‘Animals under wheels’: Wildlife roadkill data collection by citizen scientists as a part of their nature recording activities

Abstract

‘Animals under wheels’ is a citizen science driven project that has collected almost 90,000 roadkill records from Flanders, Belgium, mainly between 2008 and 2020. However, until now, the platform and results have never been presented comprehensively to the scientific community and we highlight strengths and challenges of this system. Data collection occurred using the subsite www.dierenonderdewielen.be (‘animals under wheels’) or the multi-purpose biodiversity platform observation.org and the apps, allowing the registration of roadkill and living organisms alike. We recorded 4,314 citizen scientists who contributed with at least a single roadkill record (207-1,314 active users per year). Non-roadkill records were registered by 85% of these users and the median time between registration of the first and last record was over 6 years, indicating a very high volunteer retention. Based on photographs presented with the roadkill records (n = 7,687), volunteer users correctly identified 98.2% of the species. Vertebrates represent 99% of all roadkill records. Over 145,000 km of transects were monitored, resulting in 1,726 mammal and 2,041 bird victims. Carcass encounter rates and composition of the top 10 detected species list was dependent on monitoring speed. Roadkill data collected during transects only represented 6% of all roadkill data available in the dataset. The remaining 60,478 bird and mammal roadkill records were opportunistically collected. The top species list, based on the opportunistically collected roadkill data, is clearly biased towards larger, enigmatic species. Although indirect evidence showed an increase in search effort for roadkill from 2010-2020, the number of roadkill records did not increase, indicating that roadkills are diminishing. Mitigation measures preventing roadkill could have had an effect on this, but decrease in population densities was likely to (partially) influence this result. As a case study, the mammal roadkill data were explored. We used linear regressions for the 17 most registered mammal species, determining per species if the relative proportion per year changed significantly between 2010 and 2020 (1 significant decrease, 7 significant increases). We investigated the seasonal patterns in roadkill for the 17 mammal species, and patterns per species were consistent over the years, although restrictions on human movement, due to COVID-19, influenced the seasonal pattern for some species in 2020. In conclusion, citizen scientists are a very valuable asset in investigating wildlife roadkill. While we present the results from Flanders, the platform and apps are freely available for projects anywhere in the world.

Full text

‘Animals under wheels’: Wildlife roadkill data
collection by citizen scientists as a part of their nature
recording activities
Kristijn R.R. Swinnen1, Annelies Jacobs1, Katja Claus2, Sanne Ruyts1,
Diemer Vercayie1, Jorg Lambrechts1, Marc Herremans1
1Natuurpunt Studie, Mechelen, Belgium 2Department of Environment and Spatial Development, Brussel, Belgium
Corresponding author: Kristijn R.R. Swinnen (kristijn.swinnen@natuurpunt.be)
Academic editor: Sara Santos|Received 13 August 2021|Accepted 27 December 2021|Published 25 March 2022
http://zoobank.org/55B9F61D-9889-47ED-8D39-B7B4B76C2E45
Citation: Swinnen KRR, Jacobs A, Claus K, Ruyts S, Vercayie D, Lambrechts J, Herremans M (2022) ‘Animals under
wheels’: Wildlife roadkill data collection by citizen scientists as a part of their nature recording activities. In: Santos S,
Grilo C, Shilling F, Bhardwaj M, Papp CR (Eds) Linear Infrastructure Networks with Ecological Solutions. Nature
Conservation 47: 121–153. https://doi.org/10.3897/natureconservation.47.72970
Abstract
‘Animals under wheels’ is a citizen science driven project that has collected almost 90,000 roadkill records
from Flanders, Belgium, mainly between 2008 and 2020. However, until now, the platform and results
have never been presented comprehensively to the scientic community and we highlight strengths and
challenges of this system. Data collection occurred using the subsite www.dierenonderdewielen.be (‘ani-
mals under wheels’) or the multi-purpose biodiversity platform observation.org and the apps, allowing the
registration of roadkill and living organisms alike. We recorded 4,314 citizen scientists who contributed
with at least a single roadkill record (207-1,314 active users per year). Non-roadkill records were registered
by 85% of these users and the median time between registration of the rst and last record was over 6
years, indicating a very high volunteer retention. Based on photographs presented with the roadkill re-
cords (n = 7,687), volunteer users correctly identied 98.2% of the species. Vertebrates represent 99% of
all roadkill records. Over 145,000 km of transects were monitored, resulting in 1,726 mammal and 2,041
bird victims. Carcass encounter rates and composition of the top 10 detected species list was dependent
on monitoring speed. Roadkill data collected during transects only represented 6% of all roadkill data
available in the dataset. e remaining 60,478 bird and mammal roadkill records were opportunistically
collected. e top species list, based on the opportunistically collected roadkill data, is clearly biased to-
wards larger, enigmatic species. Although indirect evidence showed an increase in search eort for roadkill
from 2010-2020, the number of roadkill records did not increase, indicating that roadkills are diminish-
Nature Conservation 47: 121–153 (2022)
doi: 10.3897/natureconservation.47.72970
https://natureconservation.pensoft.net
Copyright Kristijn R.R. Swinnen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License
(CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
RESEARCH ARTICLE
Launched to accelerate biodiversity conservation
A peer-reviewed open-access journal

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
122
ing. Mitigation measures preventing roadkill could have had an eect on this, but decrease in population
densities was likely to (partially) inuence this result. As a case study, the mammal roadkill data were
explored. We used linear regressions for the 17 most registered mammal species, determining per species
if the relative proportion per year changed signicantly between 2010 and 2020 (1 signicant decrease,
7 signicant increases). We investigated the seasonal patterns in roadkill for the 17 mammal species, and
patterns per species were consistent over the years, although restrictions on human movement, due to
COVID-19, inuenced the seasonal pattern for some species in 2020. In conclusion, citizen scientists
are a very valuable asset in investigating wildlife roadkill. While we present the results from Flanders, the
platform and apps are freely available for projects anywhere in the world.
Keywords
Citizen science, data quality, mammals, presence only data, relative trends, roadkill, structured monitor-
ing, seasonal patterns
Introduction
Roads directly impact populations and species due to vehicle induced mortality. An
estimated 29 million mammals and 194 million birds are killed annually on European
roads (Grilo et al. 2020). Worldwide, all mortality sources considered, natural or hu-
man, vehicle induced mortality was 7% for adult mammals and 1% for adult birds
(Hill et al. 2019).
Apart from direct mortality by wildlife vehicle collisions, roads and trac do have
multiple eects on ecosystems and wildlife populations including habitat loss and
habitat fragmentation (Taylor and Goldingay 2010; Whittington et al. 2019). Roads
can have genetic eects by acting as a barrier and decreasing genetic diversity (Cof-
n 2007;Holderegger and Di Giulio 2010). Furthermore, the presence of roads, and
the intensity of their use, can result in behavioural changes of individuals and species
(Mumme et al. 2000; Kerley et al. 2002; Whittington et al. 2019).
Monitoring of wildlife roadkill can, apart from the collection of the numbers being
killed, facilitate monitoring of population trends, species distribution and invasions,
animal behaviour and contaminants and disease (Schwartz et al. 2020). Volunteer
citizen scientists can collect and/or process data as part of a scientic inquiry (Silver-
town 2009) and they play an important role in the data collection of roadkill records
in projects which have been initiated worldwide http://globalroadkill.net (Shilling et
al. 2015). Globally, there are dozens of web based systems to register wildlife vehicle
collision casualties or roadkill (Shilling et al. 2015). Citizen science data on roadkill
has proven to be a valuable data source for the identication of potential roadkill
hotspots (Shilling and Waetjen 2015; Périquet et al. 2018; Engleeld et al. 2020),
temporal patterns in roadkill (Raymond et al. 2021) and species range maps (Tiede-
man et al. 2019). Long term motivation of volunteers, support for the identication
of roadkill and feedback to volunteers are of critical importance in sustaining roadkill
citizen science projects (Bil et al. 2020). e Flemish project ‘Animals under wheels’
(Dieren onder de wielen) is one of the largest citizen science driven roadkill databases

