bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1 Cascading effects of elephant-human interactions in a savanna

2 ecosystem and the implications for ecology and conservation

3

4

5

6 David Western1¶ and Victor N. Mose1¶*

7

8

9

10 1African Conservation Centre, Amboseli Conservation Program, Nairobi, Kenya.

11

12

13

14 *Corresponding author

15 Email: [email protected] (VNM)

16

17

18 ¶These authors contributed equally to this work.

19 20 21 22 23 24 25

1 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

26 Abstract

27

28 Our study monitored the changes in elephant numbers, distribution and ecological impact

29 over a fifty-year period as the free-ranging intermingled movements of wildlife and

30 traditional subsistence pastoralists across the Amboseli ecosystem were disrupted by a

31 national park, livestock ranches, farms, settlements and changing lifestyles and

32 economies.

33

34 Elephants compressed into the national park by poaching and settlement turned

35 woodlands to grassland and shrublands and swamps into short grazing lawns, causing a

36 reduction of plant and herbivore diversity and resilience to extreme events. The results

37 echo the ecological findings of high-density elephant populations in protected areas

38 across eastern and southern Africa. The impact has led to the view of elephants in parks

39 being incompatible with biodiversity and to population control measures.

40

41 In contrast to Amboseli National Park, we found woody vegetation grew and plant

42 diversity fell in areas abandoned by elephants. We therefore used naturalist and

43 exclosure experiments to determine the density-dependent response of vegetation to

44 elephants. We found plant richness to peak at the park boundary where elephants and

45 livestock jostled spatially, setting up a creative browsing-grazing tension and a patchwork

46 of habitats explaining the plant richness.

47

2 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

48 A review of prehistorical and historical literature lends support to the Amboseli findings

49 that elephants and people, the two dominant keystone species in the savannas, have

50 been intimately entangled prior to the global ivory trade and colonialism. The findings

51 point to the inadequacy of parks for conserving mega herbivores and as ecological

52 baselines.

53

54 The Amboseli study underlines the significance of space and mobility in expressing the

55 keystone role of elephants, and to community-based conservation as a way to win space

56 and mobility for elephants, alleviate the ecological disruption of compressed populations

57 and minimize population management.

58

59 Introduction

60

61 After decades of physical forces being viewed as bottom-up drivers of ecosystem

62 processes (1–3) the role carnivores and herbivores in governing community dynamics

63 from the top-down has also been well established in both ecological theory (4–8) and

64 conservation policy and management (9–11). The extermination, reintroduction,

65 displacement and compression of keystone large herbivores and carnivores can all have

66 long-term repercussions on ecosystems and landscapes (12,13)

67

68 Given the growing impact humans have on wildlife (14), population disruptions offer

69 naturalistic experiments for testing the role of megamammals as keystone species

70 precipitating trophic cascade through ecosystems (7). The impact of megafaunal

3 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

71 disruptions will, however, differ between biogeographic regions depending on their

72 histories. The heavily depleted faunal assemblages of North America during the

73 Pleistocene (15) and Madagascar during the Holocene (16), for instance, will respond

74 differently to East Africa’s megafauna which has remained relatively intact since the late

75 Pleistocene (10,17). Whereas the reintroduction of the wolf into Yellowstone’s

76 Pleistocene-depleted predator fauna caused large ripple effects on the plant and animal

77 community (11), the recolonization of wild dogs on the Laikipia Plateau in Kenya (18) in

78 contrast, caused little discernable impact due to the present of other large predators,

79 including lions and hyenas. How much greater then would be impact of reintroducing

80 megaherbivores such as elephants and rhinos into North America biomes after 11,000

81 years of faunal depletion and weakened plant herbivore defenses than the reintroduction

82 of the wolf into Yellowstone?

83

84 Africa offers an indication of the far greater biotic impact of large herbivores than

85 carnivores on ecosystems. The ecological impact of the continent’s suite of

86 megaherbivores on biomes ranging from forests to savanna has been well-documented

87 (19–22). Most studies have, however, been conducted in protected areas where the

88 impact of elephants especially has been intensified by human activity causing

89 compressed populations and curtailed seasonal movements. Wall et al. (2021) showed

90 the distribution of elephants to occupy only 17% of their potential range due to the level

91 of the human disturbance and safety in protected areas. Normalized for rainfall, censuses

92 of 34 populations showed elephant densities in East Africa to be five-times higher in

93 protected than non-protected areas (24).

4 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

94 Discussions on the keystone role of elephants have centered on their impact on woody

95 vegetation and whether the findings support equilibrium, non-equilibrium or alternative

96 multi-state theories of ecosystem dynamics (25–27). Here again, using protected areas

97 as templates exaggerates the deleterious impact of elephants on biodiversity and

98 obscures the keystone role they play in free-ranging interactions with humans. The lack

99 of such studies prior to the establishment of protected areas masks the complex and

100 shifting interplay which likely shaped African ecosystems for millennia (28). Prehistorical

101 and historical reviews of the interlinked ecologies help reveal the role elephants play in

102 ecosystems when free-ranging rather than in compressed and fragmented populations.

103

104 Elephant-human interactions

105

106 Paleo records and historical accounts point to the ancient links and changing relationship

107 between hominins and elephants since the Lower and Middle Stone Age. The earliest

108 evidence of human interactions with elephants comes from the fossil bones in

109 Olorgesailie formation 150 km east of Amboseli in the Rift Valley. Here at 1 million BP,

110 the fossilized remains of Elephas recki, a species twice the weight of Loxodonta africana,

111 is surrounded by discarded stone tools and shows multiple cut marks on the bones (29).

112 It is unclear if the disappearance of Elephas recki, which coincided with the 75% turnover

113 of the large mammal fauna in the transition from the Lower to Middle Stone Age 500,000

114 to 350,000 BP, was due to climatic changes or the emergence of more sophisticated

115 weaponry and hunting techniques or both (30). The eruption of taller grasslands following

116 the loss of the largest species across a range of herbivore taxa at a time the climate was

5 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

117 drying is, however, consistent with Owen-Smith's (1987) herbivore release hypothesis

118 explaining the growth of coarse vegetation following the extinction of megaherbivores.

119 The presence of sophisticated spears found among wild horse fossils preserved in a peat

120 bog in Germany at 350,000 BP (32) adds further evidence of projectile weapons and

121 group hunting as plausible explanations for the Olorgesailie faunal turnover. The

122 transition marks the emergence of Homo sapiens, Loxodonta africana and the modern

123 large mammal fauna (29) in which the keystone roles of elephants and humans have

124 been linked over hundreds of millennia.

125 The Venus Hohle Fels figurines dating from 35,000 to 40,000 BP testify to the role of

126 mammoth ivory in prehistoric cultural practices following the arrival of modern sapiens in

127 Europe (33) and most likely among Neanderthals far earlier. Historical accounts document

128 a flourishing North African ivory trade supplying the Greeks and Romans from as early as

129 2,500 BP, and the extirpation of elephants from their north African ranges during the

130 classical times (34). The dhow trade driven by the Indian Ocean trade winds transported

131 large quantities of ivory from Africa to the Arabian Peninsula and India from 500 AD

132 onwards, reaching a peak in the mid-19th century using slaves as portage (35–37).

133 Evidence points to the depletion of elephant populations across eastern and southern

134 African by the late 19th century (37,38), resulting in the growth of woody vegetation (39).

135 Alarmed by the outlook for elephants, the eight colonial powers led by the British and

136 German governments convened the first ever wildlife conference in 1900 to save wildlife,

137 and elephants particularly, from the extinctions happening in southern Africa (40).

138

6 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

139 Historical records of Amboseli mirror the continental picture. The writings of Joseph

140 Thomson and Count von Hoehnel, the first European explorers to pass through Amboseli

141 in the 1880s, make no mention of seeing elephants (41,42). Photographs by Schillings

142 (1906) in the early 1900s show an abundance of regenerating fever trees (Acacia

143 xanthophloea) but no mature woodlands. The elephant population rose steadily from a low

144 point at the turn of the 20th century in response to the abolition of the slave trade and

145 conservation protection by the British colonial administration, to a peak in the late 1960s

146 (37,44). A repeat photographic survey of photos taken by Martin and Osa Johnson in

147 1921 (45), shows the thick regenerating groves of fever trees and a dearth of elephant

148 dung or tree damage. In the years following the trees matured and elephant damage

149 intensified, leading to the replacement of woodlands by grasslands across much of the

150 Amboseli Basin from the 1950s onwards (46).

151

152 Gazetted as Amboseli National Reserve in 1948, the Amboseli Basin at the center of the

153 reserve was renowned for its majestic fever trees framing the backdrop of Kilimanjaro (47).

154 The film, Where No Vultures Fly (48), shot extensively in Amboseli in 1950, showed

155 growing signs of elephant damage to the fever tree groves. Early wardens reported a rising

156 number of elephants in the Amboseli Basin that period (47). By the early 1960s

157 conservationists began voicing concerns over the rapid loss of the iconic yellow fever trees

158 and pressured government to declare Amboseli a national park to protect the woodlands

159 in the belief that overstocking by Maasai livestock was the cause (49). Later studies

160 suggested other causes, including elephant damage (19), rising salinity accelerated by

161 elephant damage (50), synchronized woodland growth and ageing (51) and hydrological

7 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

162 causes (52). Long-term, experimental studies (53) and ecological monitoring (54) factored

163 out elephants as the main cause of woodland loss and broader habitat changes.

164

165 Given that elephants have contracted to a fifth of available habitat due human

166 displacement (23), how did the interaction of elephants and humans play out before the

167 industrial and colonial age (28), how has their interaction change as a result, and how

168 can the space and mobility needed to avoid compression and ecological degradation be

169 reestablished?

