The historic range and drivers of decline of the Tapanuli orangutan

The Tapanuli orangutan (Pongo tapanuliensis) is the most threatened great ape species in the world. It is restricted to an area of about 1,000 km2 of upland forest where fewer than 800 animals survive in three declining subpopulations. Through a historical ecology approach involving analysis of reports and other literature from the early 1800s to 2009, we demonstrate that historically Pongo tapanuliensis lived in a much larger area, and across a much wider range of habitat types than now. Its current Extent of Occurrence is between 2.7% and 5.0% of the historical range in the 1890s and 1940s respectively. A combination of historical fragmentation of forest habitats, mostly for small-scale agriculture, and unsustainable hunting likely drove various populations to the south, east and west of the current population to extinction. This happened prior to the industrial-scale forest conversion that started in the 1970s. Our findings indicate how sensitive P. tapanuliensis is to the combined effects of habitat fragmentation and unsustainable take-off rates. Saving this species will require prevention of any further fragmentation and killings or other removal of animals from the remaining population. Without concerted action to achieve this, the remaining populations of P. tapanuliensis are doomed to become extinct within several orangutan generations.


21
The Tapanuli orangutan (Pongo tapanuliensis) is the most threatened great ape species in the world. It is 22 restricted to an area of about 1,000 km 2 of upland forest where fewer than 800 animals survive in three 23 declining subpopulations. Through a historical ecology approach involving analysis of reports and other 24 literature from the early 1800s to 2009, we demonstrate that historically Pongo tapanuliensis lived in a 25 much larger area, and across a much wider range of habitat types than now. Its current Extent of 26 Occurrence is between 2.7% and 5.0% of the historical range in the 1890s and 1940s respectively. A 27 combination of historical fragmentation of forest habitats, mostly for small-scale agriculture, and 28 unsustainable hunting likely drove various populations to the south, east and west of the current 29 population to extinction. This happened prior to the industrial-scale forest conversion that started in the 30 1970s. Our findings indicate how sensitive P. tapanuliensis is to the combined effects of habitat 31 fragmentation and unsustainable take-off rates. Saving this species will require prevention of any further 32 fragmentation and killings or other removal of animals from the remaining population. Without concerted 33 Introduction 36 Determining the key drivers of population decline is a primary objective in conservation biology and 37 wildlife management. Many wildlife species are threatened by a range of different and often interacting 38 factors, and developing effective conservation strategies requires unravelling how these threats interact 39 [1]. This is rarely easy, because species operate in complex socio-ecological systems in which different 40 components are affected by a range of anthropomorphic factors such as habitat loss and fragmentation or 41 unsustainable harvest. Evidence-based conservation seeks to address this by quantifying the relationships 42 between conservation actions, change in threat severity and change in conservation status [2,3]. 43 Collecting evidence is, however, time-consuming, and when conservation problems are "wicked", i.e., the 44 problems change as solutions are found [4], a stable solution may not be found to a particular 45 conservation problem [5]. This often means that scientific evidence does not support clear-cut conclusions 46 in value-driven debates that characterize conservation [6]. Nevertheless, conservation advocates often 47 seek simple narratives to convince the public of the urgency of environmental problems and the need to 48 support it. 49 One way to bring more clarity in often polarized debates around simple narratives is to be more specific 50 about the system in which a particular problem plays out. For example, if the system boundaries are 51 limited to oil palm as an ecological threat to orangutan survival [7], a simple solution would be to ban 52 palm oil use and to stop its production, preventing further deforestation. If the system boundaries are 53 extended to include smallholder farmers who produce palm oil for their own needs as well as international 54 markets, a ban on palm oil would encompass broader ethical connotations as it would affect people's 55 livelihoods [8]. The use of different perspectives in complex conservation contexts may not make it easier 56 to solve them but can provide helpful insights about the system boundaries of a particular problem. Are 57 they, for example, mostly ecological, or do they involve human threats, such as hunting, or societal 58 ethics? One such perspective is history. Looking back in time on the development of a particular problem 59 may provide insights about the underlying drivers of that problem [9]. The historical ecology approach 60 uses historical knowledge on the management of ecosystems or species [10]. Referring to historical 61 evidence has, for example, provided valuable understanding about the ecology of orangutans and what 62 likely caused their decline during the Late Pleistocene and Holocene, which informs their management 63 today [11]. Here we apply an analysis of historical ecology to one particular species of orangutan, P. 64 tapanuliensis, by analysing rarely used colonial-era literature to better understand the historical 65 distribution of the species. Indonesia's colonial literature on natural history was mostly written in Dutch 66 and German, and is not commonly used by conservation scientists working in Indonesia. 67 Pongo tapanuliensis was described in 2017 as a third species of orangutan [12], 20 years after this 68 orangutan population was formally reported to modern science [13]. The species is restricted to three 69 areas of mostly upland forest in the Batang Toru area in North Sumatra (Fig 1), totalling approximately 70 1,023 km 2 [14,15]. This orangutan population had been largely overlooked by science, despite having 71 been tentatively described in the colonial literature [16]. The estimated total number of wild P. 72 tapanuliensis is currently 767 [95%: 597,14] making this the great ape species with the lowest 73 number of individuals in the wild and perhaps the most threatened in the world [17].  