Animals under wheels 123
worldwide (Waetjen and Shilling 2017). However, until now, the platform and results
have never been presented comprehensively to the scientic community. We high-
light strengths and challenges of this system, which is easily and freely available to
be deployed anywhere in the world for roadkill monitoring (and general biodiversity
monitoring as well).
Methods
We describe and analyse the roadkill data submitted to the online biodiversity data-
base https://waarnemingen.be, the local Flemish version of the international platform
https://observation.org. is platform allows for the registration of observations of all
plants, fungi and animals. Since the launch in 2008 until 2020, this resulted in more
than 26,200 species and 31,5 million observations for the 13,522 km2 of Flanders,
generating one of the densest biodiversity datasets in the world. Flanders is the north-
ern region of Belgium, situated in Western Europe. It has a very high human density
of 487 inhabitants/km2 (Statbel 2020) and 5.08 km of roads/km2, one of the densest
road systems in the whole of Europe (Vercayie and Herremans 2015). Flanders has
883 km of motorways, 6,040 km of regional roads and 64,080 km of local roads (FPS
Mobility and Transport 2011). We show the 2019 trac data since this is the last year
without a COVID-19 impact. Daily, over 70 million vehicle kilometres are driven on
Flemish motorways (Hoornaert 2019) and the monitoring of 880 motorway segments
indicated an average daily trac volume of 37,592 vehicles per segment per day (me-
dian = 32,067, min = 4,440 and max = 131,508) (Vlaams Verkeerscentrum 2021). On
regional roads, the monitoring of 127 segments showed an average daily trac volume
of 17,583 vehicles per segment per day (median = 16,666, min = 2,381 and max =
36,649). For local roads, the authors are not aware of available data. e most recent
available data from 2017 indicate the Flemish registered vehicles drive 61.1 billion
kilometre per year (Kwanten 2018).
Roadkill data in the waarnemingen.be database can be submitted using: (a) the
online platform https://waarnemingen.be, (b) the subsite www.dierenonderdewielen.
be (‘animals under wheels’) or (c) the apps ObsMapp for Android, iObs for iPhone
and recently ObsIdentify for all devices. On the online platform, the location of the
observation must be pinpointed on the map, date/time selected and species and ad-
ditional observation information ‘roadkill’ label must be selected using controlled vo-
cabulary (Waetjen and Shilling 2017). In the apps, location and time are derived from
the smartphone. Species and ‘roadkill’ must be selected using controlled vocabulary in
the appropriate data elds. Photographs and additional information can be added to
an observation but are not mandatory. e apps do also function in a voice recognition
mode to register observations, which is always useful, but essential when monitoring
during driving (Vercayie and Herremans 2015).
We analyse the number of users registering roadkill records, the active users per
year and the number of new users per year (recruitment) to show the long-term vi-

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
124
ability of the project. We investigate the number of roadkill records per user and the
distribution between users including the corresponding Gini coecient, a measure
of unevenness (0: totally equal, 1: a single person is responsible for all records) (Sau-
ermann and Franzoni 2015). We calculate the retention time per user, dened as the
time between the registration of the rst and the last roadkill per user. For all roadkill
registering citizen scientists, we examine if they also registered observations of plants,
fungi or living wildlife within the waarnemingen.be database.
Data quality
Quality control of the data is an important step in all scientic processes, and also
very important for citizen science projects (Wiggins et al. 2011). e data validation
procedure in the ‘waarnemingen.be’-database combines species specialists (experienced
volunteers) assigning a validation status to observations and an algorithm automati-
cally evaluating observations. is multi-step process depends on the proof presented
(not mandatory but possible), species status (common vs rare), location and time (was
there already a proven record of presence within a species group dependent dened
range of space and time) of the observation (Swinnen et al. 2018). Species specialists
can assign a validation status to an observation: (a) ‘Approved (based on evidence)’,
evidence can be a picture or sound, (b) ‘Approved (based on expert judgement)’, the
additional information or the knowledge of the observer makes it highly likely this is
a correct observation, (c) ‘Under review’, temporary status, no decision has been taken
yet, (d) ‘Cannot be assessed’, proof or explanation does not allow for a decision to be
made, (e) ‘Rejected’, observation was wrong and user does not correct it. e algo-
rithm can also assign a validation status: (f) ‘Automatic validation’, for a record to be
automatically validated, there need to be a number of earlier observations of the species
supported by proof (at least one or two), within a certain radius (ranging from 100 m
to 10 km) within a specied time range (60–3000 days). Remaining observations are
classied (g) ‘unveried’. e validation process is an interactive process where users
can be contacted for additional information or suggested to change the species name
or other details in case of an error. We investigate the possible error ratio by calculating
the percentage of approved observations (based on photographic evidence) which was
initially wrong but corrected by the user after interaction with a validator.
Methodology of data collection
To allow standardised data collection and a quantiable measure of search eort, two
options for data registration are oered to users. In 2013, the option to gather stand-
ardised transect data was added to the website. Users were asked to choose a specic
route, draw it online and check it at least once every two weeks, but not more than
once a day. ey were asked to ll in the survey, even if no roadkill was detected. ese
type of transects are called xed transects in this manuscript. Since 2018, smartphone
users can allow their app to register their transect while observing nature and register-

Animals under wheels 125
ing observations. When nished, users indicate per species group if their transect can
be used as a roadkill monitoring transect. Since there are no requirements for transects
to be identical, or to be repeated over time, we call them variable transects.
For the xed transects, users register the transport modus (on foot, by bike, by car).
For the variable transects, the transect is recorded by the smartphone and we derived
the speed from the track length and duration, and classied transects as 0–7 km/h as
on foot, 7–25 km/h by bike and >25 km/h by car (although another motorised vehicle
is also possible). is distinction according to speed is important because speed aects
detection probability and it is known that searching on foot is more eective than
counting while driving (Slater 2002). Data collected during standardised monitoring
contains more information but it is also more demanding for volunteers resulting in a
smaller number of participants (Bonney et al. 2009).
Waarnemingen.be is mainly used as a personal notebook by naturalists to register
and document their sightings. Although some users are aware of the additional scien-
tic advantages standardised data collection oers, the majority of all observations in
waarnemingen.be are presence only records (also known as roving records) (Vercayie
and Herremans 2015). Given the correct identication of the species, presence is con-
rmed but search eort is unknown. e absence of a record can have multiple causes:
no roadkill present, no observer present or both present but not registered by the ob-
server. We show a summary of the transect data including transect characteristics and
top 10 of recorded bird and mammal species and calculate the average distance that
needs to be covered to encounter a roadkill. For the presence only data, a top 20 for
bird and mammal casualties is presented and we compare the results with the data col-
lected during transect counts. While herpetofauna is also an important species group,
e.g. because of their worldwide threatened status (Heigl et al. 2017), we do not discuss
them here since they are only recorded at lower driving speeds, and a larger (roadkill)
database, separate from waarnemingen.be is available, calling for a specic analysis.
Case study: mammal roadkill records
e number of new observations (of all organisms) submitted to waarnemingen.be
continues to increase year after year, from 400,000 in 2008 to over 6,000,000 in 2020
(and over 8.7 million in 2021). For 2010–2020 we investigate by means of a linear re-
gression (R Core Team 2016): (a) is there an increase in mammal roadkill observations?
(b) is there an increase in mammal observations (excluding all automated observations
by camera traps and bat-detectors since they do not represent human search eort)?
(c) are both correlated?
e large majority of roadkill data is collected as presence only data. Since search
eort is unknown, absolute roadkill trends per species cannot be calculated. However,
relative trends can be calculated and give an indication of the increase or decrease of
roadkill abundance of a specic species compared to the other species killed on the
road. For this analysis all mammal roadkill records were combined (presence only and
transect data), excluding observations where observers indicated they were uncertain

