Risk factors and spatial distribution of urban rat infestations

Abstract

Urban rat infestations have multifactorial causes and may result in severe public health and environmental problems, as well as heavy economic losses. The identification of putative environmental determinants of urban rat infestations and the mapping of areas prone to experiencing such infestations (hot spots) are crucial for effectively addressing intervention efforts. We investigated the associations between a selection of environmental factors and the occurrence of rat infestations in the city of Madrid, Spain. This was done by modelling 10,956 citizen-reported rat sightings from 2002 to 2008 using generalized additive models, both at municipality (Madrid) and district (Latina) levels. Increased age and density of housing, and decreased distance to vegetated areas, markets and cat feeding stations were factors associated with an increased risk of rat infestations. Risk maps for rat infestations were also developed and recurrent hot spots of rat activity were identified. Although a better fit to the data was obtained in the model for the smaller scale and possibly more environmentally homogenous study area of Latina, modelling the spatial distribution of rat sightings was useful for identifying factors associated with an increased risk of urban rat infestations, as well as for identifying hot spots of rat activity, providing local authorities with a practical tool for effectively targeting intervention efforts to high-risk situations based on the local environmental contexts.

Full text

ORIGINAL PAPER
Risk factors and spatial distribution of urban rat infestations
Ibon Tamayo-Uria •Jorge Mateu •Francisco Escobar •
Lapo Mughini-Gras
Received: 6 June 2013 / Accepted: 12 September 2013 / Published online: 24 September 2013
ÓSpringer-Verlag Berlin Heidelberg 2013
Abstract Urban rat infestations have multifactorial cau-
ses and may result in severe public health and environ-
mental problems, as well as heavy economic losses. The
identification of putative environmental determinants of
urban rat infestations and the mapping of areas prone to
experiencing such infestations (hot spots) are crucial for
effectively addressing intervention efforts. We investigated
the associations between a selection of environmental
factors and the occurrence of rat infestations in the city of
Madrid, Spain. This was done by modelling 10,956 citizen-
reported rat sightings from 2002 to 2008 using generalized
additive models, both at municipality (Madrid) and district
(Latina) levels. Increased age and density of housing, and
decreased distance to vegetated areas, markets and cat
feeding stations were factors associated with an increased
risk of rat infestations. Risk maps for rat infestations were
also developed and recurrent hot spots of rat activity were
identified. Although a better fit to the data was obtained in
the model for the smaller scale and possibly more envi-
ronmentally homogenous study area of Latina, modelling
the spatial distribution of rat sightings was useful for
identifying factors associated with an increased risk of
urban rat infestations, as well as for identifying hot spots of
rat activity, providing local authorities with a practical tool
for effectively targeting intervention efforts to high-risk
situations based on the local environmental contexts.
Keywords Urban rodent management Rattus
norvegicus Norway rat Generalized additive model 
Risk map Geographic Information Systems
Introduction
The control of urban rat infestations, particularly those
caused by the Norway rat (Rattus norvegicus) and the black
rat (R.rattus), has primarily been seen as a matter of
protecting public health and safety, as well as preserving
urban ecosystem integrity. Rat feeding habits are highly
destructive and their nesting behaviours can compromise
the structure and functionality of infested premises. Squa-
lor and fear generated by rats are embedded in human
culture, encompassing medical, social and emotional
dimensions. Recent developments in the field of rat-borne
infections and other threats, such as allergies, injuries and
Communicated by J. Jacob.
I. Tamayo-Uria (&)
Biomedical Research Consortium for Epidemiology and Public
Health (CIBER), Madrid, Spain
e-mail: ibontama@gmail.com
I. Tamayo-Uria F. Escobar
Department of Geography and Geology, University of Alcala
´,
Madrid, Spain
I. Tamayo-Uria
Public Health Division of Gipuzkoa, BIODonostia Research
Institute, Department of Health of the regional Government of
the Basque Country, San Sebastian, Spain
J. Mateu
Department of Mathematics, University Jaume I, Castello
´n,
Spain
L. Mughini-Gras
Department of Veterinary Medical Sciences,
University of Bologna, Bologna, Italy
L. Mughini-Gras
Department of Veterinary Public Health and Food Safety,
Istituto Superiore di Sanita
`, Rome, Italy
123
J Pest Sci (2014) 87:107–115
DOI 10.1007/s10340-013-0530-x

psychological discomfort caused by the exposure to rats
and their excreta (Bonnefoy et al. 2008), have led to a
renewed interest in the environmental factors that may
increase the risk of experiencing urban rat infestations
(Langton et al. 2001; Traweger and Slotta-Bachmayr 2005;
Channon et al. 2006; Traweger et al. 2006; De Masi et al.
2009,2010; Tamayo Uria et al. 2013), as well as those
affecting the applied control measures (Patergnani et al.