170

171 The Amboseli Conservation Program (ACP) offers some insights on these questions. The

172 monitoring and research program began in 1967 when Maasai pastoralists still practiced

173 subsistence husbandry and moved seasonally across communally owned land. The early

174 study period preceded the division of the ecosystem into a national park and group

175 ranches (49,55). Here we draw on the half century of monitoring and research to

176 document the push and pull forces of human agencies on the movements, residency

177 times, densities, foraging patterns and complex ecological cascades they triggered by

178 elephants.

179

180

181

182

183

8 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

184 Methods

185

186 Study area

187

188 The 3,700 km² part of the Amboseli ecosystem (Fig 1) is defined by the seasonal

189 movements of the large wild herbivore and pastoral livestock populations lying in the rain

190 shadow of Kilimanjaro along the Kenya-Tanzania border. The wild herbivore migrants,

191 including elephants, zebra, wildebeest, hartebeest, eland and buffalo, move seasonally

192 between the wet season and the dry season range concentrated around the permanent

193 swamps of the Amboseli Basin. The ecosystem has been detailed in several publications

194 (56–59). The vegetation is dominated by bushed grassland falling within ecological Zone

195 V of Pratt, Greenway and Gwynne (60). Aquifers emanating in the northern forests of

196 Kilimanjaro drain into the dry Pleistocene lakebed of the Amboseli Basin, creating a series

197 of permanent swamps and a shallow water-table which supports hydrophilic vegetation

198 dominated by Acacia xanthophloea woodlands. Maximum temperatures range from 26 to

199 44°C and minimum temperatures from 6 to 14°C. The twice-yearly rain seasons fall

200 between October to December and March to May, averaging 350mm annually (61).

201 Despite the abundance of wildlife, the herbivore biomass is dominated by the cattle,

202 sheep and goats of the pastoral Maasai who followed the same seasonal migratory

203 patterns as wildlife until the late 1970s. In 1977 livestock were excluded from the 388 km²

204 Amboseli National Park to protect the rich wildlife concentrations of the Amboseli Basin

205 and secure their late season forage against settlement and farming. The Maasai in the

206 surrounding communal lands were given title deeds to seven group ranches, each

9 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

207 managed separately by elected representatives (55). The higher rainfall slopes of

208 Kilimanjaro were divided into small farms which spread downslope towards the

209 permanent swamps from the 1960s onwards. Beginning in the 1970s the permanent

210 swamps Namelog and Kimana lying east of Amboseli National Park were subdivided into

211 private holdings supporting irrigated farming. Subdivision of the group ranches, which

212 began in the northern Kaputei Section of eastern Kajiado District in the 1970s and

213 displaced wildlife including elephants (55), is currently underway in the group ranches

214 surrounding Amboseli National Park.

215

216 Herbivore monitoring

217

218 Sample aerial and ground counts of the 600 km2 dry season concentration area of the

219 Amboseli basin were conducted between 1967 and 1971 using the methods described in

220 Western (1973) and Pennycuick and Western (1972). Total aerial counts of elephants,

221 buffalo and human settlements in the basin area have been flown from 1975 onwards

222 using methods described by Western and Lindsay (1984). Aerial sample counts of the

223 8,500 km2 eastern Kajiado county covering the Amboseli ecosystem and bordering areas

224 in eastern Kajiado County, have been flown one to several times most years from 1974

225 and 2020. The counting methods have been described in detail elsewhere (64–67). The

226 area was divided on a UTM map projection into 5 by 5 km grids. Flight lines were flown

227 through the center of each grid in a north south-direction 90m above ground with two

228 back-seat observers counting all wild and domestic herbivores 25kg upwards, as well as

229 settlement structures within 150 to 200 meters. Herds too large to count by eye were

10 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

230 photographed and later counted under a binocular microscope. Population estimates

231 were derived from the 8% to 10% sample counts using the Jolly II equation (68). We used

232 Douglas-Hamilton and Burrill’s (1991) ratio of dead to live elephants to measure poaching

233 levels. The long-term changes in large herbivore populations from 1974 to 2020 have

234 been detailed in Western and Mose (2021).

235

236 Human activity, including settlements, the number and type of huts and presence of

237 farming in a grid, was recorded for each grid square. Traditional Maasai dung huts are a

238 useful measure of seasonally mobile settlements, and thatch and tin huts and houses of

239 permanent settlements. We use settlement type, numbers and proportion of a grid

240 covered by farms as measure of human activity rather than human population size, given

241 that livestock and farms most directly influence the numbers and distributions of wildlife

242 (65,69).

243

244 Vegetation monitoring

245

246 The detection of woodland change in the Amboseli Basin was drawn from data measuring

247 tree density in 30 randomly distributed 18ha plots demarcated on aerial photos dated

248 1950, 1961, 1967 and 1980. Tree condition was measured in 18 plots using the ratio of

249 dead to live limbs and elephant damage based on the proportion of stripped bark. The

250 methods were described by Western and Van Praet (1973). A further count of tree density

251 was conducted along the two transects using 2020 Google Earth satellite imagery. The

252 density of elephants along the transects were averaged from monthly total counts of the

11 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

253 Amboseli Basin between 1975 to 1988 using 1km2 grid. Further woodlands density and

254 associated vegetation composition studies were conducted in 1988 along transects

255 running east and west at 0.75km intervals from the center of the Amboseli Basin. Tree

256 cover and bush density by species was measured by the point-center-quarter method

257 (70) and herb layer composition by the slanting pin frame method (71). More detailed

258 descriptions of the vegetation structure, composition and long-term changes can be found

259 in Western et al. (2021).

260

261 Results

262

263 Changes in distribution and population size

264

265 The decadal changes in elephant distribution from the 1970s to 2000s are shown in Fig

266 2. The distribution maps confirm the earlier findings of Western and Lindsay (1984)

267 depicting two overlapping populations, one centered on Amboseli Basin, the other on

268 Tsavo West and Chyulus Hills National Parks. In the 1980s the two populations

269 contracted sharply and by the 1990s had split into two discrete populations concentrated

270 in Amboseli and Tsavo West National Park. By the 2000s both populations had

271 substantially recovered their 1970s ranges and overlapped once again east of Amboseli.

272 The annual range of the Amboseli sub-population based on the kernel density maps

273 declined from 7159 km2 in the 1970s to 1984 km2 in the 1980s, a 72% shrinkage. The

274 range subsequently expanded to 5916 km2 in the 1990s and 6,349 km2 in the 2000.

12 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

275 Satellite collared animals show elephants ranging into Tanzania and across to the Rift

276 Valley (73) by 2018, a far more extensive range than elephants radio tracked in the 1970s

277 (63).

278

279 Fig 2 shows the elephant numbers and carcass ratios in the 8,500km2 ecosystem based

280 on sample aerial counts from 1974 to 2020. The numbers in the 600km2 dry season range

281 are based on sample ground counts and aerial counts from 1968 to 1971, and total aerial

282 counts from 1975 onwards. The counts are averaged for each year where more than one

283 count was conducted.

284

285 Elephant numbers in the ecosystem fell from between 1,200 and 1,500 in the early 1970s

286 to under 500 between 1974 and 1978 (Fig 2). We attribute the precipitous loss to poaching

287 for several reasons. First, at the onset of the aerial counts of the ecosystem, large

288 numbers of recently dead elephants with their tusks hacked out were found across the

289 ecosystem outside Amboseli National Park. Carcass counts were added to the live animal

290 counts to document the numbers, distribution and approximate age of dead elephants.

291 Second, the population decline is inversely related to the poaching rate measured by the

292 carcass: live ratio (74) over the previous one year (푟 = ―0.37, 푃 = 0.0453). Third, the peak

293 number of elephant carcasses recorded from the aerial counts at the height of poaching

294 in 1976 numbered 536±102. Allowing for many undetected carcasses due to

295 decomposition and scattering by predators over the previous three years, and others

296 hidden under dense bush, the number of poached elephants explains most of the

297 population losses. Fourth, the recovery of the elephant population coincides with the

13 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

298 decline in poaching (Fig 2). Finally, the timing of the poaching surge matches the

299 wholesale slaughter of elephants throughout Kenya and eastern and central Africa

300 following a surging increase in ivory prices on the world market in the early 1970s (74,75).

301

302 The distribution of carcasses during the poaching surge of the 1970s (Fig 3) depicts the

303 most vulnerable locations: in the west in the dense bush country stretching to Namanga,

304 in the north and northeast on the fringes of the elephant range, and in the east on the

305 Chyulu Hills and southeast on Rombo Group Ranch, both areas where elephants spread

306 out from Tsavo East National Park. A small number of elephants were poached south of

307 the park boundary and the few carcasses were recorded in the park, largely young

308 animals which succumbed to the mid-1970s drought (54). In contrast to the widespread

309 abundant carcasses recorded in the 1970s. the number dropped to a few incidental

310 records in the 2000s (Fig 2).

311

312 We attribute the rising numbers of elephants in the park through the 1970s and early

313 1980s (Fig 2) to elephants fleeing poachers to the safety of the national park (59) where

314 the concentration of tourist vehicles and ranger forces offered protection. The safety

315 factor was also evident in the relaxed behavior of herds in the park and their clumped and

316 agitated formations in moving out of the park (63) and in response to human disturbances

317 (76). Most movements out of the park were nocturnal, a pattern common to other

318 protected area populations in response to human threats (23).

319

14 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

320 Another factor which may have added to the safety of the park was the exclusion of

321 livestock after the creation of Amboseli National Park. Though formally gazetted in 1974,

322 livestock were not banned from the park until 1978 when alternative water sources were

323 made available (77). By then, elephant numbers had been rising rapidly since the early

324 1970s (Fig 4). Although the Maasai moved back into the park in 1983 after the

325 government failed to honor an annual payment from park fees (78), elephant numbers

326 continued to rise through to the 1990s when the seasonal migrations resumed though in

327 reverse. Elephants migrated into the park during the rains and left in growing numbers

328 with the advance of the dry season. The dry season exodus continued through the 1990s

329 and into the 2000s (Fig 4). The growing population and resumption of migrations saw

330 elephants recolonized their former range (Fig 4) due largely to a drop in poaching

331 following the deployment of community rangers across the surrounding group ranches.