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
126
of species determination (1.5% of observations), and only species with a minimum of
50 roadkill individuals were withheld, resulting in 17 species (only species level records
were considered). Per species, the percentual abundance per year from these 17 species
was calculated. By using a linear regression, we determine per species if the relative
proportion per year changed signicantly between 2010 and 2020. Graphs were made
using ggplot 2 (R Core Team 2016; Wickham 2016). Based on unstructured, presence
only, citizen science data on roadkill, we propose the relative change in proportion of
roadkill victims as a means to gain insight in relative population changes as roadkill
numbers are expected to be strongly and positively associated with the local abundance
of living animals (Baker et al. 2004; George et al. 2011; Pettett et al. 2018; Schwartz
et al. 2020).
Apart from the local abundance, timing within the year does inuence the number
of victims found. Animals are sensitive to wildlife vehicle collisions during movement.
is can be daily movement while foraging or patrolling home ranges, or seasonal-
ity in mating, juvenile dispersal or migration (Taylor and Goldingay 2010; Garriga
et al. 2017; Schwartz et al. 2020). For all roadkill data combined (presence only and
transect data) we plot species specic density functions using ggplot 2 (R Core Team
2016; Wickham 2016). For this, the number of records was used, and not the number
of individuals. Overall, 98.7% of records comprises a single individual, but more than
one individual is also sometimes reported. is can reect reality, multiple individuals
killed at once or, sometimes, users combine a number of observations from a timespan
from the same location and add a single observation to the database. Analysing these
‘combination records’ as if all individuals were killed at the same time would introduce
errors in this seasonal pattern and to avoid this, the number of observations was used.
For species with more than 1,000 records, we show the annual seasonal pattern in
roadkill data. When fewer data are available, a single density plot combining the data
from 2010–2020 is shown.
Results
Within Flanders, 89,276 roadkill records were registered from 1960–2020 (Fig. 1).
Mammals (52,847), birds (23,346) and herpetofauna (11,762) represent 99% of road-
kill observations. Coleoptera (n = 499) is the invertebrate group with the most roadkill
records. One record can contain multiple individuals. Most records (93%) date from
2008 onwards, the launch of waarnemingen.be. e majority of ‘historical’ records
(79%) were added by a single account (Regional Mammal Workgroup).
A total of 4,314 citizen scientists submitted at least one roadkill record from Flan-
ders (Fig. 2). Male roadkill registering volunteers (1,547) are three times as abundant
compared to females (457). For 2,310 citizen scientists the sex is unknown. On average
881 users were active per year (range 207–1,314) and this number shows a steady in-
crease. Per year, on average 332 (range 207–465) ‘new’ users register their rst roadkill
victim with an increase of 20% in 2020 compared to the second best year (2009).

Animals under wheels 127
Contributions of users are unequal with 44.4% of users only registering a single
roadkill record (see Table 1). e median number of roadkill records per user is 2 (aver-
age 21, range 1–4,931). e Gini coecient of inequality between users is 0.87.
When including all roadkill registering users, volunteer retention time, i.e. the
median time between registration of rst and last roadkill record, is 7 days. For users
Figure 2. the number of active roadkill registering users per year in Flanders and the number of rst
time roadkill registering users per year in Flanders since 2008, the launch of https://waarnemingen.be
until 2020.
Figure 1. Roadkill observations per decade (1960-1999) or per year (2000-2020) and cumulative num-
ber of roadkill observations in Flanders, Belgium.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
128
with only a single roadkill record, we consider this single record as the rst and the last
record and the time between records was 0 days. When excluding the users with only
a single roadkill record, the median volunteer retention time increases to over 4 year
(1,501 days).
e majority of roadkill recorders (85%) did also submit non-roadkill observa-
tions to the biodiversity database and together they are responsible for 25.9 million
non-roadkill observations (on a total of 31.5 million non-roadkill records by 49,447
users registered in 2008–2020 in Flanders). is indicates that for most users, the
registration of roadkill is a natural part of their registration of nature observations, but
the focus is rarely on roadkill alone. When calculating the median volunteer retention
time of citizen scientists which registered at least a single roadkill record, based on all of
their observations, roadkill and living organisms together, this exceeds 6 years (2,318
days, range 0–5,243 days).
Data quality
In total, 38.9% of records were approved based on dierent procedures (Table 2). For
all observations approved based on the presented photographic evidence, only 139 out
of 7,687 recordings needed to be corrected by the validator. is results in an error rate
of 1.8%. In only a very small percentage of cases, users do not respond to suggestions
to change the species and the observation is then rejected.
All observations which were rejected, under review or which cannot be assessed are
removed in the following analyses.
Table 1. e amount of roadkill observations in 8 classes and the number of users in each class, including
the percentage of users per class.
Roadkill observations Users % of users
1 1,914 44.4%
2-5 1,254 29.1%
6-10 368 8.5%
11-20 267 6.2%
21-50 258 6.0%
51-100 114 2.6%
101-500 109 2.5%
501-5000 30 0.7%
Table 2. Validation status of the roadkill recordings in Flanders (1960-2020).
Validation status Number of observations (%)
Approved (based on evidence) 7,687 (8.61%)
Approved (based on expert judgement) 10,951 (12.27%)
Approved (automatic procedure) 16,062 (17.99%)
Under review 16 (0.02%)
Rejected 42 (0.05%)
Cannot be assessed 288 (0.32%)
Unveried 54,230 (60.74%)

Animals under wheels 129
Methodology of data collection
Transect data
We registered 309 xed transects online since the start in 2013 until 2020. A little
under half (148) were registered online but never monitored by the user. e remain-
ing 161 transects were monitored at least once, resulting in 2,521 records of bird and
mammal roadkill during 59,256 km of monitoring. In Table 3 we show the xed tran-
sect characteristics and results grouped per transport mode.
We registered 4,778 variable transects for bird and mammal roadkill since
2018, the year when the smartphone applications (ObsMapp and iObs) allowed
it, until the end of 2020. Each transect is considered unique since small variations
in the registration of the transect are present, resulting in no repeated counts per
transect. is resulted in 1,246 bird and mammal roadkill registrations while moni-
toring 86,235 km. In contrast with the xed transects, it is possible the user only
monitors a single species group. erefore, mammal and bird transects are shown
separately in Table 4.
When combining both transect types 3,767 roadkill records were registered. For
birds, carcass encounter rates vary from 1 carcass per 75.7 km on foot, 1 carcass per
59.3 km by car to 1 carcass per 34.6 km by bike. For mammal, carcass encounter
rates are similar, 1 carcass per 74.7 km on foot, 1 carcass per 70.7 km by car and 1
carcass per 43.5 km by bike. We show the top 10 of most frequently recorded (wild)
roadkill species for birds and mammals while monitoring transects by car (Table 5)
and bike (Table 6). We include observations not identied to species level, but they
are unranked.
Table 3. Fixed transect characteristics and results grouped per transport mode (2013-2020). * A single
transect can be monitored on foot, by bike and by car. at’s why the sum of the dierent transects diers
from 161.
Distance (km) Dierent
transect*
# counts Median # counts
per transect
Average # counts
per transect (range)
Roadkill Birds
found
Roadkill Mammals
found
By car 32,673 103 2,722 8 26 (1-484) 581 497
By bike 26,063 92 4,815 16.5 52 (1-1,204) 782 636
On foot 520 31 299 1 10 (1-70) 15 10
Table 4. Variable transect characteristics and results grouped per transport mode (2018-2020).
Distance (km) Dierent transects Roadkill victims
By car Birds 36,999 1,570 593
By car Mammals 39,910 1,723 529
By bike Birds 2,943 262 57
By bike Mammals 3,137 285 35
On foot Birds 1,600 461 13
On foot Mammals 1,646 477 19