2010; Mughini Gras et al. 2012). It is therefore becoming
increasingly evident that urban rat infestations are the
result of multiple factors that may eventually allow urban
rat colonies to establish and proliferate to a level that
exceeds human tolerance limits, resulting in the onset of a
pest problem.
Although rats may adapt to live in virtually all habitats,
including the urban milieu, some inner city areas have
shown to exhibit significantly higher amounts of rat
activity than others, the so-called ‘‘hot spots’’ (Channon
et al. 2006). This suggests that the distribution of urban rat
colonies is structured and that there are factors that make
certain areas more prone to rat infestations than others. For
instance, a strong correlation has been found between rat
infestations and areas where problems of litter, vandalism,
poor socioeconomic conditions, dishevelled green areas,
neglect and vacant buildings are widespread (Langton et al.
2001; Traweger et al. 2006; De Masi et al. 2010). Age and
density of dwellings are also important factors that may
influence urban rat infestations (Langton et al. 2001;
Bonnefoy et al. 2008). People, however, may exercise
varying degrees of control over these factors, including the
amount of food, water and shelter available to rats.
Moreover, the behaviour of rat populations may vary from
city to city due to the local environmental conditions. The
identification of factors associated with an increased risk of
urban rat infestations is important for effectively targeting
intervention efforts to high-risk situations, and in many
parts of the world this is already considered in rat man-
agement interventions. There have been several docu-
mented instances of successful application of ecologically
based integrated rodent management programmes in urban
areas. These programmes posit that control activities
should rely primarily on the understanding of relationships
between rodents and their environments, with emphasis on
the processes that may regulate rodent populations in the
managed ecosystem (Frutos 1994; Colvin and Jackson
1999; Koul and Cuperus 2007). Indeed, it has been shown
that gaining a better understanding of the environmental
factors that favour infestation is able to raise the profile of
rodent management in both rural (e.g. White et al. 1998;
Singleton et al. 2003) and urban areas (e.g. Colvin and
Jackson 1999; Ferna
´ndez et al. 2007), with associated
economic analyses indicating promising outcomes (Davis
et al. 2004). A shift from managing rats to managing
‘‘ratty’’ environments as a means for controlling rat pop-
ulations is therefore increasingly advocated.
With a focus on the urban area of the city of Madrid,
Spain, the aims of this study were to: (1) investigate pos-
sible associations between the level of rat infestation and a
selection of environmental factors that may guide future
management efforts to areas prone to experiencing rat
infestations; and (2) develop risk maps for rat infestation to
identify hot spots of rat activity on which control efforts
should primarily be targeted. This is expected to provide
local authorities with a practical tool to support the
implementation of long-term policies for urban rat control
programmes that are structured in a more effective and
sustainable way. Moreover, so far no quantitative infor-
mation on the risk factors and spatial distribution of urban
rat infestations has been made available either for the city
of Madrid or other Spanish cities. In this regard, the
abundance of information gathered by the Technical Unit
for Vector Control (TUVC) of the Municipality of Madrid
offered a previously unattainable perspective to investigate
the proneness to the rat problem of a major capital city of
Southern Europe.
Materials and methods
Study area
The study was carried out at two geographical levels: (1)
municipality level, which comprises the whole city of
Madrid (latitude 40°250North, average altitude of 655 m
above sea level, *3.2 million inhabitants and 605 km
2
of
surface area); and (2) district (borough) level, which
comprises the district of Latina (*0.26 million inhabitants
and 25 km
2
of surface area), the most problematic district
of the city of Madrid with regard to rat infestations reported
to the TUVC.
Data
In the city of Madrid, direct sightings of rats or signs of
their presence (e.g. droppings, burrows, gnaw marks, etc.)
can be reported by citizens to the TUVC by telephone, fax,
email or in person at the front desk. The person reporting
the sighting is asked to provide some personal information
(at least the name and surname, birth date, residence
address and phone number) in addition to information on
where and when the sighting occurred (exact date and
address location). Only reports from people who declare to
have themselves sighted the rat(s) or the rat sign(s) in areas
falling within the administrative borders of the Munici-
pality of Madrid are accepted. Reports are thus entered in a
database dedicated to this type of citizen complaints and
108 J Pest Sci (2014) 87:107–115
123

queued up for in situ inspection and validation by the
TUVC staff. So far, all inspections have taken place within
24 h from reporting. The TUVC staff is composed of well-
trained and equipped pest control professionals who inspect
the locations where the rat sighting has occurred. The
inspection, which may also include a direct interview of the
reporting person, is largely based on standardized protocols
(e.g. CDCP 2006) and is targeted at ascertaining the pre-
sence of rats and/or other rodent species (e.g. house mouse,
Mus musculus). Inspections can be observational only and/
or trap based when there is no other apparent evidence of
rat presence. Both the infestation and its potential causative
conditions (e.g. garbage, defective sewers, etc.) are inves-
tigated. For documentation purposes, minutes and photo-
graphs of field observations are taken during inspections. A
premise is considered infested when at least one dead or
alive rodent subject and/or its active signs of presence are
detected by the TUVC staff.
