Using spatial patterns in illegal wildlife uses to reveal connections between subsistence hunting and trade

Abstract

Although most often considered independently, subsistence hunting, domestic trade, and international trade as components of illegal wildlife use (IWU) may occurs near one another (spatially correlated). Understanding how and where subsistence and commercial uses may co-occur has important implications for the design and implementation of effective conservation actions. We analyzed patterns in the joint geographical distribution of illegal commercial and subsistence use of multiple wildlife species in Venezuela and evaluated whether available data were sufficient to provide accurate estimates of the magnitude, scope and detectability of IWU. We compiled records of illegal subsistence hunting and trade from several sources and fitted a Random Forest classification model to predict the spatial distribution of IWUs. From 1969 to 2014, 404 species and 8,340,921 specimens were involved in IWU, yielding a mean extraction rate of 185,354 ind./y. Birds were the most speciose group involved (248 spp.), but reptiles experienced the highest extraction rates (126,414 ind./y versus 3,133 ind./y for birds). Eighty eight percent of international trade records spatially overlapped with domestic trade, especially towards the north and across the coast but also in western inland areas. The distribution of domestic trade was broadly distributed along roads, suggesting that domestic trade does not depend on large markets in cities. Seventeen percent of domestic trade records overlapped with subsistence hunting, but the spatial distribution of this overlap covered a much larger area than between commercial uses. Domestic trade seems to respond to demand from rural more than urban communities. Our approach will be useful for understanding how IWU works at national scales in other parts of the world. This article is protected by copyright. All rights reserved.

Full text

Contributed Paper
Using spatial patterns in illegal wildlife uses to reveal
connections between subsistence hunting and trade
Ada S´
anchez-Mercado,∗¶ Marianne Asm¨
ussen,† Kathryn M. Rodr´
ıguez-Clark,† Jon Paul Rodr´
ıguez,†
and Wlodzimierz Jedrzejewski†
∗Centro de Estudios Bot´
anicos y Agroforestales, Instituto Venezolano de Investigaciones Cient´
ıficas (IVIC), P.O. Box 20632, Caracas,
1020-A, Venezuela
†Centro de Ecolog´
ıa, Instituto Venezolano de Investigaciones Cient´
ıficas (IVIC), P.O. Box 20632, Caracas, 1020-A, Venezuela
Abstract: Although most often considered independently, subsistence hunting, domestic trade, and inter-
national trade as components of illegal wildlife use (IWU) may be spatially correlated. Understanding how
and where subsistence and commercial uses may co-occur has important implications for the design and
implementation of effective conservation actions. We analyzed patterns in the joint geographical distribution
of illegal commercial and subsistence use of multiple wildlife species in Venezuela and evaluated whether
available data were sufficient to provide accurate estimates of the magnitude, scope, and detectability of IWU.
We compiled records of illegal subsistence hunting and trade from several sources and fitted a random-forest
classification model to predict the spatial distribution of IWUs. From 1969 to 2014, 404 species and 8,340,921
specimens were involved in IWU, for a mean extraction rate of 185,354 individuals/year. Birds were the most
speciose group involved (248 spp.), but reptiles had the highest extraction rates (126,414 individuals/year vs.
3,133 individuals/year for birds). Eighty-eight percent of international trade records spatially overlapped with
domestic trade, especially in the north and along the coast but also in western inland areas. The distribution
of domestic trade was broadly distributed along roads, suggesting that domestic trade does not depend on
large markets in cities. Seventeen percent of domestic trade records overlapped with subsistence hunting, but
the spatial distribution of this overlap covered a much larger area than between commercial uses. Domestic
trade seems to respond to demand from rural more than urban communities. Our approach will be useful
for understanding how IWU works at national scales in other parts of the world.
Keywords: poaching, random forest, subsistence hunting, Venezuela, wildlife illegal trade
Aplicaci´
on de Patrones Espaciales en los Usos Ilegales de la Vida Silvestre para Revelar las Conexiones entre la
Caza de Subsistencia y el Comercio
Resumen: Aunque muchas veces son considerados independientemente, la caza de subsistencia y el comercio
dom´
estico e internacional como componentes del uso ilegal de vida silvestre (UIVS) pueden estar correla-
cionados espacialmente. Entender c´
omo y d´
onde los usos comerciales y de subsistencia pueden co-ocurrir
tiene implicaciones importantes para el dise˜
no y la implementaci´
on de acciones efectivas de conservaci´
on.
Analizamos los patrones en la distribuci´
on geogr´
afica conjunta del uso ilegal comercial o por subsistencia de
m´
ultiples especies de vida silvestre en Venezuela y evaluamos si los datos disponibles eran suficientes para
proporcionar estimaciones correctas de la magnitud, el alcance y la detectabilidad del UIVS. Compilamos
los registros del comercio y la caza de subsistencia ilegales de varias fuentes y acoplamos un modelo de
bosques aleatorios para predecir la distribuci´
on espacial del UIVS. De 1969 a 2014, 404 especies y 8,340,921
espec´
ımenes estuvieron involucrados en UIVS, para una tasa de extracci´
on media de 185,354 individuos/a˜
no.
Las aves fueron el grupo involucrado m´
as rico en especies (248 spp.), pero los reptiles tuvieron las tasas de
extracci´
on m´
as altas (126, 414 individuos/a˜
no contra 3,133 individuos/a˜
no para las aves). El 88% de los
registros del comercio internacional se solap´
o espacialmente con el comercio dom´
estico, especialmente en el
norte y a lo largo de la costa, pero tambi´
en en las ´
areas occidentales tierra adentro. La extensi´
on del comercio
dom´
estico estuvo distribuida ampliamente a lo largo de las carreteras, lo que sugiere que el comercio dom´
estico
¶email asanchez@ivic.gob.ve
Paper submitted December 11, 2015; revised manuscript accepted April 16, 2016.
1
Conservation Biology,Volume00,No.0,1–11
C
2016 Society for Conservation Biology
DOI: 10.1111/cobi.12744

2Spatial Patterns in Illegal Wildlife Uses
no depende de un gran mercado en las ciudades. El 17% de los registros de comercio dom´
estico se solap´
ocon
la caza por subsistencia, pero la distribuci´
on espacial de este solapamiento cubri´
oun´
area mucho mayor
que entre los usos comerciales. El comercio dom´
estico parece responder a la demanda rural m´
as que a la
demanda de las comunidades urbanas. Nuestra estrategia ser´
a´
util para entender c´
omo funcionan los UIVS
a escala nacional en otras partes del mundo.
