Version of Record: https://www.sciencedirect.

com/science/article/pii/S000632071831190X
Manuscript_b346e9f7298729630be7459ec439dd34

Shorty title: Recovery of Amazonian freshwater turtles

Title:

Prospects for freshwater turtle population recovery are catalyzed by pan-

Amazonian community-based management

Authors: Darren Norris1,2,* , Carlos A. Peres3, Fernanda Michalski1,2, James P. Gibbs4

1
Ecology and Conservation of Amazonian Vertebrates Research Group, Federal
University of Amapá, Rod. Juscelino Kubitschek Km 02, 68903-419 Macapá, Amapá,
Brazil

2
Postgraduate Programme in Tropical Biodiversity, Federal University of Amapá, Rod.
Juscelino Kubitschek Km 02, 68903-419 Macapá, Amapá, Brazil

3
School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK

4
Department of Environmental and Forest Biology, State University of New York, 404
Illick Hall, 13210 Syracuse, NY, USA

* Corresponding author. Coordenação de Ciências Ambientais, Federal University of

Amapá, Rod. Juscelino Kubitschek Km 02, 68903-419 Macapá, Brazil. Email:
dnorris75@gmail.com

E-mails: dnorris75@gmail.com (DN); C.Peres@uea.ac.uk (CAP);

fmichalski@gmail.com (FM); jpgibbs@esf.edu (JPG)

Keywords: Adaptive management; Community based management; Lefkovitch matrix;

Life history; Natural resource management; Podocnemis unifilis; Population recovery;
Protected area effectiveness

Word count Abstract: 148

Word count main text: 4392 (Introduction: 732; Methods: 670; Results: 869; Dis: 2121)

Number of tables: 2; Number of figures: 6 ; Number of references: 44

1
© 2019 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
1 Abstract

2 Sustainable use as a mechanism for the conservation and recovery of exploited wildlife

3 populations remains intensely debated, including for freshwater turtles, a diverse and imperiled

4 group of aquatic reptiles that are an important food source for many residents of tropical

5 regions. Here we evaluated the geographical extent of recovery options for a heavily exploited

6 tropical freshwater turtle fauna across 8.86 M km2 of South American river catchments under

7 scenarios of Business-as-Usual (BAU), Protection (Pr) and Community-Based-Management

8 (CBM). For the widespread indicator species, Podocnemis unifilis, demographic analysis

9 showed that populations subject to moderate levels of female harvest (≤10%) can recover over

10 broad areas if concurrent headstarting of hatchlings is practiced more widely. With regional

11 strengthening of the protected area network unlikely, CBM developed with harvest frameworks

12 derived from demographic rates appropriate to tropical species could catalyze a rapid

13 continental scale recovery of Amazonian freshwater turtles within a few decades.

14 Tweetable highlight: Community-based management could ensure the rapid continental scale

15 recovery of Amazonian freshwater turtles.

16 Highlights

17 • We evaluated management scenarios for the future continental-scale recovery of a

18 conservation dependent freshwater turtle.

19 In 9 countries and 53 catchments along a cumulative river length that would stretch 5

20 times around the globe.

21 • One third (70,533 km) of river lengths are within protected areas and two-thirds of

22 these protected areas (1.97 M km2) permit human use.

23 Protected area coverage of Amazonian rivers met the Aichi 2020 target (17%) across

24 7.1 million km2 (79.8% of 8.86 M km2), reaching 50% across 2.27 M km2 (25.5% of 8.86

25 M km2) among 13 of 53 river catchments.

1
26 • Although 32.7% of rivers are protected, we estimate that most (76%, 164,971 km)

27 of these rivers are accessible to turtle harvesters, with 57.2% (123,694 km) both

28 accessible and unprotected.

29 • Widespread human access to rivers indicates that freshwater turtle populations

30 are likely to decline rapidly across much of their range.

31 Severe (≥50%) and rapid (<50 years) future losses were projected across 60% of the

32 pan-Amazonian range (5.3 M km2); in the absence of effective protection and

33 management a widespread indicator species, Podocnemis unifilis, faces imminent

34 extirpation from Bolivia, Ecuador, and Peru.

35 • Protected areas alone are clearly insufficient to ensure the future range-scale

36 viability of the species.

37 • If adult harvest is controlled then community-based-management can catalyze a

38 rapid continental scale recovery for the species.

39 If community-based management is implemented, we predict fourfold continental scale

40 population increases within 50 years across 8.86 M km2.

41 • Species demography and management experience suggest that this and related

42 freshwater turtles have potential for substantial continental scale conservation

43 gains if community-based-management can be supported over broad areas.

44

2
45 1 INTRODUCTION

46 The conservation and recovery of natural resources are global priorities if we are to avoid the

47 calamitous biodiversity losses predicted for the Anthropocene. Impacts are so widespread and

48 rapid that most “Red Listed” vertebrate species (93%) have undergone deterioration in IUCN

49 designated status, with the few successes (improvements in IUCN status) a result of direct

50 conservation interventions (Hoffmann et al. 2010). Tropical regions pose a particular challenge:

51 most biological diversity occurs in the tropics, yet rapid economic development in most tropical

52 countries is supported by only rudimentary frameworks for wildlife protection, which are often

53 ineffectively enforced. Particularly lacking for tropical regions is empirical guidance to inform the

54 development of effective polices and management guidelines for the recovery and sustainable

55 use of wild species.

56 The Amazon basin, the world’s greatest locus of continental biodiversity, encapsulates 21st

57 century conservation challenges. Centuries of exploitation combined with dramatic intensification

58 of overharvesting of aquatic species through the 20th century (Antunes et al. 2016) have

59 generated a widespread “fishing-down the food chain” process across Amazonian waterways

60 (Antunes et al. 2016; Castello et al. 2013). Among aquatic wildlife that have been subject to

61 rampant overexploitation are the once abundant freshwater turtles (Podocnemis spp.), now with

62 surviving populations manifesting dramatic declines in body size (Castello et al. 2013; Ojasti

63 1996; Smith 1979; Vogt 2008). Tropical freshwater turtles have been exploited for centuries

64 leading to widespread population declines across the Amazon through the 20th century (Castello

65 et al. 2013; Smith 1979). These turtles are also a flagship reflecting the success or failure of

66 broader efforts to protect aquatic ecosystems (Abell et al. 2007; Abell et al. 2017; Campos-Silva

67 et al. 2018; Castello et al. 2013).