The species is currently under threat of habitat loss from agriculture, hunting and conflict killing, and 79 development in the area for infrastructure, gold mining, and geothermal and hydro-energy. These threaten 80 to further reduce and fragment remaining habitat, reduce dispersal opportunities for the orangutans 81 between subpopulations, and undermine population viability through unsustainable mortality rates [14, 82 18-21]. Due to its restricted current distribution mostly centred around higher elevations ( orangutans are now extinct [24]. Given that, in the past, forest cover was also much more widespread in 95 the range of P. tapanuliensis, it is important to determine whether historically (ca. past 500 years) 96 orangutans did occur in those areas. This would help establish whether P. tapanuliensis has indeed 97 evolved to only live in the highlands and to estimate what its past distribution could have been. 98 The aim of this paper is to compile reports of orangutans occurring to the south of Lake Toba (Fig 1) with 99 the focus on determining how reliable these are, and, where feasible, provide a location for the occurrence 100 of orangutans to assess whether these are predominantly highland sites, and to assess which factors could 101 have led to their disappearance in those areas. Based on the information we develop historical distribution 102 maps as reference points for understanding historical population declines, and potentially also set long-103 term aspirational recovery targets for the species to ensure full ecological functionality [9]. 104

105
We compiled records of orangutans from historical sources by searching natural history books, scientific 106 papers, and historical newspapers from before 1940. We searched databases with location specific 107 keywords such as Sumatra, Batang Toeroe, and Tapanoeli, using Dutch spelling. We combined this with 108 searches for terms specifically referring to orangutans: Orang oetan, orang-oetan, orangutan, and also 109 mawas, mias and maias (local names for orangutan commonly used in historical literature), using a 110 variety of spellings. For the period since 1940, we used the sources from the review in Rijksen and 111 Meijaard [24] as well as scientific papers and personal communications. To determine the locations of the 112 historical sightings or captures we consulted the online Leiden University Library colonial map repository 113 (http://maps.library.leiden.edu/apps/s7). In some cases, rivers or villages were indicated which made it 114 feasible to estimate the location of the sightings quite accurately. In other cases, the area of the sighting or 115 captures was indicated in a broader area which reduced accuracy (tens of kilometres). 116 We assessed the likely vegetation types that P. tapanuliensis would have occurred in, and determined the 117 altitude at which they were reported. For this, we vectorised a high-resolution scanned copy of the first 118 official forest cover map of Indonesia [25], dated 1950 and at a scale of 1:250,000,000, which is likely 119 based on maps produced by the Netherlands-Indies cartographic service from the 1930s and 1940s. In 120 order to analyse the map and integrate with other spatial layers in a GIS, we automatically vectorised the 121 map using the ArcScan extension within ArcGIS [26]. The first step in the process was to geo-reference 122 the scanned map to the coastal boundary of Sumatra. The next step was extraction of the area of interest 123 which was then vectorized. This resulted in numerous polylines which were cleaned and edited to produce 124 polygons representing various land cover classes. The last stage of preparing the 1950s map for 125 integration with other spatial datasets in the GIS was to eliminate spatial distortion as much as possible 126 for the area of interest. Old hand-drawn maps, which in this case was an Indonesia-wide map, have 127 inherent distortion when compared to modern maps. We used a process called rubber-sheeting to make 128 small spatial adjustments in the vectorized georeferenced map to align individual parts of the map more 129 accurately with the coastline and inland features such as lakes and rivers. Although not a perfect match, 130 we are satisfied with the spatial accuracy of the vectorized 1950s map in terms of meeting the objectives 131 of this analysis. 132 While the exact location of the historical orangutan sightings cannot be determined with certainty, the 133 descriptions often provide sufficient detail through names of rivers and villages to estimate the altitude 134 and dominant vegetation where they occurred. Altitude was determined from the altitude layer in Google 135 Earth Pro. We used the vegetation map for Sumatra [27] in combination with knowledge gained by co-136 author SW during surveys in the region to assign one of the forest categories to an estimated historical 137 location. 138 To approximate the population decline of P. tapanuliensis in historical times, we mapped the historical 139 range. This provides important insights about the areas and vegetation types that the species once 140 occurred in, providing insights regarding its ecological functionality [28]. Grace et al. [9] recommend 141 using either a 1500 AD or 1750 AD target year for mapping the historical range, unless the historical data 142 are insufficiently accurate or reliable. Because there is uncertainty about all historical species data (unless 143 supported by specimens), we developed two maps: 1890s and 1940s. Each of these periods have different 144 data sources with varying reliability associated with them. By presenting these different maps, 145 conservation scientists and policy-makers can debate the merit of accepting either of these two (or a 146 different map altogether) to set a historical baseline for the species. Historical ranges were mapped by 147 modelling watersheds containing historic orangutan records of breeding females separated by large rivers.