332 The reversal in the wet season migrations typifying elephants, zebra, wildebeest and

333 livestock in the 1960s and 1970s point to explanations other than poaching. We explore

334 alternative explanations after looking at the habitat changes caused by elephant

335 compression into the park.

336

337 Elephant protection in Amboseli National Park and surrounding group ranches spurred a

338 recovery of the population from under 500 in the late 1970s to over 1,200 in the early

339 2000s. The recovery recorded by the aerial counts matches the growth in population to

340 1,500 in 2008 based on individually known animals documented by the Amboseli

341 Elephant Program, many of which move beyond the ACP monitoring area (52). Despite

342 the heavy losses of wildlife (54) and over 400 elephants in the drought of 2009 (Moss

15 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

343 pers comm), the population continued to expand to nearly 1900 by 2020 (79), higher than

344 the peak numbers recorded in the 1970s. Satellite tracking showed the enlarged

345 population moving well into Tanzania and the Rift Valley (80), far further than radio-

346 tracked elephants ranged in the 1970s (63).

347

348 Changes in seasonal migrations

349

350 From the 1960s through the late 1970s elephants spread across the ecosystem during

351 the rains and returned to the Amboseli Basin in the dry season, a regular seasonal

352 migration also followed by wildebeest, zebra, buffalo and Maasai livestock (81). The

353 seasonal movements track a plant quality gradient from the outlying bushlands during the

354 rains to the woodlands of the Amboseli Basin in the early dry season and swamp-edge

355 and swamp habitats in the late dry season (63,82) as water sources and forage declined

356 in the dispersal areas. The detailed feeding energetics of elephants studied by Lindsay

357 (2011) shows a high rate of intake on taller grasses in the rains and a shift to woody

358 vegetation and swamp sedges in the dry season. The seasonal movement of elephants

359 and other migrants along the quality gradient has been explained by optimum foraging

360 theory and allometric diet quality selection (84–86).

361

362 In a departure from the seasonal migrations which have continued unbroken by other

363 species, elephant migrations halted abruptly with the heavy ivory poaching of the 1970s.

364 The halt in wet season migrations caused a sharp rise in mean monthly population in the

365 Amboseli National Park population (Fig 4). The change in seasonal migrations is captured

16 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

366 in the monthly total counts of elephants in the Amboseli Basin. Between 1968 to 1978,

367 prior to cessation of migrations, wet season counts of the Basin were significantly lower

368 than rain season counts (W  32, P  0.0259) . From 1980 to 1983 when the migrations

369 halted, wet and dry season counts showed no significant difference (W  4, P  0.3429) .

370 From 1984 to 2005 the migrations resumed but in reverse, with the dry season

371 significantly lower than the wet season (W  42, P 0.0001) . The seasonal differences

372 became more erratic between 2006 and 2020 as peak numbers in the Basin oscillated

373 between wet and dry seasons (Fig 4).

374

375 The population slump, cessation of migrations, compression of elephants into the national

376 park and defensive herding behavior of elephants leaving the park all point to the

377 disruptive impact of poaching in the 1970s. The compression was evident in the five-fold

378 increase in elephant numbers in the park and the impact on the late season forage in the

379 swamps. The growing forage shortage coincides with steady decline in elephant numbers

380 in the park from an annual peak of 618 in 1990 to a low of 218 in 2009 during an extreme

381 drought year when over 400 elephants died during the extreme drought (Moss, pers.

382 comm), along with the majority of zebra, wildebeest and livestock (54). The decline in

383 elephant numbers in the Amboseli Basin over the twenty-year period correlates with the

384 decline in plant biomass in the basin (r  0.78, P  0.01). Elephant numbers in the park

385 rebounded with the regeneration of pastures after the drought and only began to fall again

386 during the dry years running into 2017. Numbers in the park grew once more between

387 2018 to 2020 when the swamp forage rebounded with heavy rainfall years.

388

17 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

389 Compression along the seasonal foraging gradient

390

391 The concentration of elephants in Amboseli National Park in response to poaching

392 created a knock-on effect along the seasonal habitat feeding gradient. The rising numbers

393 initially built up in the woodlands, then shifted to the swamp habitats, creating a cascade

394 of plant and animal changes. Fig 5 shows the seasonal movement from bushland through

395 woodlands to swamp documented by Western and Lindsay (1984) intensifying with the

396 surge in elephant numbers in the park during the 1970s and 1980s, rising five-fold in the

397 swamps by 1995.

398

399 The loss of Acacia xanthophloea woodlands in the park saw a proliferation of low shrubs

400 dominated by Salvadora persica and Azima tetracantha in the early 1970s. As the

401 elephant numbers continued upward, Salvadora and Azima bushes were thinned and

402 heavily pruned, causing the spread of more browse-tolerant species, including the

403 saltbush, Suaeda monoica, small herbs dominated by Dicliptera albicaulis, and the tall

404 coarse grass Sporobolus consimilis. The continued rise in elephant numbers led to

405 growing browsing pressure on Suaeda and a gradual shrinkage in Suaeda-dominated

406 shrublands in the early 1990s (87). In response to the loss of woody vegetation, elephant

407 density in the swamps rose sharply (Fig 5). The shift from woodland to swamps correlates

408 inversely with the depletion of total woody biomass in the Basin between 1975 and 1995

409 (r  0.49, P  0.01).

410 The heavy feeding and trampling of elephants in the swamps as mean monthly numbers

411 rose from 83 to 392 over the twenty years from 1975 saw fringing woodlands and 2m tall

18 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

412 thickets of Solanum incanum, Pluchia dioschordes and Withania somnifera herbaceous

413 cover replaced by a dense sward of Cynodon dactylon and Digitaria scalarum grasses.

414 The elephants also pushed deeper into the 2-3m tall Cyperus papyrus, Cyperus

415 immensus and Typha domingensis sedges in the permanent swamps, resulting in the

416 invasion of C. laevaticus and C, merkii and floating mats of herbs such as Hydrocotyle

417 ranunculoides, Ludwigia stolonifera and the Nile cabbage, Pistia stratoites (88). By the

418 early 1990s the intensive use of the swamps had opened up large stretches of open water

419 and floating banks of vegetation. By the 2000 the heavy impact of elephants on the

420 swamps caused an outward spread across the Amboseli Basin (Fig 6) and beyond (Fig

421 1), an outward spread also documented by the Amboseli Elephant Research Project (80).

422

423 Elephant impact on the Amboseli woodlands

424

425 The loss of woodlands in Amboseli mirrors the rising elephant numbers from sporadic

426 sightings in the early 1950s (47) to the large numbers presently. A study by Western and

427 Van Praet (1973) showed elephants destroying A. xanthophloea by debarking mature

428 trees, pushing over saplings and plucking seedlings. Although they attributed the cause of

429 woodland loss to salinization accelerated by elephant damage at the time, later reanalysis

430 of their data using partial correlation discounted salinization. A 20-year multifactor

431 exclusion experiment showed elephants to be the primary cause of loss and the lack of

432 regeneration (53).

433

19 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

434 The close correlation between elephant density, tree damage and tree death found by

435 Western and Van Praet (1973) makes the fever tree a useful indicator species for tracking

436 the ecological impact of elephants in the Amboseli Basin over the last seventy years using

437 aerial photographic coverage. We selected photo plots on 1950, 1980 and Google Earth

438 imagery (89) for 2020 to quantify the changes in woodlands density along transects

439 running east and west from the Basin center at 1-kilometer intervals. The results (Fig 7)

440 show fever woodlands in 1950 were dense and relatively uniform across the basin. By

441 1980 tree density had fallen steeply in the national park, corresponding to the sharp rising

442 elephant numbers in the Basin (Fig 4). Tree survivorship increased with distance from the

443 boundary and inversely proportional to elephant density (rs  0.67, P  0.0064) . No trees

444 remained standing in the park by 2020, except for two small groves, one in Ol Tukai Orok

445 partially protected by the palm tree, Phoenix reclinata, the other bordering the Enkongo

446 Narok swamp close to dense Maasai settlements (87). Despite the extinction of

447 woodlands across most of the Basin over the seventy-year period, the distant-most

448 groves outside the park remained vibrant.

449

450 A repeat measure correlation analysis shows that the declining woodland loss from the

451 park center to basin periphery correlates highly with elephant density

452 (rrm  0.62, P  0.0298) , confirming the density-dependent tree damage found by Western

453 and Van Praet (1973). The persistence of dense mature woodlands in the eastern

454 periphery of the Basin among the irrigated farms, and in the western periphery where

455 poaching and hunting across the border in Tanzania deterred elephants for several

20 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

456 decades, points to the mitigation of their impact in woodlands where they are displaced

457 by human activity.

458

459 High-level electric fences have shown how rapidly woodlands can be restored when

460 elephants are excluded. A twenty-year experiment from 1981 and 1990 showed rapid

461 tree recovery to maturity (S1 Fig) when elephants were excluded by a high-level electric

462 fence allowing in meso-browsers (53). Later experimental exclosures showed similar

463 results. These included a 0.4ha elephant exclosure built at Ilmarishari (S2 Fig) inside the

464 park by ACP in 2001, two constructed by Kenya Wildlife Service (KWS) at Simek and Ol

465 Tukai Orok in 2010, and two outside the park at Soito Nado by Ker and Downey Safaris

466 between 2000 and 2010. All three elephant exclosures grew dense groves of fever trees

467 within five years. The density of young regenerating trees in the ten-year KWS

468 exclosures far exceeded the density in the mature 1950 groves. Trees in the ACP

469 exclosure have matured, thinned and are approaching the density of the mature groves

470 counted on the 1950 aerial photos. Fifteen other exclosure across basin established by

471 KWS and tourist lodges since 1990 show strong tree regrowth against the continuing

472 loss of woodlands and woody vegetation in the park.

