We designed a hierarchical isotope sampling of caprine tooth enamel from five different-aged fossil sites (Table 2), namely, Pha Bong (400 to 200 ka), Khok Sung (217 or 130 ka), Tham Wiman Nakin (>169 ka), Tham Lod Rockshelter (32 to 12 ka), and Ban Rai Rockshelter (10 to 6 ka) (Figure 1), as well as from the modern representatives (see Supplementary Material S1 for more detailed information on geological backgrounds of aforementioned sites and see Supplementary Material S2 for the fauna list). A total of 94 bulk tooth enamel samples were extracted mostly from the second or third molars (fully formed adults) of caprine individuals (Supplementary Material S3). The bulk enamel powder was sampled along the entire length of a tooth in order to obtain average isotopic compositions over the time period of dental growth. The analyzed fossils comprise eight specimens of N. goral from Pha Bong, three specimens of C. sumatraensis from Khok Sung, four specimens of N. goral and nine specimens of C. sumatraensis from Tham Wiman Nakin, eight specimens of N. goral, 19 specimens of N. griseus, and 12 specimens of C. sumatraensis from Tham Lod Rockshelter, and six specimens of C. sumatraensis from Ban Rai Rochshelter. Modern specimens including eight specimens of wild N. goral from the Indian Himalayan region, one specimen of wild N. griseus from Northern Thailand, and 16 specimens of wild C. sumatraensis from Thailand and Sumatra were collected from several natural history museums and institutes including Chulalongkorn University Museum of Natural History (CUMNH, Bangkok), Natural History Museum (THNHM, Pathum Thani), Naturhistorisches Museum Basel (NMB, Basel), and Zoologische Staatssammlung München (ZSM, Munich). In this study, we additionally performed the stable isotope analysis of tooth enamel of other mammals from Tham Lod (12 samples) and Ban Rai (10 samples) rockshelters, used as baseline values for isotopic comparisons, in order to understand the environmental contexts of these sites (Supplementary Material S4).

To conduct isotope analysis of tooth enamel, we selected almost complete fossil tooth specimens in order to have access to morphological features that are useful for the taxonomic identification. All analyzed fossil teeth were identified based on their morphological and dimensional comparisons with modern samples. The teeth of N. goral differ from N. griseus in having a less rounded outline on P4, more distinct paracone on upper molars, more distally curved crown on M3 in labial view (straight in N. griseus), U-shaped anterior valley on p3, more prominent labial valley on p4, and more developed conids and entoconulids on lower molars (Wattanapituksakul et al., 2018). The teeth of N. goral are similar in size to those of N. griseus but are clearly smaller than those of C. sumatraensis (Suraprasit et al., 2016; Wattanapituksakul et al., 2018). In addition, Capricornis is distinguished from Naemorhedus by several dental characters: more developed cusps and labial styles on upper premolars and molars (especially the mesostyle and metastylar wing on M3) and more distinct lingual stylids on lower premolars and molars (Wattanapituksakul et al., 2018). In the case of modern samples analyzed here, we followed the taxonomic identification labeled by the field collectors and/or curators.

All caprine enamel samples were pretreated and analyzed at the Department of Geosciences, University of Tübingen, Germany (see Supplementary Material S5 for more detailed information on pretreatment and isotope protocols). Stable carbon and oxygen isotopic values are expressed as the following standard delta (δ)–notation: X = [(Rsample/Rstandard)-1]^∗1000, where X is referred to δ¹³C and δ¹⁸O values and R is equivalent to ¹³C/¹²C or ¹⁸O/¹⁶O, respectively. The recorded delta values follow the international reference standards, “Vienna PeeDee Belemnite” (VPDB) for the carbon and oxygen. Additionally, δ¹⁸O values relative to “Vienna Standard Mean Ocean Water” (VSMOW) are given. The carbonate content (CaCO₃%) was calculated using the ratio between amount of CO₂ released by the reaction, as detected from the peak intensity for mass 44 and the weight of pure carbonate used as a standard, with an analytical error of 0.3%, based on the multiple analysis of reference enamel samples.

