Content uploaded by Szymon Chmielewski
Author content
All content in this area was uploaded by Szymon Chmielewski on Oct 05, 2017
Content may be subject to copyright.
Measuring visual pollution by outdoor advertisements in an urban street
using intervisibility analysis and public surveys
Szymon Chmielewski1, Danbi J. Lee2, Piotr Tompalski3, Tadeusz J. Chmielewski4, Piotr Wężyk5
1 University of Life Sciences in Lublin, Institute of Soil Science and Environmental Engineering and
Management, Leszczyńskiego St. 7, 20-069 Lublin, Poland, gisszymon@gmail.com
2 CitySpatial, 152 Parliament St., Toronto, Canada, danbilee@gmail.com
3 Faculty of Forestry, University of British Columbia, 2424 Main Mall, Vancouver, BC V6T 1Z4,
Canada, piotr.tompalski@gmail.com
4 University of Life Sciences in Lublin Faculty of Landscape Ecology and Nature Conservation,
Dobrzańskiego St 37, 20-262, Lublin, Poland, tadeusz.chmielewski@up.lublin.pl
5 University of Agriculture in Krakow, Laboratory of Geomatics, Department of Forest Ecology, 29
Listopada St 46, 31-425 Krakow, Poland, rlwezyk@cyf-kr.edu.pl
Pre-print of published (online) version Reference:
Szymon Chmielewski, Danbi J. Lee, Piotr Tompalski, Tadeusz J. Chmielewski
& Piotr Wężyk (2015): Measuring visual pollution by outdoor advertisements in an urban
street using intervisibility analysis and public surveys, International Journal of Geographical
Information Science, DOI: 10.1080/13658816.2015.1104316
To link to this article: http://dx.doi.org/10.1080/13658816.2015.1104316
Disclaimer: The PDF document is a copy of the final version of this manuscript that was
subsequently accepted by the journal for publication. The paper has been through peer review,
but it has not been subject to any additional copy-editing or journal specific formatting (so will
look different from the final version of record, which may be accessed following the DOI above
depending on your access situation).
11. Introduction
Since the city of Sao Paulo’s radical move in 2007 to eliminate all forms of outdoor
advertising, the concept of ‘visual pollution’ has popularized globally, being adopted by scholars who
had already been raising concerns in the background on the encroaching commercialization of public
spaces and unsightly urban landscapes (e.g. Baker 2007; Koeck and Warnaby 2014). Unlike air or
water pollution for which the research is already certain on levels of harm, visual pollution remains a
loose concept tied to general irritation, emotional or psychological harming of viewers in question, and
are culturally and personally influenced terms (Enache et al. 2012; Yilmaz 2011; Nagle 2009; Penteado
2007).
Visual pollution is a compounded effect of clutter, disorder and excess of various objects and graphics
in the landscape such as outdoor advertisements (OAs), street furniture, lighting features (Falchi et al.
2012; Chalkias et al. 2006), vegetation characteristics (Ulrich 1986; Lamp and Purcell 1990; Ribeiro
2006) and other objects. The exercise of selecting and weighting the contribution of each to pollution
levels is undoubtedly a cloudy task. Widely accepted as a real concern among marketing experts (Ha
and Litman 1997), landscape designers and city planners (Gomez, 2013; Iveson, 2012; Aydin and
Nisanci, 2008), and public health specialists (Hackbarth et al 2001), from both the advertiser and
consumer perspective (design, tolerance, information absorption etc.), underdeveloped methods on
quantifying visual pollution has left it out of the conversation in cities that require measurable
evidence for decision-making. This partly has to do with a lack of awareness and when aware, a lack of
any clear and reliable measurement tools.
To begin tackling this problem, this study focused on measuring visual pollution by OAs (banners and
billboards) in a busy urban street of Lublin, Poland where OAs are very dominant and concentrated in
this landscape. By using spatial properties of OAs (location, shape and size) to measure intervisibility
(pollution exposure) and correlating it with the public opinion of OAs (pollution score), an impact
factor is proposed, the permissible visual pollution impact (pvpi), for this streetscape. As a locally-
derived threshold value, pvpi could be used to inform a permit decision for OA placements within a
streetscape (i.e. would the proposed OA contribute to visual pollution?). The intervisibility analysis can
also be used to locate visual pollution zones where stricter policies can be placed to mitigate impact
(such as higher taxation).
The study is motivated by advances in geospatial technology and growing concerns over loose
government control on the quality of public spaces in Lublin. Global concerns on the effect of visual
pollution on mental health and consumption patterns (e.g. Vardavas et al. 2009) are other strong
drivers to develop methods of rapid assessment that can enable city officials to mitigate visual
pollution. The study results contribute to research on the visual perception of urban landscapes,
spatial considerations for measuring OA visibility, and the contributions of geospatial technology to
decision-making by city planners and designers (geodesign), particularly in cities that are late-
adopters of geospatial technology.
1.1 Outdoor advertising landscapes
Iveson’s (2012) discussion on the outdoor advertising landscape around the world neatly spins
off of Naomi Klein’s (2001) manifesto against branded cultures in capitalist societies. Iveson points out
that while in-home advertising on television or magazines can be switched off and stored away, OAs
cannot, which makes public spaces extremely valuable to advertisers. And as cities struggle to invest in
infrastructure improvements through basic revenue streams, private funding initiatives for public
spaces become increasingly attractive. On the other hand, privately sponsored public infrastructure
can be seen as a slow encroachment of private control over public spaces, influencing the behaviour
and flows of city life (Baker 2007; Cronin 2006).
The proliferation of OAs in public spaces worldwide has seen many changes in the quantity, design and
materiality, which can be explained by changes in consumer attitudes, brand perceptions, innovations
in advertising, and changes in policy and enforcement (Mehta 2000; Madupu 2013). While OAs
cannot be switched off, Ha and Litman (1997) argue that too much advertising clutter in magazines
results in negative returns to the advertiser. Advertising clusters become too noisy to be understood by
consumers regardless of how clever or beautifully designed. The same could be true in public spaces,
and the threshold of visible OAs could be determined on a case-by-case basis, and is attempted in this
study.
