The mammal distributional data and patterns
Our analysis of mammalian species distributions reveals four intriguing patterns: (1) there is a peak in species range limits between the central and southern parts of the Thai–Malay peninsula (5–6° N); (2) a second peak in range limits occurs where the peninsula meets the mainland (13–16° N); (3) few species and range limits occur in the central and northern peninsula (8–12° N); and (4) fewer species than expected occur near the southern end of the peninsula (1–2° N). (Cattulo et al. (2008) independently reported the last two patterns while this paper was in press.) We will now offer explanations for these features. The issues involving the non-concordance of the northern and southern range limits of all mammals, and of limits of bats and non-volant mammals will be interesting topics for future analyses.
Overall, the mammal faunas of Indochina and Sundaland appear more similar to one another than are the avifaunas of these two subregions (Hughes et al., 2003). At higher taxonomic levels there is no marked distinction between the mammal faunas of Thailand and of peninsular Malaysia (Table 1). In fact, some species used to characterize one or other subregion occur in both; for example, the Sumatran rhinoceros, tapir and orangutan are not strictly Sundaic, as all occurred in Indochina in historical times. Nevertheless, there is a substantial turnover at the species level in many genera (see Appendix S1 and below). In this discussion we first note the limitations of our database, and then interpret our results, focusing on the historical causes of faunal differentiation.
Our database has several limitations. First, it is based on a mix of records in European and North American museums but is by no means exhaustive: a comprehensive review of all available specimens will undoubtedly change a few range limits. Moreover, as many of the records are more than 50 years old it should not surprise us if some museum specimens were misidentified. Systematic revisions and on-the-ground surveys are underway and cryptic species are to be expected, especially in some groups (Baker & Bradley, 2006; Guillen-Servent & Francis, 2006; Kingston et al., 2006). We ignored subspecific taxonomy but note that there are large numbers of nominal subspecies in some species; for example, Callosciurus erythraeus (44 nominal taxa), C. notatus (36) and Rattus rattus (> 51 in Southeast Asia alone). The vast majority of these subspecies were named c. 100 years ago and, following the arguments of Wilson & Brown (1953), probably have little merit in defining evolutionarily significant units. Some such taxa will, however, undoubtedly be elevated to species rank as a result of ongoing morphometric, karyotypic and molecular genetic analyses (e.g. Endo et al., 2004; Hayashida et al., 2007), and this will, in turn, increase the number of range limits that require consideration by perhaps 10%. However, we do not expect that changes in species designations will substantially alter our findings.
We also recognize that the transect has not been uniformly surveyed but that correction for observational effort is not feasible. Some areas have never been well studied, including the region along the Thai–Malaysian border (5°30′–6°30′ N) and the area where Myanmar and Laos share a border between 20 and 21° N. Another area has not been surveyed for over 40 years: the entire Tenasserim extension of Myanmar from 10 to 17° N. These three areas are of significance to our analyses, but, although the dearth of recent information is unfortunate, the gaps in coverage are unlikely to be filled soon. In contrast, other regions are comparatively well known, including the area around Kuala Lumpur (3° N; Medway, 1983), Khlong Saeng Wildlife Sanctuary (9° N; Nakasathien, 1989; Lynam, 1997) and Khao Yai National Park, Thailand (14° N; Kaewprom, 2004; Lynam et al., 2006; Suzuki et al., 2006, 2007).
Geographic ranges can change rapidly. Overhunting in the last few hundred years has caused range collapses of more than 1000 km in Sumatran rhino, Javan rhino, giant panda, concolor gibbon and tapir (Tougard & Montuire, 2006). Deforestation has fragmented formerly continuously distributed populations throughout the lowlands. Land clearing for rice farming in central and north-east Thailand over the last few thousand years, coupled with hunting, has contributed to the extirpation of hog deer and Eld’s deer, and to the extinction of Thailand’s endemic Schomburgk’s deer. Further south, extensive rubber plantations in Thailand and Malaysia have had similar undocumented but generally negative impacts on numerous species since the 1940s. Such rapid changes have two implications for the current analysis. First, the ranges reported herein may not be the historical ranges of interest of some taxa, and, in the few cases where historical data are available, these were considered. Second, many published range maps are misleading as they show continuously distributed populations where today there may be only a few isolated populations remaining. In this regard, some of the SAMD (2007) maps, which plot suitable habitat for a particular species, provide an important advance.