148
We divided the island of Sumatra into potential orangutan subpopulation ranges by mapping large rivers 149 [based on 29], coasts, and the 500 altitudinal contour. These boundaries were chosen with the assumption 150 that female orangutans rarely disperse across these boundaries into a neighbouring watershed. Whereas 151 subadult and adult males do disperse through areas of higher elevation and low-quality habitat, females 152 are very rarely seen in such locations [24]. This lack of inter-watershed dispersal is supported by genetic 153 studies [30,31]. We assume that if one of our subpopulations is reduced below carrying capacity or goes 154 extinct, the probability of recolonization by immigrating females is negligible. An overlay of the 155 historical distribution range (pre-commercial timber industry) and the potential subpopulations ranges 156 resulted in a map of areas with or without orangutan populations. 157

158
Historical accounts 159 We report the various historical accounts of orangutan sightings or specimens from outside the currently 160 known range in chronological order, starting with Nikolaas Tulp [32] who in 1641 reported on a specimen 161 of "Indian Satyr" he had received, which had been collected in "Angola", most likely the former district 162 of Angkola (no. 1 in Fig 2) [24], which is now part of South Tapanuli District. His descriptions and 163 drawings indicate an orangutan, "of female sex", as Tulp writes. Several authors [e.g.,33,34] argued that 164 Tulp more likely referred to the African country of Angola and that his specimen was therefore likely a 165 chimpanzee or gorilla and not an orangutan. Rijksen and Meijaard [24], however, pointed out that Tulp 166 specifically referred to his "Indian" specimen being distinct from the African species, and also mentions 167 that the species occurs of the island of Borneo. 168 Schlegel and Mueller [35] reported on two orangutan crania that were obtained by a military doctor in the 169 environs of Jambi, some 650 km south-east of the current closest orangutan population (Fig 1). It is 170 unclear whether these animals were obtained in Jambi from the wild or whether they were in captivity and 171 possibly originated from northern Sumatra. Schlegel and Mueller reported that the two skulls were housed 172 in the "Rijks-Museum", which presumably refers to the Leiden Museum of Natural History. We were 173 unable to locate these specimens in the Leiden collection. Schlegel and Mueller further wrote that on the 174 west coast of Sumatra, especially north of the equator, the orangutan was known by the name mawej, 175 although in areas further south such as Indrapura and Bengkulu, the names orang-panda or orang-pandak 176 were used. Co-author Onrizal, remembers growing up east of Padang (Sungai Dareh) listening to stories 177 about orang-pendek, human-like creatures living in the forest, which reportedly ceased to exist in the area 178 in the 1970s. Stories of such orang-pendek [or other names, such as gugu, sedapa, orang lètje or orang 179 segagang, see 36, 37] abound in other parts of southern Sumatra and there has been speculation for over a 180 century that these could be remnant orangutan populations [38][39][40][41], although evidence remains lacking. 181 For the purpose of the current paper we do not focus on the orang-pendek narratives, but recognize that 182 many of these narratives could indeed refer to orangutans, as suggested compellingly by Forth [37]. 183 Schlegel and Mueller [35] acknowledged that the orangutan was especially common in the north-east of 184 Sumatra, but that occasionally they were encountered further south and along the western shores of 185 Sumatra. Also, the species had been reported from Indrapura (in present-day Riau Province) and near 186 Padang (in West Sumatra, Fig 1), although the descriptions by Schlegel and Müller are insufficient to 187 confirm that these reports referred to orangutans. Earlier writings by Müller [42], however, clarify this, as 188 he refers to "orang-oetan" which locals named "mawej" that one could occasionally encounter in the 189 extensive forests and swamp forests in, what is now, Kampar Regency, roughly between the towns of 190 Salo and Gunung Sahilan (no. 2 in Fig 2). although the letter was 50 years old then, so approximately referring to 1842. It mentions that "monkeys, 195 and especially 'orang-oetangs'" made life difficult to people travelling inland from Padang, by throwing 196 "stones, coconuts, branches, and others" at travellers through the "Padansche Bovenlanden",197 approximately in the area that is now Bukittinggi (no. 3 in Fig 2). It specifically mentions an event during 198 which someone was attacked by orangutans half-way between Fort de Kock (in Bukittinggi) and Bonjol 199 (no. 4 in Fig 2). 200 In further description of their travels across Sumatra,Müller and Horner [44] wrote that orangutans were 201 not unknown in the Tapanuli area and especially common in "Taroemon", i.e., present-day Trumon in the 202 Singkil area (Fig 1), which is part of the P. abelii range. They report that people distinguished between 203 two types of orangutans, the maweh baroet (baroet meaning monkey in the local language) and maweh 204 orang (the 'human' orangutan). Ludeking [45], in his descriptions of West Sumatra, mentions a record of 205 a six feet tall, upright primate, possibly an adult male orangutan, that his informants had seen on Bukit 206 Gedang, which is close to present-day Bonjol (no. 5 in Fig 2). A decade later, Von Rosenberg [46] did not 207 provide much detail but similarly mentioned that orangutans were present north of "Tapanoeli" (what is 208 now Sibolga) to Singkil (Fig 1), indicating presence of the species in the coastal lowlands west of Lake 209 Toba (no. 6 in Fig 2). He saw two orangutans but did not clarify where he saw them, although Snelleman 210 [47] stated that von Rosenberg had seen "two youthful specimens" in the area between Tapanuli and 211 Singkil. 