Ecosystem Effects of the World's Largest Invasive Animal

2 DR. JONATHAN B. SHURIN (Orcid ID : 0000-0001-7870-1972)

3 DR. NATALIE JONES (Orcid ID : 0000-0001-5114-7123)

6 Article type : Articles

9 RUNNING HEAD: Hippo invasion

11 Ecosystem effects of the world’s largest invasive animal

12 Jonathan B. Shurin1,6, Nelson Aranguren Riaño2, Daniel Duque Negro2, David Echeverri 13 Lopez3, Natalie T. Jones1,4, , Oscar Laverde-R.5, Alexander Neu1 and Adriana Pedroza 14 Ramos2

15 1Section of Ecology, Behavior and Evolution 16 University of California San Diego 17 La Jolla, CA 92093-0116 18 [email protected] 19 20 2Unidad de Ecología en Sistemas Acuáticos - UDESA 21 Universidad Pedagógica y Tecnológica de Colombia 22 Avenida Central del Norte 39‑115

23 Tunja, Boyacá, ColombiaAuthor Manuscript

This is the author manuscript accepted for publication and has undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1002/ecy.2991

This article is protected by copyright. All rights reserved 24 25 3Corporación Autónoma Regional de las Cuencas de los Ríos Negros y Nare- CORNARE 26 Calle 13 # 9 -29 27 Municipio de la Unión, Antioquia, Colombia 28 29 4School of Biological Sciences 30 The University of Queensland 31 St. Lucia, QLD, Australia, 4072 32 33 5Departamento de Biología, Facultad de Ciencias 34 Pontificia Universidad Javeriana, Sede Bogotá D.C., Colombia 35 36 6Corresponding Author. Email: [email protected] 37 38 39

40 Abstract

41 The keystone roles of mega-fauna in many terrestrial ecosystems have been lost to defaunation. 42 Large predators and herbivores often play keystone roles in their native ranges, and some have 43 established invasive populations in new biogeographic regions. However, few empirical 44 examples are available to guide expectations about how mega-fauna affect ecosystems in novel 45 environmental and evolutionary contexts. We examined the impacts on aquatic ecosystems of an 46 emerging population of hippopotamus (Hippopotamus amphibus) that has been growing in 47 Colombia over the last 25 years. Hippos in Africa fertilize lakes and rivers by grazing on land 48 and excreting wastes in the water. Stable isotopes indicate that terrestrial sources contribute 49 more carbon in Colombian lakes containing hippo populations, and daily dissolved oxygen 50 cycles suggest that their presence stimulates ecosystem metabolism. Phytoplankton communities Author Manuscript 51 were more dominated by cyanobacteria in lakes with hippos, while bacteria, zooplankton and 52 benthic invertebrate communities were similar regardless of hippo presence. Our results suggest 53 that hippos recapitulate their role as ecosystem engineers in Colombia, importing terrestrial

This article is protected by copyright. All rights reserved 54 organic matter and nutrients with detectable impacts on ecosystem metabolism and community 55 structure in the early stages of invasion. Ongoing range expansion may pose a threat to water 56 resources.

57 Key words: hippopotamus, lakes, productivity, water resources, exotic species, eutrophication

59 Introduction

60 Humans have reshaped the earth’s biogeography, extirpating species and functional groups while 61 homogenizing and shifting boundaries among biotas by facilitating spread of global invaders 62 (Bernardo-Madrid et al. 2019). The extinction of diverse mega-fauna in the New World and 63 Australia at the end of the Pleistocene, potentially due to human colonization (Sandom et al. 64 2014), may have dramatically transformed ecosystems at the continental scale (Estes et al. 2011, 65 Dirzo et al. 2014, Galetti et al. 2018). For instance, the decline of terrestrial mega-fauna in North 66 America around 12,000 years ago may have precipitated a rapid shift in plant communities and 67 an increase in fire frequency (Gill et al. 2009). Wholesale disappearance of the largest animals 68 on land and in the oceans may have profoundly restructured ecosystems globally, implying that 69 human domination of ecosystem processes predates the Anthropocene.

70 One globally transformative effect of megafaunal extinction is through the functional loss of 71 their roles as vectors transporting nutrients across space and ecosystem boundaries (Doughty et 72 al. 2016). Marine mammals, migrating ungulates, anadromous fishes and other animals transport 73 and recycle nutrients horizontally and vertically, increasing productivity by fertilization and 74 precipitating state ecosystem shifts (McNaughton et al. 1997, Croll et al. 2005, Hocking and 75 Reynolds 2011). The reduction of mega-faunal abundances in the Pleistocene and following 76 industrial whaling likely impacted the global distribution of primary productivity and ecosystem 77 states (Malhi et al. 2016). Large animals often play keystone roles as nutrient vectors within 78 their present ranges; however, whether they retain their status as ecosystem engineers when

79 introduced, eitherAuthor Manuscript intentionally or unintentionally, outside their native range in entirely novel 80 ecological contexts remains unknown.

