, socially, and
contextually situated creating a nuanced dataset.
Individual factors. Individual factors. Personal character-
istics affect smell perceptions. Females have higher olfac-
tory performance than males. Age has a limiting inﬂuence
on smell performance, with 50% of people experiencing
a major loss in olfactory function over the age of 65. Fi-
nally, smoking habits reduce smell performance (Venne-
mann, Hummel, and Berger 2008).
Socio-cultural factors. Urban odor classiﬁcation is an at-
tempt to model the range of background and episodic odors
detected and reported. However, it is not an exhaustive list-
ing: cities in parts of the world with extreme climates, such
as high humidity or sub-zero temperatures. are likely to
be characterized by odors not identiﬁed in our northern
European-based classiﬁcation. Having said that, we should
Figure 8: Mid-level notes of urban smell for London
stress that most of the smell groups in the classiﬁcation
would be present in the vast majority of cities. Afterall,
where there are densely populated areas, there will always
be food, waste and materials.
Contextual factors. Urban planning and the resulting city
layout have signiﬁcant impact on odor detection in the city.
The grid layout of New York City, for example, encourages
large-scale collective odor experiences as it was designed in
a way to facilitate airﬂow using prevailing westerly winds
to dissipate the disease-carrying miasmas of the late 18th
century. In October 2005, a sweet, sirupy odor was detected
across the city. The smell was pleasant (i.e., a combination
of maple syrup and caramel), yet it resulted into hundreds
of calls to the city’s emergency services. ‘The aroma not
only revived memories of childhood, but in a city scared by
terrorism, it raised vague worries about an attack deviously
cloaked in the smell of grandma’s kitchen’ (DePalma 2005).
Also, long-distance smell detection is highly temporal, de-
pendent on weather conditions, wind patterns, and seasonal
waves of activity, with air temperature directly inﬂuencing
odor strength and volatility. Finally, in a twenty-four-hour
city, the same street will host different activities at different
times: activities associated with, for example, the cafe and
retail culture during the daytime, and drinking culture in the
Despite these limitations, our work offers new ways of fa-
cilitating olfactory interpretations of places for a variety of
Urban Planning. One hundred sites In Japan have been de-
clared as protected because of their ‘good fragrance’. How-
ever, the general situation in the rest of the world greatly
differs: urban planners to date have tended to think about
Figure 9: Mid-level notes of urban smell for Barcelona
smells in terms of management of bad odors rarely consid-
ering preserving and celebrating the smells that people like.
There are a number of ways that the urban smellscape can be
altered; manipulating the air ﬂow by changing the street lay-
out, pedestrianization to alter trafﬁc emissions (categories
from Figure 2 can be mapped onto cities to add weight to
arguments to reduce emissions), the creation of restorative
environments through the planting of trees, greenspaces and
waterways (categories from Figure 2 can be mapped over
a variety of cities to depict olfactory perception of green
spaces), and the strategic placement of car stopping points
are just a few examples. City ofﬁcials do not fully consider
the opportunities presented by the sense of smell simply be-
cause they have been the victims of a discipline’s negative
perspective. We hope that our work might help them re-think
their approaches and use olfactory opportunities to create
stimulating multi-sensory places.
Computer Science. In the near future, new way-ﬁnding
tools might well suggest not only shortest routes between
points but also short ones that are olfactorily pleasant (e.g.
runners might wish to avoid emission-infused streets). Our
methodology allows for the development of new tools to
map urban smellscapes.
Arts & Humanities. Contemporary art, design and philoso-
phy tend towards a phenomenological understanding, using
our senses to constantly rediscover the world we live in. In
this vein, one of the co-authors collects and analyses olfac-
tory data derived from smellwalks, visualizing the scents and
their possible locations in the city using a variety of creative
mapping practices. Her work exhibits internationally incor-
porating data visualisation of the smellscape and a variety of
synthetically and naturally made scents. Olfactory artists are
likely to proﬁt from our methodology, incorporating a wider
range of digital traces in their production of artwork.
Public Engagement. In addition to academic research, the
general public might also beneﬁt by contributing to the de-
velopment of a critical voice for the positive and negative
role that smell has to play in the city.
The urban smellscape is a complex set of sensorial frag-
ments, and it is debated as to whether a smellscape can ever
be fully known. For example, when we consider a landscape
it forms a continuous, integrated and deﬁned space whereas
the smellscape is a dynamic and ﬂuid entity. It is impossi-
ble for a number of humans to detect the entire smellscape
of an area as a whole at any one point in time. Smellwalks
only partly solve those problems since they suffer from two
main biases: a) sample bias (participants are not representa-
tive of the general population); and b) response bias (people
might perform tasks in the walk differently than how they
would ‘in the wild’ because of the Hawthorne effect (Mccar-
ney et al. 2007)). Social media partly reduces both biases, in
that, representative user samples can be extracted (reduced
sample bias), and data can be captured unobtrusively (with
lack of experimental demands, the response bias will be lim-
ited). In addition to being unobtrusive, social media appears
insightful for capturing elements of urban smellscapes: our
results suggest that it is possible to effectively track urban
odors from digital traces when combined with smell-related
words learned from smellwalks. This result should come as
no surprise to practitioners of mixed methods research.
We have contributed to the growing body of literature on
how people sensually experience the city. There has been
research on how we see the city and on how we hear the
city, but not much on how we smell the city. This work is
the ﬁrst in examining the role of social media in mapping
urban smell environments in an unobtrusive way. We hope
to empower designers, researchers, city managers by offer-
ing them a number of methodological tools and practical in-
sights to re-think the role of smell in their work.
