ponds, puddles, floodplains and dams in the upper xingu basin: … · 120 soil infiltrability but...

1

1 Title: Ponds, puddles, floodplains and dams in the Upper Xingu Basin: could we be witnessing

2 the ´lentification´ of deforested Amazonia?

3

4 Authors: Luis Schiesari1, Paulo R. Ilha2,3, Daniel Din Betin Negri2, Paulo Inácio Prado2,4, Britta

5 Grillitsch5

6

7 Affiliations: 1Gestão Ambiental, Escola de Artes, Ciências e Humanidades, Universidade de São

8 Paulo, São Paulo, SP, 03828-000, Brazil; 2Departamento de Ecologia, Instituto de Biociências,

9 Universidade de São Paulo, São Paulo, SP, 05508-090, Brazil; 3Present Address: Instituto de

10 Pesquisa Ambiental da Amazônia, Canarana, MT, 78640-000, Brazil; 4LAGE at Departamento de

11 Ecologia, Instituto de Biociências, Universidade de São Paulo, São Paulo, SP, 05508-090, Brazil;

12 5Department of Biomedical Sciences, University of Veterinary Medicine of Vienna, Vienna, 1210,

13 Austria.

14

15 Corresponding author: Luis Schiesari; Gestão Ambiental, Escola de Artes, Ciências e Humanidades,

16 Universidade de São Paulo, São Paulo, SP, 03828-000, Brazil; phone +55 11 992346740, email:

17 [email protected]

18

19 Short title: The lentification of deforested Amazonia?

20

21 Keywords: land use change, deforestation, soybean, pasture, dams, freshwater, fish, amphibians

.CC-BY 4.0 International licenseavailable under awas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprint (whichthis version posted January 18, 2019. ; https://doi.org/10.1101/524652doi: bioRxiv preprint

https://doi.org/10.1101/524652

http://creativecommons.org/licenses/by/4.0/

2

23 Abstract

24 Hydrological change is a conspicuous signal of land use intensification in human-dominated

25 landscapes. We hypothesized that land conversion and land use change increase the availability of

26 lentic habitats and associated biodiversity in Southern Amazonian landscapes through at least four

27 drivers. River damming promotes the formation of reservoirs, which are novel permanent lentic water

28 bodies. A rise in the water table driven by local deforestation promotes the expansion of shallow

29 riparian floodplains. Soil compaction and the deliberate construction of cattle and drainage ponds

30 promote the increase in temporary water bodies in interfluvia. We tested these hypotheses using data

31 on habitat characterization and biological surveys of amphibians and fish in forests, pastures and

32 soybean fields in the headwaters of the Xingu River in Mato Grosso, Brazil. Lentic habitat availability

33 sharply increased in deforested land, with consequences to freshwater biodiversity. Reservoir

34 formation influenced both fish and amphibian assemblage structure. Fish species ranged from

35 strongly favored to strongly disfavored by reservoir conditions. Amphibian richness and abundance

36 increased in pasture and soybean streams relative to forests, with a strong positive effect of density of

37 reservoirs in the landscape. Expansion of stream floodplains increased the abundance of

38 Melanorivulus megaroni, a fish species indicator of shallow lentic habitats. Rainwater accumulation in

39 temporary ponds and puddles, entirely absent from well-drained forested interfluvia, allowed the

40 invasion of converted interfluvia by twelve species of open-area amphibians. A literature review

41 indicates that these four drivers of hydrological change are geographically widespread suggesting that

42 we may be witnessing a major yet previously unaccounted form of habitat change in deforested

43 Amazonia.

44

45 Introduction

46 Hydrological change is a conspicuous signal of land use intensification in human-dominated

47 landscapes [1]. The importance of water as a resource for human, livestock and crops, as a source of

48 energy, and as pathway for transportation makes water management a quintessential element of

49 human activities. In addition to the deliberate management of waterbodies, interventions in the

50 terrestrial environment indirectly influence hydrology by altering the partitioning of rainfall among the

51 processes of evapotranspiration, infiltration and runoff [2].



https://doi.org/10.1101/524652


3

52 Frontier landscapes provide a relevant environmental scenario for testing the occurrence and

53 consequences of anthropogenic hydrological change both because ongoing land use change permits

54 side-by-side comparison of converted and native habitats, and because frontiers are regions where

55 biodiversity change is expected to be highest [3]. The Amazon Basin is the largest area of the world

56 currently undergoing frontier settlement and, within the Amazon Basin, the Brazilian state of Mato

57 Grosso accounts for more than 40% of deforested land [4]. Building on published hydrological studies

58 [5; 6; 7; 8], we test the hypothesis that deforestation is promoting the ´lentification´ of Southern

59 Amazonian landscapes, with consequent shifts in their freshwater fauna. The term ´lentification´ was

60 coined to the natural lentic characteristics of small, intermittent floodplain channels during low waters

61 [9] and later used to the transformation of the habitat character of rivers from lotic (i.e. flowing water)

62 to lentic (i.e. standing water) through processes such as flow regulation, unsustainable water

63 extraction and climate change [10; 11]. The term does not appear to have been embraced by the

64 scientific community though (only three papers with ´lentification´ were retrieved in a Web of Science

65 search; March 13, 2018) and lacks an operational definition. We here define ´lentification´ as an

66 increase in the availability (i.e. density and/or area cover) of lentic freshwater habitats in the

67 landscape, directly or indirectly caused by human activities. This definition is inclusive of any driver,

68 and highlights that lentification is a property of the landscape and not only of the drainage network as

69 it includes the creation and change of freshwater habitat in interfluves as well. Because few

70 processes lead to the formation of novel lotic habitats, an increase in the availability of lentic habitat

71 almost invariably implies an increase in the proportion of lentic habitats in the landscape.

72 We hypothesize that at least four drivers of lentification could be operating in the Upper Xingu

73 Basin, one of the largest tributaries of the Amazon river: stream and river damming, increased water

74 yields, soil compaction, and the deliberate construction of cattle and drainage ponds. Among these,

75 river damming is the most readily recognized driver of lentification as the Amazon Basin is the new

76 ´hydroelectric frontier´ in Brazil with 154 hydroelectric dams in operation, 21 under construction and

77 277 planned [12]. Less attention is given to the myriad of small dams found today in deforested land.

78 Yet, in the Upper Xingu Basin alone there are ~10,000 small dams, most of which were built to

79 provide water to cattle and generate electricity for local consumption [6]. Lentification is a clear

80 product of river damming as reservoirs have attributes typical of lentic systems including slow



https://doi.org/10.1101/524652


4

81 currents, high water residence times, greater depth and surface area, open canopy, high surface

82 temperatures, thermal stratification, and net accumulation of sediment and nutrients [13].

83 Deforestation could also promote lentification through increased water yields. Local

84 deforestation has little effect on precipitation but strongly reduces per-unit-area evapotranspiration

85 due to the lower height, shallower rooting depth, and lower leaf area index of grasses and crops

86 relative to trees. Thus, proportionally more rainfall becomes river discharge [2]. In the Upper Xingu

87 Basin, mean daily water yields increased four times in soybean watersheds relative to forest

88 watersheds via an increased contribution of groundwater-driven baseflow to stream discharge [7].

89 Lentification could result if deforestation-driven increased water yields promote the expansion of semi-

90 lentic riparian floodplains. Note, however, that this effect is strongly scale-dependent. When

91 deforestation reaches regional scales, the reduction in evapotranspiration feeds back into reduced

92 rainfall and therefore, reduced water yields [e.g.,14].

93 In interfluves away from streams and rivers, deforestation could also promote lentification via

94 soil compaction. In the Upper Xingu Basin soil infiltrability was markedly reduced in the conversion of

95 forests to pastures both in terms of infiltrability (from 1258 mm/h to 100 mm/h) and saturated hydraulic

96 conductivity, probably due to cattle trampling [5]. Shallow disking in the conversion of pastures to

97 soybean fields increased infiltrability (469 mm/h) but further reduced subsoil saturated hydraulic

98 conductivity [5]. Lentification could result if deforestation-driven soil compaction increases the

99 frequency in which precipitation exceeds infiltration, thereby forming ponds and puddles. An additional

100 driver of lentification in interfluvia is the deliberate construction of ´cacimbas´, artificial ponds providing

101 water to cattle and draining stormwater runoff from dirt roads.