At first glance, it might appear that we have documented yet another case for which species diversity increases towards the equator (Lomolino et al., 2006). We report 134 mammal species at 20° N and 198 species at 4° N. Parenthetically, we can also note that there are 102 mammal species at 22° N (Xishuanbanna, China) (Zhang, 2000; Smith et al., 2008) and > 222 species in Borneo (centred on the equator) (Payne et al., 1985). This gradient is not simple, however, and it deviates from a monotonic latitudinal increase in three locations. First, there appears to be a peninsula effect in the far south, resulting in a decline in diversity from 198 species at 4° N to 128 species at 1° N. A similar effect has been found in many groups of animals in other peninsulas, including Florida, Baja California, Iberia, Italy, Korea and Cape York (references in Lomolino et al., 2006). One problem with this peninsula effect hypothesis is that Bird et al. (2006) suggest that the Straits of Singapore are only 125 kyr old and the effect must therefore be very recent. Second, there is a significant reduction in species diversity in the central and northern parts of the peninsula (7–14° N). In addition, 35 species (11%) have distributional gaps in the northern or central peninsula. If there were a simple latitudinal gradient in diversity we would expect c. 175 species at 10–12° N; instead we observe only 117–118 species, which is 33% fewer than expected. The deficiency is greater among non-volant mammals than among bats. Third, there is a minor deficiency (9%) of species at 16° N on the continent mainland that is more pronounced in bats than in other mammals. This region, the hills to the west of Tak, is less well surveyed, and this deficiency may be an artefact of sampling effort. Despite these three exceptions to apparent general increase in species richness towards the equator we must note that the habitat cells being compared are not of equal area. In this first exploration of the latitudinal pattern we have not corrected for the fact that the continental sample areas are larger than some cells in the peninsula. A follow-up study in the author’s laboratory, based on the digital records of SAMD (2006) and the newly available IUCN database (Schipper et al., 2008), takes such available habitat differences into account (Luke Gibson, personal communication).
The mammalian pattern reported here is different from that described earlier for birds along a portion of the same transect (6–19° N; Hughes et al., 2003). In forest-associated birds there is a major transition between Sundaic and Indochinese faunas, with 47% of total species limits occurring between 11 and 15° N vs. only 19% in the mammals. Neither bats nor non-volant mammals show range-limit distributions similar to that of birds. One might expect bats to have patterns of range limits similar to those of birds because of their comparable vagility, but the pattern in bats is in fact more similar to that of the non-volant mammals. As noted above, 11% of the mammals have gaps in their distribution in the northern and central peninsula; these 35 species occur in continental Thailand and in peninsular Malaysia but have not been recorded in historical times in the upper half of the peninsula. Similar gaps were reported among 13% of the birds studied previously (Hughes et al., 2003; Round et al., 2003).
Repeated sea-level changes as a cause of the mammal pattern
We now turn to the larger question of what caused the divergence of the Indochinese and Sundaic faunas. It is widely believed that only significant and persistent geographic barriers to dispersal can produce faunal divergence. Woodruff (2003b) surveyed the literature and found that no geophysical or climatic barrier had ever been proposed, and hypothesized that previously unrecognized marine transgressions could account for the avian pattern described with his colleagues (Hughes et al., 2003; Round et al., 2003). He identified two periods when seaways submerged the northern and southern ends of the central peninsula: when global sea levels reached +100 m from 4 to 5 Ma and +100–150 m from 15 to 20 Ma. His reconstruction of Neogene sea levels was based on the Exxon Production Research Company (Exxon) global eustatic curve derived from proprietary seismic studies of coastal sediments (Vail et al., 1977; Vail & Hardenbol, 1979; Haq et al., 1987). Recently, Miller et al. (2005) published a new global curve that validates the Exxon curve with respect to the number and timing of sea-level events, but shows that the earlier estimates of amplitude are at least 2.5 times too high. This advance negates Woodruff’s (2003b) reconstruction and we accordingly offer the following revision.