Palabras Clave: bosques aleatorios, caza furtiva, caza por subsistencia, mercado ilegal de vida silvestre, Venezuela
Introduction
Illegal wildlife use (IWU), any use of nondomesticated
fauna that violates national or international regulations,
is a major threat to biodiversity globally and can have far-
reaching impacts on livelihoods and human food security
(Broad et al. 2003; Gavin et al. 2010). Tropical countries
in particular may experience chronic and intense IWU
when biodiversity levels are high and law enforcement
is weak. Illegal international trade in wildlife has perhaps
received the most attention, has a value of approximately
$7–10 billion/year, and affects thousands of species
worldwide (Lawson & Vines 2014). However, such
estimates exclude illegal domestic trade and subsistence
hunting. Although poorly documented, the high value
of domestic trade (approximately $175 million/year in
the Amazon basin [Nasi et al. 2008] and $24–205 million
across West and Central Africa [Gandiwa et al. 2013])
has large negative effects on biodiversity. Domestic trade
is frequently motivated by social, cultural, and economic
factors (Nasi et al. 2008). Conversely, subsistence
hunting has been seen typically as a low-impact activity
practiced mostly by indigenous communities motivated
by food insecurity, protection against crop raiding, or
tradition (MFunda & Røskaft 2010).
Legislation and conservation actions in tropical coun-
tries are based on the hypothesis that in different ar-
eas wildlife uses occur independently of one another.
For example, some countries have suspended importa-
tion of some endangered species (Di Minin et al. 2015)
under the assumption that commercial trade is higher
within cities and coastal areas, where population and
economic infrastructures are denser than in the interior
and where access to markets is higher (Maingi et al.
2012). Other countries recognize subsistence hunting as
a legitimate form of wildlife use, and local communities
are constrained by license systems that include seasons,
bag limits, and lists of species and age classes that can
be taken (Ntiamoa-Baidu 1997). The implementation of
these controls makes sense because subsistence hunting
is assumed to occur principally clustered around rural
communities, closer to areas where resource availability
is higher. However, subsistence hunting and commer-
cial trade (domestic and international) may be spatially
correlated (i.e., similar uses occur in nearby localities).
Subsistence hunting can blur into domestic trade for rural
peoples where wildlife supplements diet and income and
generate a “bushmeat crisis” (Nasi et al. 2008). Similarly,
there is a well-documented association between domes-
tic and international trade in wildlife pets and medicinal
derivatives (Santos et al. 2011).
Understanding how and where subsistence and com-
mercial uses may co-occur has important implications for
the design and implementation of effective conservation
actions such as law enforcement and environmental ed-
ucation. Evaluating how different illegal activities may
be spatially interrelated is not a simple task. First, illegal
traders and hunters may be strongly motivated to conceal
their activities, making data collection difficult (Knapp
et al. 2010). Second, many countries lack reliable mecha-
nisms to systematically compile wildlife-use records, legal
or otherwise (Dongol 2012; Challender et al. 2015), and
most data are in the grey literature. Records of IWU are
usually opportunistically collected and have substantial
geographic and temporal biases (Golden et al. 2013).
A promising recent approach to deal with these data-
collection challenges involves fitting spatially explicit
models to evaluate the relative role of anthropogenic
and ecological factors in determining the geographical
distribution of IWUs (Maingi et al. 2012; Ziegler et al.
2016). Although the incorporation of spatial dimension
makes these studies innovative, most focus on only a
single species, even though most hunters and traders deal
with multiple species (Conteh et al. 2015), and one use
(e.g., subsistence hunting or international trade). A single-
species focus ignores important correlations across taxa
and limits understanding of the scope and magnitude of
impacts. Although some hypothesize generally that uses
may be interrelated (e.g., de Merode et al. 2004; Nasi
et al. 2008; Lindsey et al. 2013), none have attempted the
rigorous analyses required to test the prediction of joint
spatial patterns in IWU.
To overcome these limitations and test this prediction,
we analyzed patterns in the geographical distribution
of illegal commercial and subsistence use of multiple
wildlife species with data from Venezuela. In contrast
to standard assumptions of independence among IWUs,
we hypothesized that domestic trade co-occurs spatially
to a large extent (but in different areas) with international
trade and subsistence hunting. We also aimed to evaluate
whether available data were sufficient to provide accu-
rate estimates of the magnitude, scope, and detectabil-
ity of IWU across Venezuela over time. To do this, we
compiled records of illegal hunting and trade from
Conservation Biology
Volume 00, No. 0, 2016

S´
anchez-Mercado et al. 3
Figure 1. (a) Relative location of
Venezuela in South America, (b) major
geographical features in the study area
(dotted lines, political divisions; solid grey
lines, primary roads; black lines,
protected areas), and (c) elevation and
illegal wildlife use records by use.
sources that varied in quality and quantity of information.
We evaluated temporal and geographical biases and then
used this information to fit a random-forest (RF) classifi-
cation model (Cutler et al. 2007) to characterize spatial
interactions among IWUs based on their association with
anthropogenic and ecological factors. Random forest is
particularly useful for supervised classifications when the
response is a categorical variable with >2 possible values.
Although commonly applied in species-distribution mod-
els and remote-sensing classification, to our knowledge
this is the first time RF has been used to evaluate IWU
patterns (Franklin 2010).
We used data from Venezuela because the country is
tropical and megadiverse, has a deep tradition of com-
mercial and subsistence wildlife use (Ojasti 1993), and
lacks a national infrastructure for wildlife-use assessment
and law enforcement (Kaufmann et al. 2011). This com-
bination of factors is not uncommon, so an approach
that works with Venezuelan data is likely to be broadly
applicable.
Methods
Study Area and Legal Framework
Venezuela is among the most urban of South American
countries; most of its approximately 30 million inhabi-
tants (Instituto Nacional de Estad´
ıstica 2010) live in cities
along the coast (Figs. 1a & b). Large-scale habitat conver-
sion has occurred north of the Orinoco River, whereas
the southern half of the country has retained much of its
precolonial vegetation (Rodr´
ıguez et al. 2010).