102 Taken together, it is apparent that there are several drivers in place capable of creating novel

103 lentic water bodies in converted Amazonian landscapes, and that such novel lentic water bodies span

104 the full spectrum of the hydroperiod gradient – from ephemeral to semi-permanent to permanent

105 waterbodies – in such a way that could consistently favor quite distinct elements of lentic freshwater

106 fauna [15].

107 In this study we test, for the first time, the hypotheses (i) that land conversion in the Upper

108 Xingu Basin is associated with an increase in the availability of lentic freshwater habitats both along

109 drainage networks (by creating reservoirs and expanding stream floodplains) and in interfluves (by

110 creating ponds and puddles), and (ii) that this landscape-level lentification increases the abundance of



https://doi.org/10.1101/524652


5

111 freshwater fauna typically associated with lentic habitats. We tested these hypotheses using data on

112 habitat characterization and biological surveys of fish and amphibians in forests, pastures and

113 soybean fields in the headwaters of the Xingu River in Mato Grosso, Brazil. Fish and amphibians are

114 excellent indicators of hydrological change [16] with high complementarity in habitat use [15]: fish are

115 usually restricted to permanent waterbodies whereas most amphibians reproduce in temporary and

116 semi-permanent lentic waterbodies. Importantly, our study was conducted at the same site where the

117 hydrological mechanisms we envision as promoting lentification were demonstrated. Hayhoe et al.[7]

118 demonstrated that deforestation increases water yields but not whether this phenomenon increases

119 the area of shallow riparian floodplains. Scheffler et al.[5] demonstrated that deforestation decreases

120 soil infiltrability but not whether this phenomenon promotes the formation of ponds and puddles.

121 Finally, Macedo et al.[6] quantified the number of impoundments in the Upper Xingu Basin as a

122 whole; we here confirm their findings at our specific study location before linking habitat with

123 biodiversity change.

124

125 Methods

126

127 Study area

128 This study was conducted in the headwaters of the Xingu River in Canarana and Querência,

129 Mato Grosso, Brazil (Fig 1). Climate is Köppen´s ‘Aw’, a tropical savanna climate with distinct wet

130 (October-April) and dry (May-September) seasons. Mean annual rainfall is 1898 mm (1987-2007;

131 Grupo AMaggi, pers.comm.). Landscape is dominated by wide interfluves where plateaus 27-62 m

132 high gently slope towards stream channels. Water table is ~20-40 m deep. Soils are ustic oxisols

133 (´latossolos vermelho-amarelos distróficos´ according to the Brazilian classification) in plateaus

134 grading into aquic inceptisols (´gleysolos´) in riparian zones [7]. Original vegetation cover is a closed-

135 canopy, evergreen seasonal forest that is transitional between the ombrophilous rainforests in the

136 north and the cerrados in the south of the Xingu Basin [17].

137 Fieldwork was conducted in and around Tanguro Ranch (Fig 1), a 82,000 ha farm covered

138 with primary closed-canopy forests (44.000 ha) and soybean plantations (38,000 ha). Neighboring

139 Cristo Rei Ranch, Lírio Branco Ranch and the INCRA Settlement were predominantly covered by



https://doi.org/10.1101/524652


6

140 pastures. At the time of sampling streams in deforested watersheds were by law bordered by a 25-m

141 wide riparian forest buffer, which structural integrity was variable across the region.

142

143 >>> Fig 1. Fish sampling sites and amphibian sampling transects in Tanguro Ranch and

144 surrounding farms (Landsat 8; July 2, 2013). Dark green represents primary closed canopy forests,

145 light green represents deforested areas (soybean fields in Tanguro Ranch, pastures in Cristo Rei

146 Ranch and in the INCRA settlement). From North to South, streams sampled for fish are Tanguro C,

147 Tanguro B, Tanguro A, APP2, APP2A, and APPM; transects in forest plateaus are Seringal, AU and

148 A1. Not represented is Lírio Branco Ranch, the third replicate pasture, located 27 km to the South.

149 The inset shows the location of the study site in Southeastern Amazon. In the Amazon Basin, green

150 represents closed-canopy forests, yellow represents deforested areas [according to 18] and uncolored

151 areas represent native savannas. Savannas outside of the Amazon Basin are not represented but

152 dominate the original vegetation cover a few tens of kilometers South and East of the study site. Note

153 the position of the study site in the agricultural frontier known as ´the Amazonian Arc of

154 Deforestation´.

155

156 Testing for a land use-driven lentification of the landscape

157 Density of impoundments in streams. To test the hypothesis that land conversion is associated with

158 an increase in the density of man-made impoundments, we defined one stream transect in each of

159 three forests (along APPM, APP2 and APP2A streams in Tanguro Ranch), three pastures (INCRA

160 Settlement, Cristo Rei Ranch, Lírio Branco Ranch) and three soybean fields (Alvorada, Mutum and

161 Cascavel fields in Tanguro Ranch) (Fig 1). One pasture (INCRA Settlement) and one soybean field

162 (Alvorada) received two additional transects each for a total of 13 transects. These ´focal´ pasture and

163 soybean fields were selected for detailed physical, chemical and biological sampling surveys (to be

164 published elsewhere). We could not select a ´focal´ forest with replicated transects due to the scarcity

165 of forest trails. Each transect was 2000 m long, and positioned at least 400 m away from other

166 transects or from other land uses. Using these criteria, transect position was randomly defined among

167 all available possibilities. In February 2012 we recorded the number of impoundments per stream

168 transect.



https://doi.org/10.1101/524652


7

169 Density of ponds and puddles in interfluvia. To test the hypothesis that land conversion is associated

170 with an increase in the density of ponds and puddles in interfluvia, we as above defined one plateau

171 transect in each of three forests (Seringal, AU and A1 in Tanguro Ranch), three pastures (INCRA

172 Settlement, Cristo Rei Ranch, Lírio Branco Ranch) and three soybean fields (Alvorada, Mutum and

173 Cascavel) (Fig 1). The focal pasture (INCRA Settlement) and soybean field (Alvorada) received two

174 additional plateau transects each for a total of 13 plateau transects. In February 2012 we recorded the

175 number and dimensions of each water body > 5 cm deep that could be sighted from the transects,

176 while estimating the maximum distance to each side of the transect in which a waterbody would be

177 visible if present.

178 Area of shallow, semi-lentic riparian floodplains. To test the hypothesis that land conversion is

179 associated with an expansion in the area of riparian floodplains, we characterized channel

180 morphometry of three first-order streams in forest (APP 2, APP2A, APPM) and three in soybean fields

181 (Tanguro A, B, and C; Fig 1). No streams were characterized in pastures. Because the putative

182 mechanism for floodplain expansion is increased baseflow, more clearly evidenced by dry season

183 water yields (Heyhoe et al. 2011), channel morphometry was characterized in October 2013. Stream

184 channel characterization was conducted in three 50-m sections at 0, 1000 and 2000 m from the

185 headwaters. For streams in converted land, we also characterized one 50-m section in one reservoir

186 contained within the same distance. Each 50-m section was crossed by 6 transects used to measure

187 channel width and water depths. We used the cut off depth of 10 cm for estimating area of shallow

188 water because this is where our indicator species of lentic riparian waters is most frequent (see

189 below). Note that sampling design for quantifying impoundments and ponds and puddles, and area of

190 shallow riparian floodplains, differed due to constraints imposed by the corresponding biodiversity

191 surveys (below): amphibian calling surveys used in the former require considerably less effort and

192 therefore permit greater replication than fish surveys used in the latter.