The Miller et al. (2005) sea-level curve is also based on the sedimentary sequence unconformities (when one erosional or non-depositional surface is replaced by another) observed in five New Jersey coastal boreholes, and spans the Phanerozoic, the last 543 Myr. Unconformities reflect changes in sea levels and local tectonics, but concordance of sequence ages in geographically disparate localities are interpreted as indicators of global eustatic change. Miller et al. used a quantitative method called backstripping to distinguish the contributions of glacioeustacy, sedimentoeustasy, tectonoeustasy (at least on passive or stable continental margins) and thermosteric (temperature) effects. Their results are concordant with data from elsewhere in the world and with those obtained from other sea-level proxies. The Miller et al. curve is radically different in appearance to the Exxon curve used by Woodruff (2003b) and many others since 1977 (Fig. 4). In particular, it provides no evidence for +100 m highstands during the last 25 Myr.
Since the analysis presented below was completed, an independent review of the Miller et al. (2005) curve (and others) based on ocean basin dynamics has appeared (Müller et al., 2008). The Müller et al. geophysical analysis corrects for substantial regional (New Jersey) subsistence over the last 70 Myr but does not negate the following discussion or conclusions. The Müller et al. analysis shows that Cretaceous sea levels were higher than estimated by Miller et al. (2005), but the two curves are reconciled over the biogeographically relevant last 25 Myr. Müller et al. (2008) do not consider the effects of global ice so their recalibration of the global curve cannot be used further here.
The Miller et al. (2005) sea-level curve for the last 9 Myr shown in Fig. 4 is based on the analysis of the oxygen isotopic records in the fossil skeletons of benthic foraminifera from two deep-sea cores (ODP 846 and 982). The ratio of 18O to 16O is temperature-dependent and serves as a proxy for ice volume and therefore sea levels (Lambeck & Chappell, 2001). The general features of this curve are concordant with oxygen isotope patterns seen in an averaged ‘stack’ of 57 deep-sea cores over the last 5 Myr (Lisiecki & Raymo, 2005). The periodic rises and falls of sea level are found consistently across many cores and oceans. Miller et al. estimated the sea-level equivalents of each of the 1800 points (stable isotopic ratios) in their curve and dated the record by curve fitting to the orbital time-scale (the Milankovitch curve based on 100,000-year cycles in the Earth’s orbit), and also to reversals in the Earth’s magnetic field and to biostratigraphy. There are errors associated with each determination (typically ± 15 m), and two recent highstands are illustrative. First, as shown in Fig. 4, the sample from core 846 for 130 ka gave a sea-level estimate of +24 m. This estimate is probably spurious as most other lines of evidence point to a +4–6 m highstand between 128 and 116 ka (MIS 5e; Siddall et al., 2007). These lower estimates are based on data from radiometrically dated fossil coral reefs elevated above sea level in the Bahamas, Barbados, Bermuda and New Guinea, and from sedimentary evidence from the mid-Atlantic coast of the USA (Wright et al., 2008). The +6 m estimate is also supported by research specifically on the last interglacial conditions (Otto-Bliesner et al., 2006;Overpeck et al., 2006; Siddall et al., 2007). Second, the reverse error probably applies to the +10 m highstand shown in Fig. 4 at c. 400 ka (MIS 11: 395–415 ka). There is evidence from Atlantic sites that sea levels rose briefly to +20–22 m at the end of the interglacial (Hearty et al., 1999; Hearty & Kaufman, 2000). This is currently the maximum highstand for which there is reasonable evidence in the last 1.5 Myr.
Miller et al.’s (2005) results, as they affect the issues under discussion here, are summarized in Table 4. The mean sea level column in Table 4 shows that four stages can be recognized in the last 63 Myr. First, for 20 Myr during the Palaeocene and Eocene (63–43 Ma) global sea levels averaged +56 m above today’s level. Second, mean sea levels then fell for 13 Myr until 30 Ma, when they reached levels within 10 m of today’s. The third stage, spanning 25 Myr from 30 to 5 Ma, was characterized by mean sea levels that fluctuated within 10 m of contemporary levels. Finally, during the fourth stage, spanning the last 5 Myr, mean sea levels have remained below today’s level, and have declined gradually to an average of −62 m during the last million years.
