Content uploaded by Ardiantiono Ardiantiono
Author content
All content in this area was uploaded by Ardiantiono Ardiantiono on Aug 18, 2023
Content may be subject to copyright.
Content uploaded by Ardiantiono Ardiantiono
Author content
All content in this area was uploaded by Ardiantiono Ardiantiono on Mar 05, 2023
Content may be subject to copyright.
Target Journal: Biological Conservation
1
Integrating social and ecological information to identify high-risk areas of
1
human-crocodile conflict in the Indonesian Archipelago
2
Ardiantiono1, Sujan M. Henkanaththegedara2, Brandon Sideleau3, Sheherazade4,5, Yogie
3
Anwar6, Iding A. Haidir7,8, A.A. Thasun Amarasinghe9,10
4
5
1Durrell Institute of Conservation and Ecology, University of Kent, Canterbury, CT2 7NZ, UK.
6
https://orcid.org/0000-0001-8398-1948
7
2Department of Biological & Environmental Sciences, Longwood University, Farmville,
8
Virginia 23909, USA https://orcid.org/0000-0003-1936-3947
9
3Croc Attack: The Worldwide Crocodilian Attack Database, 2536 Avenida de Las Plantas
10
Thousand Oaks, CA 91360, USA https://orcid.org/0000-0001-9145-6414
11
4PROGRES (Sulawesi Regional Ecological Conservation Initiative), Luwuk Banggai 94711,
12
Indonesia https://orcid.org/0000-0002-0070-250X
13
5Department of Environmental Sciences, Policy, and Management, University of California
14
Berkeley, California, USA.
15
6Wildlife Conservation Society-Indonesia Program, Bogor 16128, Indonesia
16
7Directorate General of Natural Resources and Ecosystem Conservation, Ministry of
17
Environment and Forestry of the Republic of Indonesia, Jakarta 20270, Indonesia.
18
https://orcid.org/0000-0002-1568-5650
19
8Wildlife Conservation Research Unit, Zoology Department, University of Oxford, The
20
Recanati-Kaplan Centre, Abingdon, UK
21
9Department of Biology, Faculty of Mathematics and Natural Sciences, Universitas
22
Indonesia, Depok 16424, Indonesia. https://orcid.org/0000-0002-4151-1806
23
10Herpetology Lab, Research Center for Biosystematics and Evolution, The National Research
24
and Innovation Agency (BRIN), Government of Indonesia
25
26
Corresponding authors:
27
Ardiantiono ard24@kent.ac.uk; A.A. Thasun Amarasinghe thasun.amarasinghe@ui.ac.id
28
29
30
31
32
Target Journal: Biological Conservation
2
ABSTRACT
33
Crocodile attacks on humans and subsequent retaliations are a pressing issue for saltwater
34
crocodile conservation. As human-crocodile conflict is complex, integrating social and
35
ecological information better explains the drivers and patterns of these interactions. Our
36
study aims to incorporate ecological factors associated with the intensity of crocodile
37
attacks together with social factors of mass media reports to identify high-risk areas of
38
human-crocodile conflict in Indonesia. We compiled reports of crocodile attacks in the 2010-
39
2019 period from media reports, field surveys, and local informants. The presence of attack
40
was estimated by evaluating the influence of habitat, climate, human, and reporting effort.
41
As tone of media coverage can reflect and shape reader’s tolerance about a certain issue,
42
we assessed the headline’s tone from each media article that reported crocodile attacks
43
from 2017 to 2019. A total of 665 crocodile attacks were recorded and mainly distributed in
44
western and central Indonesia. The estimated number of crocodile attacks was higher in
45
areas with lower forest biomass and human density, and wider cellular network coverage.
46
Negative media coverages were frequently reported in western Indonesia. By combining
47
social information of negative media reporting and the ecological information of crocodile
48
attacks hotspots, we identified 170,500 km2 priority risk areas in the western part of
49
Indonesia, a notable 65.8% reduction in area size compared to the attack hotspots. We
50
highlight the application of socio-ecological information in risk prioritization to address the
51
rising trends of negative human-wildlife interactions.
52
Keywords: Content analysis, Crocodylus porosus, human-wildlife conflict, risk prioritization,
53
socio-ecological framework.
54
55
56
57
58
59
60
Target Journal: Biological Conservation
3
INTRODUCTION
61
Human-wildlife conflict has become a global conservation concern when wildlife presents
62
actual or perceived threats to humans that negatively impacts people and/or wildlife (IUCN,
63
2020). As conflicts are complex interactions between human and wildlife, it is increasingly
64
evident that integrating social and ecological information is critical to better explain the
65
causes and dynamics of these interactions, and prioritize areas for intervention (Gálvez et
66
al., 2018; Struebig et al., 2018). Yet, integration between these two disciplines is limited
67
(Milner-Gulland, 2012).
68
Crocodilians are one of the major taxonomic groups that cause substantial threats to human
69
livelihood (Fukuda et al., 2014; Webb et al., 2010). Although a major cause of human
70
injuries or deaths (CrocBITE, 2020), human-crocodile conflict on a global scale has received
71
relatively little attention (Torres et al., 2018). Out of 24 Crocodilian species, the saltwater
72
crocodile (Crocodylus porosus) is responsible for the most reported attacks on humans,
73
along with the Nile crocodile (Crocodylus niloticus) (CrocBITE, 2020; Sideleau and Britton,
74
2013). The saltwater crocodile is the largest living crocodilian, potentially reaching up to 6–7
75
m in length (Webb et al. 2010). Despite its name, saltwater crocodiles are distributed in a
76
wide variety of saline and freshwater habitats that often overlap with areas of human
77
activity (Brien et al., 2017; Webb et al., 2010).
78
Attacks on humans are a pressing issue for saltwater crocodile conservation (Amarasinghe
79
et al., 2015; CrocBITE, 2020). The incidents foster increased fear and reduced tolerance
80
towards crocodiles in communities living alongside them, often leading to retaliation and
81
the removal of crocodiles which has led to the decline of crocodile populations in some
82
areas (Amarasinghe et al., 2015; Das and Jana, 2018). This is a growing concern as there has
83
been an increasing number of attacks in Southeast Asian countries with high human
84
densities, specifically Indonesia and Malaysia (CrocBITE, 2020). Pressure on saltwater
85
crocodiles (i.e. human population growth and habitat loss) is extensive in this region, as
86
crocodiles are extinct throughout almost all of mainland Southeast Asia (Cambodia,
87
Thailand, and Vietnam) (CrocBITE, 2020; Webb et al., 2010).
88
Understanding the patterns and drivers of saltwater crocodile attacks is important to
89
develop effective mitigation strategies. However, incidents of human-crocodile conflict in
90
Target Journal: Biological Conservation
4
the Indo-Malayan region are poorly studied with no peer-reviewed publications between
91
1990 and 2020, compared to Australia (11 publications) or South Asia (7) (Pooley, 2020).
92
Crocodile attacks have been reported to be associated with habitat loss, human activities,
93
and climatic factors, although the influence of these drivers may be location-specific (Webb
94
et al 2010, Amarasinghe et al. 2015). Habitat conversion (e.g. loss of riparian forest and
95
mangrove covers) can reduce the prey availability and crocodile nest quality, thus pushing
96
them to find alternative prey and explore new areas (Saragih et al. 2020; Amarasinghe et al.
97
2015). The loss of habitat often occurs in parallel with the establishment of human
98
settlements, which intensify human activities in and around water bodies inhabited by
99
crocodiles. Climatic factors such as daily temperature and precipitation may also influence
100
attacks i.e. through the combination of the increased time spend by humans in the water
101
during warmer weather and the effect of warmer temperature on crocodile physiology e.g.
102
faster digestive rates (Powell et al., 2020).
103
Incorporating ecological information with the social dimension of human-wildlife conflict
104
can be a powerful tool to identify priority areas of intervention (Gálvez et al., 2018; Struebig
105
et al., 2018). Tolerance toward wildlife is a social concept that has been widely applied in
106
studies of human-wildlife relations (Kansky et al., 2016). Tolerance is a passive acceptance
107
of wildlife populations, while intolerance occurs when wildlife becomes unacceptable,
108
leading to actions that harm or eliminate the target populations (Bruskotter and Wilson,
109
2014; Kansky et al., 2016). Tolerance concepts can be based on attitudinal aspects (e.g.
110
negative attitudes) or behavioural aspects (e.g. retaliatory killing) (Bruskotter and Wilson,
111
2014). However, assessing tolerance across a large landscape is challenging. The
112
examination of mass media reporting can address the issue as it plays key roles in both
113
reflecting and shaping the views and attitudes of the readers (Boissonneault et al., 2005;
114
Hughes et al., 2020; Sabatier and Huveneers, 2018).