193

194 Testing for a land use-driven lentification of the freshwater fauna

195 Consequences of impoundment conditions to fish assemblages. We compared the ichthyofaunas of

196 lotic and lentic (i.e. reservoir) sections in the three soybean streams described in ´Area of shallow

197 riparian floodplains´. Dipnetting was the only methodology appropriate for fish sampling in the narrow,

198 shallow streams rich in coarse woody debris; low water conductivity prevented electrofishing, and gill



https://doi.org/10.1101/524652


8

199 and seine netting were impossible. We employed a standardized sampling effort of 100 dipnetting

200 person*minutes in each stream section, with a 30-cm diameter dipnet. Because dipnetting is only

201 effective in water < 1m, fish assemblage characterization in impoundments was complemented with

202 three hauls of a seining net (9 m long, 2 m high, 5-mm mesh) and an overnight set of six nylon gillnets

203 (20 m long, 2 m high; and 3, 4, 5, 6, 7, 8 cm distance between diagonally opposite knots).

204 Consequences of shallow riparian floodplains to fish populations. We assessed the relationship

205 between area of shallow riparian floodplains and the abundance of lentic fishes by sampling

206 Melanorivulus megaroni (Cyprinodontiformes, Rivulidae) in water < 10 cm deep in all lotic and lentic

207 stream sections described in ´Area of shallow riparian floodplains ´. M. megaroni was selected as an

208 indicator of semi-lentic riparian habitat for occurring almost exclusively in shallow margins of streams,

209 swamps and lakes [19]. Preliminary surveys indicated that in marshy floodplains and streamside

210 ponds M. megaroni is the dominant fish species, occurring in flooded areas so shallow that one can

211 barely see free-standing water amid the leaf-litter (see also 20 for data on habitat use by rivulids in

212 similar habitats). All fish were euthanized with benzocaine, preserved in formalin 10%, and stored in

213 ethanol 70%.

214 Consequences of land use to amphibian assemblages. We conducted calling surveys at six sampling

215 stations (0, 400, 800, 1200, 1600 and 2000 m) along each of the 26 transects described above. Each

216 transect was sampled between 19:30h and 01:00 (+ 30 min) on three dates randomly distributed over

217 January and February 2012. Upon arrival at a sampling station we recorded, for a period of three

218 minutes, calling amphibian species in six categories of abundance (0=no individuals; I=1-2; II=3-5;

219 III=6-10; IV=11-20; V=21-50; VI=50+ individuals). For each sampling station we defined richness as

220 the number of species recorded across all sampling dates, and the abundance of each species as the

221 highest category of abundance recorded. We calculated a conservative index of aggregated (i.e. total)

222 amphibian abundance in each sampling station as the summation of the number of species in a

223 category of abundance multiplied by its lower abundance bound [(Nspp in category I * 1) + (Nspp in

224 category II * 3) + …].

225

226 Data analysis

227 All statistical analyses followed the rationale of testing a given effect by comparing nested

228 linear models [21]. For instance, to test the hypothesis that the expected value of a given response



https://doi.org/10.1101/524652


9

229 variable differs among land uses (besides all other putative effects), we compared with a log-

230 likelihood ratio test (LRT) two linear models that differed only by the inclusion of land use as a fixed

231 effect. An increase in fit was gauged by the Deviance (D), which is twice the likelihood ratio between

232 the two models, and which converges to a Chi-square distribution under the null hypothesis of no

233 improvement in fit [22]. We used Poisson or Negative Binomial models for count data that had small

234 values, and Gaussian models otherwise. In many cases we used mixed-effects models [23] to take

235 into account the dependence among sampling units (e.g. sampling stations within the same transect).

236 Compliance of fitted models to their assumptions (e.g. constant variance and normality at the link

237 scale, residuals with zero mean and without non-linear trends) where checked from the analysis of the

238 appropriate residuals. All analyses were done in R 3.3.2 [24], with the additional packages lme4 [25]

239 and DHARMa [26]. Expected values and associated standard errors for mixed-effect models were

240 estimated by model-based parametric bootstrap with the function bootMer from the lme4 package

241 [25].

242 We tested a lentification effect of land use by comparing LRT models that fit different

243 expected values of ponds and puddles/transect, numbers of dams/transect, or areas of shallow

244 riparian floodplain/stream section, with null models that fit the same expected value in all transects or

245 stream sections. We used Poisson glm in the first two cases and a mixed-effect Gaussian glm with

246 stream as a random effect in the third case, to take into account the variance shared by sections

247 within the same stream.

248 Restricting our analysis only to the dominant fish species in the assemblage (relative

249 abundance >2% [27]), we used LRT to compare mixed-effect Gaussian glm fitted to the log of catch

250 per-unit-effort (CPUE) of each species in each stream section. The first model had only the additive

251 fixed effects of fish species and habitat type (i.e. lotic versus lentic); the alternative model had an

252 additional species-by-habitat interaction term. To test the fixed effects of area of shallow riparian

253 floodplain and land use on M. megaroni abundance we used a Negative Binomial glmm because

254 preliminary analyses indicated aggregated distribution of fish counts over sampling units. In both

255 cases stream was a random effect.

256 We evaluated the effects of lentification on richness and aggregated abundance of anurans

257 recorded at each sampling station. For streamside transects we used LRT to compare mixed-effects

258 models with no fixed effect, the effect of land use, and the additive effects of land use and density of



https://doi.org/10.1101/524652


10

259 dams in a 800-m radius. For plateau transects we compared mixed-effects models with no fixed

260 effect, the effect of land use, and the additive effects of land use and density of puddles in the

261 transect. We used Poisson models for richness and Gaussian models for the logarithm of the

262 abundance index. All models included transect as a random effect to take into account the variance

263 shared by stations within the same transect.

264

265 Results

266 Testing for a land use-driven lentification of the landscape: freshwater habitat change. Number of

267 dams along stream transects increased from zero in forests to a median of 0.4 in pastures and

268 soybean fields (D=11.3, 2 df, p=0.035; Fig 2A). Area of shallow riparian floodplains roughly doubled in

269 soybean watersheds (average + 1SD = 118 + 70 m2, range 25-246 m2) relative to forested

270 watersheds (58 + 53 m2, range 0-142 m2; D=4.3, 1 df, p=0.037; Fig 2B). Density of ponds and

271 puddles increased from zero in forest plateaus to a median of 3/ha in pasture plateaus and 2/ha in

272 soybean plateaus (D=54.9, 1 df, p<0.001; Fig 2C; see S1 Figure and S2 Figure for pictures).

273

274 >>> Fig 2. Land use-driven lentification of Southern Amazonian landscapes. (a) Density of dams

275 along streamside transects (b) Area of shallow water in riparian floodplains and (c) Density of ponds

276 and puddles along plateau transects. In (b) note that no streams were sampled in pasture watersheds

277 but that in (a) and (c) no reservoirs and puddles were found along streams and in plateaus,

278 respectively, covered by primary forest.

279

280 Consequences of reservoir formation to fish assemblages. Dominant species in the fish assemblage

281 were Melanorivulus megaroni (Rivulidae; average relative abundance across all 12 stream sections

282 48.3%), Pyrrhulina australis (Lebiasinidae; 14.6%), Aequidens michaeli (Cichlidae; 11.8%), and the

283 characids Hyphessobrycon mutabilis (8.0%), Astyanax multidens (7.8%) and Moenkhausia phaeonota

284 (2.9%).

285 These six species, which together comprised 93.8% of all fish individuals dipnetted in the

286 three converted streams, responded strongly to reservoir conditions. CPUE was higher in reservoirs

287 for Aequidens michaeli, Pyrrhulina australis and Astyanax multidens, but lower for the three remaining

288 species (Fig 3). A significant habitat-by-species interaction term (D=20.5, 5 df, p=0.001) confirmed



https://doi.org/10.1101/524652


11

289 that some fish species were favored while others disfavored by stream impoundment. In fact,

290 Hyphessobrycon mutabilis and Moenkhausia phaeonota were entirely absent from reservoirs, as were

291 another ten less common species (< 0.5% in relative abundance). Nine fish species were found

292 exclusively in reservoirs (S1 Table).

293

294 >>> Fig 3. Effect of habitat (stream vs reservoir) on the abundances of the six dominant fish species

295 in streams of deforested (i.e. soybean) watersheds. Symbols and vertical bars represent, respectively,

296 observed abundance means (ln CPUE + 1) + 1 standard errors of the estimated values by a linear

297 mixed effect Gaussian model, accounting for all sources of variation (fixed and random effects).