Wildlife administration is under the Ministry for Ecoso-
cialism and Waters (previously Ministry for the Envi-
ronment), but law enforcement falls under the military
and police; the administrative agency plays a support-
ing role (Ojasti 1993). Venezuelan legislation permits
wildlife use via special licenses, including for scientific
sampling (including eggs and juveniles), pest control,
sport, and commercial hunting. Subsistence hunting is
not recognized as separate from sport hunting. Wildlife
take without a permit is punishable by fines and up to 2
years in prison; possession is penalized by seizure of the
animal (Congreso de la Rep´
ublica de Venezuela 1970;
Rep´
ublica de Venezuela 2012). Venezuela ratified CITES
in 1977 and restricts exports of all native wildlife except
for the products of captive breeding or those deriving
from government-sanctioned management plans, such
as capybara (Hydrochaeris hydrochaerus)andcaiman
(Caiman crocodrilus). We defined as illegal all take with-
out permits of undomesticated terrestrial birds, reptiles,
amphibians, and mammals from protected or private ar-
eas and uses exceeding established limits or seasons. This
included unlicensed harvest of eggs or juveniles and do-
mestic or international trade of live specimens.
Conservation Biology
Volume 00, No. 0, 2016

4Spatial Patterns in Illegal Wildlife Uses
IWU Classification and Data Sources
We classified records by use type into 3 nonexclusive
categories: subsistence hunting, the direct harvest of en-
tire specimens or their parts (meat, bones, eggs, etc.) for
consumption, traditional medicine, or other uses exclud-
ing barter or sale regardless of type of hunter or hunting
method; domestic trade, whole specimens or their parts
offered for sale or barter to local buyers either directly by
hunters or by intermediate dealers who obtained items
from hunters or other local intermediates; and interna-
tional trade, similar to domestic trade but with sales by
hunters or intermediates to buyers abroad.
Using these definitions, we compiled 4486 records of
subsistence hunting, domestic trade, and international
trade with data from enforcement agencies, the CITES
trade database, and published and unpublished literature.
We used data from 2 types of enforcement agencies: the
Venezuelan environmental ministry (VEM) and the U.S.
Fish and Wildlife Service (USFWS). Data from the VEM
were compiled by Asmussen-Soto (2010) and consisted
of 1042 records digitized from reports of seized wildlife at
the national level from 1990 to 2000. For 2001–2009, only
reports from 3 regional offices (Distrito Capital, Aragua,
and Cojedes states) were available. Personnel from VEM
strongly argued that in all cases their reports refer to na-
tional traders selling goods on international markets and
that the final destination of specimens depended on the
effectiveness and speed of response of international deal-
ers. If international dealers delay their responses, then
national traders try to sell goods on domestic markets
before specimens sicken or die (A. Mart´
ınez, personal
communication). Thus, we classified VEM’s records as
domestic and international trade to avoid underestimat-
ing international trade when the trade was detected at the
national level. Data from USFWS on all illegal shipments
entering the United States from Venezuela from May 1993
to September 1998 (333 records) was compiled by Ro-
driguez (2000). We searched the CITES trade database
(CITES 2009) for Venezuelan records from 1975 to 2015.
We obtained 1808 records of specimens and items seized
from Venezuela.
To search the published literature, we used 5 online
databases (ISI Web of Knowledge, Scirus, JSTOR, Google
Scholar, and Zoological Records). We focused on ecol-
ogy, conservation, and anthropology journals and used
keywords in English and Spanish related to wildlife use:
Venezuela,poach,hunt,harvest,consumption,trade,
and traffic. We also consulted 2 related web pages:
The Stage of the World’s Sea Turtles (SWOT 2003) and
TRAFFIC (2008). To improve detection in grey literature,
we requested support from Red Ara Venezuela (http://
red-ara-venezuela.blogspot.com/), a network of Venezue-
lan environmental nongovernmental organizations, and
asked wildlife managers and researchers we knew for
information and additional contacts who may have in-
formation about wildlife trade. This search yielded 58
documents containing 1303 IWU records (Fig. 1c & Sup-
porting Information).
We considered each record (i.e., a row in our database)
as a single observation in time of a unique combination
of the following: species involved, observation date, lo-
cation, resource extracted (specimen or part), quantity
reported (number of specimens or eggs or kilograms of
meat or parts), and reported use (domestic trade, interna-
tional trade, subsistence). Our database included all uses
described for a given report. If more than one use was
described, each use was included in the same row. Thus,
the structure of the database was based on species, which
allowed us to quantify the amount of IWU by taxonomic
class and source and to visualize temporal patterns.
For records without specific geographic coordinates,
we used the description of the location (place names,
geographic features, etc.) to assign latitude and longitude
based on gazetteers (i.e., directory of georeferenced lo-
calities; GIS Data Depo, DIVA GIS). When different coor-
dinates were provided by each gazetteer, we calculated
the mean value and error of latitude and longitude. If the
error was bigger than the cell resolution used to project
our predictions (1 km2), we discarded the record. We
used Encyclopedia of Life (www.eol.org) validate com-
mon and scientific names (considering synonyms, alter-
native spellings, and subspecies).
Scope and Magnitude of IWU
To visualize temporal variation in IWU sampling effort,
we plotted the proportion of records corresponding to
each use type by year. Records were collected oppor-
tunistically by state agencies or researchers using direct
observations or secondhand reports. Thus, data were
available only where and when nonrandom patrols and
research occurred and only for species of interest. Be-
cause of these spatial, temporal, and taxonomic biases,
truly accurate estimates of the magnitude and scope of
IWU are not possible (Gavin et al. 2010). However, we
were able to summarize the number of species and spec-
imens recorded, by taxonomic class, and calculate a bi-
ased extraction rate as the yearly average of this number
across the period considered. We also summarized the
number of records obtained from each source according
to type of use to evaluate the complementarity among
sources.
Co-Occurrence of Subsistence and Commercial IWUs
We evaluated relationships among IWUs by fitting a clas-
sification model. We used only the 1929 georeferenced
records (i.e., spatial dataset) from 1990 or later because
most of our predictive variables were built with data from
this date or later.