This article is protected by copyright. All rights reserved 81 Unintentional species introduction may provide a window into the ecosystem impacts of novel 82 mega-faunal invasion in New World ecosystems. The common hippopotamus (Hippopotamus 83 amphibius) was imported to Colombia by the notorious drug trafficker Pablo Escobar to populate 84 a private zoo at his estate, Hacienda Napoles, in the Magdalena Medio region of Antioquia. 85 After Escobar’s death in 1993, hippos escaped and formed a feral population in small artificial 86 lakes and the Magdalena River (Fig. 1). The population has expanded largely unchecked and is 87 now estimated to number between 65 and 80 animals. Breeding individuals have been observed 88 as far as 150km from Hacienda Napoles in the Magdalena River and smaller tributaries. If their 89 population is presently in exponential growth phase, their numbers could reach hundreds or 90 thousands in the coming decades in the absence of control measures or density dependent 91 regulation (Fig. 1). As no quantitative surveys have been conducted, there is no way to know 92 whether their population is growing exponentially or their carrying capacity in their introduced 93 range. The question of whether and how to restrict their ongoing population growth is 94 controversial as their potential environmental impact remains unknown (Dembitzer 2017, 95 Lundgren et al. 2018). The case of hippos in Colombia represents an example of a “social- 96 ecological mismatch” (Beever et al. 2019) where the public perception or value placed on a 97 charismatic exotic species constrains management options for removal or containing its spread in 98 order to mitigate its negative effects. Control of hippo populations is controversial in Colombia 99 and internationally, and hippos attract tourists to the region. The patchy distribution and 100 colonization of different lakes by hippos affords an unprecedented opportunity to test the impact 101 of a novel large animal invader with functional traits that are most similar to tapirs among native 102 South American fauna, but still distinct from any extinct or extant taxa.

103 Hippos in Africa function as ecosystem engineers by shaping the structure and function of 104 aquatic and terrestrial ecosystems and facilitating connectivity between the water and land 105 (Subalusky et al. 2015, Dutton et al. 2018, Stears et al. 2018). Hippos graze on grasses on land 106 at night and spend the day in lakes and rivers, excreting nutrients acquired from terrestrial 107 vegetation, fertilizing aquatic producers and providing organic material that sustains detritivores.

108 Decomposition ofAuthor Manuscript organic matter imported by hippos can result in anoxic conditions and mass 109 mortality of fishes during high flow events when material is exported downstream out of hippo 110 pools (Dutton et al. 2018), and reduction in macroinvertebrate abundances (Dawson et al. 2016). 111 Hippos also perform bioturbation, disturbing aquatic sediments and re-suspending nutrients and

This article is protected by copyright. All rights reserved 112 organic matter into the water column, thereby altering the hydrodynamics of rivers. The 113 fertilizing effect of nutrients imported on aquatic ecosystems by hippos in Africa also varies 114 seasonally. The ecosystem effects of hippos have been studied in savannah habitats of East 115 Africa where animals become concentrated in isolated pools during dry seasons of low river 116 discharge (McCauley et al. 2015, Subalusky et al. 2015, Dutton et al. 2018, Stears et al. 2018, 117 Subalusky et al. 2018, Schoelynck et al. 2019). The impact of organic matter imported by hippos 118 on aquatic ecosystems is often greatest in dry seasons when their densities are high and export 119 through river flow is low (McCauley et al. 2015, Stears et al. 2018). Hippos in Colombia occupy 120 lakes at Hacienda Napoles with much more consistent water levels and higher annual 121 precipitation than the East African sites where their effects have been studied (Appendix S1: Fig. 122 S2). The native biome of Antioquia is tropical rain forest, although the watersheds of the lakes 123 at Hacienda Napoles have been deforested for ranching. The importance of animals as resource 124 vectors across ecosystem boundaries likely depends on aspects of environmental context 125 including seasonality, resource gradients between donor and recipient habitats, and climate 126 (Subalusky and Post 2019). Hippos exert profound influence over their native African 127 ecosystems; however, their role within their invasive range in Colombia where the environment 128 is wetter and water levels less seasonally variable remains unknown.

129 Here we examine the impact of invasive hippos on small lakes in Magdalena Medio of Colombia 130 to ask whether their role as ecosystem engineers in their native Africa is recapitulated in their 131 invasive range in South America. We measured carbon and nitrogen stable isotopic signatures of 132 particulate organic matter and water chemistry parameters in 14 small lakes, including two with 133 persistent hippo populations. We also measured the daily cycle of dissolved oxygen 134 concentration in a subset of the lakes as an indicator of gross ecosystem production. Based on 135 their role in enhancing connectivity between terrestrial and aquatic habitats, and between the 136 benthic and pelagic environments within lakes, we expect lakes with hippos to show heavier 137 carbon isotopic signatures and be more eutrophic than those lacking hippos. We also 138 characterized the composition of benthic and pelagic invertebrates, phytoplankton (by

139 microscope counts),Author Manuscript and bacterioplankton (by 16S amplicon sequencing) to ask whether presence 140 of hippo results in shifts in composition of aquatic communities. We discuss the question of how 141 invasive hippo populations should be managed in Colombia in light of their environmental

This article is protected by copyright. All rights reserved 142 impacts as well as the general question of whether mega-fauna perform similar ecosystem 143 functions in an entirely novel ecological context as in their native environment.

144 Methods

145 We sampled 14 small lakes at Hacienda Napoles in Antioquia, Colombia on three field 146 campaigns (in April and July 2017, and September 2018, Appendix S1: Fig. S1). The April and 147 September sampling campaigns corresponded to the wet season and the July trip occurred during 148 the dry season (Appendix S1: Fig. S2). Seven lakes were sampled on each of the first two visits, 149 and all 14 lakes were sampled on the last. All lakes were sampled from the shore due to the 150 hazard presented by hippos. Two lakes contained persistent hippo populations, one large (Lago 151 1, 20-30 individuals, 78,000m2 lake surface area) and the other small (Lago 10, 3-5 individuals, 152 13,000m2 lake surface area, Appendix S1: Table S1). All other lakes lacked hippos at the time of 153 sampling, although transient individuals have occasionally been observed in some of the lakes. 154 All of the lakes are impoundments created to store water for ranching and cattle were present in 155 all of the catchments. We sampled all of the lakes at Hacienda Napoles with and without 156 resident hippo populations. Other populations occur in floodplain lakes along the Magdalena 157 River; however, the environment of these lakes differs considerably from the artificial water 158 bodies in our study in that they are seasonally connected to the river. The unbalanced 159 comparison between two hippo lakes and 12 non-hippo lakes is an unavoidable consequence of 160 the patchy hippo distribution at our study site.