298

299 Consequences of floodplain expansion to the abundance of Melanorivulus megaroni. M. megaroni

300 abundance was positively related to the area of shallow riparian floodplains (D=17.2 , 1 df, p<0.001;

301 Fig 4).

302

303 >>> Fig 4. Abundance of Melanorivulus megaroni as a function of the area of shallow water in each

304 50-m stream section (green solid circles: streams in forested watersheds, red open triangles: streams

305 and reservoirs in deforested watersheds). The line is that predicted by a negative binomial

306 generalized model with random effects due to streams. The grey area shows + 1 the standard errors

307 of the estimated values by the model, accounting for all sources of variation (fixed and random

308 effects).

309

310 Consequences of land use to amphibian assemblages along streams. Transects in deforested

311 watersheds had more than twice as many calling species (20 species) than transects in forested

312 watersheds (9 species). Note, however, that rarefaction curves asymptoted in pastures and to a

313 lesser degree in soybean fields, but not in forests (Fig 5A). Because more singletons and doubletons

314 were found in forest transects, Chao2 estimators of species richness were 28.8 + 16.3 for forests

315 (mean + 1 SD) but 20.3 + 0.9 for pastures and 21.6 + 2.1 for soybean fields.

316 No species was heard exclusively along streams in forest watersheds although the only

317 sightings of Phyllomedusa vaillanti (Hylidae) and Leptodactylus pentadactylus (Leptodactylidae)



https://doi.org/10.1101/524652


12

318 occurred in those streams. By contrast, 20 species were heard exclusively along streams in

319 deforested watersheds (S2 Table).

320

321 >>> Fig 5. Rarefaction curves of amphibian richness along (A) streamside transects and (B) plateau

322 transects in forests, pastures and soybean fields. Symbols represent estimated means + 1 SD. Note

323 that there were three streamside and three plateau transects in forests (therefore 18 call sampling

324 stations) but five streamside and five plateau transects in pastures and soybean fields (therefore 30

325 call sampling stations). Note also that in B no amphibian calls were recorded in forests, where ponds

326 and puddles are absent and where water bodies are invariably associated with streams and their

327 floodplains.

328

329 Amphibian species richness per sampling station increased in converted land relative to

330 forests (D=12.4, 2 df, p=0.002), and with number of dams within an 800-m radius (D=7.3, 1 df,

331 p=0.007). Predicted values confirm a much higher species richness in pastures and soybean fields

332 than in forests (Fig 6A). Amphibian aggregate abundances per sampling station also increased in

333 converted land relative to forests (D=14.2, 2 df, p=0.001), but the effect of number of dams was

334 weaker (D=3.5, 1 df, p=0.06). Still, we used the model with both predictors to estimate the expected

335 values, which again confirmed higher abundances in pastures and soybean fields relative to forests

336 (Fig 6B).

337

338 >>> Fig 6 Amphibian richness (upper row) and aggregate abundance (lower row) per sampling

339 station as a function of the density of dams along streams (left column) and puddles in plateaus (right

340 column) in forests (solid green circles), pastures (open blue squares) and soybean fields (open red

341 triangles). Curves represent trends (+ 1 standard errors of estimated values) predicted by the best

342 supported model (generalized linear models with random effects of the transects). Note that in

343 plateaus only land use had a significant effect on amphibian abundance and richness, which is

344 depicted by the horizontal lines of predicted values and standard errors. In primary forest plateau

345 transects no amphibian species were recorded, and the model was fitted only to data gathered in

346 soybean and pastures. Points are jittered along x and y axes to improve data visualization.

347



https://doi.org/10.1101/524652


13

348 Consequences of land use to amphibian assemblages in plateaus. Land use had an even stronger

349 effect on amphibian assemblages in plateaus. Transects in forest plateaus were entirely devoid of

350 calling amphibians. By strong contrast, pastures and soybean fields had respectively 12 and 6 calling

351 species (S2 Table, Fig 5B). Rarefaction curves asymptoted in pastures but not in soybean fields;

352 Chao2 estimators of species richness therefore coincided with measured richness in pastures (12.0 +

353 0 species) but not in soybean fields (8.9 + 4.3 species).

354 Because forest plateaus were devoid of calling amphibians, we could not fit models to predict

355 amphibian richness or abundances for all three land use types considered jointly. Considering

356 converted plateaus only, both amphibian richness (D=15.3, 1 df, p<0.001; Fig 6C) and aggregated

357 abundance per sampling station (D=12.2, 1 df, p<0.001; Fig 6D) was higher in pastures than in

358 soybean fields. In both cases predicted values were higher in pastures than in soybean fields. No

359 improvement of fit was obtained by including pond and puddle density.

360

361 Discussion

362 This study provides evidence, for the first time, that deforestation promotes an increase in the

363 availability of lentic freshwater habitats in Southern Amazonian landscapes. Streams in converted

364 watersheds had numerous impoundments, and riparian floodplains twice as wide than streams in

365 preserved watersheds. At the same time, converted interfluvia had numerous ponds and puddles, in a

366 striking contrast to the complete absence of waterbodies in forested interfluvia. Importantly,

367 lentification of streams and interfluvia was accompanied by strong responses by fish and amphibian

368 assemblages.

369 Reservoir conditions favored half of the six dominant fish species, and disfavored the other

370 half. The characid Astyanax multidens and the cichlid Aequidens michaeli were 8 and 2.3-fold more

371 abundant in reservoirs than in lotic reaches; several other less common cichlids were also

372 predominantly or exclusively found in reservoirs, as were the trahira Hoplias malabaricus and the

373 curimatid Steindachnerina sp. At the other extreme, not a single individual of characids

374 Hyphessobrycon mutabilis and Moenkhausia phaeonota, otherwise common in stream reaches, was

375 found in reservoirs. These patterns are consistent with large-scale studies showing that cichlids,

376 characids and other characiforms, but also erithrynids, serrasalmids, loricariids and curimatids that



https://doi.org/10.1101/524652


14

377 inhabit more lentic habitats such as backwaters, shallow river margins and lakes, are sometimes

378 favored by river damming [28].

379 Expanded stream floodplains, in turn, had a strong positive effect on the abundance of

380 Melanorivulus megaroni, a fish species associated with shallow waters (>90% of individuals were

381 sampled in water <10 cm deep) and streamside pools. Other fishes could be influenced by floodplain

382 expansion as well given that many small-sized species, or larvae and juveniles of larger species, rely

383 on these habitats as nurseries and shelters against predation and water dragging [29].

384 Conversion of forests to pastures and soybean fields had a surprising positive effect on

385 streamside amphibian abundance and richness. Both response variables were associated with

386 number of dams around calling sampling stations. Thus, landscape lentification via reservoir formation

387 is a plausible contributor to this phenomenon, although other hydrological and non-hydrological

388 changes associated with deforestation, including the expansion of stream floodplains (not quantified

389 along surveyed streamside transects) and canopy opening, could also be involved.

390 Even more striking was the influence of land conversion in interfluvia on amphibian richness

391 and abundance. No single amphibian species was heard in forested plateaus, as opposed to 12 in

392 pastures and six in soybean fields. Because calling involves not only benefits (mate attraction and

393 selection) but also costs (energetic costs, increased risk of predation and parasitism), males usually

394 call in, at or close to their breeding sites [30]. Therefore, the lack of water bodies is a sufficient

395 explanation for the absence of calling amphibians in forested plateaus, as 35 of 38 amphibian species

396 in the region depend on water for egg laying and/or premetamorphic development [31]. We suspect

397 the lack of a statistical effect of density of ponds and puddles on amphibian richness and abundance

398 results primarily from the need to exclude forested plateaus from the analysis, and secondarily from a

399 mismatch in the scale of habitat assessment (ponds that could be seen from transect, i.e., tens of

400 meters) and that of faunal assessment (frogs that could be heard from transect, i.e., hundreds of

401 meters).