Conservation Biology
Volume 00, No. 0, 2016

S´
anchez-Mercado et al. 5
We restructured the spatial data set so that is was based
on use type not species. For that, we split records with
multiple uses and added a row for every use; informa-
tion in other columns did not change (locality, species,
etc.). This artificially increased the data set (from 1929
georeferenced records to 4219 records) with identical
geographic and temporal information. We did this to
obtain a response variable for our model—use type—
and to retain geographic information and information
on multiple-use status for error analyses evaluating use
overlap.
For our spatial models, we used RF to predict spatial
patterns of wildlife use type. Although regression models
are used to predict spatial patterns in IWU by estimating
the probability of occurrence of a single use type, their
requirement of a single dependent variable make them
inappropriate for understanding the simultaneous occur-
rence of multiple uses. Classification methods such as RF
allow a dependent variable with more than 2 categories
or classes and are better able to reveal the contribution
and behavior of multiple independent variables, whose
effects may be otherwise lost in regression models (Liaw
& Wiener 2002).
We implemented RF in the randomForest package in R
(Liaw & Wiener 2002). We built each classification tree
with a training data set containing 60% of records sam-
pled randomly with replacement from the original data
and containing a random subset of predictor variables
selected from the full set of predictor variables (Breiman
2001). We resampled records to create 500 classification
trees in our RF.
We considered 8 continuous and one categorical pre-
dictor variable (Table 1). Direct, reliable measures of
household purchasing power were not available, so we
used a combination of livestock density, land cover, and
aridity index as indicators of agricultural production,
which we in turn used as a measure of average house-
hold wealth. We assumed that areas with relatively more
agricultural and arid areas were on average more rural
and poorer (Machado-Allison 2005). We evaluated the
relative importance of predictor variables with the Gini
index of variable importance values (GIVI), defined as the
averaged Gini decrease in node impurities over all trees
in the forest (Breiman 2001).
Model Evaluation
We validated the prediction of each tree in the RF with
the remaining 40% of records (i.e., out-of-bag observa-
tions [OOB]). An estimate of the misclassification error
rate was calculated for each OOB observation and aver-
aged over all trees in the forest (Cutler et al. 2007).
Our classification errors measured 2 sources of error:
misclassification and overlapping. Overlap occurs when
records from different classes have very similar or identi-
cal characteristics (i.e., values of predictive variables) that
make it difficult or impossible to assign the overlapped
record to any class (Denil & Trappenberg 2010).
To discriminate between overlap (the process of in-
terest) and misclassification (true error), we fitted an
RF model with all records, estimated OOB classification
error, and used in-bag observations to calculate an analo-
gous in-bag error. If the model misclassified in-bag obser-
vations, given that these were used to build the trees, then
we could attribute this classification error to overlap. The
difference between OOB error and in-bag error therefore
estimated the misclassification error of the model.
We visualized the spatial pattern of predictions for each
IWU category with the predict function of the random-
Forest package (Liaw & Wiener 2002) and a raster stack
of predictive variables at a resolution of 1 km2.Wealso
visualized the spatial distribution of overlapping uses by
multiplying the respective estimates of probability for
each IWU category.
Results
Scope and Magnitude of IWU
The records we found spanned 45 years (1969–2014)
and revealed 404 species (online Supporting Informa-
tion) and 8,340,921 specimens involved in IWU. Average
extraction rate was 185,354 specimens/year (Table 2).
Birds were the most speciose group involved in IWU
(248 spp.), and most bird records involved domestic and
international trade. Mammal records involved 93 spp.,
and although a similar number of records were detected
for all uses, more species were used for domestic trade
and subsistence hunting. Reptiles (58 spp.) were used
mainly for international trade but had the highest ex-
traction rates in terms of numbers of individuals (av-
erage 126,414 individuals/species and 174,572 individ-
uals/year) (Table 2).
International trade was the most frequently reported
use (3238 records), followed by domestic trade (1577
reports). We obtained 984 records of subsistence hunt-
ing. Sources of records were complementary rather than
overlapping with respect to use type. The VEM, USWFS,
and CITES sources contributed 98% of international trade
records (3183 records) and 66% of domestic trade records
(1042 records of 1577), whereas grey and published lit-
erature provided all records for subsistence hunting.
Temporal and Spatial Biases in IWU
In general, the rate of accumulation of IWU records in-
creased over time (Fig. 2a). The proportion of records
associated with each wildlife use varied before 1990, but
subsequently the proportion of reports in each use cate-
gory was roughly constant (Fig. 2b). From 1968 to 1990
the amount of international and domestic trade increased
Conservation Biology
Volume 00, No. 0, 2016

6Spatial Patterns in Illegal Wildlife Uses
Table 1. Independent variables used to fit the random-forest model of illegal wildlife use.
Original
resolution
(decimal
Variable Source degree) Type Description
Distance to road CGIAR Institute 2010 0.0274 continuous raster derivate from a shape file
retrieve from the GIS Unit of
Ecology Centre at Instituto
Venezolano de Investigaciones
Cient´
ıficas; distance in meters
calculated using distance function
from Raster package in R; rinal
resolution 2.7 km
Distance to
protected area
CGIAR Institute 2010 0.0274 continuous raster derivate from a shape file
retrieve from the GIS Unit of
Ecology Centre at Instituto
Venezolano de Investigaciones
Cient´
ıficas; distance calculated in
meters using distance function
from Raster package in R; final
resolution 2.7 km
Forest cover DiMiceli et al. 2011 0.008 continuous raster derivate from all 7 bands of the
Moderate-resolution Imaging
Spectroradiometer (MODIS)
sensor; value range 0–100% of
pixel area covered by woody
vegetation
Land cover CGIAR Institute 2010 0.008 categorical raster based on AVHRR satellites
acquired between 1981 and 1994;
analyzed to distinguish 22
land-cover classes from tree forest,
mosaic, cultivated, and managed
areas
Aridity index Trabucco & Zomer
2009
0.008 continuous data set provides raster climate data
related to evapotranspiration
processes and rainfall deficit for
potential vegetative growth
Human population
density
Center for
International Earth
Science
Information
Network 2000
0.004 continuous raster that renders global population
data; population estimates for
1990, 1995, and projected to 2000
Livestock density Ministerio del Poder
Popular de
Agricultura y Tierra
2010
0.017 continuous raster based on the number of
livestock in xkm2; derivated using
kriging interpolation function in R;
livestock data provided by national
agricultural census from 2008
steadily, but after that it remained relatively constant.