161 At each lake we collected water for analysis of chemical parameters including nutrients 162 (particulate and dissolved carbon and total nitrogen), conductivity, pH, temperature and 163 chlorophyll-a concentration. Methods for water chemistry analysis are described in Appendix 164 S1. Daily dissolved oxygen cycles were recorded with HOBO Dissolved Oxygen Loggers 165 (Onset Computer Corp., Bourne, MA, USA) placed around 50cm below the surface 2m from 166 the shore at each sampling visit. Readings were taken every 15 minutes, and the daily 167 amplitude was calculated for each day in each sampled lake as the difference between the

168 mean of all readingsAuthor Manuscript recorded within two hours of dusk (1700-1900) and dawn (0500-0700). 169 We recorded seven daily cycles in one hippo lake (Lake 1), four in the second (Lake 10) and 170 nine in three different “no hippo” lakes (Lakes 2, 8 and 13). We use the daily change in 171 dissolved oxygen concentration as an indicator of net ecosystem production (the difference

This article is protected by copyright. All rights reserved 172 between gross production and respiration). We are unable to estimate gross production or 173 respiration because we lack concurrent data on parameters associated with gas exchange like 174 wind speed and mixed layer depth (Winslow et al. 2016). However, as gas exchange was likely 175 similar among the lakes at our study site, the daily change in dissolved oxygen concentration is 176 an indicator of net production, the difference between gross production and respiration (Kalff 177 2002).

178 Taxonomic composition of phytoplankton communities were characterized by microscopy. 179 Water samples (60ml) were collected at each lake and fixed with Lugol’s Iodine solution in the 180 field. Cell counts of phytoplankton, identified to the lowest possible taxonomic level, were made 181 using the Utermöhl method (Utermöhl 1958). Sequencing of the 16S-V4 Ribosomal RNA genes 182 in environmental samples was used to describe bacterial community composition. Methods for 183 sample collection, DNA extraction and amplification, sequencing and taxonomic assignment are 184 described in Appendix S1.

185 Zooplankton samples were collected with a 30cm diameter net with a 45µm mesh size thrown a 186 known distance from the shore and drawn through the surface water. Samples were preserved in 187 the field with 70% ethanol and animals were identified to species and enumerated in the lab 188 under a dissecting microscope. Macroinvertebrates were collected in ten one meter long sweeps 189 of a small dip net drawn through the vegetation or along the sediments at the shore. Samples 190 were preserved in 70% ethanol and all organisms in each sample were identified to the lowest 191 possible taxonomic level and counted under a dissecting microscope.

192 Data analysis- Lakes were categorized according to whether hippos were present (n=2) or absent 193 (n=12) at the time of sampling. The two “hippo present” lakes differed in surface area (Lake 1: 194 78,000m2 vs. Lake 10: 13,0000m2) and hippo population sizes (Lake 1: 20-30 animals vs. Lake 195 10: 3-5 animals), but contained comparable hippo population densities (Lake 1: 253-384 km-2 vs. 196 Lake 10: 230-384 km-2). We therefore test the effects of hippo presence by comparing the two 197 hippo lakes with the 12 no hippo lakes. We also analyzed our data by classifying no hippo lakes

198 according to whetherAuthor Manuscript hippos had ever been observed in the past. We found no difference 199 between lakes where hippos had been observed in the past but were absent at the time of 200 sampling and ones where hippos were never seen in any of our measured variables. We 201 therefore present the data with only the two categories of hippo presence vs. absence at the time

This article is protected by copyright. All rights reserved 202 of sampling. Relationships of the measured lake variables to hippo density and lake surface area 203 are shown in the Supplemental Information.

204 Differences between lakes with vs. without hippos in POM stable isotope ratios (δ13C and δ15N) 205 and water chemistry variables (chlorophyll-a concentration, total C and N, pH, conductivity, 206 Secchi depth) were tested by ANOVA including lake category and date (the three sampling trips) 207 in the model with no interaction term. For the amplitude of daily dissolved oxygen change, we 208 also included Lake in the model because we had multiple daily samples from the some lakes 209 from the same sampling trip.

210 Differences in community composition of bacteria, phytoplankton, zooplankton, 211 macroinvertebrates and littoral vertebrates between lake categories were tested by 212 PERMANOVA tests with 1000 randomized permutations of the data, and community 213 composition was visualized by non-metric multidimensional scaling (NMDS). We compared 214 models including both hippo presence and sampling date (April 2017, July 2017 and September 215 2018) as predictors to determine if differences in composition were apparent after accounting for 216 seasonal variation. Differences in taxonomic richness between lake categories and sampling 217 dates were tested by two-way ANOVA after rarefying each sample to the number of individuals 218 found in the sample with the fewest individuals in order to standardize sample size among 219 samples using the “rarefy” function in the Vegan package in R (Oksanen et al. 2019).