402 Contrary to streams, where amphibian richness and abundance were similar in pastures and

403 soybean fields, in interfluvia pastures had twice as many amphibians as soybean fields. At least two

404 factors could be involved. First, several of the waterbodies in pastures were cattle ponds, which,

405 contrary to roadside puddles, are constructed to effectively hold water – i.e., they are deeper and

406 compacted with tractors. As a consequence, waterbodies in the focal pasture held water for one



https://doi.org/10.1101/524652


15

407 month longer than in the focal soybean field (personal observation), a difference in hydroperiod of

408 considerable biological significance considering that a primary cause of mortality in temporary ponds

409 is dessication [15]. Second, environmental conditions in plateaus - effectively contrasting primary

410 forests, pastures and soybean fields - are much harsher than along streams, at the time of sampling

411 protected by law by a 25-m wide vegetated buffer zone. Specifically, the intensive management of

412 soybean plantations inflicts strong, periodical disturbance to the fauna that inhabits, visits or traverses

413 interfluvia. Conversion of pastures to soybean fields is made by clearing with fire, tilling and liming,

414 whereas plantation management annually involves sowing, applying fertilizers (~300 kg NPK/ha) and

415 pesticides (13.7 kg/ha including 39 active ingredients) and harvesting [3; 32]. Considering that

416 puddles in soybean fields are subject to direct pesticide overspraying and runoff, it is surprising that

417 there are amphibian species colonizing ponds in plantations – although, we do not know whether

418 these are viable habitats for metamorph production or simply amphibian population sinks (as is the

419 case for aquatic insects; see 33).

420 In summary, the creation of new reproductive habitat is a plausible hypothesis for the

421 increased amphibian abundance in converted land. The increase in species richness is, by contrast,

422 unexpected. Differences in observed gamma diversity are likely artifacts of abundance; Chao

423 estimators predict 29 species in forests, less than the 38 recorded [31] but more than the 22 predicted

424 in converted land. Differences in alpha diversity, in turn, do not appear to be artifacts and could be

425 due to the particular biogeographical scenario of our study site and/or the spread of generalistic open

426 area species with deforestation (see also 34). Our study site is at the edge of Amazonian closed-

427 canopy forests; the savanna line is as close as ~20km south and it is likely that in the Pleistocene this

428 area has alternated open-canopy/closed-canopy phytophysiognomies [35]. The observation that

429 several species found in closed-canopy forests are more typical of open areas (Leptodactylus

430 labirinthicus, L. mystaceus, Hypsiboas albopunctatus, Rhinella schneideri) is a testimony for this

431 mixed amphibian assemblage. On top of that, converted streamside transects have the addition of at

432 least seven species exclusive of open areas (Leptodactylus fuscus, Eupemphix nattereri,

433 Physalaemus cuvieri, P. centralis, Pseudopaludicola mystacalis, Scinax fuscovarius and S.

434 fuscomarginatus). Converted watersheds could therefore be housing an enriched amphibian fauna

435 through the mixture of closed-canopy species of Amazonian origin (e.g., Osteocephalus taurinus,

436 Phyllomedusa vaillanti, Pristimantis fenestratus), open-area cerrado species capable of colonizing



https://doi.org/10.1101/524652


16

437 closed-canopy forests, and generalistic open-area cerrado species that accompany the line of

438 deforestation and are favored by landscape lentification. Of course, at this point it is impossible to

439 know whether the increased species richness in converted land is a stable endpoint or a product of

440 transient dynamics given the recent deforestation in the area (< 40 years), the even more recent

441 decrease of matrix quality via conversion of pastures to soybean fields (< 10 years), and the presence

442 of relatively large fragments of primary forest in the region and riparian buffer zones (Fig 1).

443

444 Broader geographic consequences

445 Although our research is local, it is tempting to speculate that the lentification of freshwater

446 systems and their biota could be occurring in many locations of deforested Amazonia. This is because

447 all proposed drivers of lentification are quite general. We reviewed the available evidence linking land

448 conversion and the density of dams, soil compaction and stream discharges in the Amazon Basin

449 (and its sub-basins Xingu, Tapajós, Madeira, Solimões, Negro and Amazonas) and surrounding

450 basins (Tocantins, Marajó and the NW Coastal South Atlantic to the Southeast; Orinoco and Coastal

451 North Atlantic to the North) (Table 1). A total of 179 hydroelectric dams are currently operating or

452 under construction, and another 275 planned, in four basins (Tocantins, Amazon, Orinoco, Coastal

453 North Atlantic), including all six Amazon sub-basins [12; 35]. Dramatically increased densities of small

454 impoundments in deforested land were reported in the Xingu6 and the Amazonas sub-basins [36].

455 Soil compaction, as indicated by increased bulk soil density or soil penetration resistance,

456 and/or by decreased soil porosity, infiltrability or hydraulic conductivity, were consistently observed in

457 all 19 studies we could find conducted in four basins (Coastal South Atlantic, Tocantins, Amazon,

458 Orinoco), including five of six Amazon sub-basins (Xingu, Tapajós, Madeira, Solimões and Negro)

459 [37; 38; 39; 40; 41; 42; 43; 44; 45; 46; 47; 48; 5; 49; 50; 51; 52; 53; 54; 55] (Table 1). Importantly,

460 soil compaction was observed in every type of land use including secondary and logged forests,

461 deforested land, pastures, and in soybean, maize, oil palm, coffee, and teak plantations (Table 1).

462 Among them, it is precisely in pastures – by far the prevailing land use in converted Amazonia - where

463 soil compaction is most intense. Relative to forests, infiltrability in pastures decreased from 8 to 162

464 times depending on factors such as soil type, pasture age, and cattle stocking densities. Even though

465 soil compaction only results in ponding if precipitation exceeds infiltration rates, increased overland

466 flows in pastures are commonly reported. In eastern Amazonia rainfall rarely if ever exceeds



https://doi.org/10.1101/524652


17

467 Table 1. Land-use driven changes in landscape hydrological properties - how widespread are drivers of lentification in deforested Amazonia? Arrows represent

468 an increasing trend (↑), a decreasing trend (↓), both increasing and decreasing trends (↕) and no trend (↔).

Basinsa

Amazon Basin

Response Variable Unit

dire

ctio

n of

cha

nge

cons

iste

nt w

ith le

ntifi

catio

n

Coa

stal

Sou

th A

tlant

ic

Toca

ntin

s

Mar

ajó

Xing

u

Tapa

jós

Mad

eira

Solim

ões

Neg

ro

Amaz

onas

Orin

oco

Coa

stal

Nor

th A

tlant

ic

Land use pattern and reference

Dams: Hydroelectric dams

Dams in operation N ↑ 56 6 33 43 11 1 4 2 2 [12, 35]

Dams in construction N ↑ 2 1 6 8 4 0 0 ? 0 [12, 35]

Dams planned N ↑ 101 2 73 43 47 1 8 ? 0 [12, 35]

Overall supporting evidence ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Dams: Small dams

Density of small dams N/ha↑ ↑ ↑

b,c Xingu: FOR<SOY,PAS; CER<SOY,PAS [6]; Amazonas:

FOR<DEFOR [36]

Overall supporting evidence ✓ ✓



https://doi.org/10.1101/524652


18

Soil compactiond

Bulk soil density

g/cm3

or ~

↑ ↑ ↑ ↑ ↕ ↑↑

↔↑

Coastal South Atlantic: SFOR<PAS [48], SFOR<PAS,GRAIN [49],

FOR<PAS [42e,50e]; Tocantins: CER<PAS [55]; Tapajós: FOR<LOG

[41]; Madeira: FOR<PAS, FOR>PAS [two chronosequences; 39e],

FOR<SFOR,PAS,COFFEE [54], Solimões: FOR<PAS [44e];

SFOR<DEFOR,ACROP,ALLEY,PAS [38], FOR<PALM [53],

FOR<DEFOR [37]; Negro: FOR<PAS,MAIZE,MCGRAIN; FOR=TPAS

[47], FOR<PAS; CER=PAS [51]; Orinoco: FOR<PAS [40]

Soil total porosity

m3/m3

or ~

↓ ↓ ↓ ↓ ↓↓

↔↓

Coastal South Atlantic: FOR>PAS [42e,50e]; SFOR>PAS,GRAIN [49];