In the future, we would like to conduct a more com-
prehensive multi-sensory evaluation by exploring how the
sound, visual, and olfactory aesthetics compare in the same
city. We are also interested in capturing ﬂeeting odors. As
these are localized in space and time (the high notes of the
urban smellscape), they cannot be easily identiﬁed on social
media and might require the design of new mobile apps to
facilitate crowdsourcing their collection.
[Beelen et al. 2013] Beelen et al. 2013. Development of NO2 and
NOx land use regression models for estimating air pollution expo-
sure in 36 study areas in Europe – The ESCAPE project. Atmo-
spheric Environment 72(0):10 – 23.
[Beevers et al. 2013] Beevers, S. D.; Kitwiroon, N.; Williams,
M. L.; Kelly, F. J.; Anderson, H. R.; and Carslaw, D. C. 2013.
Air pollution dispersion models for human exposure predictions in
london. Journal of Exposure Science and Environmental Epidemi-
[Blondel et al. 2008] Blondel, V. D.; Guillaume, J.-L.; Lambiotte,
R.; and Lefebvre, E. 2008. Fast unfolding of community hierar-
chies in large networks. Journal of Statical Mechanics.
[Bushdid et al. 2014] Bushdid, C.; Magnasco, M. O.; Vosshall,
L. B.; and Keller, A. 2014. Humans Can Discriminate More than
1 Trillion Olfactory Stimuli. Science 343(6177):1370–1372.
[Clifford, Richardson, and Hemon 1989] Clifford, P.; Richardson,
S.; and Hemon, D. 1989. Assessing the signiﬁcance of the cor-
relation between two spatial processes. Biomentrics 45:123–134.
[Curren 2012] Curren, J. 2012. Characterization of Odor Nuisance.
[DePalma 2005] DePalma, A. 2005. Good Smell Vanishes, But It
Leaves Air of Mystery. The New York Times October 29
[Diaconu 2011] Diaconu, M. 2011. Senses and the City: An Inter-
disciplinary Approach to Urban Sensescapes. LIT Verlag.
[Drobnick 2006] Drobnick, J. 2006. The Smell Culture Reader.
[Eeftens, M et al. 2012] Eeftens, M et al. 2012. Development
of Land Use Regression Models for PM2.5, PM2.5 Absorbance,
PM10 and PMcoarse in 20 European Study Areas; Results of
the ESCAPE Project. Environmental Science & Technology
[Engen 1991] Engen, T. 1991. Odor Sensation and Memory.
[Fortunato 2010] Fortunato, S. 2010. Community detection in
graphs. Physics Reports 486(3-5):75 – 174.
[Grifﬁth and Paelinck 2011] Grifﬁth, D., and Paelinck, J. 2011.
Non-standard Spatial Statistics and Spatial Econometrics. Ad-
vances in Geographic Information Science. Springer.
[Henshaw 2013] Henshaw, V. 2013. Urban Smellscapes: Under-
standing and Designing City Smell Environments. Routledge.
[Jacobs 1961] Jacobs, J. 1961. The Death and Life of Great Ameri-
can Cities. Random House.
[Lynch 1960] Lynch, K. 1960. The Image of the City. Publication
of the Joint Center for Urban Studies. Mit Press.
[Macdonald, Cummins, and Macintyre 2007] Macdonald, L.;
Cummins, S.; and Macintyre, S. 2007. Neighbourhood fast food
environment and area deprivation—substitution or concentration?
[Mccarney et al. 2007] Mccarney, R.; Warner, J.; Iliffe, S.; van
Haselen, R.; Grifﬁn, M.; and Fisher, P. 2007. The Hawthorne
Effect: a randomised, controlled trial. BMC Medical Research
[McGinley, Mahin, and Pope 2000] McGinley, C.; Mahin, T.; and
Pope, R. 2000. Elements of Successful Odor/Odour Laws. In
WEF Odor/VOC 2000 Specialty Conference.
[Pallasmaa 2012] Pallasmaa, J. 2012. The Eyes of the Skin: Archi-
tecture and the Senses. Wiley.
[Quercia et al. 2013] Quercia, D.; Pesce, J. P.; Almeida, V.; and
Crowcroft, J. 2013. Psychological Maps 2.0: A web engagement
enterprise starting in London. In Proceedings of ACM WWW.
[Quercia et al. 2015] Quercia, D.; Aiello, L. M.; Schifanella, R.;
and Davies, A. 2015. The digital life of walkable streets. In Pro-
ceedings of ACM WWW.
[Quercia, Schifanella, and Aiello 2014] Quercia, D.; Schifanella,
R.; and Aiello, L. M. 2014. The shortest path to happiness: Rec-
ommending beautiful, quiet, and happy routes in the city. In Pro-
ceedings of ACM HT.
[Reynolds 2009] Reynolds, R. 2009. On Guerrilla Gardening: A
Handbook for Gardening Without Boundaries. Bloomsbury Pub-
[Rosvall and Bergstrom 2008] Rosvall, M., and Bergstrom, C. T.
2008. Maps of random walks on complex networks reveal commu-
nity structure. Proceedings of the National Academy of Sciences
[Southworth 1967] Southworth, M. 1967. The Sonic Environment
of Cities. MIT Press.
[TheEconomist 2015] TheEconomist. 2015. The big smoke. Britain
needs to do more to clean up its dirty air. January 10
[Vennemann, Hummel, and Berger 2008] Vennemann, M.; Hum-
mel, T.; and Berger, K. 2008. The association between smoking
and smell and taste impairment in the general population. Journal