Table 4. Tertiary sea levels and fluctuations estimated from Miller et al. (2005: Supporting Information, Table S1, best estimate). The data comprise sea-level estimates for every 5000 years between 0 and 9.3 Ma and every 100,000 years thereafter. Means were calculated from these data and ranges may actually be greater, given the gaps in the record. Fleeting highstands of < +2 m and <1 kyr were ignored.
|Age (Ma)||Mean sea level (m)||Sea level range (m)|| Highstands > +0 m||Rapid sea level rises (m)||Duration (kyr) of lowstands of > −60 m|
|Number||Duration||% of period||> 40||> 80|
|43–63||+56||+20 to +133||1||20 Myr||100||3||1||0|
|30–43||+18||−12 to +54||4||1.3 Myr||95||4||0||0|
|20–30||−4||−36 to +39||9||3.4 Myr||34||3||0||0|
|10–20||−1||−18 to +21||7||4.0 Myr||40||0||0||0|
| 9–10||+8||−36 to +40||6||205 kyr||21||1||0||0|
| 8–9||+9||−23 to +38||14||775 kyr||78||2||0||0|
| 6–8||+1||−42 to +26||> 42||1.12 Myr||56||1||0||0|
|5.5–6||−7||−45 to +19||6||< 29 kyr||< 0.01||2||0||0|
|5–5.5||+8||−28 to +49||14||255 kyr||51||3||0||0|
|4–5||−8||−48 to +21||> 16||222 kyr||0.02||5||0||0|
|3–4||−11||−67 to +22||> 18||135 kyr||0.01||7||0||< 15|
|2–3||−16||−70 to +25||17||160 kyr||0.02||14||1||15|
|1–2||−38||−92 to +10||5||19 kyr||< 0.01||16||2||185|
|0–1||−62||−120 to +10||4||23 kyr||< 0.01||6||7||530|
Today, a highstand of +50 m would almost breach the northern end of the central peninsula between Krabi and Surat Thani (Fig. 5e), and a highstand of +70 m would also flood the southern end of the central peninsula between Kangar and Songkla. Such transgressions would connect the Gulf of Thailand to the Andaman Sea and create multiple narrow, shallow marine barriers to land mammal dispersal. Mean sea levels were +56 m during the period 43–63 Ma, and fluctuated between +20 and +133 m. From 51.7 to 53.6 Ma mean sea levels remained above +74 m, and on six other occasions during the early Tertiary exceeded +60 m. This suggests that the central peninsula comprised an island chain for significant periods during the early Tertiary and was never as extensive as it is today. Early Indochinese mammals would have had to island-hop across the central peninsula to reach the pre-Tertiary mountains of peninsular Malaysia, Sumatra (itself an archipelago until the Pliocene), Java and Borneo. This first high-sea-level stage of the Tertiary transitioned into the second stage 43 Ma, when mean global sea levels started a 13-Myr decline to levels similar to today’s. During this time span (30–43 Ma; middle and late Eocene and early Oligocene), mean sea levels were +18 m (fluctuating between −12 and +54 m) and were probably never again high enough to breach the central peninsula.
Figure 5. Schematic maps showing the extent of Sundaland when sea levels, relative to today’s, were at (a) −116 m, (b) −80 m, (c) −40 m, (d) +25 m and (e) +50 m. Map (a) shows the extent of Sundaland during the Last Glacial Maximum c. 20 ka and on two other occasions during the last million years (Fig. 4). The current positions of the Southeast Asian mainland and major islands are shown. During the last 1 Myr, sea levels fluctuated widely around −62 m and reached −80 m (map b) on at least a dozen occasions. Map (c) shows the sea level at −40 m, the average position of the coast 1–2 Ma, with half of the Sunda Shelf exposed. Map (d) shows the +25 m transgression (shaded) of numerous earlier Pliocene and late Miocene highstands. Finally, map (e) shows the extent of the brief +50 m transgression c. 5.4 Ma (land above +50 m shaded); this event, the maximal transgression during the last 10 Myr, does not appear to have breached the central peninsula. [Maps (a) to (c) redrawn from Sathiamurthy & Voris (2006) (see source and original full colour maps for details); maps (d) to (e) prepared by Kathryn Woodruff, from Woodruff & Woodruff (2008).]