473

474 The density-dependent tree damage creating an outward wave of woodland loss from the

475 basin center to periphery (Fig 7), coupled with the recovery of woodlands in the elephant-

476 proof exclosures, underscores the impact of elephants on the acacia woodlands. We next

477 look at the wider impact on habitats and species composition.

21 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

478

479 Changes in habitat and species composition

480

481 The most significant habitat changes in Amboseli since the 1950s have been described

482 in detail in Western (2007) and Croze and Lindsay (2011). From 1950 to 2017 grassland

483 habitats expanded from 28% to 40% of the Basin area in inverse proportion to the

484 contraction of woodlands from 25% to 5% (r  0.96, P  0.0001). The fever tree

485 woodlands were supplanted by a Suaeda-dominated shrubland. The heterogeneity in

486 woodland structure and species composition converged over the period from the 1970s

487 to 2017, due largely to the woody mass declining across the Basin from 600 gm-2 to 200

488 gm-2 (72). The permanent swamps increased by two-fold, switching from tall to short

489 sedges and banks of floating weed and large stretches of open water (88).

490 Both naturalistic and experimental exclusion experiments can be used to deduce the role

491 of elephants in the changing habitats and species composition since the 1950s.

492 In the naturalistic experiment we used the 1988 combined ground transects running east

493 and west from the park center to Basin periphery (Fig 7) to test whether grass and shrub

494 biomass increased with declining elephant density along the transects. We found both

2 495 the ratio of grass (푦 = ―0.6푥 + 9.3, 푅 = 0.54, 푃 = 0.0061) and shrubs (푦1

496 = ―0.28푥 + 2.6, 푅2 = 0.49, 푃 = 0.0118) to decline in proportion to tree biomass density

497 (See Appendix A1)

498 In our controlled experimental manipulation of elephant impact, we used the high-wire

499 exclosures in the former woodland areas to test whether elephant exclusion would revert

22 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

500 the Suaeda-dominated shrublands and to fever tree woodlands. The ratio of both shrub (

2 501 푦푠 = ―0.9푥푒 +1795, 푅 = 0.41, 푃 = 0.0341) and grass biomass

2 502 (푦푒 = ―7.42푥푒 + 14881, 푅 = 0.38, 푃 = 0.0158) fell sharply with the increase in tree

503 biomass (See Appendix A1). By 2015 the woodlands had formed a nearly closed canopy

504 of mature trees. The shade-intolerant Suaeda bushes grew tall and thin in the woodland

505 understory and, except for a few in the light gaps, disappeared from the maturing

506 woodlands by 2015.

507 The highwire Ilmarishari electric fence, which also enclosed a swamp heavily grazed and

508 trampled down as with other swamps in Amboseli, was used to test whether elephant

509 exclusion would reestablish the tall- sedge swamps suppressed by heavy grazing and

510 trampling as well as the fever tree woodlands. The enclosed swamp was monitored for

511 changes in plant biomass and composition between September 2002 and July 2005 (88).

512 An adjacent unfenced swamp 250 distance was monitored as a control. Despite no

513 change in the control swamp, the residual Cyperus papyrus and Cyperus immensus

514 sedges began growing back in the center of the experimental exclosure within a year,

515 even as zebra and wildebeest maintained the short grazing lawns along the shallow

516 swamp margins dominated by the grasses Cynodon dactylon and Digitaria scalarum (88).

517 The exclosure showed a full recovery of 3m-tall sedges within five years of exclosure.

518 Sakar (88) concluded that elephants play a major role in the ecological modification of

519 wetlands as well as woodlands in Amboseli and more widely in Africa.

520 The changes in plants species composition and ecological function in the Amboseli Basin

521 over the last five decades have been detailed by Western et al. (2021). The most

522 significant changes have been a reduction in habitat diversity, convergence in species

23 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

523 composition among habitats to smaller plants, and as a result, a decrease in species

524 biomass, increased turnover rate and greater species dominance by herbivore-resilient

525 species of shrubs and grasses. Plant production per unit of rainfall and resilience also

526 declined significantly.

527 The overall changes in species richness, including trees, shrubs, herbs and graminoids

528 in response to elephant density along the combined east and west transects across the

529 Basin is shown in Fig 8. Species richness increased inversely proportional to elephant

530 density from the park center to boundary (rs  0.67, P  0.0064) , before declining sharply

531 with increasing distances from the park boundary.

532

533 Herbivore ecological cascades

534

535 The changing composition of the large herbivore community in response to habitat

536 changes in Amboseli has been tracked over the 40-year period from 1964 to 2004 using

537 the surface bone assemblages which match the living population composition of 15 large

538 herbivores species (66). The results showed a strong decline in the browsing relative to

539 grazing community and the changes in guild composition to spatially track the transition

540 from woodland to grasslands and expansion of the swamp grazing lawns.

541

542 The composition of the herbivore guilds in response to the habitat changes caused by

543 intensified elephant use of the Amboseli Basin is captured by comparing ground counts

544 in the 1960s and 2010s (Fig 9). In the 1960s elephants largely used the plains early dry

545 season and woodlands in mid to late dry season. Elephants made little use of the swamps

24 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

546 which were then dominated by tall sedges and a surrounding thicket of Solanum incanum,

547 Pluchia dioschordes and Withania somnifera herbaceous cover. The seasonal shift of all

548 herbivores through the habitats in the 1960s has been described in detail elsewhere (56).

549 Browsers, including over 100 giraffe and 800 impala were regularly counted in the

550 woodlands along with small numbers of bushbuck, lesser kudu and gerenuk. By the

551 2010s, following the loss of woodlands and spread of Suaeda bushlands, zebra,

552 wildebeest and Thomson’s gazelle predominantly foraged on the swamp-edge grazing

553 lawns created by elephants (Fig 9). Fewer than 10 giraffes entering the Basin for water

554 are ordinarily recorded on current monthly counts. Impala numbers have fallen to fewer

555 than 200, and bushbuck and lesser kudu have disappeared from the Basin.

556 The shifting use of habitats by herbivores in response to the changes in the plant

557 community over the past five decades can be analyzed by the use of a Detrended

558 Correspondence Analysis (DCA) shown in Fig 10. The main changes show the grazing

559 guild dominated by zebra and wildebeest to shift from a predominant use of the plains

560 and woodlands to a heavy concentration in the swamp habitats in association with

561 elephants. The browsing guild, including giraffe and Grant’s gazelle, shifted out of the

562 basin to the bushlands. Rhinos a dominant feature of the browser guild were exterminated

563 by poachers. Livestock use of the Basin habitats, including the central swamps, declined

564 sharply following their exclusion from Amboseli National Park in the late 1970s.

565 The changes in herbivore species composition and ecological function in the Amboseli

566 Basin over the last five decades have been detailed by Western and Mose (2021).

567 Species diversity declined with the increase in grazers and decrease in browsers, causing

568 large scale changes in the macroecology of the Amboseli herbivore community. Further,

25 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

569 despite the increase in overall wildlife production due to the expansion of swamp grazing

570 lawns, the probability of extreme forage shortfalls and herbivore population collapses

571 increased due to the heavy sustained browsing and grazing pressure depressing plant

572 resilience to extreme events (90). The macroecological changes reflect the increasing

573 dominance of elephants which rose from 8% of total herbivore biomass in the Amboseli

574 Basin in the 1960s to 29%, causing a cascade of changes in the plant and animal

575 community.

576

577

578 Discussion

579

580 The compressed and growing elephant population in Amboseli National Park has

581 transformed woodlands into shrublands and expanded the area of grasslands and

582 swamps over the last seven decades. The collapse of the migrations and compression of

583 elephants into the park caused a cascade of floristic changes, including a convergence

584 in composition typified by smaller browse-tolerant species, reduced primary production,

585 an accelerated turnover rate and reduced resilience. Changes in the herbivore community

586 included a decline in browsers associated with the woodland to grassland transformation.

587 Changes in the swamp vegetation highlight the keystone role of elephants as ecological

588 engineers in opening up wetlands to a range of smaller herbivores. Sarkar (2006) has

589 documented the effects of elephant trampling, fecal deposits, water aeration and

590 sediment churning on swamp vegetation. The intensive herbivory boosted the growth rate

26 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

591 and quality of the grazing lawn pastures, explaining the attraction of the swamps in the

592 wet seasons between the 1980s and 2000s (Fig 4) when elephants would otherwise have

593 migrated to the bushlands (83). The post 1980s decline in elephant numbers in the park

594 during the dry season when elephants previously congregated in the swamps is explained

595 by the depletion of forage due to heavy grazing (54), culminating in the heavy elephant

596 and ungulate mortality in the 2009 drought (90).

597 We attribute the expansion of the swamps to the reduced transpiration rates and

598 increased puddling caused by heavy grazing and trampling of the tall sedges, by flash

599 floods into the Basin caused by sedentarization and heavily grazing in the surrounding

600 group ranches (69), and perhaps to increased discharge of aquifers into the Basin from

601 Kilimanjaro (52).

602 The impact of elephants on the Amboseli swamps mirrors Vesey-FitzGerald (1960)

603 account of the grazing succession in Lake Rukwa in Southern Tanzania. Here, elephants

604 trample and graze down tall sedges, creating a succession of smaller herbivores onto the

605 newly created grazing lawns as the dry season progresses. The Amboseli findings also

606 match reports of high-density elephant populations in protected areas. Guldemond and

607 Van Aarde (2008) in a review of 238 studies and a meta-analysis of 21 research sites

608 across Africa found high densities, amplified by low rainfall and fencing, to reduce woody

609 vegetation.