In post-socialist countries undergoing economic transformations such as Poland which since the 1990s
has been establishing a new economic order, OAs are an important tool to stimulate trade. It has
become an inseparable element of a young market economy which inevitably squeezes out any real
concern over landscape beauty and organization. Relatively low production cost combined with the
high impact of OAs and the lack of appropriate policies regulating advertising content and location
have led to an undesirable proliferation of OAs.
The significant escalation of visual pollution in Poland as observed by Dymna and Rutkiewicz (2009)
spurred deliberations on the possibility of expressing this phenomenon quantitatively. So far no
analytical method has been proposed to describe this problem although legislation (such as the
Landscape Protection Act, 2015) and local policies are forming to reorganize objects in public spaces.
In Lublin, as with other cities from Toronto to Sydney, guidelines and provisions that dictate size,
content (the advertised product), and placement (such as vertical limits) are in place, including OA
restrictive zones and proximity buffers to protect landscapes or maintain visual order (see Hillier
2009; Iveson 2012). However, the extent of illegal postings and bylaw enforcement are also critical
factors that will impact the visual landscape.
1.2 From visibility to visual pollution
In Europe one of the first studies on the visual perception of cities was undertaken by Cullen
(1961). Fundamental on the subject in Poland includes studies by Wejchert (1984), Wojciechowski
(1986), Bogdanowski (1998), and Patoczka (2000). Anthropogenic studies on landscape quality and
the fundamentals of landscape perception were also conducted by Taylor et al (1987), Zube (1987) and
by Smardon et al (1986). However, all based their work on traditional methods and techniques of
landscape studies, developed basically on the grounds of landscape architecture. The application of the
GIS techniques in studies on visual perception and the estimation of urban landscape quality is mainly
the domain of the 21st century. GIS-based-models have become an important element of politically
sensitive decision-making processes (Crosetto et al. 2002; Peccol et al. 1996), especially when a multi-
criteria evaluation procedure is applied (Malczewski 1999; Vizzari 2011).
There is no doubt visibility is valuable in advertising since it is foundational to brand exposure.
Advertisers choose locations based on maximizing visibility to target audiences, and can sometimes
have harmful effects on society. In tobacco and alcohol advertising, many issues have been raised
around the uncomfortably close proximity of OAs to vulnerable populations such as low-income
citizens, minority groups, and youth (Vardavas et al 2009). Some counties have implemented national
bans such as Canada between 1988 to 1994 (Tobacco Products Control Act), while others like the USA
or Greece, still allow for tobacco content, arguing freedom of speech and free market rules. Proximity
analyses using point data as done by Hackbarth et al (2001) in Chicago and Vardavas et al (2009) in
Athens illustrate that 2D data on OAs can be useful in measuring advertising exposure with respect to
public health. It seems simple 2D viewsheds can be quite useful in some data or time-scarce contexts.
A persons view angle and field of view should be considered when measuring visibility (Minelli et al
2014). Other obstructions such as air pollution and occluding objects like tall vegetation also impact
true visibility of an OA. Current research in visibility analysis focuses on advanced 3D methods that
attempt to measure true visibility envelopes to building facades (see Suleiman et al. 2011) and
intervisibility of public spaces (Albrecht et al 2013) that can used to measure the OA visibility more
precisely than clusters of points. Challenges lie in determining the appropriate level of detail in 3D city
models, which impacts visibility (Sander 2007).
However, even with advances in 3D visibility analysis, there is yet a clear body of research providing
guidance on how to translate visibility into visual pollution. In the aforementioned cities, there are
some proximity and content regulations for individual OAs that imply some recognition of visual
pollution, but OA landscapes are weakly described and pollution from OA landscapes are not
measured. In Portella’s (2014) book on visual pollution, she draws from cross-cultural studies in
England, Japan, and Brazil to conclude that common opinions on OA landscapes should be used to
create general design guidelines and principles for signage control. This implies that public opinion
can be predicted by certain physical manifestations of OA landscapes. Moreover, better control over
OA infrastructure can mitigate visual pollution with some global certainty.
LiDAR data has already been demonstrated as a powerful tool to measure OA infrastructure and can
assist with determining conformity to advertising regulations (Anderson, 2013). But such tools speak
little to the form and density of OAs that cause visual pollution. Projecting from Portella’s work, it is
worth experimenting with relations between public experience (or perception) of OA landscapes and
its physical model. If clearer metrics can be established that translate visibility into visual pollution,
city officials can begin answering questions like “how many is too many ads?” and “at what density
does disorder and clutter become harmful?”.
This study extends a 2D OA point dataset to 2.5D (location and height) in a preliminary experiment to
translate visibility into visual pollution. The study objective was to develop a rapid assessment of
visual pollution to be used by city officials in a time and data-scarce context, for evaluating OA
proposals and requests. It contributes to research on how well a physical model of OA landscapes can
predict visual pollution, acknowledging that OA content, and cultural perception of advertisements
strongly influence true visibility.
2. Methodology
2.1 Study Area
The research presented in this paper was carried out in Lublin, Poland (Figure 1). Lublin has
an area of about 147 km2, a population of approximately 350 000 people and is an important academic
centre in the region. It is one of the most dynamically developing cities in Poland
(PricewaterhouseCoopers Annual Report, 2011) where numerous economic investment projects are
accompanied by the development of the advertising business. In the first half of 2012 there were 231
local advertising agencies in Lublin involving not only the use of large billboards but also outdoor LED
display screens which are becoming more common, generating approximately 91,000 OAs in Lublin
(Polish Outdoor Advertising Chamber of Commerce, 2013).
The study area is located along T. Zana Street in the Rury area, one of Lublin’s central district, and has
the total area of 96.8 ha. T. Zana Street has one of the greatest concentrations of OA in Lublin
(Chmielewski 2011), which leads to the assumption that it may be affected by visual pollution. The
street is characterized by a terminating roundabout at the west and east end (roundabout A and B
respectively) with a high concentration of visible OAs. The area is relatively well separated from other
urban zones saturated with OAs, so consequently the potential influence of OA clusters located in the
neighbourhood is minimized. The isolation effect in this case is the result of characteristic topographic
features.