3
68 Much of our understanding of freshwater turtle population management comes from

69 demographic analyses of temperate species, which are characterized by late onset of sexual

70 maturity and limited capacity to cope with additive sources of adult mortality, leading to the oft-

71 made claim that “sustainable harvest” in turtles is an oxymoron (Congdon et al. 1994). There is

72 good reason to question the applicability of inference derived from temperate zone species to

73 inform harvest management guidelines for tropical turtles (Canessa et al. 2016; Spencer et al.

74 2017), given the relatively extended breeding seasons, higher somatic growth rates and earlier

75 age at first reproduction of tropical species (Moll and Moll 2004; Spencer et al. 2017). Reptiles

76 with relatively fast growth rates, young age at onset of maturity and high fecundity can persist

77 under conditions of relatively low adult survivorship (Shine and Harlow 1999). The success of

78 conservation approaches to achieve species recovery depends on a combination of adaptive

79 management, measureable objectives and realistic timelines i.e., SMART (Specific, Measurable,

80 Attainable, Relevant and Timely) criteria for which demographic analysis and population models

81 are a critical tool. Such models enable understanding the dynamics of harvested populations

82 (Morris and Doak 2002) providing a scientific foundation to inform all SMART components.

83 Myriad actions have been implemented to conserve tropical freshwater turtles.

84 Headstarting, (the rearing of eggs collected from the wild with subsequent release of juveniles

85 back to the wild), has been applied to Amazonian freshwater turtle conservation for over 40 years

86 e.g. Podocnemis expansa since the early 1970s in Brazil (Alho 1985; Páez et al. 2015; Vogt

87 2008). However, few studies to date have attempted to quantify changes in turtle abundance

88 after nest transfer and headstarting with some demonstrating increases in Podocnemididae

89 abundance in Brazil (Campos-Silva et al. 2018; Cantarelli et al. 2014; Miorando et al. 2013), Peru

90 (Harju et al. 2017) and in Ecuador (Townsend et al. 2005). Yet population projections developed

91 for freshwater turtles based on data from temperate zones suggest that headstarting cannot

92 compensate for even small reductions (e.g. 10%) in adult survivorship, which can generate rapid

4
 93 population declines (Heppell et al. 1996). However, the utility of these diagnoses for tropical

94 species is unclear based on pronounced differences in metabolic and demographic rates of these

95 ectothermic species at low versus high latitudes.

96 Here we present a spatially explicit assessment of management scenarios for the recovery

97 of a heavily exploited Amazonian freshwater turtle across 53 river catchments covering 8.86

98 million km2. We integrate demographic data appropriate to tropical freshwater turtles in

99 population projection models. We show that local community-based management (CBM) in which

100 headstarting of juveniles linked to and incentivized by moderate harvest of adults could catalyze

101 a rapid and widespread continental scale recovery trajectory for an otherwise imperiled group of

102 freshwater turtles.

103

104 2 METHODS

105

106 Figure 1. Life-cycle and associated threats in Amazonian freshwater turtles. Stage specific threats
107 (human hunting, human nest harvest and natural nest predation, e.g. by teiid lizards) are represented by
108 grey images close to the directly affected stage. The two cycles represent the differences between (A)
109 Community-Based Management (CBM) and (B) Business-as-Usual (BAU) scenarios.

5
110

111 2.1 Estimating demographic parameters

112 We explored population management scenarios for P. unifilis via a female-only, stage-based

113 (eggs/hatchlings, early juvenile, late juvenile and adults) “Lefkovitch” matrix population projection

114 model (Supplemental Material A: Demographic model development). Matrix population models

115 are appropriate for our broad-scale analysis as both turtle life stages and their associated threats

116 are readily discretized (Figure 1; (Heppell 1998; Smith 1979; Spencer et al. 2017)). To evaluate

117 plausible scenarios representing different future population growth situations for P. unifilis we

118 invoked a base (reference) projection model with vital rates obtained from the scientific literature

119 (Table 1). We then projected populations across plausible values ranges for the stage specific

120 threats i.e. 0 -90% first-year survival and 0 – 50% hunting of adult females (Table1).

121

122 Table 1. Podocnemis unifilis vital rates. Demographic vital rates synthesized from the scientific literature
123 used to model population change in Podocnemis unifilis. Full diagnostics (including sensitivity analysis)
124 are presented in Supplemental Material A. Demographic model development.

Parameter Base Projection value ranges

First year survivala 0.200 0.000 – 0.900
Early Juvenile survival 0.500 0.500
Early Juvenile durationb 2.000 2.000
Early Pic 0.333 0.333
Early Gid 0.167 0.167
Late Juvenile survival 0.400 0.400
Late Juvenile durationb 2.000 2.000
Late Pic 0.286 0.286
Late Gid 0.114 0.114
0.470, 0.700, 0.840, 0.907,
Adult survival 0.930 0.930
Annual fecunditye 11.550 11.550
Growth rate (Lamda λ) 1.020 0.465 – 1.164
a
125 Probability of an egg surviving and the hatchling “graduating” to the early juvenile stage after one year.
b
126 Years
c
127 Pi = probability of a juvenile surviving and remaining in the juvenile stage.
d
128 Gi = probability of a juvenile surviving and “graduating” to the next stage.
e
129 Clutch size X number of clutches X breeding frequency
130

6
131 2.2 Future population changes

132 Here we examine realistic exploitation scenarios to establish plausible future pan-Amazonian

133 population changes. We consider projections of management options (Protection and

134 Community-Based Management) that are widely recognized to be priorities for conservation

135 action (Harju et al. 2017; Moll and Moll 2004; Spencer et al. 2017). We conducted analysis at the

136 level of river catchments (Supplemental Material B: Population scenario mapping) across 8.85

137 million km2 as these are legally, politically, culturally and biologically relevant management units

138 (Abell et al. 2007; Abell et al. 2017; Spencer et al. 2017).

139 Modelling was developed along a four-stage process flow:

140 1. We identified all catchments where P. unifilis has been recorded (Figure 2, Figure

141 S1).

142 2. To estimate a representative value of the suitable habitat we obtained the total

143 length of all larger rivers (Stahler tributary order ≥ 6, hereafter rivers) per catchment

144 (Supplemental Material B: Population scenario mapping). We retained only larger

145 rivers as these generally provide year round navigable access for humans (i.e.

146 represent access during the low water nesting season) and also larger rivers have

147 nesting areas for P. unifilis that are known to move out from smaller tributaries to

148 find suitable nesting areas in larger river channels during the nesting season.

149 3. To enable spatially explicit modelling of population management and exploitation

150 impacts, all rivers were classified as accessible or inaccessible to humans (where

151 accessible are <49 km by river from locations with a human density ⩾3 persons per

152 km2 within each catchment, Supplemental Material B: Population scenario

153 mapping). Protected areas (including all categories from the World Database on

154 Protected Areas) were also included within the classification, providing four river

7
155 classes (accessible: protected and unprotected; and inaccessible: protected and

156 unprotected).