220 Results

221 Stable C and N isotope ratios of particulate organic matter (POM) indicate that hippo lakes have 222 significantly enriched isotopic signatures for C but not N (Fig. 2). The mean δ13C value of ‘no 223 hippo’ lakes, averaged over multiple temporal samples, was lighter (-24.9‰) than either of the 224 two lakes with hippos (Lake 1: -19.2‰, Lake 10: -22.7‰), a difference that was significant by 225 two-way ANOVA including lake category (hippos present vs. absent) and sampling trip as 226 predictors. The contrast between lakes with and without hippos in δ13C is significant whether or 227 not sampling trip is included in the model. The mean δ13C of POM from the two lakes with Author Manuscript 228 hippos present was more enriched than any of the ‘no hippo’ lakes except for a single anomalous 229 sample from one ‘no hippo’ lake that was heavily enriched in both δ13C and δ15N. The δ15N of 230 POM did not differ between the two lake categories (P=0.53).

This article is protected by copyright. All rights reserved 231 Daily dissolved oxygen cycles ranged over around 3.6 mg/L O2 in the lakes with persistent hippo 232 populations, and 2.1 mg/L in lakes without hippos (Fig. 3). Both lakes with hippos had greater 233 daily changes in dissolved oxygen than the no hippo lakes. In addition, the dissolved oxygen 234 concentration sometimes dropped below 4 mg/L at night in the two hippo lakes, the 235 concentration where fish mass mortality can occur (Dutton et al. 2018), but not in any of the 236 lakes without hippos (Appendix S1: Fig. S15). Phytoplankton biomass (measured as 237 chlorophyll-a concentration), total carbon, total nitrogen pH, conductivity and Secchi depth 238 transparency were all indistinguishable between lakes with and without hippos (Fig. 3, all 239 P>0.1). Aggregating these variables by taking the average across sampling dates or separating 240 the two lakes with hippos at the time of sampling in the analysis did not affect any of these 241 conclusions. We found no significant differences in any of the measured water chemistry 242 variables among the three sampling visits (all P>0.05).

243 Although one of the two hippo lakes (Lake 1) was the largest lake in our survey, hippo density 244 was a stronger predictor of stable C isotopic ratios and daily oxygen cycles than lake surface 245 area. Δ13C showed significant enrichment with increasing hippo density but not with greater lake 246 area (Appendix S1: Fig. S12). Daily dissolved oxygen range increased with both hippo density 247 (Appendix S1: Fig. S13a) and lake area (Appendix S1: Fig. S14a); however, model selection by 248 stepwise elimination when both were included in the same model retained the model with just 249 hippo density, eliminating lake surface area (ΔAIC = 2), indicating that hippo density is a 250 stronger predictor of lake productivity than area. Other measures of water chemistry were 251 unrelated to either lake area or hippo density (Appendix S1: Fig. S13, S14).

252 The most frequently recorded bacterial phyla were the Proteobacteria, Planctomycetes, 253 Actinobacteria, Bacteriodetes and Verrucomicrobia (Appendix S1: Fig. S3, S4). The 254 composition of bacterial assemblages did not differ between lakes with and without hippos when 255 reads were aggregated by phylum (PERMANOVA: P=0.61, Fig. 4a), class (PERMANOVA: 256 P=0.56) or lower taxonomic levels. Bacterial phylum composition differed between sampling 257 dates (P=0.003) but not hippo categories when both were included in the same PERMANOVA Author Manuscript 258 analysis. Rarefied bacterial richness by phylum was indistinguishable between lake types (Fig. 259 4b) but differed significantly among sampling dates (P<0.001).

This article is protected by copyright. All rights reserved 260 Phytoplankton communities were dominated by Chlorophyta, Cyanophyta and Charophyta (Fig. 261 4c, Appendix S1: Figs. S5-S7). Lakes with hippos had higher relative densities of Cyanophyta, 262 on average, than the lakes without (Appendix S1: Figs. S5-S7), and PERMANOVA detected 263 significant differences in phytoplankton composition between the lake categories (P=0.03). 264 When hippo presence and sampling date were included together in the analysis, both were 265 significant by PERMANOVA (date: P=0.001, hippos: P=0.02). Rarefied phytoplankton richness 266 by genus (Fig. 4d) was equivalent between the three lake categories and the three sampling dates. 267 Rarefied division richness was slightly greater in hippo lakes than no hippo lakes (a mean of 5.9 268 vs. 5.1 divisions, P=0.046) and differed among the three sample dates (P<0.001).

269 Zooplankton communities contained abundant rotifers, small cladocerans and copepods 270 (Appendix S1: Fig S8, S9). No differences in zooplankton community composition (Fig. 4e) or 271 richness (Fig. 5f) were found between the three lake categories. Zooplankton composition 272 differed among the three sampling dates by PERMANOVA (P=0.001), while rarefied 273 zooplankton species richness was equivalent among sampling dates. Macroinvertebrates were 274 dominated by the insect orders Ephemeroptera and Hemiptera, along with Annelid worms and 275 ostracods (Appendix S1: Fig. S10, S11). Composition (Fig. 4g) and richness (Fig. 5h) of 276 macroinvertebrates were similar in lakes with and without hippos. Composition, but not rarefied 277 richness, varied among the three sampling dates (PERMANOVA, P=0.001).