610 The Amboseli points to a more complicated picture than compressed populations studies

611 when elephants are free to move in response to humans at an ecosystem and larger

612 scale. Transects across the Amboseli park boundary show a loss of plant species

613 richness at low as well as high densities due to dense woodland cover suppressing

27 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

614 understory plants. Poulsen et al. (2017) also found elephant extirpation to result in plant

615 species loss and ecosystem simplification in African forests. Based on the density-

616 dependent responses of plant species richness, we deduce that protected area research

617 and compressed populations mask the larger shifting patchwork effect of elephants as

618 landscape agents (94). The impact of compression and curtailed movements on woody

619 vegetation is analogous ecologically to the impact of heavy sustained grazing on

620 grassland production in response to land subdivision and sedentarization of pastoral

621 herds on the Kaputei group ranches north of Amboseli (69,72)

622 Although Guldemond et al. (2017) and Cook and Henley (2019) found high-density

623 elephant populations do generally cause tree loss and habitat simplification, the

624 ecological cascade effects can also be positive as in the case of the grazing success in

625 wetlands. Skarpe and Ringrose (2014), in a detailed multispecies study of Chobe National

626 Park in Botswana, added to the complex view in showing elephants to cause a wide range

627 of cascade effects as a function of distance from water. Heavy localized impact of

628 elephants on vegetation near water courses and waterholes caused an expansion of short

629 grasslands, increase in meso-herbivores, large carnivores, small mammals and

630 gallinaceous birds. Further from water lower elephant densities resulted in heavy bush

631 and woodlands cover. The study also found elephant impact to be moderated by bottom-

632 up geomorphology, soils and hydrology.

633 Fritz et al. (2002) in investigating the impact of megaherbivores on the guild structure

634 across 31 African ecosystems found elephants to compete with meso-browsers and

635 mixed feeders but not grazers, a finding echoed in our study. In a yet broader review of

636 the role of megaherbivores, Bakker et al. (2016) combined paleo-data and exclosure

28 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

637 studies to show that megaherbivore extinctions and exclusions can cause large ecological

638 cascade no less than high-density populations, in this case through an expansion of

639 woodlands and increase in species dominance.

640 The focus on protected areas and a lack of studies on free-ranging populations interacting

641 with humans at a landscape scale has fostered a view of park elephants as being

642 incompatible with biodiversity conservation. Elephant reduction programs have been

643 used to alleviate the biotic impact. It is worth noting that many studies on the impact of

644 elephants on vegetation were conducted during the 1970s and 1980s when herds were

645 heavily poached and retreating to the safety of parks, as in Amboseli. The findings point

646 out the inadequacy of protected areas in conserving biodiversity due to the compression

647 of elephants and lack of large-scale movements.

648 In contrast, Hoare and Du Toit (1999) found elephants and humans coexisted across a

649 wide-range of settings below a threshold of disturbance. They note the little attention paid

650 to the ecological interaction of elephants and people moving freely in response to each

651 other. Such evidence, and the findings from Amboseli, suggest the need for context-

652 specific studies to decipher the ecological role of unrestricted elephant populations.

653 Elephants, we suggest need far larger space and the creative tension with human

654 disturbances lacking in parks to play a positive keystone role in creating the patchwork of

655 habitats and successional stages creating the mosaic of habitats supporting biodiversity

656 in the African savannas and forests.

657 Contemporary elephant populations tell us even less about the ecological role of free-

658 ranging elephants prior to colonial parks and the rapid human population growth in Africa

29 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

659 over the last century (28,98). Laws suggests elephants in pre-colonial days shifted with

660 the loci of human activity, creating a large-scale mosaic of habitats with local

661 differentiation due to browsing and grazing and human activity. Based on a multivariate

662 20-year elephant exclusion study in Amboseli, Western and Maitumo (2004) similarly

663 suggested that elephants and livestock moving freely around the landscape set up a

664 creative tension causing a shifting patchwork of habitats.

665 The creative tension is captured by the continuous “jostling” of livestock and elephants at

666 the park boundary where species richness peaks due to the patchy habitats caused by a

667 decadal ecological minuet. At the interface, elephants thin woodlands and livestock

668 reduce grass and encourage woody invasion (53). The hump-back curve of plant species

669 richness along elephant gradient (Fig 8) reflects the interplay of safety and threat

670 responses of elephants to humans. The interplay explains the hump-back curve of plant

671 species across the park boundary (Fig 6) where human-moderated elephant and livestock

672 movements create a shifting tapestry of patchy habitats.

673 The species richness along the Amboseli elephant gradient fits the Intermediate

674 Disturbance Hypothesis (99), as modified by the Milchunas-Sala-Lauenroth (MSL)

675 models (100) along a productivity gradient. Although the literature shows elephants

676 depleting woodlands and creating browse traps of dwarfed woody vegetation in high

677 density park populations (101), the impact of short-term intensive browsing has yet to be

678 studied at an ecosystem scale.

679 Our observations in Amboseli suggest that short intense browsing due to seasonal

680 elephant movements and in response to human activity may stimulate woody production

681 and diversity in much the same way that short-term intensive grazing can create browsing

30 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

682 arenas which stimulate grass production (102–104) and species richness (100). Our

683 observations of giraffe browsing lend support to this view. Heavy giraffe browsing on

684 young fever tree stands creates a dense thicket of thorns which prevent elephants

685 penetrating the barrier and debarking and trees until they reach near maturity, the apical

686 branches reach above giraffe height and the lower densely thorned branches die back.

687 The context-specific impact of elephants points to the futility of separating the ecological

688 role of elephants from people, treating parks as natural ecological systems and ignoring

689 the importance of mobility and scale in ecological cascades (7,8). The million years or

690 more of human-elephant coevolution has shaped the ecological, behavioral and cultural

691 adaptations in elephants and people. Elephants readily distinguish human responses by,

692 for example, foraging beyond park boundaries at night to avoid people (23), yet acting

693 benignly around safari vehicles, lodges and campsites in parks.

694 The response of people to elephants similarly varies, most notably in the cultural traditions

695 and familiarity with elephant behavior. In Amboseli Maasai pastoral practices and

696 tolerance of wildlife has fostered coexistence (105), whereas Maasai who have taken up

697 farming and suffer crop raiding have become fearful and intolerant of elephants (106). In

698 the same vein, human activity globally, whether promoting ivory trading for profit, or

699 elephant protection and viewing arising out of modern conservation sensibilities, have

700 become the predominant forces shaping the ecological cascades.

701 Other factors bearing on how elephants and people shape ecosystems include physical

702 factors such as rainfall (20), landscape heterogeneity (107), the size of parks, distance

703 from water (96) and seasonal migrations. Rainfall seasonality as much as total rainfall

704 may also influence the impact of elephants on vegetation in eastern Africa. Here the

31 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

705 bimodal rainfall supports shorter sparse bushlands and scattered grasslands in contrast

706 to the taller denser deciduous woodlands of southern Africa’s unimodal rainfall regime.

707 The loss of elephant mobility and range and the rising conflict has reduced the scope for

708 sharing land with people. The loss of coexistence has shifted the interaction from a

709 positive perturbation promoting heterogeneity to ecological disruption and uniformity.

710 Caughley (1988) in reviewing the causes and consequences of local overabundance in

711 mammals suggested the two exceptions to the reversion of vegetation from temporary

712 overabundance are due to the loss of mobility in livestock and elephants. O’Connor et al.

713 (2007) goes further in suggesting elephants to be predominantly grazers based on their

714 digestive physiology but increasingly browsers under range compression in protected

715 areas.

716 We propose that prior to the ivory trade the creative tension of free-ranging elephants in

717 response to humans created a patchwork of habitats and a shifting mosaic of vegetation

718 consistent with the view of elephants as a keystone species and ecosystem engineers.

719 The impact of the ivory trade over the last few centuries (37,38), a breakdown in traditional

720 lifestyles and cultural practices in land husbandry which sustained biodiversity over the

721 last 12,000 years until industrial times (109), and the growth of protected areas, has

722 subsequently compressed elephant ranges and caused a growing ecological dislocation

723 by disrupting the keystone species (110).

724 Tempering the ecological dislocations of mega-mammals caused by local overabundance

725 (111), reviving their keystone role in sustaining biodiversity (112), and reducing human-

726 wildlife conflict impeding ecological upscaling (18,113) can be achieved naturally by

727 ensuring elephants and other large herbivores and carnivores have the space needed to

32 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

728 maintain large self-sustaining populations. This is a formidable challenge though by no

729 means insurmountable.

730 The widening horizons of conservation in the last few decades have seen protected and

731 wildlife management areas win space for large mammals through community-based

732 conservation initiatives (114,115). Cumming et al. (2013) note that matching the scale of

733 human demands on an ecosystem and the services the ecosystem can deliver calls for

734 appropriate institutional frameworks for the scale-matching.

735 Examples of scaling up from park to ecosystem and the expanded institutional

736 frameworks include the cross-border Kruger link between South Africa and Mozambique

737 (117), the greater Amboseli ecosystem in Kenya (118) and the Greater Yellowstone

738 Coalition in the U.S. (119). Yet wider regional linkages include the Paseo Pantera Central-

739 South America landscape (120), the Yellowstone to Yukon landscape across the U.S.-

740 Canadian border Chester, 2015), and the four-borders Kavanga-Zambezi (KAZA)

741 landscape in southern Africa (121). The theoretical framework for widening to regional

742 and continental levels in the face of human impact and climate change, have been

743 highlighted by Soulé and Terborgh (1999), Allen and Singh (2016) and Curtin (2015).

744 Such examples hold out hope of finding the space and the mobility elephants and other

745 large herbivores and carnivores need to play vital keystone roles in sustaining biodiversity

746 in an increasing human-dominated world, in alleviating the ecological disruption of

747 compressed populations in parks, and in minimizing the need for intense species

748 population management.