Download figure to PowerPoint
The third stage of Tertiary sea-level history commenced c. 30 Ma and was characterized by mean sea levels fluctuating within 10 m of today’s level (Fig. 4). Throughout the Oligocene and Miocene the peninsula was probably similar to today’s in its basic geographic extent and physiography (Hall, 2001, 2002; Woodruff, 2003b). Although the peripheral areas of Southeast Asia (Sumatra and the Philippines) have changed dramatically during this period, the Thai–Malay Peninsula has existed for 100 Myr, and sea-level changes, rather than large-scale tectonic changes, have controlled its basic appearance. Extensive ever-wet rain forest and more seasonal forests have dominated the region since c. 20 Ma (Whitmore, 1998; Woodruff, 2003b). There was ample land and time for the dispersal and range extension of Indochinese mammals south into Malaysia and, as we will explain below, out to today’s Sundaic islands. We are more concerned with north-to-south dispersal than with the reverse, as the eastern end of Sundaland (marked by Wallace’s line) constituted a relatively hard barrier to land mammals throughout the Cenozoic (Hall, 2001, 2002). However, we also recognize that opportunities for overland dispersal in the reverse direction occurred whenever sea levels fell below c.−30 m, as they have done repeatedly since c. 6 Ma.
The fourth stage of sea-level history began c. 5 Ma and continued until very recently. During this period mean sea levels remained below today’s level and declined gradually to an average of −62 m over the last 1 Myr. Because sea level declines to below −40 m will expose half of the Sunda shelf (Sathiamurthy & Voris, 2006) (Fig. 5c), the peninsula became an insignificant fraction of the habitat available to mammals, especially during the last 2 Myr. During most of that period, the Indochinese and Sundaic faunas would have been in broad geographic contact across two million km2 of emergent surrounding plains. Only during Pleistocene highstands did anything like a peninsula appear, and then, for one with today’s coastline, for only 42 kyr or 2% of the last 2 Myr.
This substantial revision of the palaeogeographic history of the Thai–Malaysian peninsula provides no clear physical barrier to account for the divergence of Indochinese and Sundaic species pairs. The narrow seaway that occurred for a few million years 50 Ma is irrelevant to the evolution of today’s species. Similarly, there is no simple explanation for the concordance of species range limits at 5–7 or 13–14° N based on the consideration of mean sea levels during the Neogene. However, a closer examination of the eustatic curve suggests a likely explanation for both issues. We here propose that the dramatic and repeated fluctuations in sea level during the last 5 Myr will account for these observations.
To see how sea-level fluctuations might account for the differentiation of Indochinese and Sundaic species pairs and the concordance of species range limits one needs to consider the impact of each glacioeustatic cycle. In the last such cycle, beginning c. 130 ka, sea levels fell gradually and episodically from +6 m (isotopic stage 5e, 120 ka) to c.−122 m (Fig. 5a; stage 2, 18–20 ka). Since then, sea levels rose quickly until they reached today’s level c. 6 ka. This rise of > 100 m in less than 12,000 years had a very dramatic effect on the mammals inhabiting Southeast Asia. According to Sathiamurthy & Voris (2006), the total area of Sundaland above sea level shrank from 4,709,397 to 2,342,826 km2, a decrease of 50% (Fig. 5a). Although the rise was very rapid in a geological sense and exceeded 20 m kyr−1, most mammals would have had no difficulty adjusting to a 20 mm year−1 transgression. Even the maximum transgression of 20 m in 200 years c. 14,600 yr bp, or 100 mm year−1, was biologically trivial. It is unlikely that many mammal populations were ‘drowned away’, except for some trapped on the originally large South China Sea islands between peninsular Malaysia and Borneo (the Anambas, Natuna Besar and Natuna Selatan archipelagos; Meijaard, 2003). Most mammals would have had ample time to move away from the edge and inland towards higher ground. The resulting faunal compression may have led to intensified competition and, in some isolated habitat patches, to population extinction and even species extirpation. Under the equilibrium model of island biogeography we would expect a gradual loss of species with shrinking area (MacArthur & Wilson, 1967; Heaney, 1984, 1986). This may account for the observed reduction in species numbers in the narrowest parts of the peninsula today. Faced by the rising sea level, many mammals living on the vast Sundaic plains would have moved north to the continent or west to the hills of the peninsula. Those in the latter category would have found themselves restricted to very small habitat patches if they ended up in the northern or central peninsula. Those that moved into the larger and more mountainous southern peninsula would have enjoyed far more area (the peninsula was broadly connected to Sumatra until sea levels rose above −40 m) and greater topographic relief (hills exceeding 2000 m). But all would have encountered a resident fauna that had inhabited the peninsula for millions of years.