278 Discussion

279 Our survey indicates that hippos have reprised their roles as vectors of organic matter from 280 terrestrial to aquatic ecosystems in their invasive range in Colombia. Stable carbon isotopic 281 signatures of POM were more enriched with heavy carbon in lakes containing hippos, indicating 282 a greater proportional representation of terrestrial sources to the carbon budgets of the lakes (Fry 283 2006, McCauley et al. 2015). Lake net productivity, as measured by the amplitude of daily 284 cycles of dissolved oxygen, was also greater in lakes with hippos, although other trophic state 285 indicators including chlorophyll-a concentration, dissolved nutrients and light transparency were

286 indistinguishable Author Manuscript between lakes with and without hippos. Community composition of 287 phytoplankton varied between lake categories as cyanobacteria were more prevalent in hippo 288 lakes, while bacterial, zooplankton and macroinvertebrate composition were all similar between 289 lake categories. Our surveys indicate that hippos perform similar ecological roles as nutrient

This article is protected by copyright. All rights reserved 290 vectors in novel Neotropical ecosystems as in their native range in Africa with similar effects on 291 aquatic ecosystems despite dramatic differences in environmental conditions. Our results also 292 suggest that ongoing population growth and range expansion by hippos may pose a threat to the 293 quality of water resources in the Magdalena Basin of Colombia by increasing the loading of 294 terrestrial organic matter and contributing to shifts in community structure of phytoplankton 295 toward Cyanobacteria, a group associated with harmful algae blooms and eutrophication (Smith 296 1983).

297 Stable C isotopic signatures of POM indicate that hippos import terrestrial carbon into lakes, 298 supporting the hypothesis that mega-fauna redistribute elements across landscapes through their 299 migrations, stimulating productivity by recycling nutrients (Doughty et al. 2016). Terrestrial 300 plants have heavier C isotopic ratios than phytoplankton or aquatic plants (Fry 2006), therefore 301 the heavier POM signature in the two hippo lakes than the no-hippo lakes is likely due to 302 terrestrial carbon imported by hippos. The lakes at Hacienda Napoles are turbid (mean 303 chlorophyll-a = 58.3µg/L, range: 5.6-248) and dominated by pelagic primary producers 304 (phytoplankton) with few benthic vascular plants, likely due to low light transparency. 305 Phytoplankton communities also contain abundant cyanobacteria which are often indicators of 306 phosphorus pollution and eutrophic conditions in lakes (Smith 1983), and their greater relative 307 abundance in hippo lakes is consistent with their predicted effects as nutrient vectors. However, 308 other indicators of lake trophic status such as chlorophyll, nitrogen and carbon concentrations 309 were unaffected by the presence of hippos. The greater daily shift in dissolved oxygen 310 concentrations in lakes with hippos suggests that they either supply nutrients that stimulate 311 primary production or organic matter that fuels decomposition and respiration. Further study is 312 necessary to elucidate the pathways by which hippos affect ecosystem metabolism in Colombian 313 lakes.

314 Although hippo lakes showed higher net productivity on average, other indicators of trophic 315 status were similar to lakes without hippos. All of the lakes at Hacienda Napoles are shallow and 316 eutrophic, situated in an agricultural watershed with cattle grazing present. The ‘non-hippo’ Author Manuscript 317 lakes are therefore impacted by agriculture and eutrophic. In addition, hippo impacts on African 318 lakes and rivers vary seasonally with water levels and therefore hippo densities. McCauley et al. 319 (2015) found that fish and insects in hippo pools in Kenya were enriched in 13C at the end of the

This article is protected by copyright. All rights reserved 320 dry season, but not after the wet season, implying that the carbon vectored to rivers by hippos is 321 most concentrated when water levels are low. Hippo effects on aquatic ecosystems in Tanzania 322 (Stears et al. 2018) also vary seasonally, and are most pronounced in the dry season when water 323 levels are low and hippos are densely crowded in small water bodies. Anoxia driven by organic 324 matter imported by hippos in the Mara River occurs only in high density hippo pools and when 325 flood events export material downstream (Dutton et al. 2018). The Magdalena Medio of 326 Colombia is in an equatorial climate with two peaks of precipitation around May and October 327 and dryer periods around January and July (Appendix S1: Fig. S2); however annual precipitation 328 is 46-187% greater than any of the East African sites where hippo effects on aquatic ecosystems 329 have been studied (Appendix S1: Fig. S2). Water levels at Hacienda Napoles fluctuate little 330 throughout the year and hippos never become concentrated at high densities as they do in parts of 331 Africa. Their moderate effects on Colombian lakes may therefore be related to the high 332 precipitation in their invasive range compared to parts of the native range where their densities 333 become concentrated during dry seasons.

334 Another potential explanation for the lack of strong fertilization effects of hippos may be that 335 their populations are still in the initial phases of growth and well below the constraints set by 336 their environment in Colombia. Their annual per-capita growth rate over the last 26 years is 337 0.11*year-1, suggesting that hippos in Colombia may be freed from many constraints set by their 338 resources or enemies. Kanga et al. (2011) calculated an average annual growth rate of the hippo 339 population at Masai Mara National Reserve of 0.188 between 1970 and 1980 following a 340 population bottleneck, suggesting that the growth rate of the Colombian hippo population is 341 comparable to that seen in the native range. Their potential for rapid population growth implies 342 that hippos could become prevalent throughout the Magdalena River basin in a few decades. 343 The effects of hippos on the aquatic environment that we observe suggest that sustained 344 population growth poses a threat to water quality in lakes and rivers as they expand their range 345 throughout Magdalena Medio watershed and potentially colonize new regions on the Caribbean 346 slope of Colombia. Author Manuscript 347 Whether animals, including ecosystem engineers with major impacts, retain their ecological roles 348 outside of their native range in the context of a novel ecosystem remains uncertain. Our results 349 indicate that hippos function as ecosystem engineers in the Neotropics as they do in their native