Before comparing the stable isotope ratios between fossil and modern faunas, we took into account the changes in the carbon composition due to anthropogenic CO₂ emission (¹³C Suess effect) since the beginning of the industrial revolution in 1850 and ongoing today (Keeling et al., 1979). The atmospheric δ¹³C values in relation to increased CO₂ emissions have decreased approximately from −6.5‰ to −8.0‰ over the past few centuries due to fossil fuel combustion and biomass destruction (e.g., Friedli et al., 1986; Marino and McElroy, 1991). To convert δ¹³C values to modern plant isotope compositions due to the Suess effect, the δ¹³C values of modern carbonate bioapatite are therefore adjusted upward by 1.5‰ for comparing with the Pleistocene enamel carbon isotope compositions (Cerling et al., 1997a; Cerling and Harris, 1999; Passey et al., 2005). However, modern samples analyzed herein do not require the Suess effect δ¹³C correction because these wild mammals died or were collected before or around 1950 (Supplementary Material S3).

We performed statistical analyses of the isotopic data (δ¹³C and δ¹⁸O) to show a significant difference among our analyzed specimens. Based on the Shapiro–Wilk test, our samples do not correspond to a normal distribution with unequal variances. In the case for which a sample of at least five individuals is available, we used non-parametric tests to examine differences in median δ¹³C and δ¹⁸O values among our analyzed caprine samples (Kruskal–Wallis test) and between the species and the locality (Mann–Whitney U-test). Significant differences are statistically attained when p-values are equal to or less than 0.05.

The δ¹³C values of Himalayan gorals from Pha Bong (including material previously analyzed by Bocherens et al., 2017) displayed a median of −0.5‰ and ranged from −2.6‰ to +3.0‰ (Figures 2, 3A), indicating the consumption of mostly pure C₄ plants with a limited signal of mixed C₃ and C₄ diets. Compared to the previously isotopic-analyzed samples of Pha Bong Himalayan gorals by Bocherens et al. (2017), our results showed similar tooth enamel δ¹³C values (Mann–Whitney U: NS; Supplementary Material S6), supporting the occurrence of a single species in this locality. The Himalayan gorals from Tham Wiman Nakin had a similar dietary signature (a median of −1.2‰ and a range between −3.5‰ and +0.4‰) to those of Pha Bong (Figures 2, 3A) but exhibited a shift of the δ¹³C values toward the C₃/C₄ mixed signals. A trend of the δ¹³C shifts into the direction of C₃ plants is also observed from the Tham Lod Rockshelter Himalayan gorals, which yielded a median of −3.3‰ and a range between −5.3‰ and −0.1‰ (Figures 2, 3B), indicating their diet selection chiefly on both C₃ and C₄ plants and only sometimes on pure C₄ plants. In contrast, the δ¹³C values of modern Himalayan gorals displaying a median of −12.2‰ and varying from −15.2‰ to −6.0‰ (Figures 2, 3C) fall into the range of pure C₃ and mixed C₃/C₄ ecosystems, reflecting an adaptation for occupying either more closed habitats or C₃-dominated grasslands at the Indian Himalayan Region where the conditions at such an extremely high elevation are unfavorable for C₄ plants (Dobremez, 1976; Blasco et al., 1996; Galy et al., 2008; Bhattacharyya et al., 2019).

The Himalayan gorals from Pha Bong showed more positive δ¹⁸O values, compared to those from other localities (Mann–Whitney U: P < 0.05; Supplementary Material S6), ranging from −5.3‰ to +0.5‰ with a median of −3.5‰ (Figures 2, 3A), likely resulting from the ingestion of similar water plant resources mediated by evapotranspiration processes that have ¹⁸O enrichment. Some individuals, which exhibited higher δ¹⁸O values than about −2.2‰ (the first quartile, Figure 2), suggested the consumption of a significant amount of enriched ¹⁸O leaves in an open meadow. Lower δ¹⁸O values are observed in some individuals found from Tham Wiman Nakin (a median of −5.2‰ and a range from −6.5‰ to −1.7‰) and Tham Lod Rockshelter (a median of −6.4‰ and a range from −11.4‰ to −1.8‰), as well as from the modern representatives (a median of −6.2‰ and a range from −10.6‰ to −3.3‰) (Figures 2, 3), likely a consequence of the consumption of water depleted in ¹⁸O compared to those from Pha Bong. The most variable δ¹⁸O values recorded from the Tham Lod Rockshelter Himalayan gorals might be explained by more opportunistic feeding on both ¹⁸O enriched and depleted plant water. In general, the wide range of tooth enamel δ¹⁸O values observed in the Pleistocene and extant Himalayan gorals might reflect the dietary habit of each individual in response to seasonal differences because the modern population has been suggested to feed in the variable frequency on plant species between the summer and winter (Ashraf et al., 2015, 2016). Otherwise, the wide δ¹⁸O ranges in the tooth enamel of both Pleistocene and extant Himalayan gorals were possibly due to their seasonal migration across different altitudes in response to temperature changes and threats from predators and hunters (Perveen and Khan, 2013). Among the Himalayan gorals, some higher δ¹⁸O values corresponding to more dominated C₄ signals are observed, probably reflecting a consumption of ¹⁸O enriched leaf water in open habitats (Figure 3 and Supplementary Material S7).