115
This study aims to incorporate ecological data associated with saltwater crocodile attacks
116
together with the social aspect of media reporting to identify high-risk areas of human-
117
crocodile conflict in Indonesia and propose subsequent mitigation methods. Firstly, we
118
characterize the spatiotemporal patterns of crocodile attacks which consider the nature and
119
distribution of attacks, as well as the temporal trends based on the extensive 10 years of
120
Target Journal: Biological Conservation
5
crocodile attack data. Secondly, we evaluate the influences of habitat, climate, human
121
density, and reporting effort as drivers of crocodile attacks to predict the attack hotspots.
122
Third, we map negative media reporting towards crocodiles based on content analysis of
123
media reporting of crocodile attacks. Finally, we combined crocodile attack hotspots with
124
the negative reporting map to identify priority areas for future mitigation strategies. We
125
anticipate that this study will provide a much-needed spatially explicit, landscape-scale,
126
socio-ecological framework of human-crocodile interactions for future management.
127
MATERIALS & METHODS
128
Study area
129
This study focused on the current distribution range of saltwater crocodiles across
130
Indonesia, spanning an area of approximately 2.5 million km2 (See Indonesia environmental
131
profile in Supplementary Note 1). Saltwater crocodile distribution was determined using a
132
combination of recent and historical attack records, itinerant/capture records,
133
communication with local authorities, and the available literature (Supplementary Note 2).
134
Currently, saltwater crocodiles are widely distributed throughout the islands of Sumatra,
135
Kalimantan, Sulawesi, the eastern Lesser Sunda Islands, the Moluccas, and Papua with
136
exemption in most of Java and western Lesser Sunda Islands (e.g. Bali and Nusa Tenggara
137
Barat provinces) (Supplementary Figure 1). The country’s administration is divided into 34
138
provinces where crocodile attacks were reported in 29 provinces.
139
Target Journal: Biological Conservation
6
Data collection
140
Saltwater crocodile attacks
141
We compiled reports of saltwater crocodile attacks in Indonesia over the 10 years between
142
January 2010 and December 2019 by compiling information from online mass media,
143
crocodile specialists, regional wildlife authorities, and in-the-field data collection. Online
144
mass media reports were used to collect the majority of attack information using keywords
145
such as “buaya, diterkam (crocodile, attack)”, “buaya, warga (crocodile, human)”, “buaya,
146
digigit (crocodile, bite)”, and “buaya, dimangsa (crocodile, prey)”. Sometimes these phrases
147
were combined with location name abbreviations to narrow down the searches using
148
keywords such as “buaya, diterkam (crocodile, attack)”, “Kalimantan Tengah (name of
149
province)”. Regional wildlife authorities, primarily consisting of Balai Konservasi Sumber
150
Daya Alam (BKSDA/Natural Resources Conservation Center), were contacted for information
151
in various provinces, particularly in areas experiencing increased levels of conflict in the past
152
five years (e.g. Nusa Tenggara Timur, Maluku, and Kalimantan Tengah). To complement and
153
verify the abovementioned data, field data collection was conducted in Nusa Tenggara
154
Timur province in the central part of Indonesia in 2015 and 2017 through village surveys
155
(detail of methodology in Sideleau et al. (2021)).
156
We recorded encounters that resulted in non-fatal and fatal attacks (causing the death of
157
people). We excluded reports of any encounters that did not result in human injury or death,
158
attacks by captive crocodiles, and unconfirmed attacks such as victims went missing without
159
witnesses or evidence. Then we verified each attack to ensure the species involved and to
160
avoid multiple reporting of the same incident. In many cases, incidents were verified by
161
contacting local authorities, although in some cases official confirmation was not possible and
162
our best judgment was used. We determined the crocodile species responsible using a
163
combination of traits including known distribution, behavior and when possible, confirmation
164
via witness or captured/killed crocodiles (Supplementary Note 2).
165
Other details collected for each incident included the time (month, year), location
166
(coordinates and provinces), outcome of attack (fatal or non-fatal), gender of the victim, and
167
victim’s activity during the attack. All reports were uploaded to CrocBITE Worldwide
168
Crocodilian Attack Database (http://www.crocodile-attack.info/) and made available to the
169
Target Journal: Biological Conservation
7
public. The CrocBITE database was established in 2013 and has since provided open access
170
data of Crocodilian attacks that has been used for conservation management and studies of
171
human-crocodile interactions across the globe (i.e. see González-Desales et al., 2021; Pooley
172
et al., 2021; Sideleau et al., 2021).
173
Media reporting
174
To assess current media reporting, we compiled 225 online mass media reports of crocodile
175
attacks covering 2017-2019 period using Google News search tool with specific keywords in
176
Bahasa: "buaya (crocodile)", "serangan (attack)", "manusia (human)", and "konflik
177
(conflict)". We focused on Bahasa Indonesia reporting only, the national language of the
178
country, and limited the search period to a single year starting from 2017. We collected
179
every article available about crocodile attacks, including multiple reporting of the same
180
incidents. The data collection ended when there were no relevant articles after three pages
181
of search results.
182
We assessed the tone of article headline as a proxy of media attitudes towards crocodile
183
attacks, assuming it can reflect and shape readers attitudes (Dayer et al., 2017). We used
184
the headline because it is the part of the article that people read first and shown in the
185
article link. We categorized headline tone as negative, neutral, or positive based on criteria
186
provided by Dayer et al. (2017). A negative tone was assigned when the headline blame
187
crocodiles or use negative terms such as “conflict”, “scary”, “dangerous”, or “attack” e.g.
188
“WARNING! Crocodile terror is not finished” (read Wahid and Azhari (2016) for list of
189
terms). A neutral tone was assigned when the headline does not indicate any evaluation of
190
whether the events were either negative or positive terms e.g. “Crocodiles were seen in the
191
river after the incidents”. A positive tone was assigned when the headline promotes the
192
conservation of crocodiles e.g. “The importance of living together with crocodiles”.
193
194
Spatial covariate preparation
195
We generated 50 x 50 km sample grid (N = 993) across the crocodile distribution range
196
(Supplementary Figure 1). The grid cell size was determined based on the average maximum
197
distance from the centroid of 12 satellite-tracked saltwater crocodiles (Campbell et al.,
198
2013).
199
Target Journal: Biological Conservation
8
To evaluate predictors of crocodile attack, we used eight spatial covariates that represent
200
habitat: 1) water body (rivers, lake, and coastline) density (total length in km; BIG 2021), 2)
201
2010 aboveground forest biomass, and 3) difference between 2010 and 2018 aboveground
202
forest biomass (ton/hectare; Santoro & Cartus 2021); climate: 4) 2010 precipitations and 5)
203
difference between 2010 and 2019 precipitation (mm3/km2; Funk et al., 2014); human
204
activity: 6) 2010 human population density and 7) difference between 2010 and 2019 human
205
population density (people/km2; Bright et al. 2011 and Rose et al. 2020); and crocodile attack
206
reporting effort: 8) cellular network coverage (km2, OpenCelliD 2020). The preparation steps
207
of the spatial covariates and collinearity test result can be seen in Supplementary Note 3 and
208
Figure 2.
209
Statistical analysis
210
We performed statistical analysis using R statistical software version 4.0.2. (R Core Team,
211
2020). First, we summarized the number of attacks, attack outcome (fatal/non-fatal), and
212
victim demographic profiles. We ran general linear models (lm function) to identify the
213
temporal trends of crocodile attacks and any effects of human density at a nationwide scale.
214
The average numbers of monthly attacks were analyzed using one-way ANOVA to evaluate
215
the seasonality of attacks. The total number of attacks by victim's gender and activity were
216
analyzed using Pearson's Chi-squared test (chisq.test function) with null hypothesis of there
217
is no difference in crocodile attack frequency across victim’s gender and activity.
218
We then performed a negative binomial hurdle model (pscl package in R; Zeileis et al. 2008)
219
to evaluate the significant predictors of crocodile attack occurrence (0-1) and intensity
220
within sample grid cells. This class of model accommodates both binary and count data
221
within a single framework that accounts for zero inflation and overdispersion (Zeileis et al.,
222
2008). The zero hurdle part of the model analyzed the grid cells with no recorded crocodile
223
attack (zero count data) to assess the influence of covariates on the probability of attack to
224
occur in each cell. The count part of the model assessed grid cells with attack occurrence
225
(above zero count data) to identify the influence of covariates on the number of attacks
226
estimated to occur in each cell.