157 4. We established three representative management and exploitation scenarios within

158 which we applied spatially explicit population modelling (Table 2): Business-as-

159 Usual (BAU), Protection (Pr) and Community-Based Management (CBM).

160 Scenarios forecast populations within each catchment under different plausible

161 management policies.

162 Under the BAU scenario different demographic parameters were used to project

163 populations on accessible and inaccessible river sections, whereby accessible populations were

164 modelled with nest collection (first-year survival 0.1) and 10% adult female harvest and

165 inaccessible at base rates (Table 1). BAU reflects the persistent “paper parks” paradigm; i.e.,

166 protected areas are generally large, but underfunded and understaffed, lacking enforcement, and

167 relying more on relative remoteness than active management to conserve biodiversity. The

168 Protection scenario modelled the effect of effective protected areas (independent of category).

169 Accessible rivers that intersected protected areas became “inaccessible” and population

170 demographics set to base rate (Table 1). Accessible rivers outside protected areas were

171 modelled as per Business-as-Usual. The CBM scenario modelled population management along

172 accessible river sections outside protected areas with headstarting (first-year survival 0.5) and

173 10% adult female harvest. Under the CBM scenario, accessible rivers within protected areas and

174 inaccessible rivers were modelled as per Business-as-Usual.

175

176 2.3 Data analysis

177 To evaluate the likely outcomes associated with different scenarios we used decision trees

178 (Hothorn and Zeileis 2015) in the R environment for statistical computing (R Development Core

179 Team 2017). Decision trees were chosen as they provide a visual and analytical decision support

8
180 tool enabling us to identify points of change and analyze the management alternatives that were

181 likely to generate population increases or declines. All R codes and functions used to generate

182 models and run analyses are available via https://goo.gl/S8dL13 .

9
183

184 Figure 2. Basin- and country-wide patterns. Pan-Amazonian distribution of protected areas, human
185 accessibility and yellow spotted freshwater turtle (Podocnemis unifilis) populations. (A) Protected area
186 types classified from the World Database of Protected Areas (downloaded from
187 https://www.protectedplanet.net/). Indigenous lands ("Indigenous Area", "Indigenous Reserve" and/or
188 "Indigenous Territory"), Strictly Protected (categories Ia and Ib) and those with some form of use
189 (categories II to VI). Solid black lines show national boundaries and black-and-yellow dashed line indicate
190 the area covered by catchments within the species range. Percentage PA coverage of rivers in each
191 catchment (C) and country (D). Shades of red below 17%, shades of orange between 17% and 50% (Aichi
192 target) and shades of blue above 50% (“half-world” target). Percentage of waterways that are accessible

193 (<49 km) by river from locations with a human density ⩾3 persons ● km–2 within each catchment (E) and

194 country (F). Percentage of the overall range wide freshwater turtle population within each catchment (G)
195 and country (H). See Tables S1 and S2 for values per country and catchment.

10
196 3 RESULTS

197 3.1 Pan-Amazonian populations and protection

198 Table 2. Countrywide population projections. Projected yellow spotted freshwater turtle (Podocnemis
199 unifilis) population changes (“-“ decrease and “+” increase) under different exploitation scenarios across
200 nine Amazonian countries.

River Population statusb

Country Length PA Acc
Currentc BAUd Prd CBMd
(K km) (%) (%)
Bolivia 18.7 23.9 91.3 0.19 VU -EN -VU +VU
Brazil 123.2 41.1 75.0 1.23 NT -VU +NT +NT
Colombia 18.5 13.3 62.9 0.19 EN -EN +EN +EN
Ecuador 3.3 18.7 95.1 0.03 VU -CR -VU +VU
French Guiana 2.0 54.7 73.7 0.02 VU -VU +VU +VU
Guyana 5.0 10.4 66.8 0.05 VU -VU +VU +VU
Peru 23.8 12.7 90.8 0.24 VU -EN -EN +VU
Suriname 3.4 11.0 56.5 0.03 VU +VU +VU +VU
Venezuela 18.1 40.8 68.5 0.18 VU -VU +VU +VU
VU
Overall 216.0 32.7 76.4 2.16 -VU +VU +VU
(A1acd)

a
201 Summaries of rivers (Stahler tributary order ≥6 ) within each country. Cumulative river length per country
202 in thousands of km; protected area coverage (PA %) is the percentage of river length overlapping
203 protected areas of any classification from the World Database of Protected Areas (downloaded from
204 https://www.protectedplanet.net/). Accessibility (Acc %), is the percentage of river courses within 49 km of
205 any given point with at least 3 people per km2.

b
206 Current and projected threat status. Status classified following IUCN categories (NT: Near Threatened;
207 EN: Endangered; VU: Vulnerable; CR: Critically Endangered). The projected status changes are shown for
208 three scenarios, with the classification following the IUCN population size reduction criteria A3bd (projected
209 % reduction over three generations). The current status is retained when increases are projected. Cells are
210 shaded when projected population changes are greater (i.e. populations increase or reduce less) than the
211 minimum IUCN Red List criterion (30% projected reduction over three generations).

11
 c
212 Current population status: Current density expressed in millions of adult females, assuming a density of
213 10 females per river kilometer. These density values were used as a baseline reference level to obtain
214 percentage projected populations changes over 50 years. Current status is taken from existing national
215 and international (IUCN) Red List evaluations.

d
216 Projected population scenarios: Business-as-Usual (BAU), Protection (Pr) and Community-Based
217 Management (CBM). See Supplemental Material A Table A2 for population parameter values for each
218 scenario and Table S1 for population scenario values (N and % change).

219

220 With Brazil containing by far the largest proportion of rivers in lowland Amazonia, this

221 nation alone is likely to maintain 57% of the current P. unifilis population (Figure 2, Table 1), with

222 37% distributed evenly (8-10% per country) across Bolivia, Colombia, Peru and Venezuela. The

223 remaining 6% of the current population is projected to be distributed (1-2% per country) across

224 Ecuador in the west and French Guiana, Guyana and Suriname in the north-east. Protected area

225 coverage was not uniform across countries or catchments (Figure 2). French Guiana was the

226 only territory to achieve 50% protected area coverage of both terrestrial areas and river lengths

227 (Table 2). With a strong overall protected area coverage of 43.4% and 41.1% for terrestrial areas

228 and rivers, respectively the Brazilian protected area network resulted in the overall protected area

229 coverage remaining relatively high, even when neighboring nations achieved levels well below

230 17% (Figure 2, Table 2).