Once the in situ inspection is performed and the pre-
sence of a rat problem is confirmed by the TUVC staff, the
report is considered as validated, and depending on the
type and extent of the problem, appropriate control inter-
ventions are implemented accordingly. Usually such
interventions, which are based on either classical control
measures (e.g. rodenticide baits) or environmental saniti-
zation activities, remain in force and are followed up with
repeated inspections until the problem is resolved. During
the inspections the TUVC staff is often able to identify
morphologically the species involved through direct
observation of closely sought, trapped, killed or dead
subjects in the field, and/or by examining typical rodent
signs (particularly, burrows and droppings) attributable to
the different species. It is worth mentioning that during the
last 20 years, the TUVC has performed several morpho-
metric surveys based on randomized overnight trapping
designs to determine the composition of the mammalian
microfauna of the city of Madrid, and neither R. rattus nor
other rat species other than R. norvegicus have ever been
found (TUVC, unpublished data). Also via in situ inspec-
tions, the presence of rat species other than R. norvegicus
has never been detected; thus, virtually all rat sightings in
Madrid can be assumed to pertain to the Norway rat. It
follows, therefore, that the findings presented here are
particularly representative of this rat species.
For the purposes of this study, we conducted a retro-
spective spatial analysis of 10,956 validated rat sightings
reported to the TUVC of Madrid from 2002 to 2008. This
means that only those rat sightings that were validated by
the TUVC staff (i.e. the presence of one or more rats was
ascertained with at least one inspection) were retained,
whereas doubtful or incomplete reports, or reports of
rodent species other than rats, were discarded. Moreover,
repeated reports made by the same or different people from
a location where another rat sighting had previously been
validated and control interventions were still ongoing at the
time of reporting were also discarded. This prevented the
analysis from inflating with multiple reports from the same
location that actually pertain to the same infestation.
Similar to previous studies (Tamayo Uria et al. 2013),
the visible presence of one or more rats was considered
here as an indicator of evident rat infestation, i.e. an
infestation that was likely to have exceeded the average
human tolerance limits beyond which rats become pests
and control measures are usually reactively implemented.
Rat sightings (hereafter interchangeably referred to as
rat infestations) were georeferenced and mapped using
Geographical Information Systems (GIS) tools to convert
the available spatial information (addresses of the site
locations where infestations were reported) to a two-
dimensional Cartesian coordinate system (Universal
Transverse Mercator [UTM] projections, European Datum
1950, zone 30-N). Kernel density function was then applied
to the georeferenced rat sightings to estimate the level of rat
infestation (expressed as number of rat sightings per km
2
)
at a resolution of 20 920 m over the whole study area
(Simonoff 1996). The centroids of each of the 109,077
buildings registered at the Madrid’s Land Registry that
fell in the study area were used as ‘‘anchor points’’ to be
matched with the corresponding infestation level.
A comprehensive list of 31 environmental factors that
could potentially influence the presence and density of rats
in urban areas was obtained through an expert elicitation
study (Tamayo Uria et al. 2013). Spatial (mappable) data
for the study area were available for seven of these
Table 1 Environmental variables tested for association with urban
rat infestations in Madrid, Spain
Factor Unit of
measurement
Resolution
c
Distance to the closest vegetated
area
m Coord.
Surface of vegetated area
a
m
2
Coord.
Year of construction of the
building
year Coord.
Distance to the closest water
source
b
m Coord.
Density of human population No. of individuals/
km
2
Sect.
Distance to the nearest market m Coord.
Distance to the nearest cat feeding
station
m Coord.
a
Within a 150 m radius around each of the buildings, according to
the average rat home range (Bonnefoy et al. 2008)
b
Fountains and ponds
c
Resolution of the data with which the measure was computed:
Coord. Cartesian coordinates, Sect. census block
J Pest Sci (2014) 87:107–115 109
123

putatively influencing factors (Table 1). Data on human
population density (census block population/census block
area) were obtained from the Spanish National Institute of
Statistics (INE) at the resolution of census block, the
smallest geographical unit used by the Spanish census
bureau for tabulation of population data. Data on markets
(centroids), cat feeding stations and water sources were
provided by the Municipality of Madrid at the level of
Cartesian coordinates. Data on green areas were also
obtained from the Municipality of Madrid as a georefer-
enced shapefile. The year of construction of the building
was provided by the Madrid’s Land Registry together with
the georeferenced shapefile of the buildings. Values of
human density were directly assigned to the buildings.