This article is protected by copyright. All rights reserved 350 range in Africa, enhancing connectivity by vectoring organic material and nutrients across the 351 boundary separating aquatic and terrestrial ecosystems. Donlan et al. (Donlan et al. 2006) 352 propose rewilding via introduction of mega-fauna to perform particular ecological roles such as 353 seed dispersal or suppression of non-native vegetation, and caution that candidate taxa for re- 354 wilding need to be evaluated based on a thorough knowledge of their ecology. Although re- 355 wilding has been proposed as a strategy to restore ecosystems to their ancestral states (Zimov 356 2005, Hewitt et al. 2011, Galetti et al. 2017), hippos are a novel functional group in the New 357 World with no Pleistocene analog. In addition, increasing aquatic ecosystem productivity 358 through fertilization by hippos represents an ecosystem disservice in many regards. 359 Eutrophication threatens aquatic ecosystems and water resources worldwide (Carpenter et al. 360 1998), including in Colombia (Aranguren-Riaño et al. 2018), a global biodiversity hot-spot. 361 Given their rapid population growth in the initial stages of invasion and early indications of their 362 ecosystem impacts, humane approaches to population control may be warranted before their 363 numbers and impact become unmanageable.

364 Options for management of hippo populations in Colombia are constrained by conflict between 365 their potential environmental effects, their value to the local tourist economy and their status as a 366 high-profile charismatic organism. Tourists visit Hacienda Napoles in part to see hippos and 367 their connection to Colombia’s history attracts national and international interest. As such, there 368 is strong political and social resistance to any lethal means of population control, an example of a 369 “social-ecological mismatch” (Beever et al. 2019) where the societal value placed on a species 370 may conflict with its potential adverse environmental impacts. Ours is the first study of their 371 effects on aquatic ecosystems, and plans to manage their population growth need to be 372 formulated based on their role in the ecology and society of Colombia. Hippos have expanded 373 throughout the region and are now found in the Magdalena River over 150km from Hacienda 374 Napoles where they may potentially interact with native wildlife such as manatees or affect local 375 fisheries by precipitating anoxia and high fish mortality as their population grows and spreads 376 (Dutton et al. 2018). The challenge of managing invasive species that are valued economically

377 and socially but threatenAuthor Manuscript ecosystem integrity presents a dilemma for natural resource 378 management agencies.

This article is protected by copyright. All rights reserved 379 Our results have implications for assessing the lost ecosystem functions performed by extinct 380 mega-fauna. Extinction of terrestrial mega-fauna during the Pleistocene may have greatly 381 disrupted the lateral transport of nutrients and organic matter via their migrations (Doughty et al. 382 2016), while others such as gomphotheres and ground sloths may have once dispersed the seeds 383 of extant plant species (Janzen and Martin 1982). Our results indicate that invasion by hippos in 384 South America recapitulates some lost ecosystem functions by importing terrestrial carbon into 385 small lakes. Mega-fauna were lost throughout South America by the beginning of the Holocene, 386 and hippos represent an entirely novel taxonomic group in the neotropical fauna. However, other 387 extinct or extant large amphibious mammals like tapirs or capybaras may perform a similar role 388 in transporting nutrients from terrestrial to aquatic ecosystems. The only extinct large 389 amphibious mammal in South America listed in the database of Faurby and Svenning (2015) is 390 the 200kg tapir, Tapirus rondoniensis, from Brazil. Our results indicate that hippo colonization 391 increases the supply of terrestrial carbon sources in small lakes with detectable impacts on 392 system productivity and phytoplankton communities. Hippos therefore retain their role as 393 nutrient vectors that stimulate aquatic ecosystem production when introduced in an entirely novel 394 ecosystem context. Strategies to monitor their ongoing population expansion need to be 395 developed in light of their detectable impacts on Colombian ecosystems in the early stages of 396 their invasion.

397

398 Acknowledgments- We thank Andrea Barrera, Luis Antonio Gonzalez Montana, Nestor Torres 399 and Zahia Merchan for help in the field, Jorge Caro López and Hacienda Napoles Park for access 400 to field sites and logistical assistance. Funding was provided by National Geographic Waitt 401 Grant (W448-16), Fulbright Colombia U.S. Scholar Fellowship and National Science Foundation 402 (NSF-DEB Award 1457737).

403

404 REFERENCES Author Manuscript 405 Aranguren-Riaño, N. J., J. B. Shurin, A. Pedroza-Ramos, C. L. Munoz-Lopez, R. Lopez, and O. 406 Cely. 2018. Sources of nutrients behind recent eutrophication of Lago de Tota, a high 407 mountain Andean lake. Aquatic Sciences 80.