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
130
Table 5. Top 10 of birds and mammal roadkill victims encountered the most frequently by car during
transect monitoring. Observations not identied to species level are shown but not ranked.
Birds Scientic name Common name # ind.
1Columba palumbus Common wood pigeon 329
Aves unknown Bird unknown 286
2Turdus merula Common blackbird 172
3Phasianus colchicus Common pheasant 74
4Anas platyrhynchos Mallard 37
5Corvus corone Carrion crow 30
6Buteo buteo Common buzzard 20
7Pica pica Eurasian magpie 18
8Coloeus monedula Western jackdaw 17
9Gallinula chloropus Common moorhen 17
10 Strix aluco Tawny owl 16
Mammals Scientic name Common name # ind.
Mammalia unknown Mammal unknown 270
1Erinaceus europaeus Hedgehog 223
2Lepus europaeus European hare 97
3Rattus norvegicus Brown rat 79
4Oryctolagus cuniculus European rabbit 70
5Martes foina Beech marten 67
6Sciurus vulgaris Red squirrel 39
6Vulpes vulpes Red fox 39
8Mustela putorius European polecat 18
Rattus unknown Rat unknown 7
9Capreolus capreolus Roe deer 5
Mustelidae unknown Marten unknown 5
10 Talpa europaea European mole 3
Table 6. Top 10 of birds and mammal roadkill victims encountered the most frequently by bike during
transect monitoring. Observations not identied to species level are shown but not ranked.
Birds Scientic name Common name # ind.
1Turdus merula Common blackbird 256
2Columba palumbus Common woodpigeon 169
Aves unknown Bird unknown 58
3Phasianus colchicus Common pheasant 46
4Anas platyrhynchos Mallard 28
5Coloeus monedula Western jackdaw 28
6Gallinula chloropus Common moorhen 24
7Passer domesticus House sparrow 24
8Erithacus rubecula European robin 23
9Streptopelia decaocto Eurasian collared dove 22
10 Parus major Great tit 20
Mammals Scientic name Common name # ind.
1Erinaceus europaeus Hedgehog 182
2Rattus norvegicus Brown rat 144
3Oryctolagus cuniculus European rabbit 71
4Lepus europaeus European hare 52
5Sciurus vulgaris Red squirrel 29
Mammalia unknown Mammal unknown 22
6Apodemus sylvaticus Wood mouse 14
6Martes foina Beech marten 14
Muridae unknown Mouse/rat unknown 12
Soricidae unknown Shrew unknown 12
8Talpa europaea European mole 11
Rattus unknown Rat unknown 10
Rodentia unknown Rodent unknown 10
Microtidae unknown Vole unknown 8
9Vulpes vulpes Red fox 6
10 Crocidura russula Greater white-toothed shrew 5

Animals under wheels 131
Presence only data
A total of 20,638 bird victims and 39,849 mammal victims were registered in waarne-
mingen.be from 2010–2020. Consequently, 94% of all roadkill records from 2010–
2020 are presence only data. We show the top 20 in Table 7.
Table 7. Top 20 of most registered bird and mammal roadkill victims which are collected as presence only
records. Observations not identied to species level are shown but not ranked.
Birds Scientic name Common name # ind.
1Turdus merula Common blackbird 3,686
2Columba palumbus Common woodpigeon 3,624
3Anas platyrhynchos Mallard 1,411
4Phasianus colchicus Common pheasant 1,294
5Tyto alba Western barn owl 926
6Strix aluco Tawny owl 817
Aves unknown Bird unknown 766
7Gallinula chloropus Common moorhen 761
8Buteo buteo Common buzzard 728
9Pica pica Eurasian magpie 504
10 Passer domesticus House sparrow 404
11 Coloeus monedula Western jackdaw 402
12 Athene noctua Little owl 333
13 Corvus corone Carrion crow 267
14 Streptopelia decaocto Eurasian collared dove 248
15 Asio otus Long-eared owl 234
16 Erithacus rubecula European robin 213
17 Garrulus glandarius Eurasian jay 212
18 Falco tinnunculus Common kestrel 194
19 Larus argentatus European herring gull 193
20 Turdus philomelos Song thrush 175
Mammals Scientic name Common name # ind.
1Erinaceus europaeus Hedgehog 12,147
2Vulpes vulpes Red fox 5,353
3Sciurus vulgaris Red squirrel 3,779
4Martes foina Beech marten 3,619
5Mustela putorius Western polecat 2,591
6Oryctolagus cuniculus European rabbit 2,569
7Lepus europaeus European hare 2,148
8Rattus norvegicus Brown rat 2,108
9Capreolus capreolus Roe deer 855
Mammalia unknown Mammal unknown 488
10 Talpa europaea European mole 317
Mustelidae unknown Marten unknown 287
11 Meles meles Eurasian badger 283
12 Mustela nivalis Least weasel 232
13 Mustela erminea Stoat 186
Martes foina/martes Beech/Pine marten 171
14 Sus scrofa Wild boar 137
Rattus unkown Rat unknown 74
15 Castor ber Eurasian beaver 67
16 Martes martes Pine marten 65
17 Apodemus sylvaticus Wood mouse 63
18 Crocidura russula Greater white-toothed shrew 59
19 Pipistrellus pipistrellus Common pipistrelle 46
20 Mus musculus House mouse 40

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
132
Mammal case study
We compare the number of non-roadkill mammal observations (one observation can con-
tain multiple individuals) with the number of mammal roadkill observations (transect and
present only data combined) annually from 2010–2020 in Flanders, Belgium (Table 8).
Over the years, there is a signicant increase in non-roadkill mammal observations
(slope = 9106, t = 4.49, p-value = 0.00150**) but no signicant increase in roadkill
registrations (slope = 118, t = 1.88, p-value = 0.09). ere is also no signicant cor-
relation between non-roadkill and roadkill mammal observations (slope = 0.008, t =
1.379, p-value = 0.201).
Table 9 shows the 17 mammal species with more than 50 roadkill individuals, the
outcomes from the linear regression between year (2010–2020) and the percentual
abundance per year.
Table 9. Outcome of the linear regression for the 17 most registered mammal species in Flanders from
2010-2020. Signicant codes in the p-value column: <0.1 . >0.05, <0.05 * > 0.01, <0.01 ** > 0.001,
<0.001 *** For common names, see Table 7.
Rank Species N slope Std. error t-value p-value
1Erinaceus europaeus 12,262 -0.051 0.325 -0.158 0.878
2Vulpes vulpes 5,193 -0.467 0.230 -2.029 0.073 .
3Sciurus vulgaris 3,769 0.047 0.131 0.358 0.728
4Martes foina 3,566 0.425 0.121 3.526 0.006 **
5Oryctolagus cuniculus 2,578 -0.339 0.170 -1.994 0.077 .
6Mustela putorius 2,514 -0.450 0.129 -3.500 0.007 **
7Rattus norvegicus 2,268 0.141 0.159 0.884 0.400
8Lepus europaeus 2,252 0.269 0.089 3.013 0.015 *
9Capreolus capreolus 798 0.147 0.046 3.165 0.012 *
10 Talpa europaea 328 0.023 0.024 0.961 0.362
11 Meles meles 275 0.119 0.035 3.431 0.007 **
12 Mustela nivalis 226 -0.004 0.012 -0.342 0.740
13 Mustela erminea 185 -0.004 0.013 -0.306 0.767
14 Sus scrofa 103 0.057 0.009 6.007 0.0002 ***
15 Apodemus sylvaticus 74 0.020 0.004 5.389 0.0004 ***
16 Castor ber 60 0.041 0.010 3.797 0.004 **
17 Martes martes 57 0.028 0.014 1.995 0.077 .
Table 8. Mammalian roadkill and non-roadkill observations per year and the percentage of roadkill com-
pared to all mammal observations from 2010-2020 in Flanders. Obs.= observations.
Year Mammal roadkill obs. Non-roadkill mammal obs. Mammal roadkill as % of total mammal obs.
2010 3,338 20,201 14.2%
2011 2,740 21,100 11.5%
2012 2,884 30,009 8.8%
2013 2,639 27,211 8.8%
2014 4,836 46,033 9.5%
2015 4,212 35,815 10.5%
2016 4,408 51,417 7.9%
2017 3,866 108,415 3.4%
2018 4,040 123,193 3.2%
2019 4,312 73,858 5.5%
2020 3,580 88,850 3.9%