Euclidean distance was calculated to measure the distance
between each building centroid and its closest market, cat
feeding station, vegetated area and water source. The sur-
face (m
2
) occupied by green areas was calculated within a
150 m radius buffer around each of the building centroids,
according to the average rat home range (Bonnefoy et al.
2008).
The level of rat infestation at the points of buildings and
the environmental factors were treated as the dependent
and independent variables, respectively, in the analysis as
follows.
Statistical analysis
Associations between environmental factors and rat infes-
tations were investigated using generalized additive models
(GAMs) (Buja et al. 1989; Hastie and Tibshirani 1990;
Wood 2006). GAMs are a special type of statistical model
blending the properties of generalized linear models
(GLMs) with additive models, where neither the linear
structure nor the Gaussian distribution apply, and the
expected value of the response variable is linked to
covariates that are non-linear in form. Given a specified
probability distribution and a link function (g) relating the
expected value of the response variable (E[y]) to the
mpredictor variables, the model attempts to fit func-
tions f
i
(x
i
) to satisfy g(E[y]) =b
0
?f
1
(x
1
)?f
2
(x
2
)?
?f
m
(x
m
). Functions f
i
(x
i
) can be fitted using either para-
metric or non-parametric procedures, and hence better fits
to data may be achieved (Buja et al. 1989; Hastie and
Tibshirani 1990; Wood 2006). In this study, two Poisson
GAMs with a log link function were developed. One model
predicted rat infestations at the municipality level and the
other at the district level. Covariates considered were the
seven environmental factors reported in Table 1that were
included as parametric terms in the models plus a non-
parametric smoothing term fitted with 29 knots that took
into account the spatial structure of the data in terms of
xand ycoordinates. This latter term was constructed using
penalized regression splines over the coordinates combined
with the linear parametric form of the covariates, as
reported elsewhere (Friedman and Stuetzle 1981; Breiman
and Friedman 1985). The diagnosis of the models was
performed by goodness-of-fit and residual plotting.
Finally, a closely spaced set of predicted values from the
two GAMs was mapped both at the municipal and district
levels to obtain risk maps for rat infestation. This was done
by generating the fitted values from the two GAMs (i.e. the
values for the response variable that were predicted by the
models fitted to the data) on which ordinary kriging was
then applied to infer the values at unknown points as the
average of the known values at its neighbours, weighted by
the neighbours’ distance to the unknown point (Chiles and
Delfiner 1999). This allowed us to model and map the risk
of rat infestation so as to identify hot spots of rat activity.
In addition to the 2002–2008 cumulative models based
on the whole set of data (10,956 rat sightings in Madrid,
1,265 of which were in Latina), three other models were
run on the data split into three time periods: 2002–2003
(3,869 rat sightings in Madrid, 428 of which were in La-
tina), 2004–2006 (3,951 rat sightings in Madrid, 454 of
which were in Latina) and 2007–2008 (3,136 rat sightings
in Madrid, 381 of which in Latina). This temporal
arrangement was made to gain a good fit of the models to
the data while preserving the utility of this time-split
analysis in highlighting possible temporal trends in hot spot
occurrence.
The spatial analysis was carried out using ESRI ArcGIS
9.2 and the statistical analysis with R 2.11 statistical
environment. The significance level in the analysis was
a=0.05.
Results
Of the 10,956 validated rat sightings reported to the
Madrid’s TUVC from 2002 to 2008 (58 rat sightings/km
2
,
on average), 1,265 (11.5 %) were reported from the district
of Latina (108 rat sightings/km
2
, on average). In the whole
city of Madrid, an average of 4.3 rat sightings/day and 0.12
rat sightings/buildings were reported. In the district of
Latina, an average of 0.49 rat sightings/day and 0.15 rat
sightings/buildings were reported.
The b-coefficients, corresponding to 95 % confidence
intervals, and pvalues estimated by the two GAMs pre-
dicting rat infestations at the level of municipality
(Madrid) and district (Latina) are reported in Table 2. The
b-coefficients show the change in the dependent variable
(i.e. level of infestation, expressed as number of rat
sightings/km
2
) per each unit increase in the values of the
covariates in question, holding all the other covariates
included in the model constant. The variance explained by
110 J Pest Sci (2014) 87:107–115
123

the models (adjusted R
2
) was 55.5 % in the model for
Madrid and 94.2 % in that for Latina. Overall, both
models were statistically significant (z-test, pvalue
\0.0001), and all the environmental factors were signifi-
cantly associated with the outcome, with the exception of
the distance to the closest vegetated area in the model for
Latina, for which this factor was only borderline signifi-
cant (pvalue =0.0567).