212 Kramm [48] reported on a hunting expedition near "Soeroe Mantinggi", where he found several 213 orangutans and observed them for several hours. The location likely referred to Sayur Matinggi (no. 7 in 214 Fig 2), which is currently located in the Batang Gadis area, some 50 km south of the current range of P. 215 tapanuliensis. Kramm mentioned that Soeroe Mantinggi is located at a distance of 22 "palen" from 216 Padang Sidempoean. A "paal" was a measurement used in the Netherlands-Indies, equalling 1852 m on 217 Sumatra, indicating a distance of about 40 km for 22 "palen". Sayur Matinggi is currently located some 218 26 km from Padangsidempuan as the crow flies which indicates that indeed this is likely to be the location 219 where Kramm observed orangutans. Kramm was familiar with orangutans which he reported to have also 220 encountered in "Loemoet" and "Batang-Taro". We believe that the former refers to Lumut (no. 8 in Fig  221 2), just south of Sibolga (Fig 1), and that Batang-Taro is an older name for the Batang Toru area, where P. 222 tapanuliensis occurs until today. 223 Orangutans also seem to have occurred northeast of the current range of the Tapanuli Orangutan. 224 Neumann [49] described the species from "Hadjoran", which was located in the watershed of the "Batang 225 Si Ombal" and "Aek Hiloeng", and for which the following coordinates were given: N 2°1'25" and E 226 99°29', in the current district of Padang Lawas (Fig 1, no. 9 in Fig 2). This is about 50 km northeast of the 227 most eastern current range of P. tapanuliensis. The detailed description, however, suggests that the 228 species was very rare there, and the people of "Hadjoran" had not seen the orangutan there before. The 229 animal was shot, with descriptions of the local people suggesting it was at least 1 m tall, possibly 230 indicating an adult male, which are known to roam far from breeding populations. Neumann writes that 231 he travelled extensively through forest areas in the Padang Lawas area searching for orangutans but never 232 managed to encounter one. 233 Snelleman [47] mentioned a report from a government employee in Lampung (Fig 1) who had heard 234 about orangutans in that part of far southern Sumatra. Several hunts were organized to find the orangutans 235 but these were unsuccessful, and also when local people were asked they stated that they that had only 236 heard about orangutans in that area from hearsay but that they could not pinpoint where orangutans were 237 supposed to occur. We did not add these records to the maps as their reliability seems low. 238 An interesting reference to orangutans, as far south-east as Pelalawan is provided by Twiss [50]  Kampar Rivers (no. 10 in Fig 2) in 1888, and had an orangutan in his visor, but decided not to shoot it as 242 he had nothing on him to prepare the skin. Twiss [50] reported that L.H. was familiar with orangutans 243 from ones he had seen on Borneo, so the chance of a mistaken identification are small. Twiss also 244 reported that 18 to 20 years prior to his writings (i.e., around 1870), an orangutan was shot in the 245 mountains around Lake Maninjau (no. 11 in Fig 2), while older people remember seeing orangutans, 246 albeit very rarely, in forests on Bukit Silajang (no. 12 in Fig 2), a mountain near Lubuk Basung [50]. 247 Hagen [52,p. 66] stated that orangutans were known from the west coast between Tapanuli and Singkil 248 (no. 13 in Fig 2), although Singkil is in the range of P. abelii and it is not clear whether the coastal 249 Tapanuli reference referred to the area of the current range of P. tapanuliensis, or whether it referred to 250 what is now the Central Tapanuli District which extends to the southern part of Singkil, west of Lake 251 Toba. Interestingly, he referred to an orangutan from the interior of Padang (in reference to an article by abundantly, about Tapanuli Bay. Two miles up the Jaga Jaga River (no. 14 in Fig 2) some nibong palms 258 were seen that had been broken off by orangs, and also an old sarong (shelter), but the traces were old. 259 There were said to be more a few miles farther inland, particularly up the Berdiri River (no. 15 in Fig 2). 260 The natives say they always go about in pairs." Miller described the Jaga Jaga River as "a stream near the 261 south end of the Tapanuli or Sibolga Bay". We located the Berdiri River on an old map under the name 262 "Bardari River", and we located "Djaga Djaga" as well. 263 Beccari [54] reported orangutans around Rambung, in the Tapanuli region, and in the hinterland of 264 Sibolga, where he collected a specimen. We were unable to determine the location of Rambung but there 265 is a Rambong north of the Singkil area, and thus well in P. abelli range. The hinterland of Sibolga could 266 either refer to the current P. tapanuliensis population or the historical rangewe were unable to 267 determine whether this specimen still exists, and, if so, where. Beccari further stated that "in the 268 Zoological Museum at Florence is the skeleton of a young orang-utan, described as coming from 269 Palembang (Fig 1), on the east coast of Sumatra", some 800 km southeast of the nearest current orangutan 270 population. We contacted the curator of the Florence museum, who wrote in response that the specimen 271 was indeed present under specimen number MZUF-12: "The specimen was purchased in 1889 in London 272 (G.A. Frank,9 Haverstock Hill,London). It is a subadult male. suggested that