Figure 1. Study area is located along T. Zana Street in Lublin, Poland.
2.2 Field measurements and intervisibility analysis
The only spatial data on OA objects in Poland is a point layer prepared by the land surveying
services (1:500 master map), where only large, stand-alone roadside “billboards” are marked. No
attributes are attached to these points and there is no complete OA database which would describe
their spatial dimensions, frontal direction, ownership, identification numbers, etc. Therefore, an OA
inventory was created for this study. OAs inventories included billboards and advertising banners
(placed on fences or buildings), visible at pedestrian level (Figure 2). Signboards (indicating the name
of a business on the business property) and other small advertisements e.g. on busses or promotional
umbrellas, were excluded from the inventory.
Figure 2. Examples of outdoor advertisements measured in this study, in Lublin Poland.
The OA inventory was performed in 2015 with a GNSS receiver, coupled with a TruPulse 360B laser
rangefinder to determine locations of visible OAs. Thanks to the laser rangefinder, the measurements
could be done in an offset mode, i.e. without the need to be in close range to each OA. Therefore, many
visible OAs could be inventoried from a single viewing position. In order to enhance GNSS position
accuracy, real time corrections were used to achieve sub-meter accuracy.
Since Lublin advertising by-laws do not introduce OA size classification in this area, size classes were
defined to set a standard visibility range for intervisibility modelling. The majority of OAs have sizes
that fit well into three classes: A – small, no greater than 12 m2; B – medium, between 12 – 20 m2; and
C – large, greater than 20 m2. OA size measurements were made only in cases of size doubt. The
inventoried OAs were saved to a database as georeferenced point features together with the attributes
of height and class.
The intervisibility analysis was performed in ArcGIS environment (3D Analyst extension),
using viewshed modelling. A Digital Surface Model (DSM) was used that allows the analyst to inspect
whether a source and target are inter-visible. The output is a raster file whose cell values represent the
relative degree of intervisibility (De Montis and Caschili 2012). In this study, each cell contains a
binary count of whether a the top center point of an OA was intervisible given a direct line of sight at a
given observer height, within a specified visibility range. This is a standard procedure in most GIS
packages today (Llobera, 2003).
To calculate the viewshed, the observer’s eye height was assumed to be 1.6 m (Schirpke 2013, Kułaga et
al. 2011; ) with a lower and upper horizontal view angle set to -90° and 90° respectively. Visibility
ranges indicate the assumed radial area around an OA where it remains recognizeable to the observer.
The initial visibility range was zero m for all size classes and final visibility range was: A: 50 m, B: 200
m, C: 350 m. OA elevation was based on GNSS field measurements. These were based on standards by
advertising agencies and OA size classes. Each OA was treated as a 1600 visual emitter from the top
center point of the OA, and the scene viewshed was determined by accumulating radiating viewsheds
from all individual OA emission areas. A 1800 visual emitter model was not considered since it was
assumed that a rotating observer could not recognize an OA from a very acute angle of view (Figure 3).
Figure 3. Each OA (A, B, C) has a set visibility range and assumed recognition distances
based on size. An observer (1, 2, 3) at each point can experience 3600 of exposure. The 1600
visual emitter (right) was used over the 1800 visual emitter (left) since a view at an 1800
angle would not allow the OA to be recognizable or visible.
The DSM was obtained from an Airborne Laser Scanning (ALS) point cloud acquired with a density of
12 points/m2 (ISOK Project, 2012), which permitted the generation of a DSM raster with 0.5 m pixel
size. Since some OAs were hidden behind vegetation, accurate point cloud classification was crucial to
locate OAs exactly on the ground. If any OA points were to be misclassified as ground points, the
resulting DSM raster model would have higher elevations and lead to higher OA heights in the
database. To roughly and quickly account for visual obstructions by tall vegetation and buildings, a
vegetation and building mask was applied to the visibility surface raster. The outlines of high
vegetation areas and buildings were generated directly from ALS data and converted to vector format
to perform a simple clipping operation.
There are several variants of intervisibility analysis, all of which yield raster surfaces. The binary type,
which is used here, gives only a true or false output of visibility per pixel. This solution is not as
advanced as others, which may allow for more fuzzy outputs that are considered to be more realistic.
For example, the analyses proposed by Fisher (1992) takes into account weather conditions and
distance to calculate probable viewsheds (non-binary). Other work on visibility color maps
(Choudhury et al. 2014), visual metrics (Bartie et al. 2010) or the general concept of visualscapes are
emerging. Llobera (2003) demonstrated the possibility of modelling the range of visibility even though
a layer of vegetation. However, a binary analysis requires little GIS programming expertise and allows
users to quickly obtain repeatable results which can be easily understood. This is most useful in low-
tech contexts where a rapid approach can be implemented immediately.
2.3 Pollution Score
The phenomenon of visual pollution is related to cultural acceptance of advertising media as
much as it is to the way the brain processes information and understands landscape, as explained
earlier. In this study, the intent was to capture the total effect using simple metrics. By surveying
pedestrian opinion on the OA landscape in the study area, a point system (pollution score) was used to
indicate any level of visual pollution. According to Frank et al. (2013), Kearney et al. (2008) and also
Kaplan and Kaplan (1989), a five grade scale was used to rate visual pollution perceived by each of
interviewed respondents. The five grade scale was also pre-tested by the authors (October 2014,
unpublished).
Based on Portella’s (2014) close-ended questionnaire, so far the only one published on visual
pollution specifically, 3 questions referring to OAs were selected and carefully translated into Polish.