This article is protected by copyright. All rights reserved 408 Beever, E. A., D. Simberloff, S. L. Crowley, R. Al-Chokhachy, H. A. Jackson, and S. L. 409 Petersen. 2019. Social-ecological mismatches create conservation challenges in 410 introduced species management. Frontiers in Ecology and the Environment 17:117-125. 411 Bernardo-Madrid, R., J. Calatayud, M. Gonzalez-Suarez, M. Rosvall, P. M. Lucas, M. Rueda, A. 412 Antonelli, and E. Revilla. 2019. Human activity is altering the world's zoogeographical 413 regions. Ecology Letters 22:1297-1305. 414 Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth, A. N. Sharpley, and V. H. Smith. 415 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological 416 Applications 8:559-568. 417 Croll, D. A., J. L. Maron, J. A. Estes, E. M. Danner, and G. V. Byrd. 2005. Introduced predators 418 transform subarctic islands from grassland to tundra. Science 307:1959-1961. 419 Dawson, J., D. Pillay, P. J. Roberts, and R. Perissinotto. 2016. Declines in benthic 420 macroinvertebrate community metrics and microphytobenthic biomass in an estuarine 421 lake following enrichment by hippo dung. Scientific Reports 6. 422 Dembitzer, J. 2017. The Case for Hippos in Colombia. Israel Journal of Ecology & Evolution 423 63:5-8. 424 Dirzo, R., H. S. Young, M. Galetti, G. Ceballos, N. J. B. Isaac, and B. Collen. 2014. Defaunation 425 in the Anthropocene. Science 345:401-406. 426 Donlan, C. J., J. Berger, C. E. Bock, J. H. Bock, D. A. Burney, J. A. Estes, D. Foreman, P. S. 427 Martin, G. W. Roemer, F. A. Smith, M. E. Soule, and H. W. Greene. 2006. Pleistocene 428 rewilding: An optimistic agenda for twenty-first century conservation. American 429 Naturalist 168:660-681. 430 Doughty, C. E., J. Roman, S. Faurby, A. Wolf, A. Haque, E. S. Bakker, Y. Malhi, J. B. Dunning, 431 and J. C. Svenning. 2016. Global nutrient transport in a world of giants. Proceedings of 432 the National Academy of Sciences of the United States of America 113:868-873. 433 Dutton, C. L., A. L. Subalusky, S. K. Hamilton, E. J. Rosi, and D. M. Post. 2018. Organic matter 434 loading by hippopotami causes subsidy overload resulting in downstream hypoxia and 435 fish kills. Nature Communications 9. Author Manuscript 436 Estes, J. A., J. Terborgh, J. S. Brashares, M. E. Power, J. Berger, W. J. Bond, S. R. Carpenter, T. 437 E. Essington, R. D. Holt, J. B. C. Jackson, R. J. Marquis, L. Oksanen, T. Oksanen, R. T. 438 Paine, E. K. Pikitch, W. J. Ripple, S. A. Sandin, M. Scheffer, T. W. Schoener, J. B.

This article is protected by copyright. All rights reserved 439 Shurin, A. R. E. Sinclair, M. E. Soule, R. Virtanen, and D. A. Wardle. 2011. Trophic 440 Downgrading of Planet Earth. Science 333:301-306. 441 Faurby, S. and J. C. Svenning. 2015. Historic and prehistoric human-driven extinctions have 442 reshaped global mammal diversity patterns. Diversity and Distributions 21:1155-1166. 443 Fry, B. 2006. Stable Isotope Ecology. Springer, New York. 444 Galetti, M., M. Moleon, P. Jordano, M. M. Pires, P. R. Guimaraes, T. Pape, E. Nichols, D. 445 Hansen, J. M. Olesen, M. Munk, J. S. de Mattos, A. H. Schweiger, N. Owen-Smith, C. N. 446 Johnson, R. J. Marquis, and J. C. Svenning. 2018. Ecological and evolutionary legacy of 447 megafauna extinctions. Biological Reviews 93:845-862. 448 Galetti, M., M. Root-Bernsteina, and J. C. Svenning. 2017. Challenges and opportunities for 449 rewilding South American landscapes. Perspectives in Ecology and Conservation 15:245- 450 247. 451 Gill, J. L., J. W. Williams, S. T. Jackson, K. B. Lininger, and G. S. Robinson. 2009. Pleistocene 452 Megafaunal Collapse, Novel Plant Communities, and Enhanced Fire Regimes in North 453 America. Science 326:1100-1103. 454 Hewitt, N., N. Klenk, A. L. Smith, D. R. Bazely, N. Yan, S. Wood, J. I. MacLellan, C. Lipsig- 455 Mumme, and I. Henriques. 2011. Taking stock of the assisted migration debate. 456 Biological Conservation 144:2560-2572. 457 Hocking, M. D. and J. D. Reynolds. 2011. Impacts of Salmon on Riparian Plant Diversity. 458 Science 331:1609-1612. 459 Janzen, D. H. and P. S. Martin. 1982. NEOTROPICAL ANACHRONISMS - THE FRUITS 460 THE GOMPHOTHERES ATE. Science 215:19-27. 461 Kalff, J. 2002. Limnology: inland water ecosystems. Prentice-Hall, Inc., Upper Saddle River, NJ. 462 Kanga, E. M., J. O. Ogutu, H. Olff, and P. Santema. 2011. Population trend and distribution of 463 the Vulnerable common hippopotamus Hippopotamus amphibius in the Mara Region of 464 Kenya. Oryx 45:20-27. 465 Lundgren, E. J., D. Ramp, W. J. Ripple, and A. D. Wallach. 2018. Introduced megafauna are 466 rewilding the Anthropocene. Ecography 41:857-866. Author Manuscript 467 Malhi, Y., C. E. Doughty, M. Galetti, F. A. Smith, J. C. Svenning, and J. W. Terborgh. 2016. 468 Megafauna and ecosystem function from the Pleistocene to the Anthropocene.