Tocantins: CER>PAS [55]; Madeira: FOR>SFOR,COFFEE,PAS [54];

Solimões: FOR>PAS [44e], FOR>DEFOR [37];Negro:

FOR>PAS,MAIZE; FOR=TPAS,MCGRAIN [47], FOR>PAS;

CER=PAS [51]; Orinoco: FOR>PAS [40]

Soil penetration

resistance Mpa

↑ ↑↑

↔↑

Solimões: SFOR<PAS [38]; Negro: FOR<PAS,TPAS,MAIZE,

FOR=MCGRAIN [47], FOR=PAS; CER=PAS [51]; Orinoco:

FOR<PAS [40]

Infiltrability mm/h

↓ ↓↓

↔↓ ↓

Xingu: FOR>PAS,SOY [5]; Madeira: FOR>SFOR,TEAK, PAS;

FOR=DESFOR [43], FOR>PAS [46]; Solimões: SFOR>PAS [38];

Orinoco: FOR>PAS [40e]



https://doi.org/10.1101/524652


19

Hydraulic conductivity

(Ksat) mm/h

↓ ↓ ↓↓

↔↓

Coastal South Atlantic: FOR>SFOR,PAS [42e]; Xingu:

FOR>PAS,SOY [5]; Madeira: FOR>TEAK, PAS;

FOR=SFOR,DESFOR [43], FOR>PAS [46]; Solimões: FOR>PAS

[44e], SFOR>ACROP,ALLEY [38]

Overall supporting evidence✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Land-use driven changes in water level in small catchments

Discharge or water yield

mm/hr

or ~

↑ ↑ ↑↑

↔↑

Coastal South Atlantic: FOR<PAS [42e]; Xingu: FOR<SOY [7];

Madeira: FOR<PAS [46e, 56, 58e], FOR=PAS [59]; Amazonas:

FOR<PAS [57e]

Baseflow

mm/hr

or ~↑ ↑ ↑ ↑

Xingu: FOR<SOY [7]; Madeira: FOR<PAS [45e,58e]; Amazonas:

FOR<PAS [57e]

Runoff coefficient

(discharge:precipitation) %↑ ↑ ↑ ↑ ↑

Coastal South Atlantic: FOR<PAS [42e], Xingu: FOR<SOY [7];

Madeira: FOR<PAS [58e], Amazonas: FOR<PAS [57e]

Overall supporting evidence✓ ✓ ✓ ✓

469



https://doi.org/10.1101/524652


20

470 aBasins (Óttobasins Level 1´) and sub-basins (Óttobasins Level 2´) are according to Agência Nacional das Águas [55] and presented southwest from eastern Amazonia,

471 and then northeast from western Amazonia. ANA´s Tocantins Basin includes the Araguaia. Ćoastal South Atlantic´ actually refers to the northernmost sub-basin (ANA´s

472 Level 2 Ottobasin 71, Coastal Occidental Northeast) of Ottobasin Level 1 Coastal South Atlantic.

473 bStatistical results are as reported in the original literature, and always in comparison with the reference condition (i.e., differences among alternative land uses are not

474 reported).

475 cFOR=forest, CER=cerrado, SFOR=secondary forest, LOG=logged forest, DEFOR=deforested, DESFOR=recently manually cleared secondary forest, PAS=pasture,

476 TPAS=tilled pasture, TEAK=teak, COFFEE=coffee, MAIZE=maize, SOY=soybean, PALM=oil palm, ACROP=annual crop, ALLEY=alley crop, GRAIN=multiple grains,

477 MCGRAIN=grains in minimum cultivation system.

478 dResponse variables testing for soil compaction report the most superficial soil layers sampled. When there are time series (eg, pasture after 5, 10, 15 years), we report

479 significant results for the time period with strongest trend. Note that soil compaction leads to ponding if and when precipitation exceeds infiltration rates only.

480 eno statistical test performed



https://doi.org/10.1101/524652


21

481 infiltration capacities in forests; in pastures, by contrast, around ¾ of storm-averaged rainfall

482 intensities exceed infiltration capacities [42].

483 Finally, because deforestation strongly reduces evapotranspiration, a larger fraction of rainfall

484 is exported from the watershed as streamflow. This pattern, commonly observed in small catchment

485 studies in both temperate and tropical countries [56], was recorded in two basins (Coastal South

486 Atlantic and Amazon), and three of the Amazon sub-basins (Xingu, Madeira and Amazonas) [42; 56;

487 57; 58; 45; 7]. Conversion of forests to pastures and soybean fields increased water yields, discharge

488 and baseflow, and discharge-to-precipitation ratios (Table 1). Interestingly, one study that failed to

489 observe a linkage between land conversion and discharge nevertheless observed stream structural

490 changes consistent with lentification. Deegan et al. [59] found that in the conversion of forests to

491 pastures the pool-and-run structure was replaced by a narrow run of open water bordered by wide

492 marshy areas invaded by riparian grasses (mean wetted width was 5X greater in pastures than

493 forests, N=18 streams). Ground-truthed remote sensing permitted the authors to infer that the infilling

494 of stream channels by riparian grasses, which slows down water velocities and promotes the transient

495 storage of water, occurs in almost all of the small pasture streams in the Ji-Paraná Basin, a 74,000

496 square kilometer sub-watershed of the Madeira [59].

497

498 Concluding remarks

499 In this study we demonstrate that extensive hydrological change occurs very early in the

500 process of frontier settlement. The deliberate construction of reservoirs, cattle and drainage ponds

501 adds to the unintentional multiplication of puddles and expansion of floodplains to modify existing

502 habitat, or even create new habitat, for freshwater organisms. The phenomenon of freshwater

503 lentification we here propose, if indeed shown to be widespread in deforested Amazonia, could have

504 significant consequences to biodiversity and human health; the dramatic upsurge in malaria and

505 schistosomiasis following deforestation, for example, is partly a consequence of increased availability

506 of lentic freshwater habitats in the landscape [60,61].

507 It is reasonable to expect that other drivers of lentification not considered in this study could

508 be currently operating or become important in the future. For example, reservoir management may

509 promote the rise of the water table [62], potentially increasing the area of river floodplains. By

510 contrast, unsustainable water extraction, large scale deforestation [14] or climate change could all



https://doi.org/10.1101/524652


22

511 contribute to lowering water tables, reducing floodplain widths and causing flow intermittency and

512 increased stream temporality. Climate change and anomalous climatic events, in particular, could

513 strongly interact with other phenomena described in this article to magnify the intensity of freshwater

514 lentification. Such interactions could be particularly strong in the Amazonian Arc of Deforestation,

515 which is not only a region of widespread land conversion – and therefore widespread river damming,

516 change in stream discharges, soil compaction and creation of farm ponds - but also predicted to be

517 highly vulnerable to savannization [14].

518

519 Acknowledgments

520 We thank Grupo Amaggi (Fazenda Tanguro) and neighboring farmers (Assentamento do INCRA,

521 Fazenda Cristo Rei, Fazenda Lírio Branco) for granting access to their lands, and the Instituto de

522 Pesquisa Ambiental da Amazônia – IPAM (especially O. Carvalho Junior, W. Rocha, P. Brando and

523 M. Macedo), Grupo Amaggi (especially W. De Ré) and M. Rosa (Secretaria da Agricultura e Meio

524 Ambiente de Canarana) for logistical support. G. Cordeiro Junior, G. Cordeiro, V. Dimitrov, I. Calado,

525 D. Nunes, S. Rocha e R. Quintino provided invaluable help in fieldwork, and I. Calado and A. C.

526 Machado in labwork. P. Lefebvre kindly produced the map and M. Leite helped in the statistical

527 analysis. ICMBio provided collection permits (ICMBio 17559). We thank FAPESP (2008/57939-9,

528 2010/52321-7 to L.S.), CNPq (563075/2010-4 to L.S.) and the International Finance Corporation -

529 Biodiversity and Agricultural Commodities Program for funding this research (to L.S.and B.G.),

530 FAPESP (2011/20458-6 to P.I., 2010/19427-6 to D.D.B.N.) and CNPq for fellowships (Research

531 Grant to P.I.P.), and the University of Veterinary Medicine of Vienna for a travel grant (B.G.).

532

533 References

534 1. Allan, J.D. 2004. Landscapes and riverscapes: the influence of land use on stream

535 ecosystems. Annual Review of Ecolgy, Evolution and Systematics 357: 257-284.