The respondents were asked to answer the questions based on a 360 degree field-of-view (from a fixed
observation spot). The questions used for this study were:
1. How do you like the appearance of this street? (1-really like , 2-like, 3-neutral, 4- don`t like, 5-
really don`t like)
2. The number of advertisement signs (billboards and banners) on this street are: (1-very few, 2-few,
3-moderate, 4-many, 5-too many)
3. The advertisement signs make the appearance of this street: 1-very beautiful, 2-beautiful, 3-they
do not matter, 4-ugly, 5-very ugly
In all of 3 close-ended questions, the answers had been sorted from “1” (most positive
answers) to “5” (most negative answers). This approach allowed us to define pollution scores. Whereas
Dobbie (2013), Schirpke et al. (2013), Meitner (2004) used photo-based questionnaires for visual
analysis in a very controlled experiment, this study was conducted in the field to determine if simple
metrics could reasonably capture the effect of real world urban contexts.
In total, 200 measurement points were placed along T. Zana street, each spaced every 50 m
(where possible). During fieldwork, all 200 measurement points were precisely located using RTK-
GNSS measurements and marked. Additionally, an online map was prepared to help the surveyors to
navigate to exact location of each measurement point. Measurement points were numbered from 1 to
200 and grouped into five sets of 40 points numbered from A to E (examples of single point numbers:
A 38; B 62; C 102; D 155; E 187).
The survey was conducted between 11 – 25 of May 2015 by 50 students of second year of
Spatial Management at University of Life Sciences in Lublin. They were working as an surveyors in 25
two-person teams. Each measurement point was surveyed by five different teams (thus, the labels of A-
E). Surveyors were equipped with the Collector for ArcGIS application which enabled them to navigate
to a set of measurement points. After finding the exact location of given point, they introduced
themselves to passing pedestrians, gave brief information about the survey, and were asked the three
close-ended questions. The respondents were not differentiated in terms of age, sex or education,
however only adults were interviewed (18 years and older). In exceptional cases, where there was no
foot-traffic at the measurement point, the students were asked to fill out the survey as a respondent.
The survey data collected from the respondents were integrated into one spatial dataset with
the observation point locations. Since each point was surveyed by five teams, the median of the five
responses for all three questions was calculated (and considered to be the pollution score).
To create an interpolated surface (S1, S2, S3) of the pollution scores for all three questions (Q1,
Q2, Q3), a kernel interpolation by median values was used across the scene using ArcGIS Geostatistical
Analyst. The kernel method assigned a weighted semivariogram for each cell in the raster (see
Johnston et al. 2001). This method is often used to estimate the pollution which decreases as the
distance from the source increases (Krivoruchko,2011).
2.4 Permissible Visual Pollution Impact (pvpi) determination
The intervisibility surface and survey responses were analyzed for any relationships in order to
measure what the public threshold would be for visual pollution to occur in the study area.
Recognizing that the responses depend on the viewers` internal judgmental criteria that vary from
individual to individual (Shang 2000). Here we propose a metric of pvpi as the relationship between
number of visible OA and public opinion. The relationship was investigated with ANOVA (Kutner et
al. 2005, R Core Team 2014) – the analysis of the number of visible OAs against the median pollution
score. Additional post-hoc test (Tukey HSD) allowed us to define which pollution scores differed from
each other.
Although ANOVA analyzes the differences between groups, it cannot provide information about
threshold values on the number of OA for visual pollution. Therefore, the OA pollution score was
predicted with classification using regression trees (Crawley 2013). This allowed for determining the
visibility threshold values for each of the pollution scores (or the pvpi), especially to distinguish areas
with responses of pollution score 4 or 5, indicating visual pollution (negative responses such as ‘really
don’t like’). The result of the classification was assessed with overall classification accuracy, omission
and comission errors.
3. Results
3.1 Intervisibility analysis
In total, 228 OAs were inventoried (77 – class A, 139 – class B, 12 – class C). Generally, 65%
(42 - class A, 96 - class B, 10 - class C) of OAs were located along T. Zana Street, which was also the
busiest part of the study area in terms of people and cars. The visibility analysis showed that OA
visibility levels varied within the study area but the highest OA exposure was at roundabout A and B,
both areas congested during rush hour, where OA exposure time would be increased. In a relatively
small area (10.25 m2) in roundabout A, as low as 21 and high as 43 OAs were visible simultaneously
and was the highest emission zone in the study area. Over the entire study area of 96.8 ha, at least one
OA was visible from 8.78% and more than 25 OAs were visible at from over 2.52% of the area,
generally at large open areas (0.83 ha at roundabout A, 1.38 ha at roundabout B, and 0.19 ha along T.
Zana Street, Figure 4).
Figure 4. Intervisibility analysis of study area and zones where at least 25 OAs are visible.
3.2 Pollution Score
In total, 1000 survey questionnaires were collected. While 98 (9.8%) of the questionnaires
were filled out by the interviewers due to a lack of pedestrian traffic at the measurement point (e.g.
small alleys), the remaining 902 (90.2%) questionnaires were filled out by passers-by. The data
distribution revealed a relatively low flattening of the histogram (kurtosis for Q1, Q2 and Q3 was 1.89,
2.22 and 1.91 respectively) and a slight asymmetry (skewness for Q1, Q2, Q3, was 0.34, -0.03; 0.38
respectively). The mean pollution score in Q1 was 2.73, Q2=3.04 and Q3=3.43 with standard deviation
equal to 0.89, 1.15 and 0.52 for Q1, Q2 and Q3 respectively.
The three interpolated surface (S1, S2, S3) of the pollution scores (using median of responses
at each measurement point) are shown in figure 5. An averaged standard error of interpolation S1, S2
and S3 was 0.86 , 0.79, 0.50, respectively. The interpolated surfaces define spaces in the study area
that are exposed to visual pollution as perceived by respondents. Most noticeable is the clear negative
opinion by respondents in the area of the two roundabouts (“A” and “B”). This is where the highest
number of visible OAs was found in the visibility analysis.
Figure 5. Interpolated surface of public survey results: S1 - interpolation of Q1, overall appearance, S2 -
interpolation of Q2, number of OA, S3 - interpolation of Q3, number of OA on appearance.