227
We built a global model that incorporates all eight covariates on count data (attack
228
intensity) and five covariates (excluding differences of 2010 and 2019 biomass, human
229
Target Journal: Biological Conservation
9
density, and precipitation covariates) on zero data (attack occurrence). We first removed
230
each of the non-significant covariates in the zero hurdle model until the model performance
231
did not improve (based on AIC value) and then applied the same approach to the count
232
model. We then model-averaged top models with delta AIC <2. The prediction of the
233
probability of crocodile attack (zero count data) and number of attack for the ten-year
234
period (count data) in every sample grid cells was produced by back-transforming the
235
regression coefficient output of the best model.
236
We created index of negative media reporting (0-1) using the assessment result of media
237
headline tone. The index was produced by dividing the number of articles with negative
238
headline tones in each province by the maximum number of negative articles across
239
provinces. For example, if Kalimantan Tengah province had 24 negative articles and Riau
240
province had the highest number of negative articles with 37, then the index of intolerance
241
for Kalimantan Tengah is 24/37 = 0.65. The index value of every province was transferred to
242
each of the sample grids within the provincial boundary. We included only negative tones
243
on crocodile attacks as they have stronger influence on reader attitude and tolerance
244
(Jacobson et al., 2012; Kansky et al., 2016).
245
Human-crocodile interaction risk areas
246
We identified priority areas for human-crocodile conflict mitigation using a framework that
247
combines the measure of crocodile attack hotspot (the number of attacks from Hurdle
248
model) and the measure of media reporting (index of negative media reporting). The sample
249
grids were partitioned into three priority scales: high (>= 1 attack and above upper quartile
250
(top 25%) negative index), medium (>= 1 attack and below upper quartile negative index; or
251
otherwise), and low (< 1 attack and below upper quartile negative index).
252
RESULTS
253
Patterns of crocodile attacks
254
Between 2010 and 2019 (10 years), we recorded 665 attacks on humans by saltwater
255
crocodiles across the Indonesian archipelago, of which 47% were fatal and 53% were non-
256
fatal. Crocodile attacks were mainly reported in the western and central parts of Indonesia.
257
Three provinces with the highest attacks were Nusa Tenggara Timur (104 attacks),
258
Target Journal: Biological Conservation
10
Kalimantan Timur (83), and Bangka-Belitung (67) (Supplementary Figure 1; see
259
Supplementary Table 1 for the full list of provinces).
260
We detected a significant positive trend (F1,8= 323.5, p < 0.001) in the total number of
261
attacks since 2010 with an average increase of seven attacks per year (Figure 1A).
262
Additionally, we also detected a significant positive correlation (F1,8= 71.83, p < 0.001)
263
between the number of attacks and human density showing a higher number of attacks in
264
areas with increasing human densities. The total number of monthly attacks varied from as
265
low as four incidents (May) to 116 incidents (March) (Figure 1B). Similarly, the average
266
number of attacks per month varied significantly among months (F11,108= 6.774, p < 0.0001)
267
with the most incidents occurred in March and the least incidents happened in May.
268
Number of attacks in wet season (November-April; 408 attacks) were higher than in dry
269
season (May-October; 239). A total of 18 attacks were reported without information on the
270
month of incidents.
271
Higher number of incidents involved crocodile attacks on male victims (558 incidents, 84%),
272
compared to female (87, 13%) (χ2 = 775.6, df = 2, p < 0.0001). Twenty attacks (3%) did not
273
report the victims’ gender. Victim ages ranged from 4 to 90 years with an average of 37
274
years. All the attacks occurred when victims were on the edge or in the water. We detected
275
a significant difference between the total number of crocodile attacks for six categories of
276
activities of victims during attack (χ2 = 452.71, df = 5, p < 0.0001). Three main human
277
activities associated with crocodile attacks were fishing (292, 44%), followed by self-cleaning
278
(145, 22%), and working (110, 17%) (Supplementary Table 2).
279
Spatial drivers of crocodile attacks
280
The presence of crocodile attacks was strongly explained by a combination of factors
281
including habitat, climate, and reporting effort (Table 1; see Supplementary Table 3 for
282
model performance evaluation). The probability of attack occurrence (0-1) was higher in
283
areas with substantial water body (represented by water length) with less forest
284
(represented by aboveground biomass) along the edge. Crocodile attacks were also more
285
likely to occur in drier areas with lower rainfall. Areas that were well covered by cellular
286
networks were more likely to report crocodile attacks.
287
Target Journal: Biological Conservation
11
The intensity of crocodile attacks was significantly influenced by habitat characteristics,
288
human activities, and reporting efforts. The number of crocodile attacks was estimated to
289
be higher in areas with lower forest biomass and human density near water bodies with
290
wider cellular coverage. Crocodile attack hotspots were distributed in the western and
291
central parts of Indonesia: the eastern part of Sumatra and Kalimantan, Sulawesi, and Nusa
292
Tenggara Timur islands (Figure 2 Top).
293
294
Negative media reporting toward crocodile attacks
295
We compiled 225 media reports of crocodile attacks across 21 provinces published between
296
2017 and 2019. A total of 164 article headlines had negative tones (73%) and 61 used
297
neutral tones (27%). No positive headline tones were reported for this study. Intensive
298
negative media coverages were identified in the western part of Indonesia, on the island of
299
Sumatra and Kalimantan. Riau province had the highest index of negative media reporting
300
(1; 37 negative headlines) followed by Kalimantan Timur (0.65; 24 headlines) and Bangka
301
Belitung (0.57; 21 headlines) (Figure 2 Centre; see Supplementary Table 4 for the full list of
302
provinces).
303
Human-crocodile interaction risk areas
304
By combining crocodile attack hotspots with negative media reporting estimates, we
305
identified 68 grids (6.8%; 170,500 km2) in Sumatra and Kalimantan islands in west Indonesia
306
as high priority areas for future intervention (Figure 2 Bottom). Six provinces were within
307
these high priority areas were Aceh, Riau, Jambi, Bangka-Belitung, Sumatera Selatan, and
308
Kalimantan Timur. Medium priority grids (309 grids; 31.1%; 772,500 km2) were distributed
309
mostly in the Sumatra, central and eastern Kalimantan, and in the central part of Indonesia
310
including Sulawesi and Lesser Sunda Islands. About 62% of our study area (1,540,000 km2),
311
mostly in the eastern part of Indonesia, was identified as a low priority due to the lack of
312
attack and negative media coverage.
313
Target Journal: Biological Conservation
12
DISCUSSION
314
Human-crocodile conflict is a pervasive threat to human livelihoods and saltwater crocodile
315
conservation. This study demonstrates the importance of applying a socio-ecological model
316
framework to identify human and saltwater crocodile high-risk interaction areas, where
317
conservation efforts can be prioritized. We utilized the publicly available media reporting to
318
document crocodile attacks and negative media reporting towards crocodiles,
319
complemented by on-site information of the attacks. Through standard ecological
320
information on crocodile attacks, we found attack hotspots were distributed in four regions
321
in western and central Indonesia encompassing a vast area of 497,500 km2. By incorporating
322
the social information on media tolerance in our model, we identified priority risk areas in
323
the eastern part of Sumatra and Kalimantan with an area of 170,500 km2, a notable 65.7%
324
reduction in area size compared to the attack hotspots.
325
Drivers of crocodile attack
326
As an archipelagic country, there is a high dependence of local Indonesian communities on
327
water bodies for economic activities such as fishing, and daily activities such as self-cleaning
328
and cooking. We identified crocodile attacks were more likely to occur in drier regions of
329
Indonesia, notably in Nusa Tenggara Timur province in the central part of Indonesia. This
330
may be due to the higher dependence of the local communities on limited water sources
331
within the habitat of the crocodiles. Nevertheless, we also noted the extensive reporting
332
efforts may influence this pattern as we recorded most attacks distributed across wetter
333
areas in western Indonesia. Moreover, we found a seasonality of attacks where more
334
attacks were reported in the wet season during the breeding period, as adult crocodiles
335
demonstrate long distant movement to major reproduction sites and female crocodiles start
336
nesting and become more territorial (Fukuda et al., 2019, 2014).
337
Saltwater crocodile attacks were reported across Indonesia. Using the countrywide-scale
338
analysis, we showed that the increase in attacks was associated with human population
339
growth. However, spatial analysis on a finer scale (50 x 50 km grid) found that crocodile
340
attacks were spatially more numerous in areas with lower human density. The latter
341
characteristic was supported by studies that have reported a lower abundance of saltwater
342
Target Journal: Biological Conservation
13
crocodiles in human habitations due to fewer viable habitats, hunting, and pressure to
343
eliminate or translocate crocodiles seen near settlements (Fukuda et al., 2014; Pooley et al.,
344
2021; Ramdani et al., 2021). We considered that crocodile attacks may not be influenced
345
directly by the human density but because of the increased human activities. This
346
assumption is in congruence with our findings as attacks were higher in localities with lower
347
forest biomass indicating habitat loss or degradation which may reduce prey and nest
348
availability for crocodiles (Saragih et al. 2020; Amarasinghe et al. 2015). This contrasting
349
result highlights the importance of analysis of human-wildlife interactions at multiple spatial
350
scales: local-scale with 50 x 50 km grid cells and national scale.