In the model for Madrid (Table 2), factors significantly
associated with an increased risk of rat infestations were
the population density and the distance to the nearest water
source, whereas those significantly associated with a
reduced risk of rat infestations were the year of construc-
tion of the building, the distance to the nearest vegetated
area, nearest market and nearest cat feeding station, and the
surface of the nearest vegetated area.
In the model for Latina (Table 2), factors significantly
associated with an increased risk of rat infestations were
the population density, the distance to the nearest water
sources and the surface of the nearest vegetated area,
whereas factors significantly associated with a reduced
risk of rat infestations were the distance to the nearest
vegetated area, nearest market and nearest cat feeding
station.
In both Latina and Madrid models, the non-parametric
term was statistically significant (p\0.05). The non-
parametric term provides information on the spatial struc-
ture of the data. As this term was significant, the expected
value of the number of sightings depended on a set of
covariates that were spatially correlated. The total variance
explained by the three time-split models for Madrid was
53.1 % (2002–2003), 49 % (2004–2006) and 36.6 %
(2007–2008), while that explained by the three time-split
models for Latina was 95.6 % (2002–2003), 93.4 %
(2004–2006) and 88 % (2007–2008). Associations between
the considered risk factors and the level of rat infestation in
each of the time-split models were the same as those of the
2002–2008 cumulative models, so no further results are
presented. In the diagnosis of the residuals of the models, a
random distribution was observed with no residual struc-
ture not accounted for by the models. This is an indication
of goodness of fit of the models.
Hot spots of rat infestation were localized predomi-
nantly in the southwest and eastern parts of the city of
Madrid, including the district of Latina, which showed
important hot spots in its northeast part (Fig. 1). The time-
split risk maps (Figs. 2,3) showed that hot spots tend to
recur in the same inner city and inner district areas,
although with a generally decreasing intensity, a reflection
of the decreasing temporal trend in rat sightings.
Discussion
This study was performed to investigate possible associa-
tions between a set of environmental factors and urban rat
infestations in the city of Madrid and in the district of
Latina, Madrid’s most problematic district with regard to
rat infestations. Risk maps of rat infestation were also
developed to identify hot spots of rat activity.
We found that, in the city of Madrid, more recently
constructed buildings were significantly less prone to rat
infestation than older buildings. The age of housing is
known to influence rat infestations in urban areas, with rat
infestations being significantly more common in older
properties (Langton et al. 2001), and with higher infestation
rates in dilapidated structures (Battersby et al. 2002). Evi-
dence indicating that defective drains, a proxy of a generally
ageing community infrastructure, are associated with rat
infestations has also been provided (Langton et al. 2001;
Battersby et al. 2002). A coarse rule is that the older the
building, the greater is the likelihood for rats to be present.
This can be attributed to the fact that, in general, ageing
buildings usually require more intensive maintenance and
are more likely to be maintained under suboptimal condi-
tions than newly constructed buildings. Moreover, older
properties are less likely to have been built according to
effective rat-proof criteria, and the ‘‘maturity’’ of the
Table 2 Coefficients, their 95 % confidence intervals (95 % CI) and pvalues from the two GAMs for rat infestations in Madrid city and Latina
district
Madrid Latina
b-coefficient 95 % CI pvalue b-coefficient 95 % CI pvalue
Distance to the closest water source 3.92
-04
3.85
-04
3.99
-04
\2.0
-16
2.36
-04
2.08
-04
2.64
-04
\2.0
-16
Distance to the closest vegetated area -7.16
-04
-7.33
-04
-7.00
-04
\2.0
-16
-7.48
-05
-1.52
-04
-2.13
-06
0.0567
Density of human population 3.77
-04
3.72
-04
3.80
-04
\2.0
-16
1.03
-03
7.30
-04
1.33
-03
1.38
-11
Year of construction of the building -2.37
-03
-2.39
-03
-2.34
-03
\2.0
-16
-1.39
-04
-2.67
-04
-1.03
-05
0.0343
Distance to the nearest market -5.95
-04
-5.99
-04
-5.91
-04
\2.0
-16
-7.38
-04
-7.60
-04
-7.17
-04
\2.0
-16
Distance to the nearest cat feeding station -2.38
-05
-2.87
-05
-1.88
-05
\2.0
-16
-6.22
-05
-8.79
-05
-3.67
-05
1.83
-06
Surface of nearest vegetated area -3.64
-06
-3.79
-06
-3.48
-06
\2.0
-16
1.34
-06
6.76
-07
1.99
-06
7.40
-05
J Pest Sci (2014) 87:107–115 111
123

habitats around older buildings cannot be excluded to play a
role in favouring rat infestations (Langton et al. 2001).
It has been reported that the density of dwellings is
associated with rat infestations and that there may be a two-
sided effect, with both high and low densities making an
area particularly prone to rat infestation (Langton et al.
2001). In the case of Madrid and Latina, it seems that the
higher the population density (expressed as people/km
2
),
the greater is the chance of experiencing rat infestation.