231 Protected areas, including indigenous lands currently cover approximately 36% (3.19 M

232 km2) of the 53 catchments (8.86 M km2) included in our analysis. This terrestrial coverage was

233 also reflected in the coverage of larger rivers that overlapped protected area polygons. Overall

234 33% (70,533 / 215,975 km) of river length was covered by protected areas (Table 2, Table S1).

235 Yet, we estimate that 76% of rivers are accessible (Table 2, see Table S1, and Table S2 for

236 summary values per country and catchment respectively), representing approximately 164,971

237 km of the overall 215,975 km river length. With protected areas covering 25% of accessible rivers

12
238 (41,277 of 164,971 km), this leaves 57.2% (123,694 km) of river courses that are both accessible

239 and formally unprotected (Table S1, Table S2).

240

241 Figure 3. Population projections. Projected Podocnemis unifilis populations under different management
242 and female harvest scenarios. The demographic impact of stage specific threats were quantified via a
243 female-only, stage-based “Lefkovitch” matrix population projection model based on an initial population
244 abundance vector assuming 1000 adult females at year zero. The effect of reductions in adult female
245 survival (female exploitation resulting from human harvest of adult females at 5 different “hunt” levels) are
246 shown across different values of first-year survival (hatchling graduation to early-juvenile). Full details of
247 demographic parameters used and projection model sensitivity are available in Supplemental Material A:
248 Demographic model development.

249

13
250 3.2 How much protected and/or managed habitat is enough?

251 In the absence of management or exploitation the population of P. unifilis was projected to

252 increase slightly (growth rate λ = 1.02, Table 1). Exploitation (hunting) of adult females generated

253 population reductions (Figure 3), which became increasingly drastic with increasing extent of

254 harvest across river catchments (Figure 4). If all nests were removed (zero first-year survival),

255 populations declined when over 55% of all rivers were affected. These losses were intensified

256 when adult females were exploited. For example, our projections estimated that enhancing first-

257 year survival above the 0.2 base rate (i.e. to 0.3 or more) would be necessary to enable

258 population increases when 10% of adult females were depleted across more than 55% of rivers

259 (Figure 3, Figure 4). Indeed, increasing first-year survival to above 0.6 would also enable

260 populations to remain stable when up to 25% of females were extracted across more than 55%

261 of rivers (Figure 4). Population declines occurred only when scenario coverage was above 55%

262 (Figure 4). Population declines occurred in some cases with 2.5% or 10% hunting (female

263 exploitation levels) when first-year survival was ≤0.3, but above this threshold populations

264 increased even with 2.5% or 10% hunting at 95% coverage (Figure 3). In contrast, population

265 declines were predicted with >10% hunting across the range of first-year survival values (Figure

266 3, Figure 4). For example, at 25% hunting, first-year survival of 0.6 or more was required to

267 ensure populations continued to increase across the scenario coverage values (Figure 4).

14
268
269 Figure 4. Extent of impact of exploitation scenarios on future population changes of yellow spotted
270 freshwater turtles (Podocnemis unifilis). Projected population changes resulting from the extent of
271 different exploitation scenarios within 53 pan-Amazonian river catchments. Positive and negative y-axis
272 values indicate population increases and declines, respectively, and 0.5 and -0.5 represents population
273 doubling or halving time, respectively. Solid horizontal line at 0.0 shows a stable population with zero
274 population change. Full details of demographic parameters used and model sensitivity are available in
275 Supplemental Material A: Demographic model development.

276

277 Comparing scenarios that increased population sizes through enhancing first-year survival

278 above the base rate of 0.2, the population doubling time halved (reduced from 39 to 14 years)

279 when first-year survival increased from 0.2 to 0.3 (Figure 3, Figure 5). The number of adult

280 females also increased as a result of increasing first year survival, for example, when first-year

281 survival increased from 0.2 to 0.3 the adult population doubling time halved (reduced from 33 to

282 14 years) (Figure 3). Indeed, with first-year survival ≥ 0.5, the adult population was predicted to

283 double in eight or fewer years (Figure 3). Conversely, if all nests were removed (zero first-year

284 survival), populations will halve in less than 12 years and if first-year survival was reduced to half

285 of the base level (from 0.2 to 0.1), then populations would halve after 38 years (Figure 3, Figure

15
286 5). Our projections estimated that most individuals, including half of all females were lost after

287 only 16 years if first-year survival remained <0.3 but 10% of females were exploited (Figure 3,

288 Figure 5). Conversely, if first-year survival levels could be enhanced to 0.7 or more, then

289 populations could continue to increase even with a 25% harvest of adult females (Figure 5).

290 However, with 50% adult harvest we estimate that populations will decline and that increasing

291 first-year survival will only delay population collapse (halving) to 24 years (Figure 3, Figure 5).

292
293 Figure 5. Time to impact of exploitation scenarios. Years to effect substantial (50%) population
294 changes as a result of different exploitation scenarios within 53 pan-Amazonian river catchments. The time
295 (in years) to either 50% population increases or 50% population declines are shown above zero (blue
296 circles) or below zero (red circles), respectively. Full details of demographic parameters used and model
297 sensitivity are available in Supplemental Material A: Demographic model development.
298

299 3.3 Population projections of pan-Amazonian turtle populations under different

300 exploitation scenarios

301 Our BAU projections showed an overall 36% reduction in the number of adult females over 50

302 years (Tabe 2, Table S1). Reductions in the number of adult females were projected in most

303 catchments (Figure 6, Table S2), representing losses across 6.27 million km2 or 70.8% of the

304 total area. Severe (≥50%) losses were projected across 60% of the area (Figure 6, Table S2).

16
305 Both the Protection and Community-Based Management scenarios reduced projected BAU

306 losses. Although many catchments continued to experience losses under the Protection

307 scenario, CBM enabled increases across all catchments (Figure 6, Table S2).

308
309 Figure 6. Pan-Amazonian projections of population changes. Projected Podocnemis unifilis population

310 changes across 53 catchments. Columns represent results from three exploitation scenarios, from left to

311 right: Business-as-Usual, Protection and Community-Based Management. Rows show percentage change

312 in the number of adult females under different management scenarios (A –C) and river catchments

313 projected to experience either 50% (D-F) or 30% (G-I) losses of adult females.

314

315 4 DISCUSSION

17
316 Assessments of management scenarios for the recovery of exploited vertebrate populations are

317 crucial to inform meaningful long-term conservation policies over large spatial and temporal

318 scales. Here we modelled the exploitation and management scenarios for Amazonian freshwater

319 turtles and show that community-based management can catalyze a rapid and widespread

320 continental scale recovery across Amazonia. We first discuss sustainable levels of harvest, and

321 then explore how community-based management can result in rapid population recovery

322 trajectories for P. unifilis, providing future projections based on human population growth and

323 prioritized conservation actions.