Although the last glacioeustatic cycle’s impact is useful in illustrating the principal effect of a rapid sea-level rise it will alone not account for our biogeographical observations. To understand both species numbers and distributions in this area we must consider the impact of multiple rapid rises of sea level (Fig. 4, Table 4). There were seven transgressions of > 80 m in the last 1 Myr. In addition, there were six rapid rises of 40–80 m. In the fourth phase of Tertiary sea-level history, in the last 5 Myr there have been 10 rapid rises of > 80 m and 48 rapid rises of 40–80 m. We hypothesize that these 58 rapid rises, coupled with the eight that preceded them 4–6 Ma, account for the current patterns of mammal species distribution. Although we view rapid rises of > 40 m as significant for peninsular mammals, we note that there were many more smaller rapid rises of 20–40 m that may amplify the overall impact of the larger rises.
Each rapid rise began from a different depth and consequently resulted in a different reduction in habitat area. For this reason we suspect that rapid rises that began from below −40 m would have had the greatest impact. Such rapid rises from below −40 m dominate the eustatic pattern for the last 3 Myr but occurred at lower frequency back to 6.1 Ma. At sea levels of −40 m the sea would have retreated from both the east and west coasts of the present peninsula, resulting in an isthmus broadly connected to the mainland in the north and to Sumatra–Java–Borneo in the south (Fig. 5c). The area associated with today’s peninsula doubled. Both the South China Sea and the Gulf of Thailand remained extensive open water. In contrast, by −60 m the Gulf of Thailand had disappeared and by −80 m the South China Sea had retreated over 500 km and the Sunda shelf had emerged between Borneo and Vietnam (Fig. 5b). We focus here on −40 and −80 m geographic patterns because a rapid rise from the former would have halved the habitat of peninsular mammals, and most rises of > 40 m would have had an even greater impact. Sathiamurthy & Voris (2006) calculated the total additional land exposed with each 10 m drop in sea level but their estimates are based on the entire Sundaland region. A study of their reconstructions shows that when just the focal region (the peninsula) is considered, different estimates would emerge depending on how the local region’s boundaries are set. Nevertheless, a sea-level drop to −40 m would have doubled the peninsula’s area, a drop to between −50 and −60 m would have added the greatest area of land relevant to peninsular mammals, and drops to below −80 m would have had little additional local impact.
The above discussion has been premised on the idea that today’s coastlines have great antiquity and are a proper basis for palaeobiological reconstructions. In fact they are not, as biologists have long appreciated (Wallace, 1869; Woodruff & Woodruff, 2008). In addition to the seashore retreats discussed above there have been repeated marine transgressions during the last 10 Myr. Figure 4 shows 31 transgressions of +10 to 20 m during the last 4 Myr. Figure 5d shows the maximal areal extent of these brief transgressions (at +25 m) on today’s peninsula. These transgressions had only a minor impact on land area (< 10% reductions) except in the central peninsula and the far south. In the central peninsula, between the towns of Krabi, Surat Thani and Nakhon si Thammarat, the transgressions halved the peninsula’s width. Although the duration of such transgressions was very short, with each episode lasting hundreds not thousands of years, their ecological impact was locally catastrophic. The brief higher transgressions between 5 and 6 Ma (to +49 m) had little additional areal impact (Fig. 5e). The Holocene highstand of +2.5 m would have had virtually no impact. Together, these multiple brief transgressions would have amplified the general effect of the rapid sea-level rises but they themselves were not responsible for the overall divergence of the Indochinese and Sundaic faunas.
Our work shows that modern mammal distribution patterns have three features that are linked: (1) the peaks in species range limits north of the northern peninsula, (2) the peaks south of the central peninsula, and (3) the fall in species diversity in the central and northern peninsula. We hypothesize that an area effect accounts for these observations. A 70% reduction in habitat area in the north and central peninsula could account for the observed 30% decline in species number (MacArthur & Wilson, 1967; as applied to Southeast Asian mammals by Heaney, 1984, 1986). We do not know the ages of most of the extant Southeast Asian mammal species under discussion but assume that they are comparable to those in better-dated faunas, and that most species originated in the last 2–4 Myr (Avise, 2000). We hypothesize that most of the species were in place two or more million years ago, and that multiple rapid sea-level rises are the one unusual phenomenon affecting this regional fauna during this time period. We hypothesize that the > 50 episodes of significant faunal compression onto the peninsula account for the accumulation of species borders at the geographic edges of the more stable areas, both at the northern end of the large mountainous southern peninsula and at the northern end of the peninsula itself. The reduced numbers of mammal species in between these areas is probably the result of a repetitively imposed area effect.