Subsistence hunting decreased sharply prior to 1995 and
was relatively invariable after that (Fig. 2b).
Overlap between Subsistence and Commercial IWU
Classification error differed among IWU (subsistence
0.089, international trade 0.93, and domestic trade
0.203). Overlap accounted for nearly all the classification
error in the model, particularly for international trade,
for which 88% of the records overlapped with domestic
trade (in-bag error 0.88). To a lesser extent (17%), do-
mestic trade also overlapped with subsistence hunting
(in-bag error 0.168). In general, the misclassification er-
ror rate was low for all IWU categories (<5%, 0.009 for
subsistence, 0.05 for international trade, and 0.036 for
domestic trade).
The 3 most important variables to classify IWUs were
aridity index (GIVI =243.13), distance to road (183.06),
and density of human populations (154.82). Distance to
protected area, land uses, and forest cover had similar
GIVI values (127.35, 114.90, and 126.21, respectively).
Density of livestock had the lowest GIVI value (64.18).
The highest probabilities of subsistence hunting were
predicted across a wide area of the country (Fig. 3a). In
the south, subsistence hunting covered more continu-
ous areas, including the Amazon forest and nearby table
mountains, the Orinoco River delta, and the floodplains
Conservation Biology
Volume 00, No. 0, 2016

S´
anchez-Mercado et al. 7
Table 2. Number of animal species and specimens illegally extracted from Venezuela by taxon.
Extraction rateaNumber of species (number of records)b
Taxonomic Number Number per domestic international subsistence
class of species of specimens specimen in 45 years trade trade hunting
Amphibian 5 233 47 6 3 (3) 4 (10) 2 (2)
Bird 248 777,126 3,134 18,503 149 (773) 139 (1104) 95 (341)
Mammal 93 231,641 2,491 5,515 77 (408) 65 (442) 55 (448)
Reptile 58 7,335,516 126,474 174,655 48 (393 54 (1682) 27 (192)
Total 404 8,344,516 185,434
aObserved extraction rates calculated as specimens per species and specimens per year for the entire period of study (45 years).
bNumber of species and records of each taxonomic class involved in domestic trade, international trade, and subsistence hunting.
Figure 2. Temporal patterns in records of international trade, domestic trade, and subsistence hunting in
Venezuela: (a) accumulated number of records and (b) proportion of type of use reported each year.
north of the Orinoco River. In the west, subsistence hunt-
ing had a more fragmented distribution, covering high-
land areas in the major mountain ranges (the Cordillera de
M´
erida, the Coastal Cordillera, and the Sierra de Perij´
a),
but subsistence hunting also occurred in lowland areas
along the coast. In contrast, domestic trade had a more
fragmented distribution: the highest probabilities were
predicted in the north, along the coast, and in the low-
lands in the center and west (Fig. 3b). Prediction of inter-
national trade was more localized and fragmented in the
north and somewhat more focused in the west and along
the coast (Fig. 3c).
The co-occurrence of subsistence hunting and domes-
tic trade covered a wide area of the country, principally
in the central region but throughout the west and An-
dean region, the central coast, the Orinoco delta, and
the south (Fig. 3d). The co-occurrence of domestic and
international trade was much more concentrated toward
the north, along the coast but also in western inland areas
(Fig. 3e). Subsistence hunting and international trade had
virtually no co-occurrence (probabilities <0.1; Fig. 3f).
Discussion
Scope and Magnitude of Illegal Wildlife Use
To our knowledge, this is the first study to address the
overlap among IWUs from a spatial point of view that
generated a tool to evaluate hypotheses about the ge-
ographic co-occurrence of commercial and subsistence
uses. Ours is also the study to reveal spatial patterns in
subsistence and commercial IWU at a country level and
thus to allow us quantitative evaluation of the magnitude
and overlap of these activities.
Most records were associated with commercial uses,
and although enforcement agencies provided a large pro-
portion of them, this should not be taken as an indi-
cator of law-enforcement effectiveness. Particularly in
Venezuela, the quality, availability, and completeness of
IWU information from governmental agencies is far from
satisfactory, mainly because effectiveness of enforcement
patrols is low. Poorly trained and motivated staff, insuf-
ficient funds and equipment, and lack of coordination
among administrative and protection-services agencies
Conservation Biology
Volume 00, No. 0, 2016

8Spatial Patterns in Illegal Wildlife Uses
Figure 3. Spatial predictions of illegal wildlife uses (IWU) from the random-forest model: (a) subsistence hunting,
(b) domestic trade, and (c) international trade and overlap between (d) subsistence hunting and domestic trade,
(e) domestic and international trade, and (f) subsistence hunting and international trade.
(Ojasti 1993) result in a reduced capacity to obtain more
realistic, and probably higher, estimates of domestic and
international trade. However, that few records were as-
sociated with subsistence most likely reflects biased sam-
pling effort. Subsistence records came mostly from an-
thropological or conservation studies, usually conducted
at a local scale and focused on threatened species, so the
number of specimens expected is comparatively small.
The fact that subsistence hunting is not recognized in the
Venezuelan legal framework had a large influence on its
detectability, and the scarcity of records says little about
the prevalence of trade relative to commercial uses.
Although our estimate of the number of specimens
extracted illegally is necessarily an underestimate, it can
still serve to point out the importance of a threat that has
not been taken sufficiently into account in Venezuela.
The observed number of specimens extracted will al-
ways be lower than the true total; a standard estimate in
criminology is that even the most effective and well-run
enforcement programs intercept only 10% of all contra-
band (Wasser et al. 2007), such that a corrected estimate
of the number of specimens affected could be well over
83,000,000 specimens/year. Assuming this detection rate
holds in Venezuela is probably overly optimistic. A quan-
titative assessment of the environmental police in Mex-
ico indicates that the extraction rate for parrots only is
65,000–78,500 parrots/year (Cant´
uGuzm
´
an et al. 2007).
Our estimation of the number of species involved in IWU
was more precise given the accumulation of reports over
45 years. The number of species we found that were
subjected to IWU was much larger than previous lists
(153 species reported by Fergusson [1990]).