324 Adult survival is a key focus for the conservation of exploited wildlife, yet its relative

325 importance is influenced heavily by age at reproductive maturity (Bodmer et al. 1997; Campos-

326 Silva and Peres 2016; Dickey-Collas et al. 2010; Shine and Harlow 1999; Spencer et al. 2017).

327 We found that increasing first-year survival in P. unifilis could generate rapid population

328 increases and even compensate for population losses due to adult harvesting. The long-lived P.

329 unifilis is relatively fast maturing, with 5 year-old females laying eggs (Páez et al. 2015), which is

330 a similar maturation age to other commercially exploited Amazonian wildlife species such as the

331 giant freshwater fish Arapaima gigas (Campos-Silva and Peres 2016) and non-timber forest

332 products such as açai fruits (Euterpe oleraceae) (Weinstein and Moegenburg 2004). This

333 relatively short time to reproductive maturity contrasts with the slower maturing temperate and

334 larger-bodied marine turtles (Spencer et al. 2017). The potential of demographic differences to

335 drive the success of specific management actions is a clear example of where policies for

336 widespread recovery must be informed by insights generated from species-relevant projection

337 models.

338 Nesting area protection (under local CBM initiatives) can be an effective strategy to

339 enable rapid and widespread recovery for P. unifilis (Campos-Silva et al. 2018). Our scenario

340 comparison showed increased population sizes through headstarting (increasing first-year

18
341 survival), with population doubling time halved (reduced from 39 to 14 years) when first-year

342 survival increased from 0.2 to 0.3. Our models indicate that headstarting results in the number of

343 adult females increasing and with first-year survival of at least 50%, the adult population could

344 double within 8 years. Thus, under the CBM scenario it would be possible to induce rapid

345 population recovery across the pan-Amazonian distribution of this freshwater turtle. A focus on

346 species management can be more efficient than a focus on directly reducing the negative effects

347 on a specific life-history stage (Spencer et al. 2017). Moreover, P. unifilis females mature much

348 quicker than some of its congeners (Vogt 2008) and increases in first-year survival via

349 headstarting can therefore generate rapid changes in later life stages such as early-late juveniles

350 and adults. Thus, the measurable benefits of boosting first-year survival, such as a 50% increase

351 in adult females can be rapidly achieved (i.e. within a decade) for this overexploited freshwater

352 turtle. However, it is important to highlight the already known potential negative effects of rearing

353 hatchling turtles in artificially high densities in captivity that can lead to viral, bacterial, or

354 protozoan disease outbreaks (Burke 2015; Moll and Moll 2004). Thus, the adequate

355 management of headstarting projects is vital for the success of such conservation actions (Burke

356 2015; Moll and Moll 2004; Spencer et al. 2017). But even more importantly, a continuous and

357 effective screening of hatchlings before the release into natural populations must be implemented

358 to avoid transmission of diseases and parasites to wild populations (Burke 2015; Páez et al.

359 2015).

360 Headstarting has been adopted widely but is not the only method available to improve

361 first-year survival or conservation success in freshwater turtles. The use of headstarting for the

362 conservation of chelonian species continues to generate debate (Burke 2015; Spencer et al.

363 2017). For example, there remains substantial uncertainty regarding the efficacy of headstarting

364 approaches for P. unifilis and although headstarting has been widely adopted it remains poorly

365 evaluated across Brazil (Páez et al. 2015), As widely documented for chelonians, headstarting

19
366 can work reliably but needs to be implemented with due care and attention (Burke 2015). There

367 are numerous examples of unintended consequences and programs may fail if headstarted

368 hatchlings do not behave appropriately or are diseased (Burke 2015). However, considering the

369 speed at which anthropogenic impacts are estimated to reduce projected populations (our

370 findings suggest 10% hunting of adult females can halve populations in 16 years or less with nest

371 harvesting the fact that populations still remain across Brazilian Amazonia can be attributed at

372 least in part to the success of management actions including headstarting (Páez et al. 2015). Yet,

373 headstarting is obviously only one method whereby first-year survival can be increased.

374 Contrasts of conservation efforts across the species’ range clearly highlight that the most

375 important component of conservation success for the species is the involvement of local

376 communities (Campos-Silva et al. 2018; Harju et al. 2017; Spencer et al. 2017). This can include

377 various actions to protect fluvial beaches that serve as nesting habitat for turtles and many other

378 species (Campos-Silva et al. 2018) and/or limiting harvest of eggs or females to levels that can

379 achieve conservation objectives (Harju et al. 2017).

380 Both the expansion of new development frontiers and growing energy and food demands

381 can cause drastic reductions in freshwater turtle populations across Amazonia (Castello et al.

382 2013; Moll and Moll 2004; Norris et al. 2018a; Smith 1979). Evidence suggests that local peoples

383 living along Amazonian rivers may focus most of their subsistence activities close (< 1km) to their

384 houses (Norris and Michalski 2013), yet commercial activities (e.g. fishing) are associated with

385 more widespread impacts along Amazonian rivers (Rebelo et al. 2005; Tregidgo et al. 2017). For

386 example, Rebelo et al. (2005) found a clear separation between the number of P. unifilis eggs

387 collected for personal consumption (mean of 133 eggs per trip close to settlements, < 2 hour

388 travel time) and commercial trade (mean of 1007 eggs per trip > 7 hour travel time from

389 settlements). Such differences highlight the importance of understanding the role played by

20
390 social and economic factors in the harvest and consumption of both P. unifilis eggs and adult

391 females (Harju et al. 2017; Ostrom 2015; Spencer et al. 2017).

392 Increased resource demands from growing human populations can be directly linked with

393 infrastructure development, including electrification of rural communities and hydroelectric dams

394 (Castello et al. 2013; Tundisi et al. 2014). Such changes mean that physical accessibility will

395 undoubtedly change in remote Amazonian areas over the next decades. The number of urban

396 inhabitants will increase, but not all urban areas will necessarily spread/expand geographically. In

397 fact, there is little evidence to suggest substantial short to mid-term urban expansion across the

398 P. unifilis distributional range (Seto et al. 2012). However, the impacts of natural resource

399 demands from urban inhabitants can extend 1000´s of kilometers (Tregidgo et al. 2017), which

400 could affect the currently inaccessible areas across Amazonia.