3.3 Threshold of visual pollution
The relationships between the number of visible OAs and the pollution score were analyzed with
ANOVA (Figure 6). The strongest relationship was found for Q2 where ANOVA shows significant
differences between the analysed responses. The Tukey test (Tab. 1) shows that at alpha = 0.05 almost
all differences are significant with the exception of scores of 1 & 2 (p value 0,680) that correspond with
a positive assessment of the streetscape.
A decision tree (Fig. 7) was used to define the threshold number of visible OA, that would indicate the
visual pollution. The decision tree was created only for Q2 and resulted in an overall classification
accuracy of 0.52 (Tab. 2). It allowed us to state that pvpi threshold is when more than 7 OAs were
visible.
Figure 6. Boxplot of the number of visible OAs across public opinion score: a) survey response for
question 1, B) survey response for question 2, c) survey response for question 3.
The aforementioned dependency was also confirmed by ANOVA conducted for Q3 where
Tukey`s test shows that the significant difference is in fact only between 4 and 3, and 5 and 3 (Tab. 1).
A small number of responses in relation to score “1” and “2” of Q3 is understandable, generally the
respondents hadn’t noticed that a lot of OAs made the street more beautiful. To ANOVA the median
values were used and small number of low pollution scores in Q3. Only 85 of all respondents (8.5%)
declared that OAs make the appearance of the street very beautiful or just beautiful, but only 22 of
them (2.2% of all respondents) has indicated this phenomena directly in the identified pollution zones
(above 7 visible OAs). This kind of aberrations are caused by subjective character of any public survey
and may be due to other unmeasured environmental factors.
p (Q1)
p (Q2)
p (Q3)
2 – 1
0.9530733
0.6807611
-
3 – 1
0.8216191
0.0023190
-
4 – 1
0.3948080
0.0000000
-
5 – 1
0.9993973
0.0000000
-
3 – 2
0.8905291
0.0248002
0.4723579
4 – 2
0.0628607
0.0000000
0.9984999
5 – 2
0.9999980
0.0000000
0.8224431
4 – 3
0.5760691
0.0000000
0.0000000
5 – 3
0.9995174
0.0000000
0.0029642
5 – 4
0.9860775
0.0020123
0.3218346
Tab. 1. Tukey`s test results.
In the case of Q1 anova test showed that all survey response group are equal, which was also
confirmed by Tukey`s test results (all p values above 0.05). This means that OAs are not the only
factor which affect overall perception of visual pollution. Probably other factors unrelated to OAs were
being unconsciously assessed, and can explain why there was only slight tendency to negatively score
the streetscape. However, a threshold is rarely an absolute event (Shang, 2000) and our case study
experiment has shown that visual pollution can be quantified in a meaningful way that captures local
public opinion.
Fig. 7. Regression analysis of pollution scores (1 to 5) and number of visible OAs.
Table 2. Confusion matrix of predicted pollution scores (1 to 5).
4. Discussion
4.1 OA visibility and pvpi determination
Visibility modelling of various landmarks like wind turbines or solar panels is commonly and
successfully used (Molina-Ruiz et al. 2011, Chiabrando et al. 2011). OA visibility modelling is a
relatively new task described in international literature (Jamail 2015). With small and medium
mapping scales (i.e. regional), the user usually trusts the DSM accuracy and does not perform any
validation of visibility. As a kind of landmark, OAs are considerable features in rather large mapping
References
1
2
3
4
5
Total
Error of
commission
Prediction
1
19
0
2
0
0
21
9.5%
2
16
7
17
4
0
44
84.1%
3
4
3
35
21
1
64
45.3%
4
0
1
3
34
10
48
29.2%
5
0
0
1
13
9
23
60.9%
Total
39
11
58
72
20
104
Error of
omission
51.3%
36.4%
39.7%
52.8%
55.0%
Overall
accuracy=52%
scales and the question about visibility modelling accuracy may be asked. In recent literature, we have
found only Hagstroms and Messinger (2011) and Murgoitio (2014) works with high detailed visibility
modelling with voxels.
OA visibility accuracy may depend on the OA mapping approach. Here, we have presented
OAs on the map as points, but virtually all OAs are linearly shaped boards and analysing them as such
may give slightly more accurate visibility results (viewshed will treat the OA board as two edge points),
going forward with a 3D OA visibility approach (Jamail et al. 2015) it is possible to model each corner
of the OA board, but such an approach hasn’t been presented yet, perhaps because it is
computationally heavy. We have decided to present OA as point object to balance improved accuracy
with rapid computation and visualization of visual pollution zones, which will inform city policy
decisions.
In this case study we have identified a pvpi value of over seven OAs and is influenced by the
number of OA visible rather than the size of the OA. In the light of any questioning around this low
value, it should be noted that the study area consists of commercial land uses along T. Zana Street.
There are plenty of shops, a several banks and private medical practices, and generally the public
expects commercials signs in this area. Repeating this methodology for other areas should expect that
the pvpi may be different depending on the land uses and different cultural expectations compared to
Poland. The method proposed therefore yields a context-specific result in a rapid way. Repeated
studies may draw out an order of magnitude of the number of OAs that would create visual pollution,
since it seems that the excess of advertising is an unfavourable phenomenon generally.
4.2 Other parameters affecting visual pollution
During the research, the respondents were asked to evaluate the full 360° view, and the
directional character of OA board or the viewing pedestrian was not taken into account. In walking
down the street, mostly we look ahead and sideways and do not necessarily notice all visual pollution
generated by OAs. Conducting eye-tracking research (Hampp 2008) may be helpful to determine
precisely which advertisements a pedestrian looks at, and consequently, which OAs contribute to
visual pollution. It has also been shown that large-format advertisements situated along stretches of
roads distract drivers (Dukic et al. 2013, Edquist et al. 2012), so overall context could have influence
on visual pollution.
Moreover, it is conceivable that visibility is influenced by more than a point location of
observer and OA. Features of the OA itself such as height, size, shape, colour, design, readability, and
other interactive media can be measured and studied spatially (Aydin et al 2008). These features may
render the OA more or less attractive (How well does it stand out? How well can it be understood?)
especially in a crowd of other OAs (Ha and Litman 1997). Similar research is applied in wind farm
impact assessments (Minelli et al. 2014; Rodrigues 2010).