Spatial Overlap between Illegal Wildlife Uses
The widespread prevalence of domestic trade suggests
that in contrast to international trade, domestic trade does
not depend on large markets in cities but rather responds
to demand from rural communities. Given how tied in-
ternational trade appears to be to this domestic trade,
the former may be supplied by a network of domestic
trade, perhaps using local markets as collection centers,
where cross-border dealers acquire goods to transport to
neighboring countries (Arroyave Bermudez et al. 2014).
The dynamic between domestic and international trade
could differ depending on the wildlife in trade and social
and economic contexts, for example, from a highly spe-
cialized commercial chain, as for Podocnemis spp. turtles
in Amazon River in Brazil (Pantoja-Lima et al. 2014), to the
specialized bird trade in Peru (Daut et al. 2015), to simple
and opportunistic trade such as the bushmeat trade in the
Abaetetuba open-air market in Brazil (Ba´
ıa et al. 2010) and
the parrot trade in Santa Cruz’s markets in Bolivia (Pires
& Clarke 2011). Given our broad taxonomic approach,
the large extent of spatial dependence in domestic and
international trade suggests that both local organizations
and sophisticated trader networks may be involved.
Conservation Biology
Volume 00, No. 0, 2016

S´
anchez-Mercado et al. 9
In contrast to the large overlap between international
and domestic trade, the overlap between subsistence
hunting and domestic trade was smaller, yet important.
Within the areas of overlap, 2 spatial patterns probably
resulted from different anthropic pressures. The first
pattern showed a highly aggregated distribution of
subsistence hunting and domestic trade in an area where
rural and indigenous communities, logging activity (e.g.,
Imataca forest reserve and Uverito pine plantation), and
illegal mining (Ochoa 1997) converge. We suspect the
opening up of forest areas in this part of the country
by commercial logging and the development of Guri
Dam (the country’s main hydroelectric) has resulted
in an increase in wildlife harvest and trade. There are
well-documented examples in other tropical forest of
increased depletion of fauna as consequence of com-
mercial logging (Robinson 1999). We do not have direct
evidence that this is the case in southern Venezuela, but
the scarce evidence suggests that although the annual
wildlife harvest by local communities near Imataca is
comparatively low (about 12.000 kg/year [Bisbal 1994]),
a large percentage of indigenous communities (Pemon,
Warao, and Kari˜
na people) consume wildlife and trade
wildlife in local economies (Bisbal 1994).
In contrast, the more dispersed pattern of domestic
trade in the west and center may reveal a more complex
combination of cultural and economic factors, includ-
ing an increased demand for bushmeat and pets by rural
and semirural communities that is being met by a small
but widespread network of specialized restaurants and
markets (Van Vliet et al. 2014). Although no data exist
on how much of the subsistence catch is sold rather
than consumed, there are well-documented examples of
cultural factors driving the bushmeat market: the sale of
threatened marine turtle meat (Eretmochelys imbricata,
Geochelone carbonaria,G. denticulate, C. mydas)in
rural and urban markets is a longstanding tradition during
Lent in Venezuela (Rodr´
ıguez 2000). Similarly, the meat
of medium and small mammals (e.g., Tapirus terrestris,
Tayassu pecari,Agouti paca,andDasypus novemcinc-
tus) is considered particularly flavorful and is sold at ur-
ban and rural restaurants across the country at higher
prices than domestic meat (Fergusson 1990).
Implications for Conservation
International and national responses to reduce IWU have
largely focused on strengthening law enforcement efforts
and reducing international consumer demand for illegally
sourced wildlife (Gandiwa et al. 2013). However, the
overlap between domestic and international we found
suggests that much more emphasis should be placed on
the role of rural communities to determine the motiva-
tions, drivers, dynamics, and responses to IWU (Gray et al.
2015; Roe 2015; Saif et al. 2015).
If IWU has local cultural roots or economic drivers,
top-down actions like law enforcement and creation of
protected areas alone would be expected to have only
modest effects on reducing it (Knapp 2012). Particularly
in Venezuela, where 0.06% of gross domestic product is
dedicated to environmental conservation and just 350
rangers patrol nearly 150,000 km2of protected areas
(Rodr´
ıguez 2014), it seems unlikely that patrolling can
be increased in a way that will result in a meaningful
reduction of IWU over an area as large as our model pre-
dicts domestic trade to occur. We propose that bottom-
up approaches may be more efficient. Actions such as in-
creasing awareness among consumers through education
programs (Pellegrini 2001) and community-based natu-
ral resource management (Wheeler & Domingo 1997)
could more effectively reduce illegal hunting for selected
species. Understanding domestic trade may be the key
to understanding how IWU works, which underscores
the importance of implementing national strategies for
monitoring changes in supply and demand over time and
space (Challender et al. 2015; Harris et al. 2015; Taylor
et al. 2015). Our database presents an opportunity to
synthesize current and future data on IWU in Venezuela
and other South American countries and to identify infor-
mation gaps.
Acknowledgments
Funds for this research were provided by the Instituto
Venezolano de Investigaciones Cient´
ıficas. We are grate-
ful to H. Guada, P. Vernet, E. Zent, A. Mart´
ınez, A.
Fallabrino, D. Giraldon, A.S. Boher, F. Rojas-Su´
arez, R.
Calchi, E. Mujica, A. Cardozo, I. Villasmil, A.C. Molina,
P. Ortega, J. Larreal, M.D. Rinc´
on, D. Muller, K. Muller,
and T. Barros for providing technical reports, theses, and
other grey literature that was not available through online
searches. We thank J.R. Ferrer-Paris for helpful discussion
on random-forest models and 2 anonymous reviewers for
helpful comments on earlier versions of this article.
Supporting Information
A list of published and gray literature included in the
database of IWUs (Appendix S1) and a list of species
illegally used in Venezuela (Appendix S2) are available
online. The authors are solely responsible for the con-
tent and functionality of these materials. Queries (other
than absence of the material) should be directed to the
corresponding author.
Literature Cited
Arroyave Bermudez OY, Romero Goyeneche FJ, Bonilla G´
omez A,
Hurtado Heredia RG. 2014. Tr´
afico ilegal de tortugas continentales
Conservation Biology
Volume 00, No. 0, 2016

10 Spatial Patterns in Illegal Wildlife Uses
(Testudinata) en Colombia: Una aproximaci´
on desde el an´
alisis de
redes. Acta Biol´
ogica Colombiana 19:381–392.