there were no orangutans east of the Langkat River, which he thought was likely the 282 remnant of a large bay or sea connection that once separated north and south Sumatra approximately in a 283 line from Sibolga to north of Medan. He expanded on this in his work a few years later [57], in which he 284 also described additional orangutan sightings. This included a sighting in the upland area west of Lake 285 Toba at an elevation of 1,400 m asl (no. 16 in Fig 2). Referring to the same area, Kohler [58] described a 286 visit to Sibolga where the host had a young orangutan which had been caught in the forest on the west of 287 the Toba lake, indicating a breeding population there. Volz [56] also described a sighting of an orangutan 288 east of Lake Toba in the upper Kualu River area (ca. N 2°26' E 99°32'; no. 17 in Fig 2). Again, however, 289 the description of a large ape that moved 'slowly and ponderously' may suggest an adult male, and 290 because people there are not familiar with the species, possibly a wandering male outside the range of a 291 breeding population. told him that orangutans were particularly common in the area and were raiding the crops of local 295 farmers. After four weeks, they managed to catch seven orangutans. They then moved to a location four 296 hours rowing upstream, where they quickly observed a female orangutan with young. They set out cages 297 with fruit bait for capturing orangutans, but the first morning after arrival they had only managed to catch 298 some monkeys and a pig. After that they were more successful and claimed to have caught one male 299 orangutan and a female with young, and over the next few days they caught several more orangutans. 300 Delmont's stories are intriguing but strike us as somewhat fantastical, as it is unlikely that anyone could 301 catch significant numbers of orangutans with baited cages. More likely these could be Pig-tailed 302 Macaques, Macaca nemestrina, who indeed move about in groups, raid crops and can be trapped in cages.

303
We therefore do not consider this source to be reliable, and don't include this record in Fig 2 unless  304 evidence (e.g., specimens) is found for Delmont's claims in 1935. 305 The various historical accounts above where summarized in a map drawn by van Heurn [60] which shows 306 that clearly the conservation community was aware of the existence of orangutans south, west and east of 307 Lake Toba. Interestingly, though this map depicts the current Batang Toru population to be part of the 308 range where the species had become extinct, while the only extant population is a narrow band to the east 309 of Lake Toba in the Asahan District (Fig 1), where the species is not currently known. It suggests that 310 information about orangutan distribution was still rudimentary in the 1930s, which may have the reason 311 for a request to C.R. Carpenter [61] who conducted a survey on behalf of the Nederlandsch-Indische 312 Vereeniging tot Natuurbescherming. He worked mostly in the northern parts of Sumatra and sent 313 questionnaires to Dutch soldiers stationed in areas where orangutans could potentially occur. Carpenter 314 assumed that orangutan did not occur south of a line drawn from Singkil to the Sumatran east coast, thus 315 overlooking much of the historical evidence of orangutans south of Lake Toba. Carpenter's 316 questionnaires, however, included three reports of orangutans outside the known range. The first is from 317 Captain H.J. Kloprogge who had been based in Aceh, Siak, Indrapura and Pekanbaru (Fig 1) between 318 1921 and 1936, spending an average 12 days per months in the forest on patrol. He claimed to have seen 319 orangutans "two to four" times during forest patrols, and indicated their presence on the hand-drawn map 320 accompanying the questionnaire throughout Aceh and the current Batang Toru range area. Second, 321 Captain M. Kooistra reported seeing 12 orangutans in Aceh and also indicated them as present on his map 322 near Jambi (Muara Tembesi), where he had been stationed in 1925 and 1926 (Fig 1). As there is no 323 further information about this record, we do not include it in Fig 2. Third, Captain H.G.C. Pel was 324 stationed in Siak (near Pekanbaru, Fig 1) from 1933 to 1935, and reported seeing an orangutan in 325 captivity north of the town of Talu (no. 18 in Fig 2), on a tributary of the Kampar River. We consider it 326 unlikely that such a captive orangutan in a remote village would have been transported to the area, and 327 map this point as likely present. We do not map a report of orangutan presence in Batang Toru and 328 Sipirok from 1939 [62], as these are still part of the current range. 329 There seems to be a gap in records between the 1930s and 1970s, but in the early 1970s, the Indonesian 330 forester Kiras S. Depari reported orangutan sightings along the Batang Toru River, in the Sibual-buali 331 Reserve and in the Rimbu Panti Wildlife Reserve (no. 19 in Fig 2) [16]. Borner [63] also noted that a 10-332 12 year old male orangutan had been shot just outside Rimbu Panti, and that villagers had shortly before 333 seen "two other black 'orang-utans' walking on the ground", but his surveys could not find any nests. 334 Borner also interviewed timber workers in Torgamba in the South Labuhan Batu District, who said that 335 orangutans occurred in those forests but Borner's surveys could again not verify this. Two other 336 primatologists, C.C. Wilson Fig 2), although field surveys did not 346 reveal any nests. While there is significant spatial inaccuracy in the historical records of P. tapanuliensis outside the 355 current range, we can still make an educated guess of the different habitats and altitudes in which these 356 populations occurred (Table 1). Habitats in which the species once occurred included tall peat swamp 357 forest, freshwater swamp forest mosaic and secondary forest, forest on limestone, hill forest between 300 358 and 1,000 a.s.l., and submontane forest between 1,000 and 1,800 m a.s.l., indicating the full range of 359 habitats that is also used by P. abelii [16]. 