351
In this study, we included cellular network coverage to represent a reporting effort variable
352
that is rarely considered in studies of human-crocodile interaction (Brien et al., 2017;
353
González-Desales et al., 2021; Powell et al., 2020). It is relevant because most crocodile
354
attack records were from secondary data such as media reporting and local correspondents
355
which relies on cell phone communications for information exchange. As expected,
356
crocodile attacks were found to be more frequent in areas with substantial network
357
coverage. This finding may explain the interesting case of the eastern Indonesia region,
358
especially in Papua island where few attacks were reported while it is believed to be a
359
stronghold of saltwater crocodile populations (Webb et al., 2010). However, it must be
360
noted that the lower human density combined with low forest conversion rates (Allan et al.,
361
2019; Gaveau et al., 2022) may have contributed to these fewer attacks in Papua.
362
Media reporting to indicate tolerance
363
Crocodile attacks can result in retaliation against crocodiles especially when people have
364
lower tolerance toward the animal. We extended the application of utilizing media reporting
365
to estimate negative media coverage on a countrywide scale. We focused on attitudinal
366
tolerance, assuming a negative headline tone can represent and shape a negative public
367
attitude and thus a lower tolerance towards crocodile (Boissonneault et al., 2005; Jacobson
368
et al., 2012; Kansky et al., 2016). While the abovementioned assumption need to be tested
369
which is beyond our study scope, we used media reporting because of the challenges (e.g.
370
limited resources and funding) in assessing tolerance across the country using a
371
conventional method such as social surveys (Boissonneault et al., 2005; Jacobson et al.,
372
Target Journal: Biological Conservation
14
2012). The use of media reporting in human-wildlife conflict studies has been justified as
373
mass media has substantial role in reporting wildlife attacks (Athreya et al., 2015) and they
374
can influence public tolerance and actions towards wildlife management and conservation
375
(McCagh et al., 2015; Sabatier and Huveneers, 2018).
376
We showed that media reporting on crocodile attacks predominantly used negative
377
headline tones, which may have been influenced by the species involved and the type of
378
interactions. This is expected as news agencies often use sensational headlines, mostly
379
depicting wildlife in a poor light to attract the attention of the readers. Attention to
380
saltwater crocodile conservation, or crocodilians in general, is relatively lower than
381
charismatic species like large mammals (Torres et al., 2018). Thus, awareness of saltwater
382
crocodile ecology and conservation is limited and people have a substantial fear of
383
crocodiles as they associate crocodiles with predators of humans (Cavalier et al., 2022;
384
Jacobs, 2009). This emotion is exacerbated by incidents of crocodile attacks on humans,
385
which often receive a great deal of attention from the media. The frequent negative media
386
reporting of crocodile attacks coupled with a lack of awareness and intense fear of
387
crocodiles may have contributed to shaping the lower public tolerance towards crocodiles,
388
as reported in other species (Bombieri et al., 2018; Hughes et al., 2020).
389
Priority risk areas: the new capital city of Indonesia
390
Human-crocodile conflict in high-risk areas encompassed five provinces in eastern Sumatra
391
and one province in eastern Kalimantan. Among the six provinces identified as high-risk
392
areas, five of them (excluding Jambi) were within the top ten provinces with the highest
393
reported crocodile attacks. These areas are likely to experience both repeated crocodile
394
attacks on humans and potential retaliation indicated by negative media reporting. The
395
higher risk of conflict can be explained by the region's topography, human activities, and
396
economic development. The remaining crocodile habitats in high-risk areas are dominated
397
by lowland, wetland, and mangrove forests which have experienced substantial loss over
398
time due to conversion into oil palm plantations and aquaculture farms (Gaveau et al.,
399
2022). These identified risk areas are also among the most impacted regions by
400
anthropogenic pressures as these regions experience rapid human population growth and
401
Target Journal: Biological Conservation
15
economic development, mostly due to transmigration from Java Island over the last century
402
(Allan et al., 2019; BPS, 2021).
403
It is worth noting while some areas in Central Indonesia (Sulawesi Island and Nusa Tenggara
404
Timur Province) were categorized as crocodile attack hotspots, the negative media
405
reportings were much lower. While the way local news companies operate (i.e. focus on
406
other issues outside crocodile attacks) has influence, local beliefs may play a role. For
407
example, in parts of Central Sulawesi (Personal observation) and East Nusa Tenggara
408
Provinces (Paulus and Azmanajaya, 2020; Sideleau et al., 2021), there are local communities
409
who believe crocodiles are their family relatives. These communities protect the crocodiles
410
and would not blame the animals if there is an attack as they believe that incident
411
happened because of mischief done by the victims or disturbance to crocodiles and their
412
habitat. Similar beliefs are found towards other species e.g. Sumatran tiger (Struebig et al.,
413
2018) and Komodo dragons (Sunkar et al., 2020) which shape local tolerance to these
414
species.
415
We highlight Kalimantan Timur, the province that will host the new capital city, as our
416
highest-ranking risk area. The capital zone, estimated at 2566.64 km2 is within 27,500 km2 of
417
the high-risk area we identified in the province, having experienced 84 cases of saltwater
418
crocodile attacks in the ten years. Upon the establishment of the new capital, it is expected
419
to harbour 1.5 million people by 2024 and additional 0.46 million people by 2043 (Muhtar et
420
al., 2021). The subsequent rapid infrastructure development and human settlement
421
expansions will put pressure on the wildlife and remaining forest habitat (1083.64 km2) in
422
the capital zone (Mutaqin et al., 2021), potentially elevating the future risk of human-
423
crocodile conflict. The strong association between the expansion of human settlements and
424
the increase in crocodile attacks has been clearly shown in the past, followed by the
425
extirpation of local crocodile populations (Amarasinghe et al., 2015; CrocBITE, 2020).
426
It is worth noting that the level of human-crocodile conflict in the new Indonesian capital
427
will likely be determined by a combination of factors, including the level of poverty present,
428
access to safe sources of freshwater/plumbing, the level of subsistence fishing activity, and
429
the abundance of natural prey items. Considering the magnitude of human-crocodile in the
430
future capital of Indonesia that has been revealed by this study, mitigation measures and
431
Target Journal: Biological Conservation
16
strategies should have been considered by the authorities. As the Government of the
432
Republic of Indonesia aims to establish the most sustainable capital city in the world (i.e.
433
>75% area will be dedicated to green space; Mutaqin et al. 2021), we hope the
434
implementation will minimize the risk of future conflict between humans and crocodiles.
435
Study limitations and way forward
436
We acknowledge a number of limitations of our study and the methodological framework
437
that we used, and suggest that addressing these limitations would be relevant to broader
438
human-wildlife conflict studies in the future. First is the issue of underreporting conflict
439
incidents from remote areas and/or overreporting from areas with intensive surveys. While
440
we used cellular network coverage as a covariate due to a potential reporting bias
441
associated with remoteness of locations, an analytical approach like occupancy modeling
442
that explicitly accounts for imperfect detection and the influence of data collection efforts,
443
can serve to directly current for such reporting biases in human-wildlife conflict studies
444
(Athreya et al., 2015; Goswami et al., 2015).
445
Second is the potential bias in the way media report conflicts to reflect underlying public
446
attitudes. This bias can be addressed by an approach that accounts for false negative (not
447
reporting positive attitudes when the public has a positive attitude) or false positive errors
448
developed by Vasudev and Goswami (2020). Third, we did not directly evaluate the
449
association between negative media reporting and public tolerance which need to be tested
450
in future studies (Sabatier and Huveneers, 2018). Finally, although the large-scale
451
quantitative analysis in this study captures the pattern of human-crocodile conflict and
452
determines significant drivers of these interactions, a deeper qualitative research especially
453
at smaller scales is needed to better understand underlying mechanisms and tailor
454
mitigation strategy accordingly to the local context.