This article is protected by copyright. All rights reserved 469 Proceedings of the National Academy of Sciences of the United States of America 470 113:838-846. 471 McCauley, D. J., T. E. Dawson, M. E. Power, J. C. Finlay, M. Ogada, D. B. Gower, K. Caylor, 472 W. D. Nyingi, J. M. Githaiga, J. Nyunja, F. H. Joyce, R. L. Lewison, and J. S. Brashares. 473 2015. Carbon stable isotopes suggest that hippopotamus-vectored nutrients subsidize 474 aquatic consumers in an East African river. Ecosphere 6. 475 McNaughton, S. J., F. F. Banyikwa, and M. M. McNaughton. 1997. Promotion of the cycling of 476 diet-enhancing nutrients by African grazers. Science 278:1798-1800. 477 Oksanen, J., F. Guillaume Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, R. B. 478 O'Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, E. Szoecs, and H. Wagner. 2019. 479 Vegan: Community ecology package. 480 Sandom, C., S. Faurby, B. Sandel, and J. C. Svenning. 2014. Global late Quaternary megafauna 481 extinctions linked to humans, not climate change. Proceedings of the Royal Society B- 482 Biological Sciences 281. 483 Schoelynck, J., A. L. Subalusky, E. Struyf, C. L. Dutton, D. Unzue-Belmonte, B. Van de Vijver, 484 D. M. Post, E. J. Rosi, P. Meire, and P. Frings. 2019. Hippos (Hippopotamus amphibius): 485 The animal silicon pump. Science Advances 5. 486 Smith, V. H. 1983. Low nitrogen to phosphorus ratios favor dominance by blue-green algae in 487 lake phytoplankton. Science 221:669-671. 488 Stears, K., D. J. McCauley, J. C. Finlay, J. Mpemba, I. T. Warrington, B. M. Mutayoba, M. E. 489 Power, T. E. Dawson, and J. S. Brashares. 2018. Effects of the hippopotamus on the 490 chemistry and ecology of a changing watershed. Proceedings of the National Academy of 491 Sciences of the United States of America 115:E5028-E5037. 492 Subalusky, A. L., C. L. Dutton, L. Njoroge, E. J. Rosi, and D. M. Post. 2018. Organic matter and 493 nutrient inputs from large wildlife influence ecosystem function in the Mara River, 494 Africa. Ecology 99:2558-2574. 495 Subalusky, A. L., C. L. Dutton, E. J. Rosi-Marshall, and D. M. Post. 2015. The hippopotamus 496 conveyor belt: vectors of carbon and nutrients from terrestrial grasslands to aquatic Author Manuscript 497 systems in sub-Saharan Africa. Freshwater Biology 60:512-525. 498 Subalusky, A. L. and D. M. Post. 2019. Context dependency of animal resource subsidies. 499 Biological Reviews 94:517-538.

This article is protected by copyright. All rights reserved 500 Utermöhl, H. 1958. Zur Vervollkommnung der quantitative Phytoplankton-Methodik. 501 Mitteilungen des Internationalen Limnologie, 9:1-38. 502 Winslow, L. A., J. A. Zwart, R. D. Batt, H. A. Dugan, R. I. Woolway, J. R. Corman, P. C. 503 Hanson, and J. S. Read. 2016. LakeMetabolizer: an R package for estimating lake 504 metabolism from free-water oxygen using diverse statistical models. Inland Waters 505 6:622-636. 506 Zimov, S. A. 2005. Pleistocene park: Return of the mammoth's ecosystem. Science 308:796-798. 507

508

509 Figure Legends:

510 Figure 1- Observed and projected growth of the Colombian hippo population. Four individuals 511 were present at the time of Pablo Escobar’s death in 1993, and the population is presently 512 estimated to number between 65 and 80. The increase represents an exponential rate of 513 population growth of 11% per year (ln(65)-ln(4))/26 years). A population growing at this 514 exponential rate in the absence of density dependence or control measures would number 785 in 515 2040 and 7,089 in 2060. Photos by Luis Antonio Gonzalez Montana.

516 Fig. 2- Stable C and N isotopic ratios for POM samples collected from lakes without (orange) 517 and with (green) resident hippo populations. Solid circles are the mean values of multiple 518 samples collected at different times from the same lake, and open circles are the individual 519 observations from each sampling date. P-values are the effects of hippo presence from two-way 520 ANOVA including hippo presence and sampling trip as predictor variables.

521 Figure 3- Physical-chemical conditions in lakes with vs. without hippos. (A) The amplitude of 522 daily oxygen profiles (the difference between dusk and dawn), (B) concentrations of chlorophyll- 523 a, (C) total carbon, (D) total nitrogen, (E) pH and (F) conductivity. Only daily changes in 524 oxygen concentration differed between the two categories of lakes. The difference in daily

525 oxygen cycle amplitudeAuthor Manuscript remains significant if we separate the two lakes containing hippos in a 526 one-way ANOVA with all three lake categories.

This article is protected by copyright. All rights reserved 527 Fig. 4- Community ordinations (non-metric dimensional scaling biplots, left column) and 528 rarefied richness for bacteria (by phylum, A and B), phytoplankton (by division, C and D), 529 zooplankton (by species, E and F), and macroinvertebrates (by families or genera, G and H). 530 Significant differences in composition were tested by PERMANOVA with 1000 randomizations 531 of the lake-by-taxon matrix. Richness was not different among any of the groups by ANOVA 532 (all P >0.1). Author Manuscript

r=0.11

r r= = 0.1 0.11 =785 =7089 =65 =4 2040 2000 4000 6000 2060 N

2019 N 1993

N ippo population size N 0 H 2000Manuscript 2020 2040Author 2060 Year