360  (Fig 3). We also present an alternative historical baseline based on the distribution map drawn 376 by van Heurn in 1935 [60], with additional records from our own dataset (Fig 3). The historical range of P. tapanuliensis 382 Our historical ecology analysis of P. tapanuliensis indicates that the species occurred beyond its current 383 range until quite recently, and has rapidly declined in the past 100 to 150 years. Breeding populations 384 occurred in the Batang Gadis area (Fig 1), probably through much of today's South Tapanuli and 385 Mandailing Natal Regencies, and further south-east in the Kampar River area. The historical records to 386 the north of the current P. tapanuliensis range make it difficult to judge whether these are part of P. abelii 387 or P. tapanuliensis. For example, the records of orangutans west of Lake Toba, could also refer to 388 populations To what extent do these historical data allow us to accurately determine the former range of P. 394 tapanuliensis. We know from evidence of Late Pleistocene orangutan fossils in the Padang area (Lida 395 Ajer, Ngalau Sampit, Ngalau Gupin) [72], that orangutans lived in this part of Sumatra at least until some 396 50,000 years ago. What we do not know is whether this was P. abelii, P. tapanuliensis, or a species 397 different from both, as suggested by Drawhorn [73]. No specimens of P. tapanuliensis were collected by 398 any of the historical sources, except those reported by Hagen [52] and Beccari [54], but these have not yet 399 been genetically analysed. There is therefore no robust evidence as to whether the orangutans reported 400 from outside the current P. tapanuliensis range were P. tapanuliensis or P. abelii. Further genetic study of 401 the specimens reportedly originating from Padang (RMNH.MAM.544) and Palembang (MZUF-12), and 402 also of fossil teeth from the Padang Caves area (e.g., through proteomic analysis) could shed light on the 403 taxonomic status of the orangutans in Central Sumatra, and their relationship to P. tapanuliensis. Based 404 on distribution range patterns, with P. abelii clearly restricted to the northern parts of Sumatra [12,65], 405 however, we consider it most likely that all historical orangutan populations south and south-east of the 406 current range of P. tapanuliensis were also P. tapanuliensis. If correct, this would indicate a historical 407 extent of occurrence [74] of about 37,511 km 2 in the 1890s, and about 21,313 km 2 in the 1940s. If we 408 compare this to the current distribution range of 1,023 km 2 [14,15], it suggests that the species currently 409 retains some 2.7% of the 1890s range and 5.0% of the 1940s range. It might even be possible to map the 410 range going back further in time, e.g., 1750 AD as recommended by Grace et al. [9]. As populations 411 dwindled and encounters with orangutans became increasingly rare, this may have resulted in the folklore 412 regarding mythical creatures (orang-pendek, gugu, sedapa etc.) [36,37] in Bengkulu, Jambi, and South 413 Sumatra, indicating an even larger range. We leave it to other conservation stakeholders to determine 414 which historical range is most appropriate (and reliable) for use as a historical baseline for P. 415 tapanuliensis, and also for setting an aspirational recovery target for the long-term future (e.g., 100 years 416 from now) towards full ecological functionality of the species [9,28,67]. 417 Possible drivers of historical declines 418 While our findings indicate that orangutans disappeared from much of their historical range, it is less 419 clear why the species declined and became locally extinct. Some populations such as those in Lumut, seen 420 by Neumann in 1885 and by SW in 2001, but not seen since, may have become extinct quite recently 421 because of forest loss. Other populations likely disappeared sometime in the 20 th century. There have 422 been no recent confirmed records from areas west or southwest of Lake Toba, nor from the Batang Gadis 423 region, south of the current populations. Also, while there are a number of alleged records from the area 424 east of Padang, there are no confirmed recent records. It thus appears that a lot of these populations 425 disappeared around a time when forests were still extensive and the commercial exploitation of forest for 426 timber (starting in the 1970s) or their conversion to plantations (starting in the 1990s) had not yet 427 decimated available habitats. Nevertheless, there had been significant historical deforestation prior to 428 1950 as shown in Fig 2, mostly for small-holder agriculture and livestock, firewood and timber, and as 429 result of wars and fires [75,76]. For example, the colonial-era district of Tapanoeli (now North, South 430 and Central Tapanuli) had an estimated forest cover of 19000 km 2 in a total area of 39,481 km 2 (48% 431 forested) in the 1930s [75]. In 1824, one of the first European visitors to the region was astonished to see 432 that "the plain [north of Batang Toru] was surrounded by hills from five hundred to one thousand feet 433 high, in a state of cultivation; and the whole surrounding country was perfectly free from wood, except 434 the summits of two or three mountains" [77]. Some orangutan populations therefore appear to have 435 become isolated in historical times, when early agricultural development created large grassland areas. 