Animals under wheels 133
Graphs showing percentual abundance per year per species are shown in Appendix
A. Mustela putorius is the only species with a signicant decreasing relative trend from
2010–2020. ere are seven species with an increasing relative trend, ordered here
from steepest to gentlest slope: Martes foina, Lepus europeaus, Capreolus capreolus, Meles
meles, Sus scrofa, Castor ber and Apodemus sylvaticus. Graphs showing seasonal patterns
in relative density per species for each year (2010–2020) are added to Appendix B.
Seasonal patterns in roadkill recordings dier clearly from species to species with most
species showing a bi- or unimodal pattern. When comparing the pattern from a single
species over multiple years, the consistency within the patterns is (very) good. Also the
species with fewer observations show mostly a clear seasonal pattern.
Discussion
e detected and registered roadkill observations are only the tip of the iceberg. Even a
structured daily roadkill census underestimates the death rate (of smaller victims) with a
factor 12–16 (Slater 2002). Apart from the eect that roadkill has on wildlife (popula-
tions) there is also an economic cost. ere are no numbers available for Flanders, or the
whole of Europe, but wildlife-vehicle collisions in Spain cost 105 million € yearly (Sáenz-
de-Santa-María and Tellería 2015) while the animal-vehicle accidents with ungulates in
Sweden resulted in a cost of 275 million € in 2015 (Gren and Jägerbrand 2019).
For Flanders, Capreolus capreolus, Sus scrofa, Canis lupus and Castor ber are among
the heaviest wild mammals, but injury or even death of drivers or passengers can also
occur when crashing into, or trying to avoid, smaller animals (Langbein 2007). A bet-
ter understanding of roadkill is therefore in the best interest of wildlife and humans.
e amount of roadkill records increased heavily since the launch of https://
waarnemingen.be in 2008 and together, over 4,300 citizen scientists collected almost
90,000 roadkill records. Similar to crowd science user contribution patterns, a small
number of users contributed most of the recordings and the Gini coecient of 0.87
is very similar to the average crowd science Gini coecient of 0.85 Sauermann and
Franzoni (2015) calculated for 7 crowd science projects. e registration of roadkill
seems to be an integrated part of the nature observation and registration, for most
volunteers, since 85% of users did also register non-roadkill observations. e use of a
multi-purpose biodiversity platform has a positive eect on the retention time, which
is over 6 years for roadkill recorders in waarnemingen.be. is long volunteer retention
time indicates that allowing the registration of all species groups, roadkill or not, us-
ing the tools the users are already familiar with, is a successful alternative, and possibly
even preferable to a single purpose data platform focussing on roadkill alone.
Some scientists may be sceptical about the data quality of records collected by
citizen scientists, although they have the potential to produce data with an accuracy
at least equal to professionals (Kosmala et al. 2016). We report a species identication
accuracy of roadkill recordings with photographs of 98% (n = 7,687) which is nearly
identical to the 97% presented by Waetjen and Shilling (2017). is high propor-

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
134
tion of correct species identication is an indication of the quality of the database.
However, we suspect species identication accuracy to be lower for records without
photographs since many of these identications are from driving vehicles. Although
more than 60% of observations are unveried, the majority of these observations are
‘common’ species, which are mainly registered by a limited group of experienced na-
ture observers, and there is no reason to assume ‘a priori’ that these records contain
more errors. Depending on the purpose of the analysis, dierent data selections can
be made but the increase in data quality by eliminating all possible errors does not
always compensate for the loss in data quantity (Van Eupen et al. 2021). Continuous
communication on the importance of photographs when registering roadkill aims to
increase the amount of veriable records in the future.
Differences in the most registered species depending on data collection method
In order to determine which species is killed the most in trac, standardised monitor-
ing is necessary. Our results indicate that for birds and mammal species, searching at
an intermediate speed from 7 to 25 km/h results in the highest number of carcasses
found. is is somewhat unexpected given that a slower speed should increase detec-
tion rates (Slater 2002). We suggest that the searching for roadkill carcasses was tted
into the routine of a number of people in the past years and that biking happens more
frequently next to busy roads, where more carcasses are present compared to walking,
which is more likely along calmer roads. Driving by car resulted in roughly the same
encounter rates of birds and mammals carcasses compared to walking, however due
to the higher speed, corpses not identied to species level are more numerous. Stop-
ping safely to identify the species is often not possible in Belgium and stopping on
motorways is forbidden (and dangerous) (minimum speed 70 km/h, maximum speed
120km/h). At this speed, identication at species level is frequently impossible.
e quality of transect data (with a standard protocol) is higher but it is more dif-
cult to nd volunteers to collect them (Bonney et al. 2009; Vercayie and Herremans
2015). As a consequence, they only represent 6% of all available roadkill data from
Flanders. Although informative for local situations, currently, this is too sparse for
region-wide analysis. e variable transects are promising in this respect because they
can be monitored anywhere and anytime, but they are currently not yet widely enough
adopted by the user community. It is also too early for a detailed analysis since they
were only launched in 2018. Additional promotion and awareness in the user commu-
nity of the applicability could boost the popularity of these variable transects.
ere is a clear dierence between the rank list of most observed species during
transects and the rank list of most observed species in the opportunistic data. When
comparing data collected by car and bike, it is clear that only larger species are regis-
tered from cars and a higher proportion was not identied on species level. For the
mammal data, all rank lists of most observed species are led by Hedgehogs (with the
exception of unidentied mammals which outrank them in species lists collected from
cars). Hedgehogs are frequently reported as trac victims in Western Europe (Huijser

Animals under wheels 135
and Bergers 2000; Pettett et al. 2018) and road mortality of Hedgehogs is expected to
be an important factor in their decline (Wright et al. 2020). Common blackbirds are
ranked third by monitoring from the car, but rst in the other lists. is is not unex-
pected since they had the highest predicted roadkill rate, 12 individuals/km/year, in the
model of Grilo et al. (2020) and are among the most frequently killed bird species in
Western Europe (Erritzoe et al. 2003). Even transect data must be interpreted with care.
Carcass persistence times and detection depend on size, with smaller animals being re-
moved faster by scavengers (Santos et al. 2011; Teixeira et al. 2013; Ratton et al. 2014).
Detection probability of larger mammals can also be inuenced since they are more
likely to be removed by maintenance workers or during police intervention at the site
of an accident. Data collected by these services can be an important addition to the data
collected by citizen scientists. Although proven to be a valuable data source (Grilo et al.
2009) additional steps need to be taken in Flanders to collect and centralise this data.
As expected, the ranking of victims collected as presence only data diers from the
rankings in the transect data: presence only data show a clear bias to larger species, but
possibly also species which are perceived as more interesting. Number two in the presence
only data ranking is Red fox, which ranks only 6th in transects by car, and 9th in transects
by bike. Foxes are infrequently seen alive, so, an encounter with a dead fox is for many
people special enough to report. e number three, Red squirrel ranked 6th in transects
by car and 5th in transects by bike. e Brown rat, the species encountered most frequent-
ly as roadkill (with exception from the Hedgehog) in transects by bike was only ranked 8th
in the presence only data list. is indicates that due to reporting bias the presence only
data should not be used to determine which species are killed the most in trac.
Mammal case study
From 2010–2020 there is a strong increase in the number of non-roadkill mammal
observations registered on waarnemingen.be but no signicant increase in registered
roadkill mammal observations. It is known that retention of volunteers can be chal-
lenging (Pocock et al. 2014; Shilling et al. 2015, 2020) but the number of observers
registering roadkill has never been higher than the past 3 years (see Fig. 2) and their
retention time on the waarnemingen.be platform exceeds 6 year. Volunteer participa-
tion depends also on repeated communication about the project. Over the last 3 years,
our own communication channels mentioned the project ‘animals under wheels’ in 23
newsletters, we provided 15 contributions to written magazines, made 2 promotion
videos and contributed to 10 national symposiums. Mainstream media wrote 47 arti-
cles about the project, and we gave 20 radio and 3 TV interviews (overview in Jacobs et
al. (2021)) on the subject. is indicates that the absence in increase in registered road-
kill mammals is not due to a reduction in observers/search eort but we believe that
this is a strong indication that the number of roadkill is diminishing. Additional stand-
ardised collected data could conrm/refute this hypothesis. If this reduction is caused
by eective road mitigation such as fencing, when possibly combined with crossing
structures or animal detection systems (Rytwinski et al. 2016) this reduction does not