401 The establishment and maintenance of protected areas has been the priority for pan-

402 Amazonian biodiversity conservation, yet it can take decades for newly established protected

403 areas to meet minimum management effectiveness criteria. Although actions to encourage social

404 integration are recognized as key for conservation success, outside of Indigenous lands social

405 integration remains a secondary/complementary component to protected area establishment and

406 maintenance. For example, after an investment of hundreds of millions of dollars in the creation

407 and consolidation of protected areas across the Brazilian Amazon, the median overall

408 management effectiveness score obtained for 114 PAs (covering 592,000 km2 of Brazilian

409 Amazonia) was 0.62 (WWF 2017), which represents management described as “basic but with

410 major deficiencies” (Leverington et al. 2008). Perhaps even more surprisingly, within this globally

411 important multi-million dollar program there remain only recommendations for the inclusion of

412 “specific indicators of social and economic impact for incorporation into ARPA’s monitoring

413 protocol” (WWF 2017).

21
414 Our finding that CBM can be more effective than the more widely extolled protected area

415 model suggests that a reassessment of priorities for the development of effective pan-Amazonian

416 conservation actions is necessary. Our Business-as-Usual scenario reflects the pessimistic

417 evaluation that neither protected areas, national water resource legislation, nor community-based

418 management schemes can adequately protect Amazonian freshwater ecosystems against

419 current pressures (Castello et al. 2013). Yet, projections from both Protection and Community-

420 Based Management scenarios clearly demonstrate the potential of conservation policies to

421 generate widespread recoveries of managed species across their freshwater ecosystems. Indeed

422 additional conservation and development benefits are likely if Protection and Community-Based

423 Management approaches can be effectively integrated.

424 The continued challenges to improve protected area effectiveness mean that CBM is an

425 increasingly necessary requirement for conservation success. The recent (29 November 2018),

426 adoption by the UN Convention on Biological Diversity of a new conservation designation “other

427 effective area-based conservation measures” (CBD/COP/14 2018) is perhaps the clearest

428 recognition of the importance of community-based management initiatives for achieving future

429 biodiversity conservation and sustainable development goals. We found that under BAU

430 scenarios populations can collapse within the decades necessary for PAs to become effectively

431 managed. The Peruvian Amazon and Venezuelan Llanos were clear examples of regions that

432 were both largely accessible and poorly protected. As predicted by our models, the adoption of

433 CBM in Peruvian protected areas (Pacaya Samiria National Reserve) is a clear demonstration of

434 the effectiveness of integrating Protection and CBM approaches to generate rapid population

435 recoveries and conservation success (Harju et al. 2017; Sinovas et al. 2017). But how can

436 Community-based management effectively operate across 123,694 km of rivers that are

437 accessible and unprotected?

22
438 With 57% (representing a cumulative length that would stretch three times around the

439 globe) of rivers accessible and unprotected CBM becomes vital for pan-Amazonian conservation

440 efforts. A diverse range of policies, governance regimes and management actions have been

441 used successfully for the local scale conservation of freshwater turtles outside of protected areas

442 (Campos-Silva et al. 2018; Moll and Moll 2004; Sinovas et al. 2017; Spencer et al. 2017;

443 Townsend et al. 2005). The major challenge is to translate these local conservation gains to

444 more widespread recoveries. As found in many species, P. unifilis has a clustered distribution

445 along rivers especially during the nesting season when females congregate around suitable

446 nesting areas (Coway-Gomez 2007; Escalona and Fa 1998; Ferreira and Castro 2010; Moll and

447 Moll 2004). This clustered distribution pattern during key reproductive periods enables

448 strategically localized and targeted actions to generate widespread impacts. For example, CBM

449 at only two nesting areas enabled the successful protection of 75% of nests along a 33 km

450 stretch of river (Norris et al. 2018b). The CBM coverage required to generate population

451 recoveries can therefore be achieved by relatively simple actions at strategic locations that

452 represent a fraction of the overall river length. Additionally, the fact that CBM is necessary along

453 accessible pan-Amazonian rivers creates substantial opportunities for widespread and integrated

454 conservation solutions. Advancing education is a priority across all nine countries and more

455 accessible rivers are closer to academic and research centers. Therefore this physical proximity

456 immediately generates opportunities for adaptive multilevel management options.

457 Conducting analyses that are relevant over broad scales requires that our results are

458 generalizations and there remain important considerations before more localized applications can

459 be informed. Moreover, our demographic analysis was parameterized from a variety of sources,

460 which could potentially lead to inaccuracies. Yet, any possible bias induced by potential errors in

461 the demographic estimation, as well as our focus on a single species as an indicator of a larger

462 suite complex of congeners, are likely to be inconsequential in relation to the scale and

23
463 magnitude of the effects on Amazonian turtle populations being wrought by widespread human

464 impacts (Moll and Moll 2004; Smith 1979; Vogt 2008). Notably, we selected scenario values

465 based on their robust quantification across their full potential range (e.g. 0 - 90% first-year

466 survival). Although adopting different parameter values would obviously generate moderately

467 different outcomes, this would not change our conclusions that community-based management

468 can potentially catalyze freshwater turtle population recovery across Amazonia. Nevertheless,

469 modeled projections will only be improved as further demographic data become available for

470 these and other tropical freshwater turtle species.

471 We examined the conservation potential of three scenarios for future projections of a

472 tropical freshwater turtle species ubiquitously distributed across Amazonia. Our community-

473 based management scenario could increase first-year survival and consequently number of

474 adults, which in turn would result in rapid increases in population levels across the distributional

475 range. Protected areas can be successful relative to other conservation targets, but they will

476 often still fall short of meeting minimum criteria for sustaining natural resources and biodiversity

477 retention. Local community engagement is necessary and perhaps the only way forward to

478 ensure long-term conservation objectives for overexploited wildlife species in low-governance

479 tropical regions like the Amazon. Our projections of widespread demographic population

480 increases of freshwater turtles associated with community-based management reveals a road to

481 recovery for imperiled turtles and other aquatic vertebrates, outlining a potential success for 21st

482 century biodiversity conservation.