5. Conclusion
In recent years geodesign and other spatial design practices (Wilson, 2014) as well desktop
GIS (Slager et al. 2013), point out visibility analysis as integral to evidence-based spatial planning. As a
preliminary experiment, the results are promising and could be implemented in a context of data or
time-scarce contexts. Measuring visual pollution can be an input into new policies to mitigate pollution
and to monitor OA policy impacts. Using intervisibility analysis and the concept of OA visibility range,
the OA pollution zone can be rapidly delineated as a special taxation zone. Visual pollution zones can
inform city zoning by-laws and possibly urban design guidelines for heritage management.
The development of GIS technology in the field of shaping OA management principles seems
to be promising. We believe developing 2.5D and 3D GIS-based methods will be critical to asset
management and by-law enforcement as well (particularly in Lublin where illegal OAs are commonly
erected despite set by-laws). The spatial data infrastructure which was created for this study offers
numerous other options. For example, imagine if the OA database was accessed through a GNSS-
enabled mobile device of a by-law officer on-site in a high-pollution zone, which was identified through
the intervisibility analysis. OAs could be assigned with a unique code (e.g. barcode) attached to a
database identifying ownership as well as the permit status.
Visual pollution by OAs in public spaces is a growing concern in cities with emerging market
economies like Lublin, Poland, especially with growing awareness of visual pollution impacts on public
behaviour, health, and concerns around the privatization of public spaces. This paper demonstrated a
rapid method to measure visual pollution by combining GIS-based intervisibility analysis and
pollution scoring (public surveys). By using spatial properties of OAs (location, size, and visibility
range) and relating it with the public opinion of OAs (pollution score), a permissible visual pollution
impact (pvpi) was identified for this specific streetscape (more than 7 OAs). We believe this is one of
very few attempts in research to specify a metric for visual pollution that could translate into practice
and expect future research to focus on applications and 3D methods.
References
Albrecht, F., Moser, J., and Hijazi, I., 2013. Assessing FAÇADE Visibility in 3d City Models for City
Marketing. ISPRS-International Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences, 1(2), 1-5.
Anderson, B. 2013. Is CityScan the Future of Code Enforcement? Urban Land, Market Trends.
Retrieved from: http://urbanland.uli.org/economy-markets-trends/is-cityscan-the-future-of-
code-enforcement/.
Aydin, C., and Nisanci R., 2008. Environmental harmony and evaluation of advertisement billboards
with digital photogrammetry technique and GIS capabilities: a case study in the City of Ankara.
Sensors 8(5), 3271-3286.
Baker, L. E., 2007. Public sites versus public sights: The Progressive response to outdoor advertising
and the commercialization of public space. American Quarterly, 59(4), 1187–1213.
doi:10.1353/aq.2007.0077
Bartie, P., Mills, S., and Kingham, S., 2008. An Egocentric Urban Viewshed: A Method for Landmark
Visibility Mapping for Pedestrian Location Based Services [in] Moore A., Drecki I (Ed.),
Geospatial Vision New Dimensions in Cartography Selected Papers from the 4th National
Cartographic Conference GeoCart’2008 New Zealand. Springer. Retreived from:
http://link.springer.com/chapter/10.1007/978-3-540-70970-1_4.
Beza, B., 2010. The aesthetic value of a mountain landscape: A study of the Mt. Everest Trek.
Landscape and Urban Planning, 97(4), 306–317. doi: 10.1016/j.landurbplan.2010.07.003
Bogdanowski, J., 1998. Konserwacja i ochrona krajobrazu kulturowego, ewolucja i metody
(Conservation and protection of cultural landscape-evolution and methods). Teki Krakowskie.
(6)1, 1-224.
Breiman, L. 2001. Random forests. Machine Learning, 45(1), 5–32.
Chalkias, C., Petrakis, M., Psiloglou, B., and Lianou, M. 2006. Modelling of light pollution in
suburban areas using remotely sensed imagery and GIS. Journal of Environment Management,
79 (1), 57-63. doi- org/10.1016/j.jenvman.2005.05.015.
Chmielewski, Sz., 2011. Analiza widoczności billboardów reklamowych w Lublinie (Billboard visibility
analysis in Lublin). ArcanaGIS, Warsaw: Esri Poland Press. Retrieved from:
http://www.esri.pl/images/Acana_GIS_pdf/arcana_09_2011.pdf
Chiabrando, R., Fabrizio, E., and Garnero, G., 2011. On the applicability of the visual impact
assessment OAISPP tool to photovoltaic plants. Renewable and Sustainable Energy Reviews
(15)1, 845–850. doi:10.1016/j.rser.2010.09.030.
Choudhury, F., Ali, M., Masud, S., Nath, S., and Rabban, I., 2014. Scalable visibility colour map
construction in spatial databases. Information Systems (42), 89–106.
dx.doi.org/10.1016/j.is.2013.12.002
Crawley, M.J., 2013. The R Book, Second. ed., UK: Wiley. doi: 10.1002/9781118448908
Cronin, A. M., 2006. Advertising and the metabolism of the city: Urban space, commodity rhythms.
Environment and Planning D: Society and Space, 24(4), 615–632. doi:10.1068/d389t
Crosetto, M., Crosetto, F., and Tarantola, S. 2002. Optimized Resource Allocation for Gis Based
Model Implementation. Photogrammetric Engineering & Remote Sensing 68(3), 225-232
Cullen, G. 1961. The Concise Townscape. Kidlington, GB: Elsevier Ltd.
De Montis, A., and Caschili, S. 2012. Nuraghes and landscape planning: Coupling viewshed with
complex network analysis. Landscape and Urban Planning, 105 (3), 315–324. doi:
10.1016/j.landurbplan.2012.01.005
Dobbie, M., Green, R. 2013. Public perceptions of freshwater wetlands in Victoria, Australia.