High human population densities are usually associated
Fig. 1 Risk maps of rat infestation as predicted by the two GAMs for the city of Madrid (a) and for the district of Latina (b). The colorimetric
scale represents the number of rat sightings per km
2
. The district of Latina is highlighted in black in the map of Madrid (a)
Fig. 2 Risk maps of rat infestation as predicted by the three GAMs for the city of Madrid in 2002–2003, 2004–2006 and 2007–2008. The
colorimetric scale represents the number of rat sightings per km
2
Fig. 3 Risk maps of rat infestation as predicted by the three GAMs for the district of Latina in 2002–2003, 2004–2006 and 2007–2008. The
colorimetric scale represents the number of rat sightings per km
2
112 J Pest Sci (2014) 87:107–115
123

with high densities of dwellings, and the higher the density
of dwellings the more likely it is that a nearby dwelling
acts as a source of rat infestation (Langton et al. 2001),
especially since the home range of rats may well encom-
pass more than one dwelling at a time and dispersal is more
likely to be successful over short distances (Dickman and
Doncaster 1987). Moreover, high human densities could be
representative of other important determinants of rat
infestation, such as increased amount of refuse and possi-
bly even low socioeconomic status, as more impoverished
areas tend to be more densely populated.
We also found that increasing distance to markets and
vegetated areas decreases the risk of rat infestations. It is
clear that food availability is an important factor in
attracting rats to a given area as well as to allow them to
proliferate. A characteristic of markets is that they usually
produce large amounts of waste food. Although rats are
omnivorous, opportunistic feeders and rarely find it hard to
obtain enough food to survive, especially in urban envi-
ronments, an easy access to food sources such as those
provided by markets may well encourage infestation.
Regarding vegetated areas, it is known that vegetation may
regulate the local microclimate, in particular vegetation
filters the direct solar radiation, reduces wind speed and
moderates local temperatures. Vegetation also contributes
to the decrease of noise generated by human activity, as
well as shielding any visual intrusion, making it less dis-
turbing for rodents (Bolund and Hunhammar 1999; Pat-
ergnani et al. 2010). Interestingly, the association between
rat infestations and the surface of the nearest vegetated
areas was positive in Latina and negative in Madrid. In
Latina, where the general socioeconomic context is rela-
tively low, most vegetated areas consist of dishevelled or
neglected gardens, and in such circumstances the wider the
garden, the more likely it is to provide rats with forage and
harbourage. Moreover, compost heaps may be more com-
mon or more substantial in larger gardens, providing
readily accessible food supply alongside the soft soil and
sloping terrain suitable for constructing secure burrow
(Calhoun 1963; Langton et al. 2001). On the other hand,
green areas of the city of Madrid as a whole, particularly in
the central districts, are generally well maintained and
mainly consist of large city parks with frequent attendance
by people, especially during daytime, a factor which may
explain the observed negative association. It is indeed
conceivable that a person may be more likely to report a rat
sighting if this actually occurred in his/her private garden
than in a public space.
It has been suggested that the need for water would
aggregate rats to areas in proximity to water sources such
as fountains, ponds, streams and rivers. A particularly
striking result from our study is that the closer the water
source, the lower is the risk of sighting a rat. Possibly, this
may be related to the structure of the sewage system in
proximity to water sources that may be less likely to pro-
vide rats with favourable harbourage. In fact, this part of
the sewage system tends to be particularly well maintained
due to the weightier water flow that the system itself usu-
ally has to deal with.
We found that increasing distance to cat feeding stations
decreases the risk of rat infestation. Historically, cats have
a reputation for being effective predators of small mam-
mals, including rats. However, there is little evidence of
their effectiveness, and studies of the diet of urban cats
reveal virtually no predation of rats (Baker et al. 2005),
with considerably stronger impact on bird predation instead
of on rodent (Nogales et al. 2004). Moreover, in urban
areas there may be a positive relationship between the
presence of cats and rats, perhaps because of a common
benefit derived from access to waste food (Jackson 1951),
or because, as it is likely in our case, rats could take
advantage of the food provided to cats.
The models for the city of Madrid explained approxi-
mately half of the variance encountered in the data (except
for the 2007–2008 model which was based on less data),
whereas the model for Latina explained almost all the
variance. This may be due to the larger variation of the
geographical features of the whole city compared to the
smaller study area of Latina. The output of GAM model-
ling reflects the magnitude of the individual covariates
present in the model. Any slight variation in their magni-
tudes implies a variation in the response variable, so that if
the covariates take different values in the whole Madrid
and in the smaller area of Latina, we should also expect
having different outputs of the response variable and so of
the intensity function. This lends weight to the suggestion
that, although pest managers can benefit from the appli-
cation of GAM methodologies, their usefulness may
depend on the size of the area to be managed. Specifically,
by applying GAMs to large areas there is a risk of losing
fine-scale information that may eventually be important to
determine what is needed to be done for keeping rodent
populations at safe levels for the community. Yet, while a
municipality-level map may not be representative of the
sum of all its parts, its utility depends on the understanding
of the scale on which the data are represented. For instance,
the municipality-level maps show the more problematic
areas relative to all areas present within the city of Madrid,
while the district-level maps show the same matter but
relative to the all areas within the district of Latina.