749

33 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

750 Appendix A1 751

752 The ratio of grass to tree biomass density (푦) reduced by 0.6 for every additional kilometer

753 (푥) away from the centre of the park, (푦 = ―0.6푥 + 9.3, 푅2 = 0.54, 푃 = 0.0061) . For

754 shrubs, the ratio (푦1) reduced by 0.28 with a unit increase in distance (푦1 = ―0.28푥 + 2.6,

755 푅2 = 0.49, 푃 = 0.0118).

756 In the controlled experimental manipulation of elephant impact, the ratio of grass to tree

2 757 biomass density (푦푒) reduced by 7.42 over the years (푥푒) , (푦푒 = ―7.42푥푒 +14881, 푅

2 758 = 0.38, 푃 = 0.0158) . That of shrubs (푦푠) reduced by 0.9 (푦푠 = ―0.9푥푒 +1795, 푅

759 = 0.41, 푃 = 0.0341).

760

761 Acknowledgments

762

763 We wish to thank the many wardens and the staff of Kenya Wildlife Service for their

764 support of the Amboseli Conservation Program over the years. David Maitumo assisted

765 in data collection for the experimental exclosures and plant monitoring. Eric Ochwangi,

766 Rebecca Kariuki and Caroline Mburu helped analyze elements of the long-term data set

767 we draw on in this paper. Winfridah Kemunto has compiled much of the field data for

768 computer analysis. We thank Shirley Strum for her review of the manuscript.

769

770

34 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

771 Author Contributions

772

773 David Western has designed, planned and directed the field programs of the Amboseli

774 Conservation Program since its inception in 1967. Victor N. Mose has been responsible

775 for data management, statistical analysis and graphical presentation the data used in this

776 paper. Both authors contributed to the conceptualization of the paper and divided writing

777 assignments. Western was responsible for the literature review and final manuscript

778 writing, Mose for the preparation for publication.

779

780 Financial disclosure

781

782 The ACP study has been funded since its inception by many organizations, principally

783 Ford Foundation (www.fordfoundation.org), Wildlife Conservation Society

784 (www.wcs.org), and the, Liz Claiborne Art Ortenberg Foundation (www.lcaof.org),. We

785 thank them all for supporting specific components of the study and the long-term support

786 of the ACP field operations. The funders have had no role in study design, data collection

787 and analysis, decision to publish, or preparation of the manuscript.

788

789 Conflict of interest

790

791 The authors have declared that no competing interests exist.

35 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

792 Data Availability Statement

793

794 The extensive data used in this paper was collected and is retained by the Amboseli

795 Conservation Program at the African Conservation Centre, Nairobi, and can be

796 requested via www.amboseliconservation.org or through the director and Database

797 Administrator of the Amboseli Conservation Program at ([email protected]|

798 [email protected]).

799

800 Ethics statement 801

802

803 Kenya Wildlife Service (KWS) issued permits to conduct field work inside the protected

804 Amboseli National Park (latitude: -2.626610; longitude: 37.2544060). For study sites

805 located in the surrounding group ranches (Fig 1), no endangered plant species were

806 involved. Aerial sample surveys of large herbivores were conducted in line with the Kenya

807 Government, Department of Resource Surveys and Remote Sensing (DRSRS)

808 methodology.

809

810

811

812

813

36 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

814 References

815

816 1. Lindeman RL. The trophic‐dynamic aspect of ecology. Ecology. 1942;23(4):399–417.

817 2. Odum HT. Trophic structure and productivity of Silver Springs, Florida. Ecol Monogr.

818 1957;27(1):55–112.

819 3. Whittaker RH, Likens GE. The biosphere and man. In: Primary productivity of the biosphere.

820 Springer; 1975. p. 305–28.

821 4. Paine RT. A note on trophic complexity and community stability. Am Nat. 1969;103(929):91–3.

822 5. Oksanen L, Fretwell SD, Arruda J, Niemela P. Exploitation ecosystems in gradients of primary

823 productivity. Am Nat. 1981;118(2):240–61.

824 6. Schoener TW, Spiller DA. Effect of lizards on spider populations: manipulative reconstruction of a

825 natural experiment. Science (80- ). 1987;236(4804):949–52.

826 7. Terborgh J, Holt RD, Estes JA, Terborgh J, Estes J. Trophic cascades: what they are, how they

827 work, and why they matter. Trophic cascades predators, prey, Chang Dyn Nat. 2010;1:18.

828 8. Curtin CG, Allen TFH. Complex ecology: foundational perspectives on dynamic approaches to

829 ecology and conservation. Cambridge University Press; 2018.

830 9. Seip DR. Factors limiting woodland caribou populations and their interrelationships with wolves

831 and moose in southeastern British Columbia. Can J Zool. 1992;70(8):1494–503.

832 10. Sinclair ARE, Norton-Griffiths M, editors. The eruption of the ruminants. In: Serengeti: Dynamics of

833 an ecosystem. 1979. p. 82–103.

834 11. Eisenberg C. The wolf’s tooth: keystone predators, trophic cascades, and biodiversity. Island

835 Press; 2013.

836 12. Owen-Smith RN. Megaherbivores: the influence of very large body size on ecology. Cambridge

37 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

837 university press; 1988.

838 13. Bakker ES, Gill JL, Johnson CN, Vera FWM, Sandom CJ, Asner GP, et al. Combining paleo-data

839 and modern exclosure experiments to assess the impact of megafauna extinctions on woody

840 vegetation. Proc Natl Acad Sci. 2016;113(4):847–55.

841 14. Barrett M, Belward A, Bladen S, Breeze T, Burgess N, Butchart S, et al. Living planet report 2018:

842 Aiming higher. 2018;

843 15. Grayson DK. The chronology of North American late Pleistocene extinctions. J Archaeol Sci.

844 1989;16(2):153–65.

845 16. Burney DA, Robinson GS, Burney LP. Sporormiella and the late Holocene extinctions in

846 Madagascar. Proc Natl Acad Sci. 2003;100(19):10800–5.

847 17. Reid RS. Savannas of our birth: people, wildlife, and change in East Africa. Univ of California

848 Press; 2012.

849 18. Woodroffe R, Thirgood S, Rabinowitz A. The future of coexistence: resolving human-wildlife

850 conflicts in a changing world. Conserv Biol Ser. 2005;9:388.

851 19. Laws RM. Elephants as agents of habitat and landscape change in East Africa. Oikos. 1970;1–15.

852 20. Fritz H, Duncan P, Gordon IJ, Illius AW. Megaherbivores influence trophic guilds structure in

853 African ungulate communities. Oecologia. 2002;131(4):620–5.

854 21. Toit JT, Moe SR. Elephants and ecological cascades. 2015;96(January):309–10.

855 22. Guldemond RAR, Purdon A, Van Aarde RJ. A systematic review of elephant impact across Africa.

856 PLoS One. 2017;12(6):e0178935.

857 23. Wall J, Wittemyer G, Klinkenberg B, LeMay V, Blake S, Strindberg S, et al. Human footprint and

858 protected areas shape elephant range across Africa. Curr Biol. 2021;31(11):2437–45.

859 24. Western D. The Ecological Role of Elephants in Africa. Pachyderm. 1989;

860 25. Caughley G. Overpopulation. In ‘Problems in Management of Locally Abundant Wild

38 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

861 Mammals’.(Eds PA Jewell, S. Holt and D. Hart.) pp. 7–19. Academic Press: New York; 1981.

862 26. Walker BH. Stability properties of semiarid savannas in southern African game reserves. In:

863 Problems in management of locally abundant wild mammals. Academic Press; 1981. p. 57–67.

864 27. Dublin HT, Sinclair ARE, McGlade J. Elephants and fire as causes of multiple stable states in the

865 Serengeti-Mara woodlands. J Anim Ecol. 1990;1147–64.

866 28. Laws RM. Large mammal feeding strategies and related overabundance problems. In: Problems

867 in management of locally abundant wild mammals. Academic Press New York; 1981. p. 217–32.

868 29. Potts R, Behrensmeyer AK, Faith JT, Tryon CA, Brooks AS, Yellen JE, et al. Environmental

869 dynamics during the onset of the Middle Stone Age in eastern Africa. Science (80- ).

870 2018;360(6384):86–90.

871 30. Potts R, Dommain R, Moerman JW, Behrensmeyer AK, Deino AL, Riedl S, et al. Increased

872 ecological resource variability during a critical transition in hominin evolution. Sci Adv.

873 2020;6(43):eabc8975.

874 31. Owen-Smith N. Pleistocene extinctions: the pivotal role of megaherbivores. Paleobiology.

875 1987;351–62.

876 32. Balter M. The killing ground. Sci (New York, NY). 2014;344(6188):1080.

877 33. Conard NJ. A female figurine from the basal Aurignacian of Hohle Fels Cave in southwestern

878 Germany. Nature. 2009;459(7244):248–52.

879 34. Manuella R. Orientalising ivories from Italy. In: British Archaeological Reports, Proceedings of

880 International Symposium of Mediterranean Archaeology, February 24–26, 2005. Università degli

881 Studi di Chieti; 2005. p. 111–8.

882 35. Spinage CA. A review of ivory exploitation and elephant population trends in Africa. Afr J Ecol.

883 1973;11(3‐4):281–9.

884 36. Alpers EA, Alpers EA. Ivory and Slaves: changing pattern of international trade in East Central

39 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

885 Africa to the later nineteenth century. Univ of California Press; 1975.

886 37. Coutu AN, Lee-Thorp J, Collins MJ, Lane PJ. Mapping the elephants of the 19th century East

887 African ivory trade with a multi-isotope approach. PLoS One. 2016;11(10):e0163606.

888 38. Håkansson NT. The human ecology of world systems in East Africa: the impact of the ivory trade.

889 Hum Ecol. 2004;32(5):561–91.

890 39. Leuthold W. Recovery of woody vegetation in Tsavo National Park, Kenya, 1970‐94. Afr J Ecol.

891 1996;34(2):101–12.