436 So, why did these populations become extinct? This appears to have been a combination of habitat loss 437 and population fragmentation, and mortality rates that exceeded reproductive replacement rates. 438 Several authors have suggested that orangutan density and range on both Borneo and Sumatra were 439 primarily determined by the ability of people to access areas and hunt orangutans [78,79]. For example, 440 Jentink (1889) writes that orangutans in Sumatra are only common in swamp areas like those in Singkil 441 which are so inaccessible that they are rarely "stepped on by human feet", apparently quoting von 442 Rosenberg [ account about orangutans, describes how generally orangutans were shot with poison darts, after which 457 they fell out of the trees and were killed with spears. Alternatively, they were caught alive and killed later. 458 The whitish meat was generally grilled over a fire, and was described as soft and sweet [90]. This is also 459 suggested by the use in local language of juhut bontar, or white meat, to describe orangutan age classes, running 500 iterations with populations of 250 orangutans. In the best quality orangutan 471 habitats, i.e., mosaic landscapes of swamp, riparian and hill forests [24], annual hunting rates of 1% did 472 not cause population extinction, but did decrease population size. In less than optimal habitats, e.g., 473 forests at higher elevation, a 1% level of hunting caused declines to extinction irrespective of initial 474 population size. Higher rates of hunting were unsustainable even in the highest quality habitats [93]. 475 These models were conducted for P. pygmaeus, but the authors thought that hunting effects would be 476 similar for Sumatran orangutans. The best orangutan habitats would like be those with the highest soil 477 fertility, which at levels of intermediate rainfall would also be the best areas for agriculture [94]. It is thus 478 likely that historically P. tapanuliensis occurred in suboptimal habitats, where the removal of one animal 479 from a population of 100 per year, would drive such a population to extinction. Given the available 480 information, we consider it likely that P. tapanuliensis was hunted to extinction in the increasingly 481 fragmented parts of its former range and only survived in the remote and rugged Batang Toru mountains 482 which may have provided orangutans with a refuge from hunting. At the same time, we recognize that in 483 the complex socio-ecological system from which orangutans disappeared many processes may have 484 contributed, and simple, linear cause-effect reasoning may not apply. 485 The biases and constraints of a historical perspective 486 While the use of historical data provides useful insights about the likely historical range of orangutans on 487 Sumatra, and possible drivers of their decline and local extinction, there is uncertainty in the data. Few of 488 the records are based on specimens that provide evidence for the veracity of claims, and some records are 489 only based on hearsay or alleged sightings of orangutans without further evidence. While we critically 490 examined each record, there is subjectivity in interpreting their reliability. For example, we decided not to 491 incorporate the many records of orang-pendek in our analysis, even though some of them could well have 492 referred to orangutans. Then again, we did accept sources that reported orangutans seen by local people 493 but not the source, e.g., Neumann [49]. Had the people in Hadjoran mentioned "orang-pendek" rather 494 than orangutan, we would have rejected the information, even though it would have related to the same 495 animal. The potential bias of this approach is obvious. 496 Another form of bias in our study is the literature accessible to us in this study. Most of our information 497 sources are colonial-era explorers, naturalists and hunters, for which we were able to find information in 498 the large numbers of books, newspapers and journals that have been digitized and can be searched and 499 accessed online. This means that we are missing out on two potentially valuable data sources: 1) Local 500 folklore about orangutans among people that live in the orangutan's historical range; and 2) Post-501 independence publications and grey literature from government (e.g., forest inventories), universities 502 (e.g., student survey reports), companies (environmental impact assessments), and local media. There is 503 much information about orangutan folklore from Borneo (especially Sarawak), but such information does 504 not seem to have been recorded in the anthropological literature for Sumatra. Post-independence writings 505 from the 1950s to ca. 1980s are likely to contain many references to records of orangutans from their 506 historical range, but such information has not yet been captured electronically and remains beyond our 507 reach. 508 Future studies that would include more local and socio-culturally specific information, would put 509 conservationists in a better position (with the help of local experts, anthropologists, etc.) to understand 510 local drivers of extinction and formulate more targeted interventions. For example, information from 511 indigenous groups that hunt and consume orangutans, versus conflict-related killing of orangutans, versus 512 Muslim taboos against eating orangutans, versus groups that may or may not have specific ritual relations 513 with orangutans [95], could result in locally specific management strategies for reducing killing, harming 514 and capture of orangutans. There are also contextual specificities, e.g., transmigrants from Java or 515 tsunami refugees from Nias Island having very different experiences of forest life, land rights and 516 reactions to orangutans compared to indigenous people in the orangutan's range. All these nuances 517 relevant to species management require that we go beyond the confines of the data sources used for the 518 current study. There is thus value in the historical ecology approach but there are also limitations, or in 519 the words of the statistician George Box "All models are wrong, but some are useful" [96]. We hope our 520 historic models add useful information to the conservation debate regarding P. tapanuliensis. 521 indicative of a lowest minimum number of killings, as they represent only criminal acts that have been 536 detected and acted upon, which is a fraction of the total orangutan-related wildlife crime [101,102]. 537