483

484

24
485 References

486 Abell, R., Allan, J.D., Lehner, B., 2007. Unlocking the potential of protected areas for
487 freshwaters. Biological Conservation 134, 48-63.
488 Abell, R., Lehner, B., Thieme, M., Linke, S., 2017. Looking Beyond the Fenceline: Assessing
489 Protection Gaps for the World's Rivers. Conservation Letters 10, 384-394.
490 Alho, C.J.R., 1985. Conservation and management strategies for commonly exploited
491 amazonian turtles. Biological Conservation 32, 291-298.
492 Antunes, A.P., Fewster, R.M., Venticinque, E.M., Peres, C.A., Levi, T., Rohe, F., Shepard, G.H.,
493 2016. Empty forest or empty rivers? A century of commercial hunting in Amazonia.
494 Science Advances 2, e1600936.
495 Bodmer, R.E., Eisenberg, J.F., Redford, K.H., 1997. Hunting and the likelihood of extinction of
496 Amazonian mammals. Conservation Biology 11, 460-466.
497 Burke, R.L., 2015. Head-starting turtles: learning from experience. 10, 299-308.
498 Campos-Silva, J.V., Hawes, J.E., Andrade, P.C., Peres, C.A., 2018. Unintended multispecies
499 co-benefits of an Amazonian community-based conservation programme. Nature
500 Sustainability 1, 650.
501 Campos-Silva, J.V., Peres, C.A., 2016. Community-based management induces rapid recovery
502 of a high-value tropical freshwater fishery. Scientific Reports 6, 34745.
503 Canessa, S., Genta, P., Jesu, R., Lamagni, L., Oneto, F., Salvidio, S., Ottonello, D., 2016.
504 Challenges of monitoring reintroduction outcomes: Insights from the conservation
505 breeding program of an endangered turtle in Italy. Biological Conservation 204, 128-133.
506 Cantarelli, V.H., Malvasio, A., Verdade, L.M., 2014. Brazil's Podocnemis expansa Conservation
507 Program: Retrospective and Future Directions. Chelonian Conservation and Biology 13,
508 124-128.
509 Castello, L., McGrath, D.G., Hess, L.L., Coe, M.T., Lefebvre, P.A., Petry, P., Macedo, M.N.,
510 Renó, V.F., Arantes, C.C., 2013. The vulnerability of Amazon freshwater ecosystems.
511 Conservation Letters 6, 217-229.
512 CBD/COP/14, 2018. Protected areas and other effective area-based conservation measures, In
513 Agenda item 24. ed. F.m. Conference of the parties to the convention on Biological
514 Diversity, Sharm El-Sheikh, Egypt, 17-29 November 2018, pp. 1-19. UN Convention on
515 Biological Diversity.
516 Congdon, J.D., Dunham, A.E., Sels, R.V.L., 1994. Demographics of common snapping turtles
517 (Chelydra serpentina): implications for conservation and management of long-lived
518 organisms. American Zoologist 34, 397-408.
519 Coway-Gomez, K., 2007. Effects of human settlements on abundance of Podocnemis unifilis
520 and P. expansa turtles in Nortbeastern Bolivia. Chelonian Conservation and Biology 6,
521 199-205.
522 Dickey-Collas, M., Nash, R.D.M., Brunel, T., van Damme, C.J.G., Marshall, C.T., Payne, M.R.,
523 Corten, A., Geffen, A.J., Peck, M.A., Hatfield, E.M.C., Hintzen, N.T., Enberg, K., Kell, L.T.,
524 Simmonds, E.J., 2010. Lessons learned from stock collapse and recovery of North Sea
525 herring: a review. Ices Journal of Marine Science 67, 1875-1886.
526 Escalona, T., Fa, J.E., 1998. Survival of nests of the terecay turtle (Podocnemis unifilis) in the
527 Nichare-Tawadu rivers, Venezuela. Journal of Zoology 244, 303-312.
528 Ferreira, P.D., Castro, P.T.A., 2010. Nesting ecology of Podocnemis expansa (Schweigger,
529 1812) and Podocnemis unifilis (Troschel, 1848) (Testudines, Podocnemididae) in the
530 Javaes River, Brazil. Brazilian Journal of Biology 70, 85-94.
531 Harju, E., Sirén, A.H., Salo, M., 2017. Experiences from harvest-driven conservation:
532 Management of Amazonian river turtles as a common-pool resource. Ambio 47, 327.