Reconstructing Neogene mammal distribution patterns
Although the effects of sea-level change on habitat area appear highly significant to us, we recognize that other factors must also be considered in explaining today’s mammal distribution patterns. We conclude this discussion by noting the need for consideration of four additional factors likely to be important in the final solution of the issues raised: palaeoenvironments, the role of individual species ecology, the fossil record, and the genetic variation and phylogeography of today’s populations. Future discussions will also have to begin with the premise that the present geography of this region is not a simple key to the past: today’s high sea levels and warm-wet climates are highly unusual and have prevailed for less than 3% of the last few million years.
In the above palaeogeographic reconstruction we have implied that the emergent Sunda shelf constituted an equal opportunity for range expansion of all mammals: it certainly did not. Forest-associated mammals would be far more attuned to the differences between perhumid rain forest, and seasonal, monsoonal and mangrove forest than we have allowed. The current determinants of the distribution of representative members of these plant associations have only recently come under investigation (Baltzer et al., 2007, 2008). The past distribution and species composition of these plant associations would have changed with each cool-dry hypothermal and warm-wet hyperthermal phase, and these climatic cycles have been becoming progressively cooler since the Pliocene. Furthermore, much of the flat Sunda plain may have been vegetated by Pinus savanna woodland or grassland, and not by rain forest (Whitmore, 1987, 1998; Heaney, 1991; Taylor et al., 1999; Morley, 2000, 2007; Kershaw et al., 2001, 2007; Gathorne-Hardy et al., 2002; Hope et al., 2004; Hope, 2005). Mammals restricted to primary rain forest could have crossed these plains along riparian forest corridors or could have become isolated by ecological barriers (Gorog et al., 2004). The areal extent and distribution of savanna and rain forest (and rain forest refugia) are still controversial (Sun et al., 2000, 2003; Taylor et al., 2001; Meijaard, 2003; Bird et al., 2005; Tougard & Montuire, 2006; Kershaw et al., 2007; Morley, 2007), but the implications of these palaeoenvironmental changes need to be coupled with the impact of the area changes. Habitat requirements are clearly significant to some species’ distribution patterns and responses to change (see Meijaard, 2003), and should be taken into account in future analyses.
Although habitat availability (for example limestone outcrops) may determine the distribution of some species, habitat selection is rarely a fixed trait in mammals, and palaeogeographic reconstructions based on the ecology of living populations can be misleading. Ecological niches, fundamental and realized, can change over time. Similarly, niches of widely distributed species may vary geographically. An example of the latter involves the lesser gymnure, Hylomys suillus, which ranges from China to Borneo: Bornean populations are exclusively montane (> 1000 m) but elsewhere the species ranges down to sea level. Similar niche shifts have been reported in several species of Indochinese birds that range south towards the equator (Round et al., 2003). A full understanding of current species range limits requires the investigation of the plasticity or conservatism of each species niche (Wiens & Graham, 2005). It must also address the denial of the niche concept in neutral theories (e.g. Hubbell, 2001). Some of these ideas are being tested by comparing niches of conspecific populations in species-poor and species-rich areas of the transect (L. Gibson and D.S.Woodruff, in preparation).
Repetitive range compression will also leave a mark on the genetic structure of today’s populations. Genetic drift in small isolated populations will rob them of their genetic variability. Interestingly, several authors have reported evidence consistent with repeated demographic bottlenecks in a variety of taxa in this region (Inger & Voris, 1993; Cannon & Manos, 2003). As proposed for the avifauna (Woodruff, 2003b), a molecular clock approach could be used to date the divergence of the Indochinese and Sundaic faunas. Candidate mammalian genera with species-pairs appropriate for such an analysis include Nycticebus, Tupaia, Macaca, Callosciurus, Hylopetes, Rattus, Niviventer, Hipposideros and Prionailurus. Our prediction that Indochinese–Sundaic species pairs diverged following repeated range disruption in the central and northern peninsula 1–4 Ma can be tested. The resolving power of phylogeographic analyses is well illustrated by recent examples in Southeast Asian rodents (Mercer & Roth, 2003; Gorog et al., 2004), primates (Brandon-Jones, 1996; Tosi et al., 2002; Harrison et al., 2006; Steiper, 2006; Ziegler et al., 2007; Roos et al., 2008), bats (Thabah et al., 2006), frogs (Emerson et al., 2000; Inger & Voris, 2001), arachnids (Su et al., 2006; Warrit et al., 2006), insects (Pramual et al., 2005; Quek et al., 2007) and freshwater prawns (De Bruyn et al., 2005; De Bruyn & Mather, 2007).