Pleistocene Chinese gorals documented only in Tham Lod Rockshelter yielded a median δ¹³C values of −0.4‰ with a range between −7.1‰ and +1.9‰ (Figures 2, 3B), wider than those of Himalayan gorals from Pha Bong and Tham Wiman Nakin as well as in the same locality (although non-statistical significance tested by Mann–Whitney U; Supplementary Material S6) (Figure 2). This implies that the Chinese gorals had the ability to forage more exclusively on some C₃ plants and/or to occupy more closed habitats than the Himalayan gorals. However, this was possibly due to the sampling bias for which more or less specimens were analyzed among these localities. One modern sample of the wild Chinese goral individual collected from the region of Northern Thailand, displays a δ¹³C value of −3.3‰ (Figure 3C), suggesting the consumption of mixed C₃ and C₄ plants, similar to some Pleistocene populations in Tham Lod Rockshelter (Figure 2).

The Tham Rod Rochshelter Chinese gorals had a median δ¹⁸O value of −5.6‰, similar to the modern sample (−5.7‰), and exhibited broad δ¹⁸O values varying from −8.9‰ to −0.3‰ (Figures 2, 3B), suggesting that they derived their water from highly variable food or local resources, similar to what is observed in the congeneric species (N. goral) in the same locality. Otherwise, it is conceivable that the Chinese gorals might have ingested water from local resources in different altitudes in relation to their seasonal habitat use or migration, as seen in the modern population (Chen et al., 2009, 2012).

The δ¹³C values of Khok Sung Sumatran serows including three samples in this study and one previously analyzed material from Suraprasit et al. (2018, Table 1) displayed a median of −13.9‰ and a range between −15.1‰ and −11.8‰ (Figures 2, 3A), indicating diet and habitat preferences in a closed C₃-dominated environment. The Sumatran serows recovered in Tham Wiman Nakin and Tham Lod Rockshelter showed a median of −11.0‰ and −12.2‰ (Mann–Whitney U: NS; Supplementary Material S6) and exhibited considerably wide δ¹³C values ranging from −12.8‰ to +1.7‰ and from −14.3‰ to +1.9‰, respectively (Figure 2). The Sumatran serows from both localities inhabited the various types of tropical ecosystems from exclusively closed C₃ to open C₄ vegetation landscapes. The carbon isotope compositions of tooth enamel of these Pleistocene Sumatran serows clearly support a broader foraging niche than those for Naemorhedus (Mann–Whitney U: P < 0.05; Supplementary Material S6). However, the Early Holocene Ban Rai and modern populations showed more limited diets of either pure C₃ or C₃/C₄ mixed vegetation (Figure 3C) and more restricted habitat use in either closed environments or areas adjacent to forest/woodland landscapes based on the median δ¹³C values of −14.1‰ with a narrow range between −16.6‰ and 12.3‰, for the former site and on the median δ¹³C values of −13.3‰ with a narrow range between −17.8‰ and −8.8‰ for the latter (Mann–Whitney U: NS; Supplementary Material S6).