Landscape and Urban Planning 110 143– 154. dx.doi.org/10.1016/j.landurbplan.2012.11.003
Dukic, T., Ahlstrom, Ch., Patten, Ch., Kettwich, and C., Kircher, K. 2013. Effects of Electronic
Billboards on Driver Distraction, Traffic Injury Prevention, 14(5), 469-476.
dx.doi.org/10.1080/15389588.2012.731546
Dymna, E. and Rutkiewicz, M. 2009. Polish Outdoor. Warsaw, Klucze Press.
Edquist, J., Horberry, T., Hosking, S., and Johnston, I., 2011. Effects of advertising billboards on
simulated driving. Applied Ergonomics and Transportation Safety 42 (4), 619-626.
dx.doi.org/10.1016/j.apergo.2010.08.013
Enache, E., Morozan, C., and Purice, S., 2012. Visual pollution: A new axiological dimension of
marketing. Conference proceedings: The 8th Edition of the International Conference “European
Integration – New Challenges” EINCO2012, May 25 – 26 2012, Oradea, Romania: 2046-2051.
Retrieved from: http://anale.steconomiceuoradea.ro/volume/2012/proceedings-einco-2012.pdf.
Falchi F., Cinzano, P., Elvidge, C.,D., Keith, D.,M.,and Haim, A. 2011. Limiting the impact of light
pollution on human health, environment and stellar visibility. Journal Environment
Management, 92(10), 2714-22. doi-10.1016/j.jenvman.2011.06.029.
Fisher, P. F., 1992. First experiments in viewshed uncertainty: Simulating fuzzy viewsheds.
Photogrammetric Engineering and Remote Sensing, 58(3), 345 – 352.
Frank, S., Furst, Ch., Koschke, L., Witt, A., Majeschin, F. 2013. Assessment of landscape aesthetics—
Validation of a landscape metrics-based assessment by visual estimation of the scenic beauty.
Ecological Indicators (32) 222– 231 Doi: 10.1016/j.ecolind.2013.03.026
Gomez, J. E. A,. 2013. The Billboardization of Metro Manila. International Journal of Urban and
Regional Research 37(1), 186-214. doi:10.1111/j.1468-2427.2011.01098.x
Ha, L., and Litman, B. R., 1997. Does advertising clutter have diminishing and negative returns?.
Journal of Advertising, 26(1), 31-42.
Hackbarth, D. P., Schnopp-Wyatt, D., Katz, D., Williams, J., Silvestri, B., and Pfleger, M. 2001.
Collaborative research and action to control the geographic placement of outdoor advertising of
alcohol and tobacco products in Chicago. Public Health Reports, 116(6), 558-567.
Hagstroms, S., and Messinger, D., 2011. Line of sight analysis using voxelized discrete LIDAR. Proc.
SPIE 8037, Laser Radar Technology and Applications XVI, 80370B (June 01, 2011).
doi:10.1117/12.884049
Hampp, A., 2008. Traffic audit bureau readies release of new metrics: Moving audience data from
“might see” to “did see”. Advertising Age 11 December (adage.com/print/133209) (accessed
5.03.2013)
Hillier, A., Cole, B. L., Smith, T. E., and Yancey, A. K. 2009. Clustering of unhealthy outdoor
advertisements around child-serving institutions: Acomparison of three cities. Health & Place
15(4), 935-45. doi:10.1016/j.healthplace.2009.02.014
ISOK., 2012. IT System of the Country's Protection Against Extreme Hazards
(http://www.isok.gov.pl/pl/); source of LiDAR-ALS data (data acquired in 2012).
Iveson, K., 2012. Branded cities: outdoor advertising, urban governance, and the outdoor media
landscape. Antipode 44(1), 151–174. doi:10.1111/j.14678330.2011.00849.x
Jamali, B., Sadeghi-Niaraki, A., and Arasteh, R. 2015. Application of geospatialanalysis and
argumented reality visualization in indoor advertising. International Journal of Geography and
Geology, 4(1), 11-23.
Johnston, K., Jay, M., Hoef, V., Krivoruchko, K., and Lucas, N., 2001. Using ArcGIS Geostatistical
Analyst. USA: Esri Press.
Kaplan, R., Kaplan, S., 1989. The experience of nature: A psychological perspective. Cambridge
University Press, Cambridge, UK.
Kearney, A., et al. (2008) Public perception as support for scenic quality regulation in a nationally
treasured landscape. Landscape and Urban Planning (87) 117-128.
doi:10.1016/j.landurbplan.2008.05.005
Klein, N. 2001. No Logo. London: Flamingo 2000 .
Koeck, R., and Warnaby, G., 2014. Outdoor advertising in urban context: spatiality, temporality and
individuality. Journal of Marketing Management, (ahead-of-print), 1-21.
Kutner, M.H., Nachtsheim, C.J., Neter, J., and Li, W., 2005. Applied Linear Statistical Models, Fifth
ed. Chicago: McGraw-Hill.
Krivoruchko, K., 2011. Spatial Statistical Data Analysis for GIS User. Redlands: Esri Press (DVD
Book)
Kułaga, Z., et al., 2011. Polish 2010 growth references for school-aged children and adolescents.
European Journal of Pediatrics 170 (5), 599–609. doi:10.1007/s00431-010-1329-x
Lamp, R., J., Purcell, A.,T., 1990. Perception of naturalness in landscape and its relationships to
vegetation structure. Landscape and Urban Planning 19 (4), 333-352. doi-10.1016/0169-
2046(90)90041-Y
Landscape Protection Act, 2015 (15 May; Dz. Law Gazette No 774):
http://isap.sejm.gov.pl/DetailsServlet?id=WDU20150000774 (accessed 01.07.2015).
Llobera, M., 2003. Extending GIS-based visual analysis: the concept of visualscapes. International
Journal of Geographical Information Science 17 (1), 25-45. doi: 10.1080/13658810210157732
Madupu, V., and Ranganathan, S., 2013. The impact of visual structure complexity on ad liking,
elaboration and comprehension. Marketing Managemen 23(2), 58-70.
Malczewski, J. 1999. GIS and multicriteria decision analysis. New York: John Wiley & Sons Inc.