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
136
reect a decrease in population but a decrease in wildlife victims due to the mitigation
measures. However, it might also reect a reduction in abundance of (a number of )
mammal species in Flanders that are most prone to being killed by vehicles.
Our species specic linear regression models indicate that 8 out of 17 mammal
species have a signicant change in proportion of roadkill victims through time. e
number of reported roadkill victims of Mustela putorius, the Western polecat, declines,
with the steepest signicant slope of all species (slope = -0.450). e polecat is sus-
pected to be in decline in Belgium, and also in most neighbouring countries (Croose
et al. 2018) and there are indications this decline was already present from 1998–2010
(Van Den Berge and Gouwy 2012).
e proportion of victims of the seven other species are increasing over the years.
Two species are (recently) recolonising (parts of) Flanders after a period of absence:
Eurasian beaver (Swinnen et al. 2017) and Wild boar (Rutten et al. 2019). Roe deer
has increased in range and numbers signicantly since the 70’s (Casaer and Huysen-
truyt 2016), Beech marten, is doing the same the last decades (Van Den Berge 2016)
and more recently, Badgers are also expanding from their last stronghold (Van Den
Berge et al. 2017). Although the increase in population density is not quantied, we
assume that this translates in higher relative roadkill numbers. e increase of the
Eurasian hare was unexpected since the species was recently added as vulnerable to the
red list of the Netherlands (bordering Flanders) (van Norren et al. 2020). However, for
Flanders no monitoring scheme is in place. For Wood mouse we have no knowledge
of population monitoring. is is a small-bodied species resulting in low carcass reten-
tion times (Santos et al. 2011; Ratton et al. 2014) and they were recorded relatively
infrequently indicating that these results have to be interpreted cautiously. Remark-
able is that the number of reported European hedgehog roadkill remains stable from
2010–2020. Until 2018, a strong decrease was occurring, but in 2019 and 2020 the
proportion abruptly increased and was again at the 2010 level. is increase is current-
ly unexplained but a fast recovery of the populations seems unlikely. ere are reports
of an unknown disease the last few years in Hedgehogs, possibly this also inuences
behaviour and making Hedgehogs more sensitive to being killed by cars.
Species distribution maps can be consulted at www.waarnemingen.be and addi-
tional info in Verkem et al. (2003). Linear regression models were also performed
for the period of 2010–2019 since the global pandemic of the coronavirus disease
(COVID-19) in 2020 resulted also in Flanders in connement measures which are
expected to have aected the search eort and the number of animals killed (Bíl et al.
2021; Driessen 2021). All trends remained similar, with the exception of the European
hare, where the increase became non-signicant.
Although the seasonal patterns are based on the rough data, without any correc-
tion for search eort within or between years, patterns of the same species are (highly)
consistent. We expect that the large amount of data smoothens smaller inter- and
within-year variation in search eort of individual observers. However, major events
are detectable. In Flanders, there was a strict ban on non-necessary (car)travel from the
18th of March 2020 to the 8th of June 2020 due to the COVID-19 pandemic. Apart
from the lives of wildlife this would have saved (Bíl et al. 2021; Driessen 2021), also

Animals under wheels 137
very few observers were on the road to quantify this eect. Determining which of both
factors was the most important is not possible using presence only data. For species in
which the peak period of kills overlaps with the connement measures, such as West-
ern polecat, the seasonal pattern of 2020 is clearly aected. Knowing the roadkill pat-
terns can help to protect specic species of interest by using specic warning signs, and
(temporal) road closure can even increase habitat quality (Whittington et al. 2019).
Although no age or sex of the individuals was recorded in most cases, most peaks in
roadkill density are presumed to be linked to increased movement because of mating or
juvenile movements and dispersal (Carvalho et al. 2018; Raymond et al. 2021).
We show that roadkill monitoring using citizen scientists can generate informa-
tive results. However, this is not the endpoint. Data collected during the ‘animals
under wheels’ project also contributed to the mitigation of local mortality hotspots.
Furthermore, the data can be consulted by policy makers and a number of questions
were asked in the Flemish Parliament concerning wildlife roadkill, indicating that the
problem is acknowledged at the political level.
Conclusion
Large quantities of roadkill records are collected by citizen scientists in Flanders, Bel-
gium. Volunteers remain engaged for a long period of time, probably due to the use of
a multi-purpose platform which also allows the registration of living organisms. Species
identication accuracy is high. Data collected using a standardised protocol is present,
however, data quantities are currently too low for nation-wide analysis. Currently, 94%
of all roadkill data are presence only records. Our results indicate that the amount of
mammal roadkill is diminishing in Flanders, possibly due to mitigation measures or
due to reduced population densities. We show that the citizen science data can be used
to detect trends in percentual abundance of roadkill per species per year and to show
seasonal patterns in relative roadkill density. Additional research to identify and conse-
quently mitigate roadkill hotspots, minimise and correct for biases and the comparison
between roadkill and population trends remains to be done. An increased eort to
convince observers to collect standardised transect data and photographs of roadkill
will increase the value of the dataset even further. We conclude that citizen scientists are
playing an important role in roadkill research and will continue to do so in the future.
Acknowledgements
e authors like to thank all citizen scientists for their records and the species experts for
the validation. Without your contributions, roadkill in Flanders would be a black box.
We thank Dominique Verbelen for his work on the bird species names. We thank the
IENE 2020 conference organising committee for the possibility to publish in the Special
issue: Linear Infrastructure Networks with Ecological Solutions. We thank the editor and
the anonymous referee for their contribution in signicantly improving the manuscript.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
138
References
Baker PJ, Harris S, Robertson CPJ, Saunders G, White PCL (2004) Is it possible to monitor
mammal population changes from counts of road trac casualties? An analysis using Bris-
tol’s red foxes Vulpes vulpes as an example. Mammal Review 34(1–2): 115–130. https://doi.
org/10.1046/j.0305-1838.2003.00024.x
Bil M, Heigl F, Janoska Z, Vercayie D, Perkins SE (2020) Benets and challenges of collaborating
with volunteers : Examples from National Wildlife Roadkill Reporting Systems in Europe.
Journal for Nature Conservation 54: e125798. https://doi.org/10.1016/j.jnc.2020.125798
Bíl M, Andrášik R, Cícha V, Arnon A, Kruuse M, Langbein J, Náhlik A, Niemi M, Pokorny B,
Colino-Rabanal VJ, Rolandsen CM, Seiler A (2021) COVID-19 related travel restrictions
prevented numerous wildlife deaths on roads: A comparative analysis of results from 11 coun-
tries. Biological Conservation 256: 109076. https://doi.org/10.1016/j.biocon.2021.109076
Bonney R, Cooper CB, Dickinson J, Kelling S, Phillips T, Rosenberg KV, Shirk J (2009) Citi-
zen Science: A Developing Tool for Expanding Science Knowledge and Scientic Literacy.
Bioscience 59(11): 977–984. https://doi.org/10.1525/bio.2009.59.11.9
Carvalho F, Lourenço A, Carvalho R, Alves PC, Mira A, Beja P (2018) e eects of a mo-
torway on movement behaviour and gene ow in a forest carnivore: Joint evidence from
road mortality, radio tracking and genetics. Landscape and Urban Planning 178: 217–227.
https://doi.org/10.1016/j.landurbplan.2018.06.007
Casaer J, Huysentruyt F (2016) Reewildbeheer zonder kennis van dichtheden – een realiteit in
Vlaanderen. Vakblad natuur bos landschap. Mei 26–29.
Con AW (2007) From roadkill to road ecology: A review of the ecological eects of
roads. Journal of Transport Geography 15(5): 396–406. https://doi.org/10.1016/j.
jtrangeo.2006.11.006
Croose E, Duckworth JW, Ruette S, Skumatov DV, Kolesnikov VV, Saveljev AP (2018) A
review of the status of the Western polecat Mustela putorius: A neglected and declining
species? Mammalia 82(6): 550–564. https://doi.org/10.1515/mammalia-2017-0092
Driessen MM (2021) COVID-19 restrictions provide a brief respite from the wildlife roadkill
toll. Biological Conservation 256: 109012. https://doi.org/10.1016/j.biocon.2021.109012
Engleeld B, Starling M, Wilson B, Roder C, McGreevy P (2020) e Australian roadkill re-
porting project— Applying integrated professional research and citizen science to monitor
and mitigate roadkill in Australia. Animals (Basel) 10(7): 1–20. https://doi.org/10.3390/
ani10071112
Erritzoe J, Mazgajski TD, Rejt Ł (2003) Bird Casualties on European Roads— A Review. Acta
Ornithologica 38(2): 77–93. https://doi.org/10.3161/068.038.0204
FPS Mobility and Transport (2011) Length of the Belgian road network. Available from: http://
mobilit.belgium.be/nl/publicaties/stat [accessed: 2021/12/06]
Garriga N, Franch M, Santos X, Montori A, Llorente GA (2017) Seasonal variation in verte-
brate trac casualties and its implications for mitigation measures. Landscape and Urban
Planning 157: 36–44. https://doi.org/10.1016/j.landurbplan.2016.05.029
George L, Macpherson JL, Balmforth Z, Bright PW (2011) Using the dead to monitor the
living: Can road kill counts detect trends in mammal abundance? Applied Ecology and
Environmental Research 9(1): 27–41. https://doi.org/10.15666/aeer/0901_027041