Future discussions must also take into account the growing body of information on fossil mammals and their distribution. Although we did not incorporate this evidence in our analysis as so many of the fossils are undated, we recognize that this situation is changing rapidly. Chaimanee (1998), Chaimanee & Jaeger (2000a,b), Bacon et al. (2006) and Tougard & Montuire (2006) have made significant progress in elucidating the Pliocene–Pleistocene evolution of rodents and should be consulted for references to the broader palaeontological literature. Murine rodents are important regional palaeoenvironment indicators and their data can be interpreted to reconstruct the past extent of forest and savanna communities. The fossil record shows that the transition between Indochinese and Sundaic rodents lay south of the Isthmus of Kra during parts of the Pleistocene, and that more than a dozen species previously ranged further north or further south than they do today. Chaimanee also identified five genera of rodents no longer present in the area. Tougard (2001) and Tougard & Montuire (2006) have reviewed the evidence for mammal movements between the Indochinese and Sundaic subregions during the Pleistocene, and these analyses again underscore how different mammal communities were in this region in the not so distant past. The reconstruction of the history of today’s mammals in this area will also be incomplete until the impacts of the disappearance of the megafauna are understood (see, for example, Tougard et al., 1996; Van den Bergh et al., 2001; Corlett, 2006, 2007; Hill et al., 2006; Louys et al., 2007; Louys, 2008). The megafauna of the region included proboscideans (Stegodon and Palaeloxodon), hippopotamus (Hexaprotodon), hyenas (Crocuta and Pachycrocuta), giant panda (Ailuropoda), tapirs (Tapirus and Megatapirus), rhinoceroses (Rhinoceros and Dicerorhinus), giant pangolin (Manis) and giant primates (Pongo and Gigantopithecus). Their regional extirpation or extinction is less related to the arrival of Homo erectus (c. 1.9 Ma) than it is to that of H. sapiens (c. 70 ka). Similarly, a meteor’s impact 793 ka just east of our transect (near Ubon Ratchathani in eastern Thailand) has to be assessed as it may have killed mammals over an area half the size of Sundaland (Hope, 2005; Meijaard & Groves, 2006). The cataclysmic eruption of Mt Toba, Sumatra, c. 74 ka, albeit along the western edge of the region, appears to have had a limited effect on mammals (Louys, 2007). The ever-improving dating of the region’s fossils, coupled with the demonstrable association of orbital forcing with climate and sea levels (Vrba, 1992; Dynesius & Jansson, 2000; Van Dam et al., 2006), provide a better framework for the consideration of these complex interactions.
Finally, several contrasting hypotheses that were developed to explain species-richness patterns elsewhere (reviewed by Hortal et al., 2008) should now be examined for their applicability to the situation on the Thai/Malay peninsula. Competitive replacement models (Case & Taper, 2000), which appear relevant to the avian transition, are clearly not going to explain the mammal pattern. By contrast, the habitat theory hypotheses developed by Vrba (1992) to explain the Central American mammal interchange appear highly relevant to the Indochinese–Sundaic mammal transition, albeit on a lesser taxonomic scale. Habitat theory models, in which habitats shift, shrink or disappear in response to cyclical climatic change, lead to a set of evolutionary hypotheses that can be tested by comparing, for example, the differential survival of extant rain forest, savanna, grassland, lowland and montane species. Similarly, the differential dispersal of Sundaic and Indochinese species merits attention: were the overland dispersal opportunities from Indochina to Borneo comparable to those from Sundaland back to the mainland, as suggested for two groups of primates (Ziegler et al., 2007; Roos et al., 2008)? In turn, the answers to such questions have implications for projecting the response of today’s species to ongoing climate changes. Interesting as such discussions will be, they lie beyond the scope of the present paper.