Highly variable δ¹⁸O values are observed among most of the Pleistocene and modern Sumatran serows including a median of −5.7‰ and a range between −8.9‰ and −3.6‰ for Tham Wiman Nakin, a median of −6.3‰ and a range between −8.1‰ and +0.1‰ for Tham Lod Rockshelter, and a median of −7.0‰ and a range between −10.5‰ and −1.9‰ for the modern population (Mann–Whitney U: NS; Supplementary Material S6) (Figure 2). Most of the Pleistocene and extant Sumatran serows with the exception of Khok Sung samples thus indicated their independent habits in consuming various leaf water or ingesting meteoric water across variable local resources. The narrower range of the δ¹⁸O values, with a median of −5.0‰ and a range between −6.4‰ and −3.7‰ (Figure 2), for Khok Sung serows might be explained by fewer isotopic-analyzed samples from this locality. Compared to other localities, C. sumatraensis from Ban Rai Rockshelter likely displayed more limited and negative δ¹⁸O values with a median of −9.0‰ and a range between −11.5‰ and −8.3‰ (Figure 2) (Mann–Whitney U: P < 0.05; Supplementary Material S6), corresponding to its occupation of a closed forest canopy with a more humid condition. Some C₄ grazers tended to have higher δ¹⁸O values compared to C₃ browsers, consistent with their dietary intake and habitat use (Figure 3 and Supplementary Material S7).

Analyzing stable carbon isotope composition of fossil caprines from different time periods shows that none of the analyzed Pleistocene N. goral revealed a significant shift in their tooth enamel δ¹³C values over time prior to the Holocene (Figure 4). N. goral therefore had an ecological persistence on the diet of exclusively C₄ plants but sometimes mixed C₃/C₄ vegetation and on the habitat in open environments through the Middle to latest Pleistocene. In contrast, stable carbon isotope compositions of modern samples of N. goral showed that their δ¹³C values became lower, indicating a diet with significantly more C₃ food items in more closed habitats or C₃-dominated grasslands (Figure 4). This isotopic shift corresponds to the extant Indian Himalayan population that is surviving in the extreme condition of highland or mountain climate refugia where the C₃-grasslands are dominant (Dobremez, 1976; Blasco et al., 1996; Galy et al., 2008; Bhattacharyya et al., 2019) and where the anthropic effects are presumably weaker.

The isotopic analysis of tooth enamel of N. griseus from the Late Pleistocene of Tham Lod Rockshelter indicated its occupation of resources in the same habitats of N. goral (Mann–Whitney U: NS; Supplementary Material S6). Although only one modern isotope sample of N. griseus has been analyzed for this study due to the inaccessibility and rareness of wild individual specimens, the Chinese goral usually inhabits steep landscapes and plateaus in mountainous areas or sometimes subtropical mixed and evergreen-deciduous forests near cliffs based on the field observation (Duckworth et al., 2008a; Chen et al., 2009, 2012; Groves and Grubb, 2011; Buranapim et al., 2014; Buranapim and Sitasuwa, 2014). This information recorded in the field indicates an ecological adaptation struggling to live in a more closed habitat for the extant population of N. griseus, despite having mixed C₃/C₄ plants in diet (as seen in the single analyzed sample from this study), different from what has initially been observed in some Pleistocene inhabitants that mainly occupied an open habitat.

Our stable isotope results also indicate that C. sumatraensis has been well-adapted to variable environmental conditions expanding from closed C₃ to open C₄ landscapes since the Middle Pleistocene and has been a generalist with a greater ecological flexibility in diet and habitat than both Pleistocene and extant Naemorhedus (Figures 2, 4). Although some isotopic data from well-dated localities such as Khok Sung (MIS7 or MIS5) (Suraprasit et al., 2018; Duval et al., 2019; this study) and Nam Lot (MIS5) (Bacon et al., 2018b) as well as from the Early Holocene of Ban Rai Rockshelter have documented the preferred habitat restriction of the Sumatran serows in closed forest environments (Figure 2), this might be explained by the sampling bias for which few specimens are available or analyzed or by the ecological adaptation of C. sumatraensis in response to more homogeneous and closed conditions during interglacial stages. In the heterogeneous environments during the Pleistocene, C. sumatraensis preferably occupied in both closed forest and open grassland landscapes but rarely in an intermediate area between these two canopies. The rareness of C. sumatraensis as being a mixed C₃/C₄ feeder in the intermediate habitat area was possibly a consequence of its competition avoidance for food resources with other local wildlife such as megaherbivores (e.g., Stegodon) and omnivores (e.g., Sus) (Figure 4). Previous isotope studies have revealed that Stegodon was mainly a C₃/C₄-mixed feeder and significantly held an intermediate canopy landscape in some coexisting area (e.g., Khok Sung, Suraprasit et al., 2018) (Figure 4). Moreover, the preferred habitat of suids in the intermediate habitat canopy between closed and open ecosystems remained relatively stable over time (e.g., Pha Bong, Khok Sung, Tham Wiman Nakin, Nam Lot, and Boh Dambang) (Pushkina et al., 2010; Bocherens et al., 2017; Bacon et al., 2018b, c; Suraprasit et al., 2018) (Figure 4 and Supplementary Material S4). The limited δ¹³C range of numerous modern enamel samples suggests that the ecological flexibility of C. sumatraensis has become restricted after the latest Pleistocene (Figures 3C, 4). This rests on the assumption that the present-day wildlife habitat in Thailand is likely constrained by the Holocene climate change and the human impact on the ecosystem restricted its population to more closed canopy habitats such as dense forests in higher altitude areas.