According to TUVC historical data, areas identified as at
increased risk of rat infestation do realistically reflect the
hot spots of the city. These hot spots also appeared to recur
over time, suggesting that the use of such information may
well help us in identifying potentially (and possibly pre-
viously undocumented) problematic areas in the near
J Pest Sci (2014) 87:107–115 113
123

future. The identification of these hot spots, which reflects
the balance of the foregoing limiting factors, will allow the
focusing of resources to control or eliminate the problem of
having rats and inter alia a potential source of diseases
close to human housing.
An apparent limitation of this study was the use of cit-
izens’ complaints as the basis for analysis, especially since
this response variable has not yet been validated with
respect to real rat abundance in field conditions. Indeed,
tolerance to rat presence depends on several factors related
to the local environmental context, propensity for human
outdoor/indoor activities, socioeconomic status and indi-
vidual sensitivity. Moreover, the level of complaints about
rats reported to the TUVC may not always in itself mean an
increased rate of infestation, or an increase in the overall
rat population, but merely reflect more frequent rat sight-
ings during daytime as a result of changes in local rat
behaviour or reduced tolerance by some members of the
public (Battersby et al. 2002). Other limitations associated
with the use of citizens’ complaints as a proxy for rat
abundance are due to the difficulty in objectively assessing
the extent to which the local environment influences rat
populations, which are only possibly related to the sight-
ings of rats by citizens. The same may be true for the
impact of environment on food availability and rat popu-
lation size. For instance, it is possible that the positive
association between population density and rat infestation
was merely the result of more reports per km
2
(instead of
more rats per km
2
) in densely populated areas. Given this,
the hypotheses generated by this study need to be sup-
ported by further (field) investigations based upon more
independent data, particularly through the implementation
of ad hoc rat surveys and/or more in depth observational
studies. Besides the limitations of the dependent variable,
there are limitations related to the use of environmental
(independent) variables that are probably proxies for other
factors that directly influence rat abundance. This prevents
us from fully understanding those micro-environmental
factors that actually support or deter rat infestations and
therefore prevent direct modification of such factors.
In conclusion, although this study has thrown up many
questions on the need of further investigation, a selection
of environmental factors was found to be significantly
associated with urban rat infestations, providing insights
about target environments that should be priorities for
management efforts because they are more likely to
experience rat infestations. Although our selection of risk
factors was constrained by data availability for the study
area and some may argue that they are obvious predictors
of rat infestations, they are in turn rather generalizable and
can well predict the susceptibility of an area to rat infes-
tations. Besides examining which micro-environmental
factors the independent variables are actually proxying for,
it would also be interesting to take into consideration the
possible link between sewers and rat infestation by
including, for instance, the types, materials and age of
pipelines.
The GAM showed a better fit to the data at the smaller-
scale study area of Latina, probably because of more
homogenous surroundings. Moreover, there was a decreas-
ing temporal trend in rat sightings over the study period, a
possible reflection of the rat control efforts made by TUVC
in the city of Madrid. Although our maps show the risk of
urban rat infestations in a relatively static way, their infor-
mation is expected to have important implications for future
rodent management practices, particularly in the monitoring
of long-term responses to rat control.
Acknowledgments The authors are grateful to all professionals who
provided data, information and help, with particular thanks to the
Madrid City Council (Madrid Salud).
References
Baker PJ, Bentley AJ, Ansell RJ, Harris S (2005) Impact of predation
by domestic cats Felis catus in an urban area. Mammal Rev
35:302–312
Battersby SA, Parsons R, Webster JP (2002) Urban rat infestations
and the risk to public health. Int J Environ Health Res 1:57–65
Bolund P, Hunhammar S (1999) Ecosystem services in urban areas.