455
In broader context, our study findings can be extended to improve human-wildlife conflict
456
management. We recommend focused mitigation intervention in the identified high-risk
457
areas where frequent wildlife attack incidents and negative reportings overlap. In these
458
areas, we suggest stakeholders to develop local and case specific strategies to reduce the
459
number of attacks, for example through habitat enrichment and restoration, physical barrier
460
Target Journal: Biological Conservation
17
or buffer zone establishment, signboards installation, and/or translocation of problematic
461
animals following established guidelines (e.g. IUCN/SSC, 2013). We also strongly encourage
462
collaboration with mass media to increase public tolerance by publishing objective and
463
comprehensive reports on human-wildlife conflict (Ardiantiono et al., 2022). The reporting
464
could explain the ecology and conservation of focus species, drivers and detailed chronology
465
of attacks, and mitigation approaches to be taken.
466
The incorporation of social dimensions with ecological data in managing and mitigating
467
human-wildlife conflicts will result in more effective and practical solutions to promote
468
coexistence. Our study provides a socio-ecological framework that utilizes publicly available
469
data to identify priority areas for future conservation interventions. We encourage
470
conservation scientists and practitioners to adopt, verify, and expand our framework in
471
studies of human-animal interaction to better understand the dynamics involved and
472
effectively allocate resources to promote coexistence.
473
Acknowledgements
474
We thank the Ministry of Environment and Forestry (KLHK) and Ir Wiratno, M.Sc, the
475
Director-General of Conservation of Natural Resources and Ecosystems (KSDAE) of the
476
Republic of Indonesia for their support. Associate editor Varun. R. Goswami for his support
477
and invaluable suggestions to improve this manuscript. Three anonymous reviewers for
478
their constructive feedback. Ricky K. Atmadja, Anastasia Wardhani, Yanuar Ishaq, Reza
479
Septian, Rhemawati, and Afrizal Alfarisi for their assistance in the data collection.
480
Additionally, we thank Nicolas Deere for his insightful statistical advice, Katie Spencer for
481
her valuable review on the early manuscript, and Jatna Supriatna and the staff of the
482
Research Center for Climate Change, University of Indonesia for their support to AATA. ARD
483
was funded by the Leverhulme Trust's Tropical Defaunation Hub at University of Kent.
484
Authors’ contribution
485
AR, AATA, and SMH conceived the ideas and designed the study; AR and BS collected the
486
data; AR, SMH, BS, and SH analyzed the data; AR, AAT, and SMH led the writing of
487
manuscript. All authors contributed to the drafts and gave final approval for publication.
488
489
Target Journal: Biological Conservation
18
Data accessibility
490
The data associated with the manuscript are available upon formal request to the first
491
author.
492
493
References
494
Allan, J.R., Watson, J.E.M., Di Marco, M., O’Bryan, C.J., Possingham, H.P., Atkinson, S.C.,
495
Venter, O., 2019. Hotspots of human impact on threatened terrestrial vertebrates.
496
PLoS Biol. 17, 1–18. https://doi.org/10.1371/journal.pbio.3000158
497
Amarasinghe, A.A.T., Madawala, M.B., Suranjan Karunarathna, D.M.S., Manolis, S.C., De
498
Silva, A., Sommerlad, R., 2015. Human-crocodile conflict and conservation implications
499
of Saltwater Crocodiles Crocodylus porosus (Reptilia: Crocodylia: Crocodylidae) in Sri
500
Lanka. J. Threat. Taxa 7, 7111–7130. https://doi.org/10.11609/JoTT.o4159.7111-30
501
Ardiantiono, Alfarisi, A.M., Ishaq, Y., Wijaya, R., Septian, R., Hadi, A.N., Surya, R.A., Rahmi, T.,
502
2022. Analisis konten pemberitaan interaksi manusia-satwa predator di Indonesia. J.
503
Belantara 5, 153–168. https://doi.org/10.29303/jbl.v5i2.874
504
Athreya, V., Srivathsa, A., Puri, M., Karanth, K.K., Kumar, N.S., Karanth, K.U., 2015. Spotted in
505
the news: Using media reports to examine leopard distribution, depredation, and
506
management practices outside protected areas in southern India. PLoS One 10.
507
https://doi.org/10.1371/journal.pone.0142647
508
BIG, 2019. Peta rupabumi digital Indonesia [WWW Document]. Badan Inf. Geospasial
509
Republik Indones. URL http://tanahair.indonesia.go.id/portal-web/ (accessed 4.1.21).
510
Boissonneault, M.-F., Gladstone, W., Scott, P., Cushing, N., 2005. Grey nurse shark human
511
interactions and portrayals: A study of newspaper portrayals of the grey nurse shark
512
from 1969-2003. Electron. Green J. 1.
513
Bombieri, G., Nanni, V., Delgado, M. del M., Fedriani, J.M., López-Bao, J.V., Pedrini, P.,
514
Penteriani, V., 2018. Content analysis of media reports on predator attacks on humans:
515
Toward an understanding of human risk perception and predator acceptance.
516
Bioscience 68, 1–8. https://doi.org/10.1093/biosci/biy072
517
BPS, 2021. Berita resmi statistik hasil Sensus Penduduk 2020, Badan Pusat Statistik. Jakarta.
518
Brien, M.L., Gienger, C.M., Browne, C.A., Read, M.A., Joyce, M.J., Sullivan, S., 2017. Patterns
519
of human-crocodile conflict in Queensland: A review of historical estuarine crocodile
520
Target Journal: Biological Conservation
19
(Crocodylus porosus) management. Wildl. Res. 44, 281–290.
521
https://doi.org/10.1071/WR17011
522
Bright, E.A., Coleman, P.R., Rose, A.N., Urban, M.L., 2011. LandScan 2010. LandScan.
523
Bruskotter, J.T., Wilson, R.S., 2014. Determining where the wild things will be: Using
524
psychological theory to find tolerance for large carnivores. Conserv. Lett. 7, 158–165.
525
https://doi.org/10.1111/conl.12072
526
Campbell, H.A., Dwyer, R.G., Irwin, T.R., Franklin, C.E., 2013. Home range utilisation and
527
long-range movement of estuarine crocodiles during the breeding and nesting season.
528
PLoS One 8. https://doi.org/10.1371/journal.pone.0062127
529
Cavalier, R., Pratt, E.N., Serenari, C., Rubino, E.C., 2022. Human dimensions of crocodilians: a
530
review of the drivers of coexistence. Hum. Dimens. Wildl. 27, 380–396.
531
https://doi.org/10.1080/10871209.2021.1953195
532
CrocBITE, 2020. The worldwide crocodilian attack database [WWW Document]. Big Gecko.
533
URL http://www.crocodile-attack.info (accessed 4.1.20).
534
Das, C.S., Jana, R., 2018. Human-crocodile conflict in the Indian Sundarban: An analysis of
535
spatio-temporal incidences in relation to people’s livelihood. Oryx 52, 661–668.
536
https://doi.org/10.1017/S0030605316001502
537
Dayer, A.A., Williams, A., Cosbar, E., Racey, M., 2017. Blaming threatened species: media
538
portrayal of human–wildlife conflict. Oryx 1–8.
539
https://doi.org/10.1017/S0030605317000783
540
Fukuda, Y., Manolis, C., Appel, K., 2014. Management of human-crocodile conflict in the
541
Northern Territory, Australia: Review of crocodile attacks and removal of problem
542
crocodiles. J. Wildl. Manage. 78, 1239–1249. https://doi.org/10.1002/jwmg.767
543
Fukuda, Y., Webb, G., Manolis, C., Lindner, G., Banks, S., 2019. Translocation, genetic
544
structure and homing ability confirm geographic barriers disrupt saltwater crocodile
545
movement and dispersal. PLoS One 14, 1–20.
546
https://doi.org/10.1371/journal.pone.0205862
547
Funk, C.C., Peterson, P.J., Landsfeld, M.F., Pedreros, D.H., Verdin, J.P., Rowland, J.D.,
548
Romero, B.E., Husak, G.J., Michaelsen, J.C., Verdin, A.P., 2014. A quasi-global
549
precipitation time series for drought monitoring. U.S. Geol. Surv. Data Ser. 832, 4.
550
Gálvez, N., Guillera-Arroita, G., St. John, F.A.V., Schüttler, E., Macdonald, D.W., Davies, Z.G.,
551
Target Journal: Biological Conservation
20
2018. A spatially integrated framework for assessing socioecological drivers of
552
carnivore decline. J. Appl. Ecol. 55, 1393–1405. https://doi.org/10.1111/1365-
553
2664.13072
554
Gaveau, D.L.A., Locatelli, B., Salim, M.A., Husnayaen, Manurung, T., Descals, A., Angelsen, A.,
555
Meijaard, E., Sheil, D., 2022. Slowing deforestation in Indonesia follows declining oil
556
palm expansion and lower oil prices. PLoS One 17, e0266178.