As Naemorhedus and Capricornis did coexist in several Thai fossil sites during the Pleistocene, niche partitioning between them might have occurred on other aspects that are not discernible through the stable carbon isotope analysis.

Regarding the dental morphology, these three caprine taxa possess a distinctly selenodont appearance with similar occlusal patterns. In the evolutionary history of mammals, the development of hypsodont teeth allows them to feed on more abrasive food such as grasses and to therefore expand the range of dietary resources (e.g., Mendoza and Palmqvist, 2008; von Koenigswald, 2011). Both Naemorhedus species have the same degree of hypsodonty as an index of 4.03 (Mendoza and Palmqvist, 2008), while C. sumatraensis has comparatively lower (but still high) crowned teeth according to the hypsodonty index of 3.39 (Kaiser et al., 2013). The dental morphologies of these three caprine species are congruent with our stable carbon isotope results, which indicate their life history adaptations to grazing during the Pleistocene.

Additionally, differences in the average body size between Naemorhedus and C. sumatraensis, as exemplified from the tooth dimensions on fossils (e.g., Suraprasit et al., 2016; Wattanapituksakul et al., 2018) and from the observation of recent populations (e.g., Nowak, 1999; Grubb, 2005) may permit them to feed at different elevations. The body size of N. goral and N. griseus is identical and ranges from approximately 22–35 kg for an adult individual (Nowak, 1999; Grubb, 2005). The Sumatran serows, in general, having a larger body size (about 50–140 kg) (Nowak, 1999), are more capable of foraging the higher level of plant bodies, such as leaves of shrubs or small trees, than the Himalayan and Chinese gorals. In an open habitat landscape around Tham Lod Rockshelter, C. sumatraensis might indeed have had more intense competition with other similar-sized grazers such as an Eld’s deer Rucervus eldii (about 95–150 kg; Nowak, 1999) than these two smaller gorals, as suggested by our carbon isotope data (Figure 4 and Supplementary Material S4).

Although the tooth morphology supports a grazing adaptation for both genera, differences in body size between Naemorhedus and C. sumatraensis might have led to niche partitioning and minimized intergeneric competition, which have likely contributed to their prevalent coexistence during the Late Middle Pleistocene (i.e., Tham Wiman Nakin) and during the Last Glacial Maximum (LGM) (i.e., Tham Lod Rockshelter).

Today Chinese gorals and Sumatran serows are surviving in the wild despite having been considered as being vulnerable species due to a population decline (Duckworth et al., 2008a, b), while Himalayan gorals went locally extinct from the Thai territory likely at the end of the Pleistocene based on the lack of their fossil remains in the Holocene stratigraphic sequence records, as deduced from the rockshelters of Tham Lod and Ban Rai in Mae Hong Son, Northwestern Thailand. We thus proposed other possible extinction-related causes that are discernible through these two archeological sites.