Ecol Econ 29:293–301
Bonnefoy X, Kampen H, Sweeney K (2008) Public health signif-
icance of urban pests. World Health Organization, Copenhagen
Breiman L, Friedman JH (1985) Estimating optimal transformations
for multiple regression and correlation. J Am Statist Assoc
80:580–598
Buja A, Hastie T, Tibshirani R (1989) Linear smoothers and additive
models. Ann Stat 17:453–555
Calhoun JB (1963) The ecology and sociology of the Norway rat. US
Department of Health, Education and Welfare, Washington, DC
CDCP-Centers for Disease Control and Prevention (2006) Integrated
pest management: conducting urban rodent surveys. US Depart-
ment of Health and Human Services, Atlanta
Channon D, Channon E, Roberts T, Haines R (2006) Hotspots: are
some areas of sewer network prone to re-infestation by rats
(Rattus norvegicus) year after year? Epidemiol Infect 134:41–48
Chiles J-P, Delfiner P (1999) Geostatistics: modeling spatial uncer-
tainty. Wiley series in probability and statistics. Wiley, Hoboken
Colvin BA, Jackson WB (1999) Urban rodent control programs for
the 21st century. In: Singleton GR, Hinds LA, Leirs H, Zhang Z
(eds) Ecologically based management of rodent pests. Australian
Centre for International Agricultural Research, Camberra,
pp 242–257
Davis S, Leirs H, Pech R, Zhang Z, Stenseth N (2004) On the
economic benefit of predicting rodent outbreaks in agricultural
systems. Crop Prot 23:305–314
De Masi E, Villac¸a P, Razzolini MTP (2009) Environmental
conditions and rodent infestation in Campo Limpo district, Sao
Paulo municipality, Brazil. Int J Environ Health 19:1–16
De Masi E, Pino FA, Santos MDS, Genehr L, Albuquerque JOM,
Bancher AM, Alves JCM (2010) Socioeconomic and environ-
mental risk factors for urban rodent infestation in Sao Paulo,
Brazil. J Pest Sci 83:228–238
114 J Pest Sci (2014) 87:107–115
123

Dickman CR, Doncaster CP (1987) The ecology of small mammals in
urban habitats. I. Populations in a patchy environment. J Anim
Ecol 56:629–640
Ferna
´ndez MS, Cavia R, Cueto GR, Sua
´rez OV (2007) Implemen-
tation and evaluation of an integrated program for rodent control
in a Shantytown of Buenos Aires city, Argentina. EcoHealth
4:271–277
Friedman JH, Stuetzle W (1981) Projection pursuit regression. J Am
Stat Assoc 76:817–823
Frutos J (1994) Biologı
´a y control de plagas urbanas. McGraw-Hill
Interamericana Editores, Madrid
Hastie T, Tibshirani R (1990) Generalized additive models. Chapman
& Hall/CRC, Boca Raton
Jackson WB (1951) Food habits of Baltimore, Maryland cats in
relation to rat populations. J Mammal 32:458–461
Koul O, Cuperus GW (2007) Ecologically-based integrated pest
management. CABI publishing, Wallingford
Langton SD, Cowan DP, Meyer AN (2001) The occurrence of
commensal rodents in dwellings as revealed by the 1996 English
house condition survey. J Appl Ecol 38:699–709
Mughini Gras L, Patergnani M, Farina M (2012) Poison-based
commensal rodent control strategies in urban ecosystems: some
evidence against sewer-baiting. EcoHealth 9:75–79
Nogales M, Martina A, Tershy BR, Donlan CJ, Veitch D, Puerta N,
Wood B, Alonso J (2004) A review of feral cat eradication on
islands. Conserv Biol 18:310–319
Patergnani M, Mughini Gras L, Poglayen G, Gelli A, Pasqualucci F,
Farina M, Stancampiano L (2010) Environmental influence on
urban rodent bait consumption. J Pest Sci 83:347–359
Simonoff JS (1996) Smoothing methods in statistics. Springer series
in statistics. Springer, New York
Singleton GR, Sudarmaji Tuan NP, Sang PM, Huan NH, Brown PR,
Jacob J, Heong KL, Escalada MM (2003) Reduction in chemical
use following integrated ecologically based rodent management.
Int Rice Res Notes 28(2):33–35
Tamayo Uria I, Mateu Mahiques J, Mughini Gras L (2013) Temporal
distribution and weather correlates of Norway rat (Rattus
norvegicus) infestations in the city of Madrid, Spain. EcoHealth
10:137–144
Traweger D, Slotta-Bachmayr L (2005) Introducing GIS-modelling
into the management of a brown rat (Rattus norvegicus Berk.)
(Mamm. Rodentia Muridae) population in an urban habitat.
J Pest Sci 78:17–24
Traweger D, Travnitzky R, Moser C, Walzer C, Bernatzky G (2006)
Habitat preferences and distribution of the brown rat (Rattus
norvegicus Berk.) in the city of Salzburg (Austria): implications
for an urban rat management. J Pest Sci 79:113–125
White J, Horskins K, Wilson J (1998) The control of rodent damage
in Australian macadamia orchards by manipulation of adjacent
non-crop habitats. Crop Prot 17:353–357
Wood SN (2006) Generalized additive models: an introduction with
R. Chapman & Hall/CRC, Boca Raton
J Pest Sci (2014) 87:107–115 115
123