557
https://doi.org/10.1371/journal.pone.0266178
558
González-Desales, G.A., Sigler, L., García-Grajales, J., Charruau, P., Zarco-González, M.M.,
559
Balbuena-Serrano, Á., Monroy-Vilchis, O., 2021. Factors influencing the occurrence of
560
negative interactions between people and crocodilians in Mexico. Oryx 1–9.
561
https://doi.org/10.1017/s0030605319000668
562
Goswami, V.R., Medhi, K., Nichols, J.D., Oli, M.K., 2015. Mechanistic understanding of
563
human-wildlife conflict through a novel application of dynamic occupancy models.
564
Conserv. Biol. 29, 1100–1110. https://doi.org/10.1111/cobi.12475
565
Hughes, C., Foote, L., Yarmey, N.T., Hwang, C., Thorlakson, J., Nielsen, S., 2020. From human
566
invaders to problem bears: A media content analysis of grizzly bear conservation.
567
Conserv. Sci. Pract. 2, 1–11. https://doi.org/10.1111/csp2.176
568
IUCN/SSC, 2013. Guidelines for reintroductions and other conservation translocations,
569
Version 1. ed. IUCN Species Survival Commission, Gland, Switzerland.
570
IUCN, 2020. IUCN SSC Position Statement on the Management of Human-Wildlife Conflict.
571
Jacobs, M.H., 2009. Why do we like or dislike animals? Hum. Dimens. Wildl. 14, 1–11.
572
https://doi.org/10.1080/10871200802545765
573
Jacobson, S.K., Langin, C., Carlton, J.S., Kaid, L.L., 2012. Content analysis of newspaper
574
coverage of the florida panther. Conserv. Biol. 26, 171–179.
575
https://doi.org/10.1111/j.1523-1739.2011.01750.x
576
Kansky, R., Kidd, M., Knight, A.T., 2016. A wildlife tolerance model and case study for
577
understanding human wildlife conflicts. Biol. Conserv. 201, 137–145.
578
McCagh, C., Sneddon, J., Blache, D., 2015. Killing sharks: The media’s role in public and
579
political response to fatal human-shark interactions. Mar. Policy 62, 271–278.
580
https://doi.org/10.1016/j.marpol.2015.09.016
581
Milner-Gulland, E.J., 2012. Interactions between human behaviour and ecological systems.
582
Target Journal: Biological Conservation
21
Philos. Trans. R. Soc. B Biol. Sci. 367, 270–278. https://doi.org/10.1098/rstb.2011.0175
583
Muhtar, R., Fitriadi, Y., Janna, J., Febriani, I., 2021. Tinjauan kewilayahan dan harmonisasi
584
fungsi ruang kawasan calon ibu kota negara dengan Kabupaten Penajam Paser Utara
585
sebagai wilayah pendukung, in: Seminar Nasional Geomatika. pp. 347–354.
586
Mutaqin, D.J., Muslim, M.B., Rahayu, N.H., 2021. Analisis konsep Forest City dalam rencana
587
pembangunan ibu kota negara. Bappenas Work. Pap. 4, 13–29.
588
https://doi.org/10.47266/bwp.v4i1.87
589
OpenCelliD, 2020. Global cell tower database [WWW Document]. URL
590
https://opencellid.org/ (accessed 5.6.21).
591
Pooley, S., 2020. Croc Digest: A bibliography of human- crocodile conflict research and
592
reports, Third edit. ed. Simon Pooley, London.
593
Pooley, S., Siroski, P.A., Fernandez, L., Sideleau, B., Ponce‐Campos, P., 2021. Human–
594
crocodilian interactions in Latin America and the Caribbean region. Conserv. Sci. Pract.
595
1–13. https://doi.org/10.1111/csp2.351
596
Powell, G., Versluys, T.M.M., Williams, J.J., Tiedt, S., Pooley, S., 2020. Using environmental
597
niche modelling to investigate abiotic predictors of crocodilian attacks on people. Oryx
598
54, 639–647. https://doi.org/10.1017/S0030605319000681
599
R Core Team, 2020. R: A language and environment for statistical computing.
600
https://doi.org/10.1038/sj.hdy.6800737
601
Ramdani, Kusrini, M.D., Prasetyo, L.B., 2021. Mapping the distribution of saltwater crocodile
602
(Crocodylus porosus) and risks of human-crocodile conflicts in settlements around
603
Kutai National Park , East Kalimantan. Media Konserv. 26, 52–62.
604
https://doi.org/10.29244/medkon.26.1.52-62
605
Rose, A.N., McKee, J.J., Sims, K.M., Bright, E.A., Reith, A.E., Urban, M.L., 2020. LandScan
606
2019. LandScan.
607
Sabatier, E., Huveneers, C., 2018. Changes in media portrayal of human-wildlife conflict
608
during successive fatal shark bites. Conserv. Soc. 16, 338–350.
609
https://doi.org/10.4103/cs.cs-18-5
610
Santoro, M., Cartus, O., 2021. ESA Biomass Climate Change Initiative (Biomass_cci): Global
611
datasets of forest above-ground biomass for the years 2010, 2017 and 2018, v2 [WWW
612
Document]. Cent. Environ. Data Anal.
613
Target Journal: Biological Conservation
22
https://doi.org/doi:10.5285/84403d09cef3485883158f4df2989b0c.
614
Saragih, G.S., Kayat, Hidayatullah, M., Hadi, D.S., 2020. A preliminary study on the
615
population and habitat of saltwater crocodile (Crocodylus porosus) in Timor Island, East
616
Nusa Tenggara. IOP Conf. Ser. Earth Environ. Sci. 591. https://doi.org/10.1088/1755-
617
1315/591/1/012044
618
Sideleau, B., Sitorus, T., Suryana, D., Britton, A., 2021. Saltwater crocodile (Crocodylus
619
porosus) attacks in East Nusa Tenggara, Indonesia. Mar. Freshw. Res.
620
https://doi.org/10.1071/MF20237
621
Sideleau, B.M., Britton, A.R.C., 2013. An analysis of crocodilian attacks worldwide for the
622
period of 2008 - July 2013, in: Proceedings of the 22nd Working Meeting of the IUCN-
623
SSC Crocodile Specialist Group. pp. 110–113.
624
Struebig, M.J., Linkie, M., Deere, N.J., Martyr, D.J., Millyanawati, B., Faulkner, S.C., Le
625
Comber, S.C., Mangunjaya, F.M., Leader-Williams, N., McKay, J.E., St. John, F.A. V.,
626
2018. Addressing human-tiger conflict using socio-ecological information on tolerance
627
and risk. Nat. Commun. 9, 3455. https://doi.org/10.1038/s41467-018-05983-y
628
Sunkar, A., Kusrini, M.D., Ramadhani, F.S., 2020. Role of culture in the emotional response
629
towards komodo dragon in Komodo and Rinca Islands of Komodo National Park. BIO
630
Web Conf. 19, 00021. https://doi.org/10.1051/bioconf/20201900021
631
Torres, D.F., Oliveira, E.S., Alves, R.R.N., 2018. Conflicts between humans and terrestrial
632
vertebrates: A global review. Trop. Conserv. Sci. 11, 1–15.
633
https://doi.org/10.1177/1940082918794084
634
Vasudev, D., Goswami, V.R., 2020. A Bayesian hierarchical approach to quantifying
635
stakeholder attitudes toward conservation in the presence of reporting error. Conserv.
636
Biol. 34, 515–526. https://doi.org/10.1111/cobi.13392
637
Wahid, D.H., Azhari, S.N., 2016. Peringkasan sentimen ekstraktif di Twitter menggunakan
638
Hybrid TF-IDF dan Cosine Similarity. Indones. J. Comput. Cybern. Syst. 10, 207–218.
639
Webb, G.J.W., Manolis, S.C., Brien, M.L., 2010. Saltwater Crocodile Crocodylus porosus, in:
640
Manolis, S.C., Stevenson, C. (Eds.), Crocodiles. Status Survey and Conservation Action
641
Plan. Crocodile Specialist Group, Darwin, pp. 99–113.
642
Zeileis, A., Kleiber, C., Jackman, S., 2008. Regression models for count data in R. J. Stat.
643
Softw. 27, 1–25. https://doi.org/10.18637/jss.v027.i08
644
645
Target Journal: Biological Conservation
23
Table
646
Table 1. Hurdle model describing the drivers of saltwater crocodile (Crocodylus porosus)
647
attacks based on model averaging of top models with ΔAICc <2. The zero hurdle part
648
represents a binary response (zero attack versus presence of attack per grid cell >=1) and
649
the count part represents number of crocodile attacks per grid cell (data above zero).