536 2. Wohl, E., A. Barros, N. Brunsell, N.A. Chapell, M. Coe, T. Giambelluca, et al. 2012. The

537 hydrology of the humid tropics. Nature Climate Change 2: 655-662.

538 3. Schiesari, L., A. Waichman, T. Brock, C. Adams, and B. Grillitsch. 2013. Pesticide use and

539 biodiversity conservation in the Amazonian agricultural frontier. Philosophical Transactions of

540 the Royal Society 368:20120378.



https://doi.org/10.1101/524652


23

541 4. Nepstad, D.C., C.M. Stickler, and O.T. Almeida. 2006. Globalization of the Amazonian soy

542 and beef industries: opportunities for conservation. Conservation Biology 20:1595-1603.

543 5. Scheffler, R., C. Neill, A. V. Krusche, and H. Elsenbeer. 2011. Soil hydraulic response to land-

544 use change associated with the recent soybean expansion at the Amazon agricultural frontier.

545 Agriculture, Ecosystems and Environment 144:281–289.

546 6. Macedo, M. N., M. T. Coe, R. DeFries, M. Uriarte, P. M. Brando, C. Neill, et al. 2013. Land-

547 use-driven stream warming in southeastern Amazonia. Philosophical Transactions of the

548 Royal Society 368:1–9.

549 7. Hayhoe, S. J., C. Neill, S. Porder, R. McHorney, P. Lefebvre, M. T. Coe, et al. 2011.

550 Conversion to soy on the Amazonian agricultural frontier increases stream flow without

551 affecting storm flow dynamics. Global Change Biology 17:1821–1833.

552 8. Neill, C., M.T. Coe, S.H. Riskin, A.V. Krusche, H. Elsenbeer, M.N. Macedo, et al. 2013.

553 Watershed responses to Amazon soya bean cropland expansion and

554 intensification. Philosophical Transactions of the Royal Society 368:20120425

555 9. Drago, E. C., I. E. de Drago, O. B. Oliveros, and A. R. Paira. 2003. Aquatic habitats, fish and

556 invertebrate assemblages of the middle Paraná river. Amazoniana 17:291–341.

557 10. Sabater, S. 2008. Alterations of the global water cycle and their effects on river structure,

558 function and services. Freshwater Reviews 1:75–88.

559 11. Sabater, S., X. Timoner, C. Borrego, and V. Acuña. 2016. Stream biofilm responses to flow

560 intermittency: from cells to ecosystems. Frontiers in Environmental Science 4:1–14.

561 12. Castello, L., D. G. McGrath, L. L. Hess, M. T. Coe, P. A. Lefebvre, P. Petry, et al. 2013. The

562 vulnerability of Amazon freshwater ecosystems. Conservation Letters 6:217–229.

563 13. Henry, R. 1999. Ecologia de reservatórios: estrutura, função e aspectos sociais.

564 FAPESP/FUNDIBIO, São Paulo, SP.

565 14. Nobre, C. A., P. J. Sellers, and J. Shukla. 1991. Amazonian deforestation and regional

566 climate change. Journal of Climate 4:957–988.

567 15. Wellborn, G. A., D. K. Skelly, and E. E. Werner. 1996. Mechanisms creating community

568 structure across a freshwater habitat gradient. Annual Review of Ecology and Systematics

569 27:337–363.



https://doi.org/10.1101/524652


24

570 16. Hering, D., R. K. Johnson, and A. Buffagni. 2006. Linking organism groups – major results

571 and conclusions from the STAR project. Hydrobiologia 566:109–113.

572 17. Velasquez, C., H. Q. Alves, and P. Bernasconi. 2010. Fique por dentro: a Bacia do Rio Xingu

573 em Mato Grosso. Instituto Socioambiental, São Paulo, SP.

574 18. Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, et al.

575 2013. High-resolution global maps of 21 st-century forest cover change. Science 342:850–

576 853.

577 19. Costa, W. J. E. M. 2011. Phylogenetic position and taxonomic status of Anablepsoides,

578 Atlantirivulus, Cynodonichthys, Laimosemion and Melanorivulus (Cyprinodontiformes:

579 Rivulidae). Ichthyological Exploration of Freshwaters 22:233–249.

580 20. Espírito-Santo, H.M.V., M.A. Rodriguez, and J. Zuanon. 2013. Reproductive strategies of

581 Amazonian stream fishes and their fine-scale use of habitat are ordered along a hydrological

582 gradient. Freshwater Biology 58: 2494-2504.

583 21. Crawley, M. J. 2012. The R book. John Wiley & Sons, Chichester, UK.

584 22. McCullagh, P, and J. A. Nelder. 1989. Generalized Linear Models. Monographs on Statistics

585 and Applied Probability 37. Chapman & Hall/CRC, New York, NY.

586 23. Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H. Stevens, et al.

587 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in

588 Ecology & Evolution 24:127–135.

589 24. R Core Team. 2016. R: A language and environment for statistical computing. R Foundation

590 for Statistical Computing, Vienna, Austria.

591 25. Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models

592 Using lme4. Journal of Statistical Software 67:1–48.

593 26. Hartig, F. 2016. DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed)

594 Regression Models. R package version 0.1.6., CRAN/GitHub. Available from https://CRAN.R-

595 project.org/package=DHARMa

596 27. Ilha, P., S. Rosso, and L. Schiesari. 2019. Effects of deforestation on headwater stream fish

597 assemblages in the Upper Xingu River Basin, Southeastern Amazonia. Neotropical

598 Ichthyology 17: e180099



https://doi.org/10.1101/524652


25

599 28. Agostinho, A. A., L. C. Gomes, and F. M. Pelicice. 2007. Ecologia e manejo de recursos

600 pesqueiros em reservatórios do Brasil. Eduem, Maringá, PR.

601 29. Schiemer, F., H. Keckeis, and E. Kamler. 2003. The early life history stages of riverine fish:

602 ecophysiological and environmental bottlenecks. Comparative Biochemistry and Physiology

603 133:439–449.

604 30. Duellman, W.E., and L. Trueb. 1996. Biology of amphibians. McGraw Hill.

605 31. Bitar, Y. O. C., L. P. C. Pinheiro, O. S. Abe, and M. C. Santos-Costa. 2012. Species

606 composition and reproductive modes of anurans from a transitional Amazonian forest, Brazil.

607 Zoologia 29:19–26.

608 32. Schiesari, L., and B. Grillitsch. 2011. Pesticides meet megadiversity in the expansion of

609 biofuel crops. Frontiers in Ecology and the Environment 4:215–221.

610 33. Negri, D.D.B. 2015. Desempenho de larvas de anfíbios e libélulas em uma paisagem

611 agroindustrial. Master’s Thesis. University of São Paulo, Brazil.

612 34. Bitar, Y. O. C., L. Juen, L. C. Pinheiro, and M. C. Santos-Costa. 2015. Anuran beta diversity

613 in a mosaic anthropogenic landscape in transitional Amazon. Journal of Herpetology 49:75–

614 82.

615 35. Macedo, M. N., and L. Castello. 2015. State of the Amazon: Freshwater connectivity and

616 ecosystem health. In D. Oliveira, C. C. Maretti, and S. Charity, editors. WWF Living Amazon

617 Initiative. WWF, Brasilia, DF.

618 36. Pocewicz, A., and E. Garcia. 2016. Deforestation facilitates widespread stream habitat and

619 flow alteration in the Brazilian Amazon. Biological Conservation 203:252–259.

620 37. Woodward, C. L. 1996. Soil compaction and topsoil removal effects on soil properties and

621 seedling growth in Amazonian Ecuador. Forest Ecology and Management 82:197–209.

622 38. Alegre, J. C., and D. K. Cassel. 1996. Dynamics of soil physical properties under alternative

623 systems to slash-and-burn. Agriculture, Ecosystems and Environment 58:39–48.