25
533 Heppell, S.S., 1998. Application of life-history theory and population model analysis to turtle
534 conservation. Copeia, 367-375.
535 Heppell, S.S., Crowder, L.B., Crouse, D.T., 1996. Models to evaluate headstarting as a
536 management tool for long‐lived turtles. Ecological Applications 6, 556-565.
537 Hoffmann, M., Hilton-Taylor, C., Angulo, A., Böhm, M., Brooks, T.M., Butchart, S.H.M.,
538 Carpenter, K.E., Chanson, J., Collen, B., Cox, N.A., Darwall, W.R.T., Dulvy, N.K.,
539 Harrison, L.R., Katariya, V., Pollock, C.M., Quader, S., Richman, N.I., Rodrigues, A.S.L.,
540 Tognelli, M.F., Vié, J.-C., Aguiar, J.M., Allen, D.J., Allen, G.R., Amori, G., Ananjeva, N.B.,
541 Andreone, F., Andrew, P., Ortiz, A.L.A., Baillie, J.E.M., Baldi, R., Bell, B.D., Biju, S.D.,
542 Bird, J.P., Black-Decima, P., Blanc, J.J., Bolaños, F., Bolivar-G., W., Burfield, I.J., Burton,
543 J.A., Capper, D.R., Castro, F., Catullo, G., Cavanagh, R.D., Channing, A., Chao, N.L.,
544 Chenery, A.M., Chiozza, F., Clausnitzer, V., Collar, N.J., Collett, L.C., Collette, B.B.,
545 Fernandez, C.F.C., Craig, M.T., Crosby, M.J., Cumberlidge, N., Cuttelod, A., Derocher,
546 A.E., Diesmos, A.C., Donaldson, J.S., Duckworth, J.W., Dutson, G., Dutta, S.K., Emslie,
547 R.H., Farjon, A., Fowler, S., Freyhof, J., Garshelis, D.L., Gerlach, J., Gower, D.J., Grant,
548 T.D., Hammerson, G.A., Harris, R.B., Heaney, L.R., Hedges, S.B., Hero, J.-M., Hughes,
549 B., Hussain, S.A., Icochea M., J., Inger, R.F., Ishii, N., Iskandar, D.T., Jenkins, R.K.B.,
550 Kaneko, Y., Kottelat, M., Kovacs, K.M., Kuzmin, S.L., La Marca, E., Lamoreux, J.F., Lau,
551 M.W.N., Lavilla, E.O., Leus, K., Lewison, R.L., Lichtenstein, G., Livingstone, S.R.,
552 Lukoschek, V., Mallon, D.P., McGowan, P.J.K., McIvor, A., Moehlman, P.D., Molur, S.,
553 Alonso, A.M., Musick, J.A., Nowell, K., Nussbaum, R.A., Olech, W., Orlov, N.L.,
554 Papenfuss, T.J., Parra-Olea, G., Perrin, W.F., Polidoro, B.A., Pourkazemi, M., Racey,
555 P.A., Ragle, J.S., Ram, M., Rathbun, G., Reynolds, R.P., Rhodin, A.G.J., Richards, S.J.,
556 Rodríguez, L.O., Ron, S.R., Rondinini, C., Rylands, A.B., Sadovy de Mitcheson, Y.,
557 Sanciangco, J.C., Sanders, K.L., Santos-Barrera, G., Schipper, J., Self-Sullivan, C., Shi,
558 Y., Shoemaker, A., Short, F.T., Sillero-Zubiri, C., Silvano, D.L., Smith, K.G., Smith, A.T.,
559 Snoeks, J., Stattersfield, A.J., Symes, A.J., Taber, A.B., Talukdar, B.K., Temple, H.J.,
560 Timmins, R., Tobias, J.A., Tsytsulina, K., Tweddle, D., Ubeda, C., Valenti, S.V., Paul van
561 Dijk, P., Veiga, L.M., Veloso, A., Wege, D.C., Wilkinson, M., Williamson, E.A., Xie, F.,
562 Young, B.E., Akçakaya, H.R., Bennun, L., Blackburn, T.M., Boitani, L., Dublin, H.T., da
563 Fonseca, G.A.B., Gascon, C., Lacher, T.E., Mace, G.M., Mainka, S.A., McNeely, J.A.,
564 Mittermeier, R.A., Reid, G.M., Rodriguez, J.P., Rosenberg, A.A., Samways, M.J., Smart,
565 J., Stein, B.A., Stuart, S.N., 2010. The Impact of Conservation on the Status of the World’s
566 Vertebrates. Science 330, 1503-1509.
567 Hothorn, T., Zeileis, A., 2015. partykit: A Modular Toolkit for Recursive Partytioning in R. Journal
568 of Machine Learning Research 16, 3905-3909.
569 Leverington, F., Hockings, M., Costa, K.L., 2008. Management effectiveness evaluation in
570 protected areas: a global study. World Commission on Protected Areas University of
571 Brisbane, Australia.
572 Miorando, P.S., Rebêlo, G.H., Pignati, M.T., Pezzuti, J.C.B., 2013. Effects of Community-Based
573 Management on Amazon River Turtles: A Case Study of Podocnemis sextuberculata in
574 the Lower Amazon Floodplain, Pará, Brazil. Chelonian Conservation and Biology 12, 143-
575 150.
576 Moll, D., Moll, E.O., 2004. The ecology, exploitation and conservation of river turtles. Oxford
577 University Press on Demand.
578 Morris, W.F., Doak, D.F., 2002. Quantitative conservation biology. Sinauer, Sunderland,
579 Massachusetts, USA.
580 Norris, D., Michalski, F., 2013. Socio-economic and spatial determinants of anthropogenic
581 predation on Yellow-spotted River Turtle, Podocnemis unifilis (Testudines:
582 Pelomedusidae), nests in the Brazilian Amazon: Implications for sustainable conservation
583 and management. Zoologia (Curitiba) 30, 482-490.

26
584 Norris, D., Michalski, F., Gibbs, J.P., 2018a. Beyond harm’s reach? Submersion of river turtle
585 nesting areas and implications for restoration actions after Amazon hydropower
586 development. PeerJ 6, e4228.
587 Norris, D., Michalski, F., Gibbs, J.P., 2018b. Community involvement works where enforcement
588 fails: conservation success through community-based management of Amazon river turtle
589 nests. PeerJ 6, e4856.
590 Ojasti, J., 1996. Wildlife utilization in Latin America: current situation and prospects for
591 sustainable management. Food & Agriculture Org.
592 Ostrom, E., 2015. Governing the commons. Cambridge University Press, Cambridge.
593 Páez, V.P., Lipman, A., Bock, B.C., Heppell, S.S., 2015. A Plea to Redirect and Evaluate
594 Conservation Programs for South America’s Podocnemidid River Turtles. Chelonian
595 Conservation and Biology 14, 205-216.
596 R Development Core Team, 2017. R: A language and environment for statistical computing. R
597 Foundation for Statistical Computing, Vienna, Austria.
598 Rebelo, G.H., Pezzuti, J.C.B., Lugli, L., Moreira, G., 2005. Pesca artesanal de quelonios no
599 Parque Nacional do Jau (AM). Boletim do Museu Paraense Emı´lio Goeldi, serie Ciencias
600 Humanas 1, 111-127.
601 Seto, K.C., Güneralp, B., Hutyra, L.R., 2012. Global forecasts of urban expansion to 2030 and
602 direct impacts on biodiversity and carbon pools. Proceedings of the National Academy of
603 Sciences 109, 16083.
604 Shine, R., Harlow, P.S., 1999. Reticulated pythons in Sumatra: biology, harvesting and
605 sustainability. Biological Conservation 87, 349-357.
606 Sinovas, P., Price, B., King, E., Hinsley, A., Pavitt, A., 2017. Wildlife trade in the Amazon
607 countries: an analysis of trade in CITES listed species, p. 112. Technical report prepared
608 for the Amazon Regional Program, Cambridge, UK.
609 Smith, N.J.H., 1979. Aquatic turtles of Amazonia: an endangered resource. Biological
610 Conservation, 165–176.
611 Spencer, R.J., Dyke, J.U., Thompson, M.B., 2017. Critically evaluating best management
612 practices for preventing freshwater turtle extinctions. Conservation Biology 31, 1340-1349.
613 Townsend, W.R., Borman, R., Yiyoguaje, E., Mendua, L., 2005. Cofan Indians' monitoring of
614 freshwater turtles in Zabalo, Ecuador. Biodiversity and Conservation 14, 2743-2755.
615 Tregidgo, D.J., Barlow, J., Pompeu, P.S., de Almeida Rocha, M., Parry, L., 2017. Rainforest
616 metropolis casts 1,000-km defaunation shadow. Proceedings of the National Academy of
617 Sciences 114, 8655-8659.
618 Tundisi, J.G., Goldemberg, J., Matsumura-Tundisi, T., Saraiva, A.C.F., 2014. How many more
619 dams in the Amazon? Energy Policy 74, 703-708.
620 Vogt, R., 2008. Tartarugas da Amazonia. INPA, Manaus.
621 Weinstein, S., Moegenburg, S., 2004. Acai palm management in the Amazon Estuary: Course
622 for conservation or passage to plantations? Conservation and Society 2, 315.
623 WWF, 2017. ARPA For Life: Amazon Region Protected Area Phase III Report 2017. World
624 Wildlife Fund, Washington, DC USA.

625

27