650
Model-averaged coefficients (β) and standard error (SE) represent the strength and
651
direction of influence. All statistically significant variables (p < 0.05) are highlighted in
652
asterisk (*).
653
654
Figure captions
655
Figure 1: (Top) The total number of attacks by saltwater crocodile (Crocodylus porosus) in
656
each year has been increasing significantly since 2010 (F1,8= 323.5, p < 0.001). (Bottom) The
657
temporal distribution of the total of saltwater crocodile attacks on humans in each month
658
across the 2010-2019 period (Blue-coloured bars indicate wet season).
659
Figure 2: (Top) The estimated number of saltwater crocodile (Crocodylus porosus) attacks
660
based on the output of the best hurdle model. Areas that were likely to experience repeated
661
Zero hurdle model coefficients
β
SE
z value
p
Intercept*
-1.51
0.09
-16.28
<0.001
Water body length*
0.49
0.09
5.50
<0.001
Mean aboveground forest biomass 2010*
-0.32
0.09
-3.43
<0.001
Mean rainfall 2010*
-0.66
0.11
-6.03
<0.001
Proportion of area covered by cellular network*
0.38
0.08
4.61
<0.001
Count model coefficients
β
SE
z value
p
Intercept
-0.58
0.69
-0.83
0.405
Mean aboveground forest biomass 2010*
-0.50
0.14
-3.57
<0.001
Mean human density 2010*
-0.36
0.18
-2.00
<0.045
Difference of mean human density 2010 and
2018
0.06
0.12
0.51
0.61
Proportion of area covered by cellular network*
0.40
0.13
3.00
0.003
Target Journal: Biological Conservation
24
attacks (> 1) were represented in red grid cells. (Centre) The distribution of index of
662
negative media reporting based on the number of negative headline tones. Provinces with
663
higher negative coverages were indicated in red grid cells. (Bottom) High-risk areas of
664
negative human-saltwater crocodile interactions. Six priority provinces were highlighted:
665
Aceh (AC), Riau (RI), Jambi (JA), Sumatera Selatan (SS), BB (Bangka Belitung), and Kalimantan
666
Timur (KI).
667
Figure 3: Prioritization of sample grids based on attack counts and negative media reporting
668
information partitioned into three priority scales: high (>= 1 estimated attack and above
669
upper quartile negative media reporting index), medium (>= 1 attack and below upper
670
quartile negative index; or otherwise), and low (< 1 estimated attack and below upper
671
quartile negative index).
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
Target Journal: Biological Conservation
25
Figure 1
688
689
690
691
692
693
694
695
696
Target Journal: Biological Conservation
26
Figure 2
697
698
Target Journal: Biological Conservation
27
Figure 3
699
700
701
Supplementary materials
702
Integrating social and ecological information to identify high-risk areas of
703
human-crocodile conflict in the Indonesian Archipelago
704
705
Supplementary Note 1. Indonesia environmental profiles
706
Indonesia is an archipelagic country comprising over 17,000 islands with a total shoreline of
707
108,000 km long (MoEF, 2018). It harbours the fourth largest human population in the world
708
with 270.2 million people and a density of 141 people/km2 (BPS, 2021a). Indonesian climate
709
is mostly hot and humid with a mean annual temperature of 26.12oC and mean annual
710
precipitation of 2856.5 mm (CCKP, 2021). Based on Koppen-Geiger Classification, the
711
country is dominated by tropical forest climate, while Java and Lesser Sunda regions in the
712
central south are dominated by tropical monsoon and savanna climate (CCKP 2021). About
713
63% (~1.21 million km2) of Indonesia’s land cover is forest, of which 18% (0.22 million km2)
714
Target Journal: Biological Conservation
28
is designated for conservation (MoEF, 2018). Forest cover loss in Indonesia is the largest
715
among ASEAN nations with annual deforestation rate of 6,600 km2 between 2011-2017
716
period (MoEF, 2018; Stibig et al., 2013). Less than half of Indonesian coastal mangrove
717
forest areas (17,400 km2; 48%) are categorized as in good condition or intact (BPS, 2020).
718
719
Supplementary Note 2. Saltwater crocodile current distribution and species identification
720
To determine the current Saltwater crocodile Crocodylus porosus (C. porosus) range we used
721
a combination of CrocBITE attack records, non-attack related media reports, and
722
populations known from surveys or local knowledge. Unfortunately, when C. porosus is
723
present there is usually some level of human-crocodile conflict (HCC). Even in areas where
724
HCC is currently rare (e.g., Palau, Peninsular Malaysia, and Singapore), reports of sightings
725
and, sometimes, captures, are frequent. If an area has no recent conflict or non-conflict
726
records and no recent survey data, the species has likely been extirpated from the area or
727
the area is too remote to expect such records to be attainable. These methods, paired with
728
the available literature and personal communications with local contacts, has allowed us to
729
reliably label C. porosus as extinct in Cambodia, Vietnam, and probably extinct in Thailand.
730
In Indonesia movement of crocodiles between islands appears to be frequent and
731
widespread, so recolonization of former habitat is much more likely. In Bali and most of Java
732
Island, for example, despite reliable recent records of vagrant crocodiles, the species is
733
almost certainly extinct. This is because little suitable habitat remains and what is left is
734
heavily populated by humans. When using non-conflict media reports of crocodiles as
735
evidence there is always the possibility that the crocodile in question is a captive release or
736
escapee, rather than a wild individual. We attempt to minimize this possibility by deeming
737
large crocodiles, particularly individuals found along the coast or in estuaries, as much more
738
likely to be wild animals than escapees. In instances where photos of the captured or
739
sighted animal are provided, we can positively ID the physical attributes of C. porosus (as
740
opposed to a hybrid or a non-native species) and look for red flags indicating a potential
741
escapee (e.g., deformation of the snout due to calcium deficiency).
742
We used information from current C. porosus to help in the species identification.
743
Crocodylus porosus is the only species present throughout most of Indonesia. It is the only
744
Target Journal: Biological Conservation
29
species present in Sulawesi, East & West Nusa Tenggara, the Moluccas, the Riau Islands, and
745
Bangka-Belitung. In Sumatera and Kalimantan, both C. porosus and Tomistoma schlegelii are
746
present. In the case of coastal attacks, C. porosus is always the most likely culprit and the
747
crocodile is often seen by witnesses or identified by victims. We also looked for what
748
species are being captured or photographed in these areas for a better idea of which species
749
is responsible. For inland areas, we rely on known distributions (based on surveys),
750
photographs, captures, and expert advice (local authorities). In the upper Batanghari River
751
of Jambi Province in Sumatera, for example, T. schlegelii is the only species known to be
752
present- all photos, captures, and surveys have revealed T. schlegelii and no sign of C.
753
porosus. In the upper Indragiri River of Riau Province in Sumatera, on the other hand, only C.
754
porosus has been seen, photographed, and captured. We use this information to determine
755
which species is responsible. In the rare case where this information does not exist, we label
756
an incident as “undetermined species” (only a few of these incidents exist in Indonesia, but
757
they are numerous in Sri Lanka where C. porosus and C. palustris are sympatric and both
758
attack people).
759
760
761
Supplementary Note 3. Preparation of spatial covariates
762
We extracted the spatial covariates using Quantum GIS software version 3.12.3-București
763
(QGIS 2021). We created two buffer distances from the water body edge. First, we used 500
764
m buffer for the forest biomass layer, an optimal riparian forest width to support terrestrial
765
and aquatic biodiversity (Deere et al., 2021). Second, we used a 5000 m buffer for human
766
population density based on the estimated radius of human activity (Jarchow & Carnes,
767
2021). To create cellular network area coverage, we created an 8 km radius buffer
768
(Kanchwala, 2021) from cellular tower locations using the open-source tower database
769
OpenCellID (https://www.opencellid.org/). Cellular network coverage and other covariates
770
such as water body density and precipitation were extracted within the land area in 50 x 50
771
km grid cells. All covariates were standardized and tested for collinearity (cut-off value 0.7).
772
Target Journal: Biological Conservation
30
Supplementary Figure 1. The map of the study area with 50 x 50 km sampling grids based on the current saltwater crocodile (Crocodylus
773
porosus) distribution range (blue grid cells). Five provinces with the highest number of reported attacks are highlighted: Riau (RI), BB (Bangka
774
Belitung), Kalimantan Timur (KI), Nusa Tenggara Timur (NT), and Maluku (MA).
775
776
777
Target Journal: Biological Conservation
31
Supplementary Figure 2. Spearman collinearity test results for eight spatial covariates for model development (cut-off value 0.7). All covariates
778
were standardized (Z-score normalization).