624 39. Moraes, J.F. L., B. Volkoff, C. C. Cerri, and M. Bemoux. 1996. Soil properties under Amazon

625 forest and changes due to pasture installation in Rondônia, Brazil. Geodema 70:63–81.

626 40. Martínez, L. J., and J. A. Zinck. 2004. Temporal variation of soil compaction and deterioration

627 of soil quality in pasture areas of Colombian Amazonia. Soil & Tillage Research 75:3–17.

628 41. Keller, M., R. Varner, J. D. Dias, H. Silva, P. Crill, R. C. de Oliveira Jr, et al. 2005.



https://doi.org/10.1101/524652


26

629 Soil–atmosphere exchange of nitrous oxide, nitric oxide, methane, and carbon dioxide in

630 logged and undisturbed forest in the Tapajos National Forest, Brazil. Earth Interactions 9:23.

631 42. Moraes, J. M., A. E. Schuler, T. Dunne, R. O. Figueiredo, and R. L. Victoria. 2006. Water

632 storage and runoff processes in plinthic soils under forest and pasture in eastern Amazonia.

633 Hydrological Processes 20:2509–2526.

634 43. Zimmermann, B., H. Elsenbeer, and J. M. de Moraes. 2006. The influence of land-use

635 changes on soil hydraulic properties: implications for runoff generation. Forest Ecology and

636 Management 222:29–38.

637 44. Araújo, E. A. 2008. Qualidade do solo em ecossistemas de mata nativa e pastagens na

638 região leste do Acre, Amazônia Ocidental. PhD Thesis, Universidade Federal de Viçosa, MG.

639 45. Germer, S., C. Neill, T. Vetter, J. Chaves, A. V. Krusche, and H. Elsenbeer. 2009.

640 Implications of long-term land-use change for the hydrology and solute budgets of small

641 catchments in Amazonia. Journal of Hydrology 364:349–363.

642 46. Germer, S., C. Neill, A. V. Krusche, and H. Elsenbeer. 2010. Influence of land-use change on

643 near-surface hydrological processes: Undisturbed forest to pasture. Journal of Hydrology

644 380:473–480.

645 47. Sousa, M. I. L. 2010. Qualidade físico-hídrica de um argissolo vermelho-amarelo sob

646 agroecossistema e floresta natural em Roraima. Master’s Thesis, Universidade Federal de

647 Roraima, RR.

648 48. Comte, I., R. Davidson, M. Lucotte, C. J. R. Carvalho, F. A. Oliveira, B. P. Silva, et al. 2012.

649 Physicochemical properties of soils in the Brazilian Amazon following fire-free land

650 preparation and slash-and burn practices. Agriculture, Ecosystems and Environment.

651 156:108–115.

652 49. Guedes, E. M. S., A. R. Fernandes, H. V. de Lima, A. P. Serra, J. R. Costa, and R. S.

653 Guedes. 2012. Impacts of different management systems on the physical quality of an

654 amazonian oxisol. Revista Brasileira de Ciência do Solo 36:1269–1277.

655 50. Braz, A. M. S., A. R. Fernandes, and L. R. F. Alleoni. 2013. Soil attributes after the conversion

656 from forest to pasture in amazon. Land degradation & development 24:33–38.

657 51. Cruz, D. L. S., J. F. V. Júnior, P. L. S. Cruz, A. B. S. Cruz, and P. P. R. R. Nascimento. 2014.



https://doi.org/10.1101/524652


27

658 Atributos físico-hídricos de um argissolo amarelo sob floresta e savana naturais convertidas

659 para pastagem em Roraima. Revista Brasileira de Ciência do Solo 38:307–314.

660 52. Hunke, P., E. N. Mueller, B. Schröder, and P. Zeilhofer. 2015. The Brazilian Cerrado:

661 assessment of water and soil degradation in catchments under intensive agricultural use.

662 Ecohydrology 8:1154–1180.

663 53. van Straaten, O., M. D. Corre, K. Wolf, M. Tchienkoua, E. Cuellar, R. B. Matthews, et al.

664 2015. Conversion of lowland tropical forests to tree cash crop plantations loses up to one-half

665 of stored soil organic carbon. Proceedings of the National Academy of Sciences of the United

666 States of America 112(32):9956–9960.

667 54. Melo, V. F., A. G. Orrutéa, A. C. V. Motta, and S. A. Testoni. 2017. Land use and changes in

668 soil morphology and physical-chemical properties in Southern Amazon. Revista Brasileira de

669 Ciência do Solo 41:e0170034.

670 55. Nóbrega, R. L. B., A. C. Guzha, G. N. Torres, K. Kovacs, G. Lamparter, R. S. S. Amorim, et

671 al. 2017. Effects of conversion of native cerrado vegetation to pasture on soil hydro-physical

672 properties, evapotranspiration and streamflow on the Amazonian agricultural frontier. PLoS

673 ONE 12(6): e0179414.

674 56. Neill, C., L. A. Deegan, S. M. Thomas, C. L. Haupert, A. V. Krusche, V. M. Ballester, et al.

675 2006. Deforestation alters the hydraulic and biogeochemical characteristics of small lowland

676 Amazonian streams. Hydrological Processes 20:2563–2580.

677 57. Trancoso, R. 2006. Mudanças na cobertura da terra e alterações na resposta hidrológica de

678 bacias hidrográficas na Amazônia. Master’s Thesis, Instituto Nacional de Pesquisa da

679 Amazônia, AM.

680 58. Chaves, J., C. Neill, S. Germer, N. S. Gouveia, A. Krusche, and H. Elsenbeer. 2008. Land

681 management impacts on runoff sources in small Amazon watersheds. Hydrological Processes

682 22:1766–1775.

683 59. Deegan, L.A., C. Neill, C.L. Haupert, M.V.R. Ballester, A.V. Krusche, R.L. Victoria, et al. 2011.

684 Amazon deforestation alters small stream structure, nitrogen biogeochemistry, and

685 connectivity to large rivers. Biogeochemistry 105: 53-74.

686 60. Vittor, A. Y., W. Pan, R. H. Gilman, J. Tielsch, G. Glass, T. Shields, et al. 2009. Linking

687 deforestation to malaria in the Amazon: characterization of the breeding habitat of the



https://doi.org/10.1101/524652


28

688 principal malaria vector, Anopheles darlingi. The American Society of Tropical Medicine and

689 Hygiene 81:5–12.

690 61. Myers, S. S. 2009. Global environmental change: the threat to human health. Worldwatch

691 Report, Washington, DC.

692 62. Rains, M. C., J. F. Mount, and E. W. Larsen. 2004. Simulated changes in shallow

693 groundwater and vegetation distributions under different reservoir operations scenarios.

694 Ecological Applications 14:192–207.

695

696 Supporting Information Captions

697

698 S1 Figure. Streams in forested (a) and deforested (b,c) watersheds. (c) is a man-made reservoir.

699 Pictures by Paulo Ilha.

700

701 S2 Figure. Ponds and puddles in plateaus converted to pastures (a-d) and soybean plantations (e-g).

702 No ponds or puddles were found in forested plateaus. (a) and (b) are man-made cattle ponds; all

703 other puddles (c-g) appear to be formed solely by soil compaction by cattle trampling, road

704 construction and machinery traffic. In the rainy season ponds and puddles like these were inhabited

705 by up to 5 and 3 species of amphibians in pastures and soybean fields, respectively. Those in

706 pastures contained in addition high abundances of predatory insects including dragonflies, bugs and

707 beetles.Pictures by Luis Schiesari and Victor Dimitrov.

708

709 S1 Table. Pooled number of fishes sampled in lotic and lentic (i.e. reservoir) stream sections in three

710 streams in deforested watersheds, ranked by abundance.

711

712 S2 Table. Incidence matrix of amphibian species recorded in streamside and plateau transects in

713 forests, pastures and soybean fields (1= calling, +=observed).



https://doi.org/10.1101/524652




https://doi.org/10.1101/524652


ponds, puddles, floodplains and dams in the upper xingu basin: … · 120 soil infiltrability but...

Documents