Asmussen-Soto M. 2010. Estimaci´
on del comercio ilegal de fauna sil-
vestre a tres escalas espacio-temporales: Global, regional y nacional.
Instituto Venezolano de Investigaciones Cient´
ıficas. Centro de Estu-
dios Avanzados, Caracas.
Ba´
ıa JPC, Guimaraes DA, Le Pendu Y. 2010. Non-legalized commerce in
game meat in the Brazilian Amazon: a case study. Revista de Biolog´
ıa
Tropical 58:1079–1088.
Bisbal FJ. 1994. Consumo de la fauna silvestre en la zona de Imataca,
estado Bol´
ıvar, Venezuela. Interciencia 19:28–33.
Breiman L. 2001. Random forest. Machine Learning 45:5–32.
Broad S, Mulliken T, Roe D. 2003. The nature and extent of legal and
illegal trade in wildlife. Pages 3–22 in Oldfield S, editor. The trade
in wildlife: regulation for conservation. Earthscan, London.
Cant´
uGuzm
´
an JC, S´
anchez Salda˜
na ME, Grosselet M, Silve Gamez J.
2007. The illegal parrot trade in Mexico. A comprehensive assess-
ment. Defenders of Wildlife, Mexico, DF.
Center for International Earth Science Information Network (CIESIN).
2000. Global human population density. Available from http://
sedac.ciesin.columbia.edu/data/sets/browse?facets=theme:
population (accessed September 2014).
CGIAR Institute. 2010. DIVA-GIS environmental and physic variables.
Available from http://www.diva-gis.org/ (accessed October 2014).
Challender DWS, Harrop SR, MacMillan DC. 2015. Understanding mar-
kets to conserve trade-threatened species in CITES. Biological Con-
servation 187:249–259.
CITES (The Convention on International Trade in Endangered Species
of Wild Fauna and Flora). 2009. CITES trade database, UNEP
World Conservation Monitoring Centre, Cambridge, UK. Available
from http://www.unep-wcmc.org/citestrade/trade.cfm (accessed
September 2015).
Congreso de la Rep´
ublica de Venezuela. 1970. Ley de protecci´
on a la
fauna silvestre. Gaceta Oficial Nro 29.289. Rep´
ublica de Venezuela,
Caracas.
Conteh A, Gavin MC, Solomon J. 2015. Quantifying illegal hunting:
a novel application of the quantitative randomised response tech-
nique. Biological Conservation 189:16–23.
Cutler DR, Edwards TC, Beard KH. 2007. Random Forests for classifica-
tion in ecology. Ecology 88:2783–2792.
Daut EF, Brightsmith DJ, Mendoza AP, Puhakka L, Peterson MJ. 2015.
Illegal domestic bird trade and the role of export quotas in Peru.
Journal for Nature Conservation 27:44–53.
de Merode E, Homewood K, Cowlishaw G. 2004. The value of bushmeat
and other wild foods to rural households living in extreme poverty in
Democratic Republic of Congo. Biological Conservation 118:573–
581.
Denil M, Trappenberg T. 2010. Overlap versus imbalance. Lecture Notes
in Computer Science 6085:220–231.
DiMiceli C, Carroll ML, Sohlberg RA, Huang C, Hansen, MC. 2011.
Annual global automated MODIS vegetation continuous fields
(MOD44B) at 250 m spatial resolution for data years beginning day
65, 2000 - 2010, Collection 5 Percent Tree Cover. Available from
http://glcf.umd.edu/data/vcf/ (accessed September 2014).
Di Minin E, Laitila J, Montesino-Pouzols F, Leader-Williams N, Slotow
R, Goodman PS, Conway AJ, Moilanen A. 2015. Identification of
policies for a sustainable legal trade in rhinoceros horn based on
population projection and socioeconomic models. Conservation
Biology 29:545–555.
Dongol Y. 2012. A policy gap analysis of Convention on International
Trade in Endangered Species of Wild Fauna and Flora (CITES) imple-
mentation in Nepal. Environmental Management DOI 10.100:62.
Fergusson A. 1990. Aprovechamiento de la fauna silvestre en Venezuela.
Cuadernos Lagoven, Caracas.
Franklin J. 2010. Mapping species distributions. Spatial inference and
prediction. Cambridge University Press, Cambridge.
Gandiwa E, Heitkonig IMA, Lokhorst AM, Prins HHT, Leeuwis C. 2013.
Illegal hunting and law enforcement during a period of economic
decline in Zimbabwe: a case study of northern Gonarezhou National
Park and adjacent areas. Journal for Nature Conservation 21:133–
142.
Gavin MC, Solomon JN, Blank SG. 2010. Measuring and monitoring
illegal use of natural resources. Conservation Biology 24:89–100.
Golden CD, Wrangham RW, Brashares JS. 2013. Assessing the accuracy
of interviewed recall for rare, highly seasonal events: the case of
wildlife consumption in Madagascar. Animal Conservation 16:597–
603.
Gray CL, Bozigar M, Bilsborrow RE. 2015. Declining use of wild re-
sources by indigenous peoples of the Ecuadorian Amazon. Biological
Conservation 182:270–277.
Harris JBC, Green JMH, Prawiradilaga DM, Giam X, Giyanto, Hikmatullah
GD, Putra CA, Wilcove DS. 2015. Using market data and expert
opinion to identify overexploited species in the wild bird trade.
Biological Conservation 187:51–60.
Instituto Nacional de Estad´
ıstica. 2010. Censo poblacional. Ministerio
del Poder Popular de Planificaci´
on y Finanzas, Caracas, Venezuela.
Available from http://www.ine.gov.ve/ (accessed May 2014).
Kaufmann D, Kraay A, Mastruzzi M. 2011. The worldwide governance
indicators: a methodology and analytical issues. Hague Journal on
theRuleLaw3:220–246.
Knapp EJ. 2012. Why poaching pays: a summary of risks and benefits
illegal hunters face in Western Serengeti, Tanzania. Tropical Con-
servation Science 5:434–445.