892 40. Simon N. Between the Sunlight and the Thunder: The Wild Life of Kenya. 1st ed. Boston:

893 Houghton Mifflin; 1963.

894 41. Thomson J. Through Masai land: a journey of exploration among the snowclad volcanic mountains

895 and strange tribes of eastern equatorial Africa. Low, Marston, Searle, & Rivington; 1887.

896 42. von Höhnel LR. Discovery of Lakes Rudolf and Stefanie: A Narrative of Count Samuel Teleki’s

897 Exploring & Hunting Expedition in Eastern Equatorial Africa in 1887 & 1888. Vol. 1. Longmans,

898 Green; 1894.

899 43. Schillings CG. With Flashlight and Rifle: A Record of Hunting Adventures and of Studies in Wild

900 Life in Equatorial East Africa. Vol. 2. London: Hutchinson; 1906.

901 44. Skarpe C, Aarrestad PA, Andreassen HP, Dhillion SS, Dimakatso T, du Toit JT, et al. The return of

902 the giants: ecological effects of an increasing elephant population. AMBIO A J Hum Environ.

903 2004;33(6):276–82.

904 45. Johnson M. Over African Jungles: The Record of a Glorious Adventure Over the Big Game

905 Country of Africa 60,000 Miles by Airplane. New York: Harcourt; 1935.

906 46. Western D. People, elephants, and habitat: detecting a century of change using repeat

907 photography. In: Repeat photography: methods and applications in the natural sciences. Island

908 Press Washington, Covelo, London; 2010.

40 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

909 47. Lovatt-Smith D. Amboseli: nothing short of a miracle. Nairobi: Kenway Publications; 1997.

910 48. Watt H, Balcon M. Where No Vultures Fly. United Kingdom: General Film Distributors (UK); 1951.

911 49. Western D. Ecosystem conservation and rural development: the case of Amboseli. In: Western D,

912 Wright RM, Strum SC, editors. Natural connections: perspectives in community-based

913 conservation. Island Press, Washington, DC; 1994.

914 50. Western D, Van Praet C. Cyclical changes in the habitat and climate of an East African

915 ecosystem. Nature. 1973;241(5385):104–6.

916 51. Young TP, Lindsay WK. Role of even‐age population structure in the disappearance of Acacia

917 xanthophloea woodlands. Afr J Ecol. 1988;26(1):69–72.

918 52. Croze H, Lindsay WK. Amboseli Ecosystem Context: Past and Present. In: The Amboseli

919 Elephants. University of Chicago Press; 2011. p. 11–28.

920 53. Western D, Maitumo D. Woodland loss and restoration in a savanna park: A 20-year experiment.

921 Afr J Ecol. 2004;42(2):111–21.

922 54. Western D, Mose VN. The changing role of natural and human agencies shaping the ecology of

923 an African savanna ecosystem. Ecosphere. 2021;12(6):e03536.

924 55. Kimani K, Pickard J. Recent trends and implications of group ranch sub-division and fragmentation

925 in Kajiado District, Kenya. Geogr J. 1998;202–13.

926 56. Western D. The structure, dynamics and changes of the Amboseli Ecosystem. University of

927 Nairobi; 1973.

928 57. Western D, Dunne T. Environmental aspects of settlement site decisions among pastoral Maasai.

929 Hum Ecol. 1979;7(1):75–98.

930 58. Western D, Nightingale DL. Environmental Change and the Vulnerability of Pastoralists to

931 Drought: A Case Study of the Maasai in Amboseli, Kenya. In: Unidas N, Programa para el

932 desarrollo, United Nations Environmental Program, editors. Africa Environment Outlook Case

41 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

933 Studies: Human Vulnerability to Environment Change. London: Earthprint Limited; 2004. p. 31–50.

934 59. Moss CJ, Croze H, Lee PC. The Amboseli elephants: a long-term perspective on a long-lived

935 mammal. University of Chicago Press; 2011.

936 60. Pratt DJ, Greenway PJ, Gwynne MD. A classification of East African rangeland, with an appendix

937 on terminology. J Appl Ecol. 1966;369–82.

938 61. Altmann J, Alberts SC, Altmann SA, Roy SB. Dramatic change in local climate patterns in the

939 Amboseli basin, Kenya. Afr J Ecol. 2002;40(3):248–51.

940 62. Pennycuick CJ, Western D. An investigation of some sources of bias in aerial transect sampling of

941 large mammal populations. Afr J Ecol. 1972;

942 63. Western D, Lindsay WK. Seasonal herd dynamics of a savanna elephant population. Afr J Ecol.

943 1984;22(4):229–44.

944 64. Western D. An aerial method of monitoring large mammals and their environment (with a

945 description of a computer program for survey analyses)-Project working document 9. 1976.

946 65. Western D, Groom R, Worden J. The impact of subdivision and sedentarization of pastoral lands

947 on wildlife in an African savanna ecosystem. Biol Conserv. 2009;142(11):2538–46.

948 66. Western D, Behrensmeyer AK. Bone assemblages track animal community structure over 40

949 years in an African savanna ecosystem. Science (80- ). 2009;324(5930):1061–4.

950 67. Mose VN. Mathematical modelling of the dynamics of migrations for large mammal populations in

951 the Amboseli National Park, Kenya [Internet]. Paris 6; 2013. Available from:

952 http://www.theses.fr/2013PA066161

953 68. Jolly GM. Sampling methods for aerial censuses of wildlife populations. East African Agric For J.

954 1969;34(sup1):46–9.

955 69. Groom RJ, Western D. Impact of land subdivision and sedentarization on wildlife in Kenya’s

956 southern rangelands. Rangel Ecol Manag. 2013;66(1):1–9.

42 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

957 70. Cottam G, Curtis JT. The use of distance measures in phytosociological sampling. Ecology.

958 1956;37(3):451–60.

959 71. Jonasson S. Evaluation of the point intercept method for the estimation of plant biomass. Oikos.

960 1988;101–6.

961 72. Western D, Mose VN, Maitumo D, Mburu C. Long-term changes in the plant ecology of an African

962 savanna landscape and the implications for ecosystem theory and conservation management.

963 Ecol Process [Internet]. 2021;10(1):15. Available from: https://doi.org/10.1186/s13717-021-00286-

964 5

965 73. Osipova L, Okello MM, Njumbi SJ, Ngene S, Western D, Hayward MW, et al. Fencing solves

966 human‐wildlife conflict locally but shifts problems elsewhere: A case study using functional

967 connectivity modelling of the African elephant. J Appl Ecol. 2018;55(6):2673–84.

968 74. Douglas-Hamilton I. African elephants: population trends and their causes. Oryx. 1987;21(1):11–

969 24.

970 75. Caldwell JR. Recent developments in the raw ivory trade of Hong Kong and Japan. Traffic Bull.

971 1984;6(2):16–20.

972 76. Kangwana KF. Elephants and Maasai: conflict and conservation in Amboseli, Kenya. University of

973 Cambridge; 1993.

974 77. Western D. In the dust of Kilimanjaro. Island Press; 1997.

975 78. Lindsay WK. Integrating parks and pastoralists: Some lessons from Amboseli. Conserv Africa

976 Peoples, Policies Pract. 1987;149.

977 79. Moss C, Lee P. Amboseli Trust for Elephants Annual Research Report 2020. 2020.

978 80. Croze H, Moss CJ. Patterns of Occupancy in Time and Space. In: The Amboseli Elephants.

979 University of Chicago Press; 2011. p. 89–106.

980 81. Western D. Water availability and its influence on the structure and dynamics of a savannah large

43 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

981 mammal community. Afr J Ecol. 1975;13(3‐4):265–86.

982 82. Lindsay WK. Habitat selection and social group dynamics of African elephants in Amboseli,

983 Kenya. University of British Columbia; 1982.

984 83. Lindsay WK. Habitat Use, Diet Choice, and Nutritional Status in Female and Male Amboseli

985 Elephants. In: Moss C, Croze H, Lee PC, editors. The amboseli elephants. University of Chicago

986 Press; 2011. p. 51–73.

987 84. Illius AW, Gordon IJ. Modelling the nutritional ecology of ungulate herbivores: evolution of body

988 size and competitive interactions. Oecologia. 1992;89(3):428–34.

989 85. Kshatriya M. African savanna herbivore communities: a model based on competition for food

990 resources. University of Miami; 1999.

991 86. Mose VN, Nguyen-Huu T, Auger P, Western D. Modelling herbivore population dynamics in the

992 Amboseli National Park, Kenya: application of spatial aggregation of variables to derive a master

993 model. Ecol Complex. 2012;10:42–51.

994 87. Western D. A half a century of habitat change in Amboseli National Park, Kenya. Afr J Ecol.

995 2007;45(3):302–10.

996 88. Sarkar S. Long-and short-term dynamics of the Wetlands in the Amboseli Savanna Ecosystem,

997 Kenya. University of Waterloo; 2006.

998 89. Lisle RJ. Google Earth: a new geological resource. Geol today. 2006;22(1):29–32.

999 90. Western D, Mose VN, Worden J, Maitumo D. Predicting extreme droughts in savannah Africa: A

1000 comparison of proxy and direct measures in detecting biomass fluctuations, trends and their

1001 causes. PLoS One. 2015;10(8).

1002 91. Vesey-FitzGerald DF. Grazing succession among East African game animals. J Mammal.

1003 1960;41(2):161–72.

1004 92. Guldemond R, Van Aarde R. A meta‐analysis of the impact of African elephants on savanna

44 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1005 vegetation. J Wildl Manage. 2008;72(4):892–9.

1006 93. Poulsen JR, Koerner SE, Moore S, Medjibe VP, Blake S, Clark CJ, et al. Poaching empties critical

1007 Central African wilderness of forest elephants. Curr Biol. 2017;27(4):R134–5.

1008 94. Davies AB, Gaylard A, Asner GP. Megafaunal effects on vegetation structure throughout a

1009 densely wooded African landscape. Ecol Appl. 2018;28(2):398–408.