Animals under wheels 139
Gren IM, Jägerbrand A (2019) Calculating the costs of animal-vehicle accidents involving un-
gulate in Sweden. Transportation Research Part D, Transport and Environment 70: 112–
122. https://doi.org/10.1016/j.trd.2019.03.008
Grilo C, Bissonette JA, Santos-Reis M (2009) Spatial-temporal patterns in Mediterranean
carnivore road casualties: Consequences for mitigation. Biological Conservation 142(2):
301–313. https://doi.org/10.1016/j.biocon.2008.10.026
Grilo C, Koroleva E, Andrášik R, Bíl M, González-Suárez M (2020) Roadkill risk and popula-
tion vulnerability in European birds and mammals. Frontiers in Ecology and the Environ-
ment 18(6): 1–6. https://doi.org/10.1002/fee.2216
Heigl F, Horvath K, Laaha G, Zaller JG (2017) Amphibian and reptile road-kills on tertiary
roads in relation to landscape structure: Using a citizen science approach with open-access
land cover data. BMC Ecology 17(1): 1–12. https://doi.org/10.1186/s12898-017-0134-z
Hill JE, DeVault TL, Belant JL (2019) Cause-specic mortality of the world’s terrestrial ver-
tebrates. Global Ecology and Biogeography 28(5): 680–689. https://doi.org/10.1111/
geb.12881
Holderegger R, Di Giulio M (2010) e genetic eects of roads: A review of empirical evidence.
Basic and Applied Ecology 11(6): 522–531. https://doi.org/10.1016/j.baae.2010.06.006
Hoornaert S (2019). Verkeersindicatoren Snelwegen Vlaanderen 2019: 1–147.
Huijser MP, Bergers PJM (2000) e eect of roads and trac on hedgehog (Erinaceus eu-
ropaeus) populations. Biological Conservation 95(1): 111–116. https://doi.org/10.1016/
S0006-3207(00)00006-9
Jacobs A, Swinnen K, Vercayie D (2021) Registratie van faunaslachtoers in het verkeer : “Dieren
onder de wielen 3.0”. Natuurpunt Studie i.o.v. Vlaamse Overheid, Departement Omgeving.
Kerley LL, Goodrich JM, Miquelle DG, Smirnov EN, Quigley HB, Hornocker MG (2002) So-
ciety for Conservation Biology Eects of Roads and Human Disturbance on Amur Tigers.
Conservation Biology 16(1): 97–108. https://doi.org/10.1046/j.1523-1739.2002.99290.x
Kosmala M, Wiggins A, Swanson A, Simmons B (2016) Assessing data quality in citizen science.
Frontiers in Ecology and the Environment 14(10): 551–560. https://doi.org/10.1002/
fee.1436
Kwanten M (2018) Directoraat-generaal Duurzame Mobiliteit en spoorbeleid. Kilometers af-
gelegd door Belgische voertuigen in 2017.
Langbein J (2007) e Deer Initiative Ltd National Deer-Vehicle Collisions Project England
(2003–2005). Wrexham, UK.
Mumme RL, Schoech SJ, Woolfenden GE, Fitzpatrick JW (2000) Life and Death in the Fast
Lane : Demographic Consequences of Road Mortality in the Florida Scrub-Jay. Conserva-
tion Biology 14(2): 501–512. https://doi.org/10.1046/j.1523-1739.2000.98370.x
Périquet S, Roxburgh L, le Roux A, Collinson WJ (2018) Testing the value of citizen science
for roadkill studies: A case study from South Africa. Frontiers in Ecology and Evolution 6:
1–9. https://doi.org/10.3389/fevo.2018.00015
Pettett CE, Johnson PJ, Moorhouse TP, Macdonald DW (2018) National predictors of hedge-
hog Erinaceus europaeus distribution and decline in Britain. Mammal Review 48(1): 1–6.
https://doi.org/10.1111/mam.12107
Pocock MJO, Chapman DS, Sheppard LJ, Roy HE (2014) Choosing and Using Citizen Science:
a guide to when and how to use citizen science to monitor biodiversity and the environment.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
140
R Core Team (2016). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. https://www.R-project.org/
Ratton P, Secco H, da Rosa CA (2014) Carcass permanency time and its implications to
the roadkill data. European Journal of Wildlife Research 60: 543–546. https://doi.
org/10.1007/s10344-014-0798-z
Raymond S, Schwartz ALW, omas RJ, Chadwick E, Perkins SE (2021) Temporal patterns
of wildlife roadkill in the UK. PLoS One 16(10): 1–17. https://doi.org/10.1371/journal.
pone.0258083
Rutten A, Casaer J, Swinnen KRR, Herremans M, Leirs H (2019) Future distribution of wild
boar in a highly anthropogenic landscape : Models combining hunting bag and citizen science
data. Ecological Modelling 411: 108804. https://doi.org/10.1016/j.ecolmodel.2019.108804
Rytwinski T, Soanes K, Jaeger JAG, Fahrig L, Findlay CS, Houlahan J, van der Ree R, van
der Grift EA (2016) How Eective Is Road Mitigation at Reducing Road-Kill? A Meta-
Analysis. PLoS One 11(11): e0166941. https://doi.org/10.1371/journal.pone.0166941
Sáenz-de-Santa-María A, Tellería JL (2015) Wildlife-vehicle collisions in Spain. European Jour-
nal of Wildlife Research 61(3): 399–406. https://doi.org/10.1007/s10344-015-0907-7
Santos SM, Carvalho F, Mira A (2011) How long do the dead survive on the road? Carcass
persistence probability and implications for road-kill monitoring surveys. PLoS One 6(9):
e25383. https://doi.org/10.1371/journal.pone.0025383
Sauermann H, Franzoni C (2015) Crowd science user contribution patterns and their implica-
tions. Proceedings of the National Academy of Sciences of the United States of America
112(3): 679–684. https://doi.org/10.1073/pnas.1408907112
Schwartz ALW, Shilling FM, Perkins SE (2020) e value of monitoring wildlife roadkill. Euro-
pean Journal of Wildlife Research 66(1): 18. https://doi.org/10.1007/s10344-019-1357-4
Shilling FM, Waetjen DP (2015) Wildlife-vehicle collision hotspots at US highway extents:
Scale and data source eects. Nature Conservation 11: 41–60. https://doi.org/10.3897/
natureconservation.11.4438
Shilling F, Perkins SE, Collinson W (2015) Wildlife/Roadkill Observation and Reporting Sys-
tems. Handbook of Road Ecology, 492–501. https://doi.org/10.1002/9781118568170.ch62
Shilling F, Collinson W, Bil M, Vercayie D, Heigl F, Perkins SE, MacDougall S (2020) Design-
ing wildlife-vehicle conict observation systems to inform ecology and transportation stud-
ies. Biological Conservation 251: 108797. https://doi.org/10.1016/j.biocon.2020.108797
Silvertown J (2009) A new dawn for citizen science. Trends in Ecology & Evolution 24(9):
467–471. https://doi.org/10.1016/j.tree.2009.03.017
Slater FM (2002) An assessment of wildlife road casualties - e potential discrepancy between
numbers counted and numbers killed. Web Ecology 3(1): 33–42. https://doi.org/10.5194/
we-3-33-2002
Statbel (2020) StatBel. Available from: https://statbel.fgov.be/nl/themas/bevolking/bevolkings-
dichtheid [accessed: 2021/12/06]
Swinnen KRR, Strubbe D, Matthysen E, Leirs H (2017) Reintroduced Eurasian beavers (Cas-
tor ber): Colonization and range expansion across human-dominated landscapes. Biodi-
versity and Conservation 26(8): 1863–1876. https://doi.org/10.1007/s10531-017-1333-9
Swinnen KRR, Vercayie D, Vanreusel W, Barendse R, Boers K, Bogaert J, Dekeukeleire D,
Driessens G, Dupriez P, Jooris R, Asten K, Van Neucker T, Van Den Dorsselaer P, Van