Knapp EJ, Rentsch D, Schmitt J, Lewis C, Polasky S. 2010. A tale of three
villages: choosing an effective method for assessing poaching levels
in western Serengeti, Tanzania. Oryx 44:178–184.
Lawson K, Vines A. 2014. Global impacts of the illegal wildlife trade.
The costs of crime, insecurity and institutional erosion. Chatham
House (The Royal Institute of International Affairs), London.
Liaw A, Wiener M. 2002. Classification and regression by randomforest.
RNews2:18–22.
Lindsey PA, et al. 2013. The bushmeat trade in African savannas:
Impacts, drivers, and possible solutions. Biological Conservation
160:80–96.
Machado-Allison, C. 2005. Perfil de la agricultura venezolana. Acta
Biol´
ogica Venezuelica 25:7–18.
Maingi JK, Mukeka JM, Kyale DM, Muasya RM. 2012. Spatiotemporal
patterns of elephant poaching in south-eastern Kenya. Wildlife Re-
search 39:234–249.
MFunda I, Røskaft E. 2010. Bushmeat hunting in Serengeti, Tanzania: an
important economic activity to local people. nternational Journal of
Biodiversity and Conservation 2:263–272.
Ministerio del Poder Popular para la Agricultura y Tierras. 2010. Censo
agr´
ıcola. Available from http://www.mat.gob.ve/CensoAgricola/
(accessed April 2014).
NasiR,BrownD,WilkieD,BennettE,TutinC,vanTolG,Christo-
phersen T. 2008. Conservation and use of wildlife-based resources:
the bushmeat crisis. Technical Series no. 33. Secretariat of the Con-
vention on Biological Diversity, Montreal, and Center for Interna-
tional Forestry Research (CIFOR), Bogor.
Ntiamoa-Baidu Y. 1997. Wildlife and food security in Africa. Food and
Agriculture Organization of the United Nations, Rome.
Ochoa J. 1997. Sensibilidades potenciales de una comunidad de
mam´
ıferos en un bosque productor de maderas de la Guayana.
Interciencia 22:112–122.
Ojasti J. 1993. Utilizaci´
on de la fauna silvestre en Am´
erica Latina.
Situaci´
on y perspectiva para un manejos sostenible. Gu´
ıa FAO Con-
servaci´
on 25. Organizaci´
on de las Naciones Unidas para la Agricul-
tura y la Alimentaci´
on, Rome.
Pantoja-Lima J, Aride PHR, De Oliveira AT, F´
elix-Silva D, Pezzuti JCB,
Rebelo GH. 2014. Chain of commercialization of Podocnemis spp.
turtles (Testudines: Podocnemididae) in the Purus River, Amazon
Conservation Biology
Volume 00, No. 0, 2016

S´
anchez-Mercado et al. 11
basin, Brazil: current status and perspectives. Journal of Ethnobiol-
ogy and Ethnomedicne 10:2–10.
Pellegrini N. 2001. Educaci´
on ambiental en el sistema de parques
nacionales de Venezuela. T´
opicos en Educaci´
on Ambiental 3:
55–69.
Pires SF, Clarke RV. 2011. Sequential foraging, itinerant fences and
parrot poaching in Bolivia. British Journal of Criminology 51:314–
335.
Rep´
ublica de Venezuela. 2012. Ley penal del ambiente. Gaceta Oficial
Nro 39.913. Caracas.
Robinson JG. 1999. Wildlife harvest in logged tropical forests. Science
284:595–596.
Rodr´
ıguez JP. 2000. Impact of the Venezuelan economic crisis on wild
populations of animals and plants. Biological Conservation 96:151–
159.
Rodr´
ıguez JP. 2014. Public investment in biodiversity conservation.
Interciencia 39:538-538.
Rodr´
ıguez JP, Rojas-Suárez F, Giraldo Her´
andez D. 2010. Libro rojo de
los escosistemas terrestres de Venezuela. Provita, Compa˜
n´
ıa Shell
de Venezuela, Lenovo, Caracas.
Roe D. 2015. Conservation, crime and communities: Case studies
of efforts to engage local communities in tackling illegal wildlife
trade. International Institute for Environment and Development,
London.
Saif S, Russell AM, Nodie SI, Inskip C, Lahann P, Barlow A, Barlow CG,
Islam A, MacMillan DC. 2015. Local usage of tiger parts and its role
in tiger killing in the Bangladesh sundarbans. Human Dimensions of
Wildlife 1209:1–16.
Santos A, Satchabut T, Vigo Trauco G. 2011. Do wildlife trade bans
enhance or undermine conservation efforts? Applied Biodiversity
Perspective Series 1:1–15.
SWOT. 2003. The state of the world’s sea turtles. Available from
http://www.seaturtlestatus.org (accessed February 2014).
Taylor G, et al. 2015. Synthesizing bushmeat research effort in West
and Central Africa: A new regional database. Biological Conservation
181:199–205.
Trabucco A, Zomer R. 2009. Global aridity index (Global-Aridity)
and global potential evapo-transpiration (Global-PET). Geospa-
tial Database. Available from http://www.csi.cgiar.org/ (accessed
September 2014).
TRAFFIC. 2008. TRAFFIC the wildlife trade monitoring network. Avail-
able from http://www.traffic.org/ (accessed February 2014).
Van Vliet N, Quiceno-Mesa MP, Cruz-Antia D, Jonhson L, De Aquino
N, Moreno J, Nasi R. 2014. The uncovered volumes of bushmeat
commercialized in the Amazonian trifrontier between Colombia,
Peru & Brazil. Ethnobiology and Conservation 7:1–11.
Wasser SK, Mailand C, Booth R, Mutayoba B, Kisamo E, Clark B, Stephens
M. 2007. Using DNA to track the origin of the largest ivory seizure
since the 1989 trade ban. Proceedings of the National Academy of
Sciences of the United States of America 104:4228–4233.
Wheeler JC, Domingo HR. 1997. Community participation, sustainable
use, and vicu˜
na conservation in Peru. Mountain Research and De-
velopment 17:283–287.
Ziegler S, Fa J, Wohlfart C, Streit R, Jacob S, Wegmann M. 2016. Mapping
bushmeat hunting pressure in Central Africa. Biotropica 48:405–
412.
Conservation Biology
Volume 00, No. 0, 2016