It is likely that the environments around the Mae Hong Son Province where the three caprine taxa did coexist became more homogeneous and closed after the end of the Pleistocene based on the possible δ¹³C shift of C. sumatraensis and Bubalus arnee into more exclusively C₃-dietary signals (Figure 4), and the high diversity and abundance of primates (Supplementary Material S2) in the Early Holocene of Ban Rai Rockshelter. Although an open landscape might have existed around the region of Ban Rai Rockshelter based on the C₄ signals of Sambar deer Rusa unicolor (Figure 4), large bovid remains are very scarce in Ban Rai Rockshelter and cervid subfossils are much less abundant than those found in Tham Lod Rockshelter. Supporting this argument, the sequential δ¹⁸O records of fresh water bivalves from both localities have indicated that substantially increased precipitation and decreased climate variability occurred in Northwestern Thailand after 9,800 BP (Marwick and Gagan, 2011). The climatic and environmental changes related to an increasing development of rainforests starting at the Late Pleistocene/Holocene transition likely contributed to the local extinction of grassland-dwelling taxa as Naemorhedus but other ecological factors also need to be considered.

The major role of anthropic activities was characterized by the accumulation of teeth and bones in Tham Lod Rockshelter where the animal remains have been modified (e.g., traces of fire, stone artifacts, and cut marks) by human hunter-gatherers at that time. Carnivores such as Asiatic black bear Ursus thibetanus and tiger Panthera tigris were major predators in the Tham Lod Rockshelter area but likely did not play a significant role in modifying or accumulating the skeletal elements in the assemblage. According to the observed fire modification of bones, it is assumed that prehistoric humans also hunted these carnivores for food and resources (Wattanapituksakul et al., 2018). Wattanapituksakul et al. (2018) demonstrated that prime adult individuals of Naemorhedus and Capricornis were preferentially selected by the hunter-gatherers based on the mortality profile analysis. The highest number of caprine individuals in the accumulation of the Tham Lod Rochshelter assemblages is observed in N. griseus (represented by the minimum number of individuals: MNI = 40), followed by C. sumatraensis (MNI = 15) and N. goral (MNI = 13). This evidence may provide general information on the abundance of caprine populations in this area (N. griseus > C. sumatraensis > N. goral), reflecting that N. goral was likely more scarce than N. griseus in the area and probably started decresing in its population prior the latest Pleistocene. Moreover, the mortality profile analyzed in Tham Lod caprines implies a hunter-gatherer subsistence strategy with selective hunting of easy preys such as these gorals and serows (Wattanapituksakul et al., 2018). As exemplified by the Tham Lod Rockshelter assemblage, we suggest that the impacts of the human hunting on the Himalayan and Chinese goral community were possibly one of the major factors leading to their population decline in that region during the latest Pleistocene.

Concerning the effects of predation pressure on these gorals, the diversity of carnivores in Southeast Asia appears to have been higher in the closed canopy than the open area throughout the Pleistocene based on numerous fossil records and previous carbon isotope data so far (e.g., Pushkina et al., 2010; Bocherens et al., 2017; Bacon et al., 2018b, c) (Figure 4). Our isotope results analyzed on the Tham Lod carnivores indicate their habitat preferences in closed forest ecosystems far away from the ones where the Himalayan and Chinese gorals lived at that time (Figure 4). Accordingly, the predation probably had a direct effect on which both of the goral species were forced to avoid closed habitats occupied by more diverse predators. Other predators that occupied an open habitat have been poorly known from the Late Pleistocene of Southeast Asia since spotted hyaenas possibly went extinct prior to MIS2 (see discussion in Suraprasit et al., 2015, 2019). These gorals consequently maintained their preferred habitats in the different area as an open landscape where predation pressure was lower or minimal.

Regarding the coexistence patterns, other sympatric herbivore species such as large bovids and cervids mainly occupied the open habitat around Tham Lod Rockshelter based on our isotope results (Figure 4). Accordingly, they possibly competed for available food resources with these three caprine taxa. However, this study has demonstrated that the interspecific exploitative competition between sympatric N. goral and N. griseus in Tham Lod Rockshelter was more intense than the intergeneric overlap for Capricornis and Naemorhedus due to the inflexibility of shared food resources and habitats in an open landscape for both goral species. Indeed, N. griseus might have been better adapted to a broader realized niche than N. goral did according to our stable carbon isotope analysis (Figures 2, 3B). The considerable competition for resources between these two closely related species that have a similar dental morphology and body mass and that occupied the same habitat might have led to their reduced population over time and to an extirpation of one species (i.e., N. goral) at the end of the Pleistocene.