Implications for species conservation
Records of an adult male killed in 2013 (OIC pers. comm. 2020), a male severely injured by humans in 538 2019 [103], and another male captured and translocated twice in the past 12 months due to complaints 539 about crop raiding from local community members [104] suggest that killings have been ongoing in 540 recent years, although prior to 2017 most detected incidents would have been recorded as P. abelii. While 541 translocation has been used as a response to orangutans in conflict with humans, translocated animals are 542 not monitored beyond a few days following release, sometimes not at all, and their long-term survival is 543 not known. Behavioural traits of female site fidelity and male territoriality, and adaption issues P. 544 pygmaeus released in unfamiliar habitats indicate that translocation risks are high and survival rates may 545 be low [101,105]. 546 Long-term protection of P. tapanuliensis requires that mortality rates of <1% per year are maintained 547 over long (decadal) time frames across the species' range. This also means that that all subpopulations 548 have to remain connected, because once connections between populations are lost this should result in 549 higher extinction risks for the remaining subpopulations, as was modeled for P. pygmaeus [30]. Within 550 the subpopulations, the prevention of killing and translocation or rescues is urgently needed, which 551 requires innovative management of crop conflicts [106,107], and effective law enforcement and 552 awareness campaigns. Such campaigns have so far had insufficient impact on reducing orangutan losses 553 and new approaches may be required [108]. This could include, for example, direct conditional payments 554 to rural communities for maintaining habitats and preventing any deaths or harm, i.e., orangutan 555 guardians [109] or support for "buffer gardens" to concentrate crop losses from orangutan foraging into 556 areas acceptable to communities [110]. Viable conservation solutions that prevent the extinction of P. 557 tapanuliensis require an awareness of the specific problem posed by small-scale anthropogenic factors 558 that have driven historical declines. Addressing these factors requires more targeted interventions, for 559 example, through a conservation plan that is tailored specifically to the needs and characteristics of P. 560 tapanuliensis and the different socio-ecological drivers of its decline, rather than a generic national-level 561 approach that encompasses a huge range of contexts and all three species [111]. 562 Currently, P. tapanuliensis is rated Critically Endangered A4bcd on the IUCN Red List [112] an 563 "observed, estimated, inferred, projected or suspected population size reduction of ≥80% over three 564 generation periods (i.e., 75 years), where the time period must include both the past and the future, and 565 where the reduction or its causes may not have ceased or may not be understood or may not be reversible, 566 based on (b) an index of abundance appropriate to the taxon; (c) a decline in area of occupancy, extent of 567 occurrence and/or quality of habitat; and (d) actual or potential levels of exploitation [22]. The 568 information from the current information makes it likely that a similar decline population size reduction 569 of ≥80% has occurred over the past 75 years, based on the estimated reduction of the Extent of 570 Occurrence [see 74] of 95-97 % over 100 to 150 years. This would qualify the species as Critically 571 Endangered A4bcd and A2cd. 572 Given the high extinction risks, it is important that a comprehensive plan of action is developed for the 573 species that accurately determines how many animals remain, the level of gene flow between 574 subpopulations, current loss rates (including removal of animals in rescues and translocations), and works 575 towards full and permanent protection of all remaining habitat and enforcement of zero unnatural losses. 576 Such a conservation management plan would need clarity about long-term funding, organizational 577 responsibilities, and clear, science-based plan to allow the P. tapanuliensis population to stop declining, 578 or better, increase to safer population numbers. Without such concerted and coordinated action, the 579 remaining populations of P. tapanuliensis are doomed to follow their historical predecessors on their path 580 to rapid extinction. 581

582
We thank Herman Rijksen for inspiring to access the historical literature on orangutans, Gabriella 583 Fredriksson, Graham Usher, and Liana Chua for providing comments on an earlier version of this 584 manuscript, Matthew Minarchek for sharing Carpenter's interview data, which allowed us to update the 585 historical range maps, and Molly Grace and Mike Hoffmann for suggesting how we could map the 586 historical ranges. We also thank three reviewers for their constructive comments. 587