Animals under wheels 141
Vooren P, Van Wysmantel N, Gielen K, Desmet P, Herremans M (2018) Waarnemingen.
be – Non-native plant and animal occurrences in Flanders and the Brussels Capital Region,
Belgium. BioInvasions Records 7(3): 335–342. https://doi.org/10.3391/bir.2018.7.3.17
Taylor BD, Goldingay RL (2010) Roads and wildlife: Impacts, mitigation and implications
for wildlife management in Australia. Wildlife Research 37(4): 320–331. https://doi.
org/10.1071/WR09171
Teixeira FZ, Coelho AVP, Esperandio IB, Kindel A (2013) Vertebrate road mortality estimates:
Eects of sampling methods and carcass removal. Biological Conservation 157: 317–323.
https://doi.org/10.1016/j.biocon.2012.09.006
Tiedeman K, Hijmans RJ, Mandel A, Waetjen DP, Shilling F (2019) e quality and contribu-
tion of volunteer collected animal vehicle collision data in ecological research. Ecological
Indicators 106: 105431. https://doi.org/10.1016/j.ecolind.2019.05.062
Van Den Berge K (2016) Advies over de populatiedynamiek en predatorcontrole van de
steenmarter.
Van Den Berge K, Gouwy J (2012) Bunzing op de dool. Meander 10: 4–8.
Van Den Berge K, Gouwy J, Berlengee F, Vansevenant D (2017) Verspreiding van de das (Me-
les meles) in Vlaanderen : recente evoluties. Rapporten van het Instituur voor Natuur-en
Bosonderzoek 2017(34). Instituurt voor Natuur-en bosonderzoek, Brussel. https://doi.
org/10.21436/inbor.12435597
Van Eupen C, Maes D, Herremans M, Swinnen KRR, Somers B, Luca S (2021) e im-
pact of data quality ltering of opportunistic citizen science data on species distribution
model performance. Ecological Modelling 444: 109453. https://doi.org/10.1016/j.ecol-
model.2021.109453
van Norren E, Dekker J, Limpens H (2020) Basisrapport Rode Lijst Zoogdieren 2020.
Vercayie D, Herremans M (2015) Citizen science and smartphones take roadkill monitoring to
the next level. Nature Conservation 11: 29–40. https://doi.org/10.3897/natureconserva-
tion.11.4439
Verkem S, De Maeseneer J, Vandendriessche B, Verbeylen G, Yskout S (2003) Zoogdieren in
Vlaanderen. Ecologie en verspreiding van 1987 tot 2002.
Vlaams Verkeerscentrum (2021) Verkeersindicatoren. http://indicatoren.verkeerscentrum.be/
vc.indicators.web.gui/indicator/index [accessed: 2021/12/06]
Waetjen DP, Shilling FM (2017) Large Extent Volunteer Roadkill and Wildlife Observation
Systems as Sources of Reliable Data. Frontiers in Ecology and Evolution 5: 1–10. https://
doi.org/10.3389/fevo.2017.00089
Whittington J, Low P, Hunt B (2019) Temporal road closures improve habitat quality for wild-
life. Scientic Reports 9(1): 1–10. https://doi.org/10.1038/s41598-019-40581-y
Wickham H (2016) ggplot2: Elegant Graphics for Data Analysis. https://doi.org/10.1007/978-
3-319-24277-4
Wiggins A, Newman G, Stevenson RD, Crowston K (2011) Mechanisms for data quality and
validation in citizen science. Proceedings - 7th IEEE International Conference on e-Science
Workshops, eScienceW 2011: 14–19. https://doi.org/10.1109/eScienceW.2011.27
Wright PGR, Coomber FG, Bellamy CC, Perkins SE, Mathews F (2020) Predicting hedgehog
mortality risks on British roads using habitat suitability modelling. PeerJ 2020: e8154.
https://doi.org/10.7717/peerj.8154

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
142
Appendix A
For the 17 mammal species with more than 50 roadkill individuals, we show the linear
regression gures between year (2010–2020) and the percentual abundance per year.
Signicant regressions are shown with a black line, non-signicant with a grey line.
Figure A1.
Figure A2.

Animals under wheels 143
Figure A3.
Figure A4.
Figure A5.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
144
Figure A6.
Figure A7.
Figure A8.

Animals under wheels 145
Figure A9.
Figure A10.
Figure A11.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
146
Figure A12.
Figure A13.
Figure A14.

Animals under wheels 147
Figure A15.
Figure A16.
Figure A17.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
148
Appendix B
For the 17 most recorded mammal species we show the variation in the roadkill pattern
within Flanders. For species with more than 1000 recordings, we show the pattern of
each individual year (2010-2020). For species with fewer than 1000 recordings all data
are combined to generate a general pattern.
Figure B1.
Figure B2.

Animals under wheels 149
Figure B3.
Figure B4.
Figure B5.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
150
Figure B6.
Figure B7.
Figure B8.

Animals under wheels 151
Figure B10.
Figure B11.
Figure B9.

Kristijn R.R. Swinnen et al. / Nature Conservation 47: 121–153 (2022)
152
Figure B12.
Figure B13.
Figure B14.

Animals under wheels 153
Figure B15.
Figure B16.
Figure B17.