Overall, we propose that these four factors: (1) the loss or reduction of grassland habitats after the latest Pleistocene when rainforests became dominant, (2) the negative effects of human hunting, (3) the predation pressure, and (4) the high interspecific competition between both goral species were possibly main causes driving the patterns of local extinction of N. goral in Thailand.

The pace of decline population of extant threatened Himalayan and vulnerable Chinese gorals are accelerating and their distribution ranges are constantly decreasing (Mead, 1989; Duckworth and MacKinnon, 2008) due to the habitat loss caused by several disturbance factors of natural processes (such as forest fire) and human activities (such as deforestation, hunting, and agricultural and commercial use) (Lydekker, 1907; Anwar, 1989; Duckworth and MacKinnon, 2008; Duckworth et al., 2008a; Ashraf et al., 2015; Shakeel et al., 2015). Some studies on the feeding biology of modern Himalayan and Chinese gorals directly targeted at the field have indicated that these two species are both grazers and browsers based on the food availability in their stomach, gizzard, and rumen content (e.g., Anwar and Chapman, 2000; Awasthi et al., 2003; Junaid et al., 2012; Buranapim and Sitasuwa, 2014). However, the most preferred food of both Himalayan and Chinese gorals are grasses, followed by shrubs, trees, twigs, nuts, and herbs (Duckworth and MacKinnon, 2008; Duckworth et al., 2008a; Fakhar-I-Abbas et al., 2008, 2009; Buranapim et al., 2014; Buranapim and Sitasuwa, 2014). Our stable isotope results of tooth enamel of the Himalayan gorals reflected that this refugee species usually consumes the C₃ vegetation or grasses nowadays on the Himalayan regions where no or few C₄ grasses are able to develop under such an extremely high altitude condition due to an increase of atmospheric CO₂ and humidity levels after the LGM (Galy et al., 2008). As is the case for N. griseus, the current geographic distribution clearly shows the habitat restriction of this species in the high-elevation mountain areas (e.g., the regions of Northwest Thailand and South China). Although the natural or anthropogenic grasslands may exist somewhere in these regions, most of which are covered by dense forests such as deciduous dipterocarp and coniferous forests (Duckworth et al., 2008a; Groves and Grubb, 2011; Buranapim et al., 2014) (Table 1). In subtropical regions where Chinese gorals are living and where vegetation types vary along altitudinal gradients, alpine shrubs and grasslands generally occur at the highest altitudinal zonation of mountainous regions (e.g., Sichuan in China), followed by coniferous and mixed evergreen and deciduous forests in the lower-elevation canopy landscapes (Chen et al., 2009, 2012) (Figure 5). In contrast, based on paleobotanical evidence, the plant zonation in Southeast Asia during Pleistocene glacials has been suggested to drop about 1,000 m in elevation lower than today, implying the expansion of grasslands in low-altitude mountain ranges and lowland floodplains (e.g., Kershaw et al., 2001; Morley, 2012) (Figure 5). A long-term ecological stasis in the occupation of both N. goral and N. griseus based on our isotope analysis of tooth enamel through several periods of glacial-interglacial cycles therefore provides a dietary consistency on (C₃ or C₄) grass components but a contrasting traditional view of their habitat use during the Pleistocene when the anthropic impacts on the ecosystems were lower than today. This study has demonstrated that the populations of both N. goral and N. griseus were C₄ grazers in lower altitude grasslands on mountainous areas or lowland floodplain during the Pleistocene glacials (e.g., Pha Bong, Tham Wiman Nakin, and Tham Lod Rockshelter) (Figure 5). It is thus presumable that, during the Holocene, the warming climate and intensely anthropic effects have forced these two goral species to retreat upslope in pursuit of their comfort zone with the food availability (grasses). Particularly, the critical anthropic effects still happen on a regular basis upon the modern wildlife goral community with or without intention and may drive these species to the population decline or extinction risk in the future. Besides minimizing the ecological effects caused by human activities, we propose an alternative conservation-concern plan about the sustainable biodiversity management by supplying an optimal open lowland area with the availability of grass resources to the rest of Himalayan and Chinese goral populations.