1010 95. Cook RM, Henley MD. The management dilemma: Removing elephants to save large trees.

1011 Koedoe African Prot Area Conserv Sci. 2019;61(1):1–12.

1012 96. Skarpe C, Ringrose S. The Chobe Environment. Elephants Savanna Woodl Ecosyst. 2014;7–29.

1013 97. Hoare RE, Du Toit JT. Coexistence between people and elephants in African savannas. Conserv

1014 Biol. 1999;13(3):633–9.

1015 98. Caughley G. A projection of ivory production and its implications for the conservation of African

1016 elephants. CSIRO, Division of Wildlife and Ecology; 1988.

1017 99. Roxburgh SH, Shea K, Wilson JB. The intermediate disturbance hypothesis: patch dynamics and

1018 mechanisms of species coexistence. Ecology. 2004;85(2):359–71.

1019 100. Gao J, Carmel Y. Can the intermediate disturbance hypothesis explain grazing–diversity relations

1020 at a global scale? Oikos. 2020;129(4):493–502.

1021 101. Staver AC, Bond WJ. Is there a ‘browse trap’? Dynamics of herbivore impacts on trees and

1022 grasses in an African savanna. J Ecol. 2014;102(3):595–602.

1023 102. McNaughton SJ. Grassland-herbivore dynamics. Serengeti Dyn an Ecosyst Univ Chicago Press

1024 Chicago. 1979;19796:46–81.

1025 103. Georgiadis NJ, Ruess RW, McNaughton SJ, Western D. Ecological conditions that determine

1026 when grazing stimulates grass production. Oecologia. 1989;81(3):316–22.

1027 104. De Mazancourt C, Loreau M, Abbadie L. Grazing optimization and nutrient cycling: when do

1028 herbivores enhance plant production? Ecology. 1998;79(7):2242–52.

45 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1029 105. Kangwana K, Browne-Nuñez C. The Human Context of the Amboseli Elephants. In: The Amboseli

1030 Elephants. University of Chicago Press; 2011. p. 29–36.

1031 106. Western D, Nightingale DLM, Mose VN, Sipitiek JO, Kimiti KS. Variability and change in Maasai

1032 views of wildlife and the implications for conservation. Hum Ecol. 2019;47(2):205–16.

1033 107. du Toit JT. Large herbivores and savanna heterogeneity. Kruger Exp Ecol Manag savanna

1034 Heterog. 2003;292–309.

1035 108. O’Connor TG, Goodman PS, Clegg B. A functional hypothesis of the threat of local extirpation of

1036 woody plant species by elephant in Africa. Biol Conserv. 2007;136(3):329–45.

1037 109. Ellis N, Smith SJ, Roland Pitcher C. Gradient forests: Calculating importance gradients on physical

1038 predictors. Ecology. 2012;93(1):156–68.

1039 110. Power ME, Tilman D, Estes JA, Menge BA, Bond WJ, Mills LS, et al. Challenges in the quest for

1040 keystones: identifying keystone species is difficult—but essential to understanding how loss of

1041 species will affect ecosystems. Bioscience. 1996;46(8):609–20.

1042 111. Jewell PA. Problems in management of locally abundant wild mammals. Elsevier; 2012.

1043 112. Estes JA, Terborgh J, Brashares JS, Power ME, Berger J, Bond WJ, et al. Trophic downgrading of

1044 planet Earth. Science (80- ). 2011;333(6040):301–6.

1045 113. Conover MR. Resolving human-wildlife conflicts: the science of wildlife damage management.

1046 CRC press; 2001.

1047 114. Anderson D, Grove R. The scramble of for Eden: Past, present and future in African conservation.

1048 In: Anderson D, Grove R, editors. Conservation in Africa: People, Policies and Practice.

1049 Cambridge, UK: Cambridge University Press; 1987. p. 1–13.

1050 115. Western D, Wright M, editors. Natural Connections: Perspectives In Community-Based

1051 Conservation. Island Press; 1994.

1052 116. Cumming GS, Olsson P, Chapin FS, Holling CS. Resilience, experimentation, and scale

46 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1053 mismatches in social-ecological landscapes. Landsc Ecol. 2013;28(6):1139–50.

1054 117. Mabunda D, Pienaar DJ, Verhoef J. The Kruger National Park: a century of management and

1055 research. In: du Toit JT, Rogers KH, Harry C. Biggs, editors. The Kruger experience: Ecology and

1056 management of savanna heterogeneity. Island Press Washington, DC; 2003. p. 3–21.

1057 118. KWS. Amboseli National Park Management Plan, 2020-2030 [Internet]. Nairobi; 2020. Available

1058 from: https://kws.go.ke/content/amboseli-national-park-management-plan-2020-2030

1059 119. Keiter RB. The Greater Yellowstone Ecosystem Revisited: Law, Science, and the Pursuit of

1060 Ecosystem Management in an Iconic Landscape. U Colo L Rev. 2020;91:1.

1061 120. Lambert JD, Carr MH. The Paseo Pantera Project: A Case Study Using GIS to Improve

1062 Continental-Scale Conservation Planning. In: GIS Methodologies for Developing Conservation

1063 Strategies. Columbia University Press; 1998. p. 138–48.

1064 121. Stoldt M, Göttert T, Mann C, Zeller U. Transfrontier conservation areas and human-wildlife conflict:

1065 the case of the Namibian component of the Kavango-Zambezi (KAZA) TFCA. Sci Rep.

1066 2020;10(1):1–16.

1067 122. Soulé ME, Terborgh J. Conserving nature at regional and continental scales—a scientific program

1068 for North America. Bioscience. 1999;49(10):809–17.

1069 123. Allen AM, Singh NJ. Linking movement ecology with wildlife management and conservation. Front

1070 Ecol Evol. 2016;3:155.

1071 124. Curtin CG. The science of open spaces: Theory and practice for conserving large, complex

1072 systems. Island Press; 2015.

1073

1074

1075

1076

47 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1077

1078 Figure Captions

1079

1080 Fig 1: Kernel Density Distributions of Elephants in Amboseli Ecosystem, Kenya 1974 to

1081 2016.

1082 Fig 2: Elephant numbers in the 8,500 km2 of eastern Kajiado spanning the Amboseli

1083 ecosystem (black) and 700 km2 dry season range (grey) plotted against carcass ratios as

1084 a measure of poaching levels (red).

1085 Fig 3: The density distribution of elephant carcasses in the 1970s at the height of

1086 poaching and after the 2000s when elephants recolonized their former range (Fig 1).

1087 Fig 4: Mean annual wet and dry season population counts of the Amboseli Basin showing

1088 a drop in migrations in late 1970s and early 1980s in response to a surge in poaching

1089 outside the national park, followed by a reversal of migrations through to 2006 when

1090 movements became more erratic in response to wet and dry years.

1091 Fig 5: Elephant habitat use in response to poaching shifted progressively along the

1092 selectivity gradient from bushlands to woodland, intensifying the use of the late season

1093 swamp-edge habitat and permanent swamp habitats in the Amboseli Basin.

1094 Fig 6: The compression of elephant into the swamp-edge and swamp habitats (Fig. 5)

1095 depleted pasture biomass, causing herds to forage more widely within and beyond (Fig.

1096 4) the Amboseli Basin by the early 1990s.

48 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1097 Fig 7: Density of fever trees in transects from center to periphery of Amboseli National

1098 Park from 1950 to 2020. Woodland density was uniform across the basin in the 1950s

1099 and declined steeply with the influx of elephants into the park in the 1970s. The pattern

1100 of woodland loss and regeneration reflects elephant impact which declined with distance

1101 from park boundary. Elephant-proof exclosures have reestablished dense regenerating

1102 groves.

1103 Fig 8: Plant species richness in relation to elephant densities along combined east and

1104 west transects from the center to periphery of the Amboseli Basin. Plant richness

1105 increases linearly with declining elephant density to the national park boundary then

1106 declines sharply at increasing distances.

1107 Fig 9: The loss of woodlands and spread of Sueada shrubland shifted elephant foraging

1108 to heavy year-round use the swamps, resulting in the tall sedges and shrubs giving way

1109 extensive grazing lawns. Browser numbers fell in response to woodland loss and grazers

1110 numbers increased as they moved onto the newly created swamp grazing lawns opened

1111 by elephants.

1112

1113 Fig 10: Detrended Correspondence Analysis (DCA) for the Amboseli large herbivores in

1114 the 1960s and 2010s. The swamps, which in the 1960s were used mostly by waterbuck,

1115 cattle and buffalo, are now dominated by elephant, wildebeest and zebra due to the

1116 transformation of the swamps from tall sedges and herbaceous cover to grazing lawns

1117 dominated the Cynodon dactylon. Browsing species have shifted to the bushlands

49 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1118 surrounding Amboseli along with livestock which have been excluded from the Basin due

1119 to the establishment of Amboseli National Park.

1120

1121 Supporting information captions

1122 1123 S1 Fig: Exclusion trench and high-wire fence established in 1981 to test the impact of

1124 elephant removal on fever tree regeneration (A). The fever trees had grown to a mature

1125 woodland by 2003 (B) and the high-wire fence expanded in 1992 to include the lodges

1126 in the background.

1127

1128 S2 Fig: A high-wire electric fence set up in 2001 at Ilmarishari exclosure (A) to test the

1129 impact of elephant removal on woodland recovery. By 2008 the regenerating fever trees

1130 had shaded out the Suaeda-dominated shrublands which supplanted the woodlands

1131 destroyed by elephants in the 1970s. The aerial view of the exclosure in 2005 (B) showing

1132 extensive fever tree regeneration and regrowth in the swamps relative to the control open-

1133 water in the foreground. By 2018 the fever trees formed a dense canopy closing out

1134 Suaeda shrub (C).

50 bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. bioRxiv preprint doi: https://doi.org/10.1101/2021.08.18.456886; this version posted August 18, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.