Minelli, A., Marchesini, I., Taylor, F. E., De Rosa, P., Casagrande, L., and Cenci, M. 2014. An open
source GIS tool to quantify the visual impact of wind turbines and photovoltaic panels.
Environmental Impact Assessment Review, (49), 70-78. doi:10.1016/j.eiar.2014.07.002
Mehta, A. 2000. Advertising attitudes and advertising effectiveness. Journal of advertising research
(40)3, 67-71.
Meitner, M., 2004. Scenic beauty of river views in the Grand Canyon: Relating perceptual judgments
to locations. Landscape and Urban Planning, 68 (1), 3–13.
Molina-Ruiz, J., Martínez-Sánchez, M., Pérez-Sirvent, C., Tudela-Serrano, M., García-Lorenzo, M.,
2011. Developing and applying a GIS-assisted approach to evaluate visual impact in wind farms.
Renewable Energy 36 (3), 1125-1132. doi:10.1016/j.renene.2010.08.041
Murgoitio, J., et al. 2014. Airbirne LiDAR and terrestial laser scanning derived vegetation obstruction
factor for visibility models. Transaction in GIS, 18(1), 147-160. doi: 10.1111/tgis.12022.
Nagle, J., 2009. Cell phone towers as visual pollution, 23 Notre Dame J.L. Ethics & Pub. Policy 23(2),
537-568.
Polish Outdoor Advertising Chamber of Commerce, 2013. Report IV 2013 (www.igrz.com.pl)
(accessed 7 Jan 2014)
Patoczka, P., 2000. Ściany i bramy w krajobrazie (The walls and gates of the landscape). Krakow:
Politechnika Krakowska.
Peccol, E., Bird, A.,and Brewer, T.R., 1996. GIS as a tool for assessing the influence of country side
designations and planning policies on landscape change. Journal of Environmental
Management, 47(4), 355-367. doi-10.1006/jema.1996.0059
Penteado, C., 2007. A Sign of Things to Come? Advertising Age, 1 October (adage.com/print/120839)
(accessed 5.03.2013)
Portella, A., 2014. Visual Pollution: Advertising, Signage and Environmental Quality. UK: Ashgate
Publishing.
PricewaterhouseCoopers Annual Rapport 2011; http://www.pwc.pl/pl/wielkie-miasta-
polski/raport_Lublin_2011.pdf (accessed 27.10.2013)
R Core Team (2014). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/ (accessed 06.02.2015)
Ribeiro, L., and Barao, T. 2006. Greenways for recreation and maintenance of landscape quality: five
case studies in Portugal. Landscape and Urban Planning, 76 (1-4), 79-97. doi-
10.1016/j.landurbplan.2004.09.042
Rodrigues, M., Montañés C., and Fueyo, N,. 2010. A method for the assessment of the visual impact
caused by the large-scale deployment of renewable-energy facilities. Environmental Impact
Assessment Review,30(4), 240–246. doi:10.1016/j.eiar.2009.10.004
Sander, H. A., and Manson, S. M. 2007. Heights and locations of artificial structures in viewshed
calculation: How close is close enough? Landscape and urban planning, 82(4), 257-270.
Slager, C. T .J., Vries de., B., 2013. Landscape generator: Method to generate landscape configurations
for spatial plan-making. Computers, Environment and Urban Systems 39, 1–11.
http://dx.doi.org/10.1016/j.compenvurbsys.2013.01.007
Schirpke, U., Tasser E., Tappeiner, U. 2013. Predicting scenic beauty of mountain regions. Landscape
and Urban Planning 111 (1), 1-12. doi:10.1016/j.landurbplan.2012.11.010
Shang, H., Bishop, I., 2000. Visual thresholds for detection, recognition and visual impact in
landscape setting. Journal of Environmental Psychology (20), 125-140
doi:10.1006/jevp.1999.0153
Smardon, R. C., Palmer, J. E., and Felleman, J. P. 1986. Foundations For Visual Project Analysis.
New York: John Wiley and Sons.
Suleiman, W., Joliveau, T., and Favier, E., 2011. 3D urban visibility analysis with vector GIS data.
Presented at the GISRUK 2011, University of Portsmouth, UK, 27-29.
Taylor, J. G., Zube, E. H., and Sell,. J. L., 1987. Landscape assessment and perception research
methods, In: Bechtel, R.B., Marans, R.W., and Michelson, W., Methods in Environmental and
Behavioral Research. New York: Van Nostrand Reinhold Company, 361-393.
Ulrich, R.S., 1986. Human responses to vegetation and landscape. Landscape and Urban Planning
13(1), 26-44. doi-10.1016/0169 2046(86)90005-8
Vardavas, C. I., Connolly, G. N., and Kafatos, A. G. 2009. Geographical information systems as a tool
for monitoring tobacco industry advertising. Tobacco control, 18(3), 190-196.
doi:10.1136/tc.2008.026856
Vizzari, M. 2011. Spatial modelling of potential landscape quality. Applied Geography 31(1), 108 –118.
doi-10.1016/j.apgeog.2010.03.001
Wejchert, K. 1984. Elementy kompozycji urbanistycznej (Elements of urban composition). Warsaw:
Arkady.
Wilson, M., 2014. On the criticality of mapping practices: Geodesign as critical GIS? Landscape
Urban Planning, (in press), http://dx.doi.org/10.1016/j.landurbplan.2013.12.017
Wojciechowski, K. H., 1986. Problemy percepcji i oceny estetycznej krajobrazu (Perception and
aesthetic evaluation of landscape). Lublin: UMCS Press.
Yilmaz, D., 2011. In the Context of Visual Pollution: Effects to Trabzon City Center Silhoutte. Asian
Social Science 7(5), 98-109 doi:10.5539/ass.v7n5p98
Zube, E. H., 1987. Perceived land use patterns and landscape values. Landscape Ecology 1(1), 37-45.
doi-10.1007/BF02275264
This research was funded by Polish National Science Centre (NCN) Grant No. DEC-2012/07/D/HS-
4/01569