779
780
Target Journal: Biological Conservation
32
Supplementary figure 3. Effect plots of relationship between estimated number of
781
crocodile attacks and drivers of attack based on model averaging of top models with ΔAICc
782
<2. Grey area represents 95% confidence interval.
783
784
785
786
787
788
789
790
791
792
793
Target Journal: Biological Conservation
33
Supplementary Table 1. Number of crocodile attacks in 2010-2019 period per province.
794
Province
Code
2019
population
density
(people/km2)*
Area size
(km2)*
Fatal
Non-
fatal
Total
attack
Nusa Tenggara Timur
NT
112
48718.1
63
41
104
Kalimantan Timur
KI
29
129066.64
45
38
83
Bangka-Belitung
BB
91
16424.06
26
41
67
Riau
RI
80
87023.66
24
41
65
Maluku
MA
38
46914.03
21
21
42
Lampung
LA
244
34623.8
10
20
30
Aceh
AC
93
57956
8
21
29
Sumatera Selatan
SS
92
91592.43
23
5
28
Maluku Utara
MU
39
31982.5
17
9
26
Kalimantan Tengah
KT
18
153564.5
7
17
24
Kalimantan Utara
KU
10
75467.7
15
6
21
Sumatera Barat
SB
130
42012.89
6
15
21
Sulawesi Tenggara
SG
71
38067.7
6
11
17
Kepulauan Riau
KR
267
8201.72
3
13
16
Sulawesi Tengah
ST
49
61841.29
7
7
14
Kalimantan Barat
KB
34
147307
2
12
14
Sumatera Utara
SU
200
72981.23
6
7
13
Sulawesi Selatan
SN
189
46717.48
8
2
10
Jambi
JA
72
50058.16
3
6
9
Papua Barat
PB
9
102955.15
6
2
8
Bengkulu
BE
100
19919.33
2
4
6
Sulawesi Barat
SR
82
16787.18
3
2
5
Sulawesi Utara
SA
181
13892.47
3
3
Kalimantan Selatan
KS
110
38744.23
3
3
Papua
PA
11
319036.05
1
1
2
Nusa Tenggara Barat
NB
273
18572.32
2
2
Jawa Tengah
JT
1058
32800.69
1
1
Jawa Timur
JI
831
47803.49
1
1
Gorontalo
GO
107
11257.07
1
1
*Human density data from BPS (2022) and province area size from (BPS, 2021b)
795
796
797
798
Target Journal: Biological Conservation
34
Supplementary Table 2. Total number of saltwater crocodile (Crocodylus porosus) attacks
799
on humans by the activity of victim during 2010-2019 period (N = 665).
800
Activities
Total
attack
%Fatal
%Non-
fatal
Fishing
292
51.37
48.63
Self-cleaning (bathing, urinating, defecating)
145
42.76
57.24
Working (repairing boat, collecting woods, washing)
110
39.09
60.91
In river (on boat, swimming, crossing rivers)
55
52.73
47.27
Others (disposing garbage, provoking crocodiles)
16
25.00
75.00
Unknown (no witness, victims were found after the
attack)
47
53.19
46.81
801
802
Supplementary Table 3. Hurdle model performances based on the AIC values. Water
803
represents water length in km; difference_ represents the mean covariate value difference
804
between 2018/2019 and 2010 period.
805
Model and covariates
AIC
ΔAIC
logLik
df
zero hurdle: water + biomass.2010 + rain.2010 +
network.area
count: biomass.2010 + difference_biomass +
human.2010 + network.area
1730.76
0
-855.4
10
zero hurdle: water + biomass.2010 + rain.2010 +
network.area
count: biomass.2010 + human.2010 +
difference.human + network.area
1731.57
0.812
-854.8
11
zero hurdle: water + biomass.2010 + rain.2010 +
network.area
count: biomass.2010 + difference_biomass +
human.2010 + difference.human + network.area
1733.42
2.658
-854.7
12
Target Journal: Biological Conservation
35
zero hurdle: water + biomass.2010 + rain.2010 +
network.area
count: water + biomass.2010 + difference_biomass +
human.2010 + difference.human + network.area
1735.33
4.564
-854.7
13
zero hurdle: water + biomass.2010 + rain.2010 +
network.area
count: water + biomass.2010 + difference_biomass +
human.2010 + difference.human + rain.2010 +
difference.rain + network.area
1739.28
8.519
-854.6
15
(Global model)
zero hurdle: water + biomass.2010 + human.2010 +
rain.2010 + network.area
count: water + biomass.2010 + difference_biomass +
human.2010 + difference.human + rain.2010 +
difference.rain + network.area
1740.32
9.559
-854.2
16
806
Supplementary Table 4. Media reporting headline tone per province. Index of negative
807
media reporting was calculated by dividing the number of articles with negative headline
808
tones in each province by the maximum number of negative articles across provinces.
809
Province
Code
Headline tone
Index of negative
media reporting
Negative
Neutral
Riau
RI
37
10
1
Kalimantan Timur
KI
24
11
0.65
Bangka-Belitung
BB
21
8
0.57
Kalimantan Tengah
KT
16
8
0.43
Aceh
AC
14
4
0.38
Jambi
JA
9
5
0.24
Sumatera Selatan
SS
9
1
0.24
Kalimantan Barat
KB
7
0.19
Maluku
MA
5
2
0.14
Nusa Tenggara
Timur
NT
4
2
0.11
Target Journal: Biological Conservation
36
Sulawesi Utara
SA
4
2
0.11
Sulawesi Tenggara
SG
3
1
0.08
Sumatera Utara
SU
3
1
0.08
Kalimantan Utara
KU
2
4
0.05
Lampung
LA
2
0.05
Maluku Utara
MU
1
0.03
Sulawesi Selatan
SN
1
0.03
Sulawesi Tengah
ST
1
0.03
Sumatera Barat
SB
1
0.03
Bengkulu
BE
1
0
Papua Barat
PB
1
0
810
References
811
BPS. (2020). Statistik sumber daya laut dan pesisir 2020. In Badan Pusat Statistik. Badan
812
Pusat Statistik.
813
BPS. (2021a). Berita resmi statistik hasil Sensus Penduduk 2020. In Badan Pusat Statistik
814
(Issue 27). https://www.bps.go.id/pressrelease/2021/01/21/1854/hasil-sensus-
815
penduduk-2020.html
816
BPS. (2021b). Luas daerah dan jumlah pulau menurut provinsi, 2021. Badan Pusat Statistik.
817
https://www.bps.go.id/indikator/indikator/view_data_pub/0000/api_pub/UFpWMmJZ
818
OVZlZTJnc1pXaHhDV1hPQT09/da_01/1
819
BPS. (2022). Kepadatan penduduk menurut provinsi (jiwa/km2), 2019-2021. BPS.
820
https://www.bps.go.id/indicator/12/141/1/kepadatan-penduduk-menurut-
821
provinsi.html
822
CCKP. (2021). Indonesia climate change overview. Climate Change Knowledge Portal.
823
https://climateknowledgeportal.worldbank.org/country/indonesia
824
Deere, N. J., Bicknell, J. E., Mitchell, S. L., Lewis, O., & Slade, E. (2021). Riparian buffers can
825
help mitigate biodiversity declines in oil palm agriculture. Frontiers in Ecology and the
826
Environment.
827
Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J. R. G.,
828
Gruber, B., Lafourcade, B., Leitão, P. J., Münkemüller, T., Mcclean, C., Osborne, P. E.,
829
Reineking, B., Schröder, B., Skidmore, A. K., Zurell, D., & Lautenbach, S. (2013).
830
Collinearity: A review of methods to deal with it and a simulation study evaluating their
831
performance. Ecography, 36(1), 027–046.
832
Jarchow, M., & Carnes, M. (2021). Incorporating social values and wildlife habitats for
833
Target Journal: Biological Conservation
37
biodiversity conservation modeling in landscapes of the Great Plains. Landscape
834
Ecology, 7.
835
Kanchwala, H. (2021). What are cell towers and how do they work? Science ABC.
836
https://www.scienceabc.com/innovation/cell-tower-work.html
837
MoEF. (2018). The state of the Indonesia’s Forests 2018. Ministry of Environment and
838
Forestry, Republic of Indonesia.
839
QGIS.org. (2021). QGIS Geographic Information System. QGIS Association.
840
http://www.qgis.org
841
Stibig, H.-J., Achard, F., Carboni, S., Raši, R., & Miettinen, J. (2013). Change in tropical forest
842
cover of Southeast Asia from 1990 to 2010. Biogeosciences Discussions, 10(8), 12625–
843
12653.
844
845
846
