Savanna in equatorial Borneo during the late Pleistocene

Equatorial Southeast Asia is a key region for global climate change. Here, the Indo-Pacific Warm Pool (IPWP) is a critical driver of atmospheric convection that plays a dominant role in global atmospheric circulation. However, fluctuating sea-levels during the Pleistocene produced the most drastic land-sea area changes on Earth, with the now-drowned continent of Sundaland being exposed as a contiguous landmass for most of the past 2 million years. How vegetation responded to changes in rainfall that resulted from changing shelf exposure and glacial boundary conditions in Sundaland remains poorly understood. Here we use the stable carbon isotope composition (δ13C) of bat guano and High Molecular Weight n-alkanes, from Saleh Cave in southern Borneo to demonstrate that open vegetation existed during much the past 40,000 yrs BP. This location is at the southern equatorial end of a hypothesized ‘savanna corridor’ and the results provide the strongest evidence yet for its existence. The corridor would have operated as a barrier to east-west dispersal of rainforest species, and a conduit for north-south dispersal of savanna species at times of lowered sea level, explaining many modern biogeographic patterns. The Saleh Cave record also exhibits a strong correspondence with insolation and sea surface temperatures of the IPWP, suggesting a strong sensitivity of vegetation to tropical climate change on glacial/interglacial timeframes.


Supplementary information
The age model for the Saleh Cave guano deposit A total of 10 samples were submitted for radiocarbon dates (Supplementary Table 1). Radiocarbon dates were measured at the Waikato Radiocarbon Dating Laboratory and the Australian Nuclear Science and Technology Organisation (ANSTO). Radiocarbon measurements were calibrated to calendar years using SHCal 1 . An age model for each profile was constructed using Bacon 2.2, and a probability distribution is determined as a function of depth ( Supplementary Fig. 1).
Two samples were identified as outliers by Bacon 2.2, and were repeated by isolating pyrogenic carbon using hydrogen pyrolysis (hypy), a new technique for determining reliable radiocarbon measurements 2,3 . Briefly, samples were immersed in 2M HNO3 for 3 hours followed by 30% peroxide overnight to remove inorganic and labile carbon, then loaded with a Mo catalyst using an aqueous/methanol (1:1) solution of ammonium dioxydithiomolybdate [(NH4)2MoO2S2]. Catalyst weight was ~5 % sample weight for all samples to give a nominal loading of 1% Mo. Catalyst loaded samples were then lyophilized and placed in the hypy reactor, pressurized with hydrogen to 15 GPa with a sweep gas flow of 5 L min -1 , then heated using a pre-programmed temperature profile. We used the recommended temperature program previously optimized for pyrogenic carbon quantification where samples are initially heated at rate of 300 °C min -1 to 250 °C, then heated at a rate of 8 °C min -1 until the final hold temperature of 550 °C for 2 min 2 .

Estimation of tropical grass biomass contribution.
In order to estimate tropical grass (e.g., C4) biomass contribution, two independent estimations were assessed in order to provide a possible range in values: (1) Empirically derived C4 plant estimations determined from a study on insectivorous bat guano 6 and (2) assuming a simple mass balance model.

Empirically derived estimates
An empirically derived equation for bat guano from the southwest US has been published 6 , and subsequently modified to assume that sites with less than 25 mm precipitation/year have no contribution of C4 biomass (Eq 1) 7 . δ 13 Cguano = 10.8 • (C4 relative abundance) -26. 8 (1)

Mass balance
The abundance of C4 biomass was determined using a simple mass balance model where: where δ 13 Cguano is the δ 13 C value of guano, fC4 is the proportion of C4 biomass, δ 13 CC4 is the average δ 13 C value of C4 biomass, δ 13 CC3 is the average δ 13 C value of C3 biomass and e is the fractionation between dietary plant biomass and insect cuticles (which may be different between C4 and C3 vegetation). C4 biomass can be estimated to be -12.5‰ and C3 can be estimated to be -27.5‰ 7,8 . Under different atmospheric δ 13 CO2 conditions, it is possible to compensate for changed plant endmember values by simply adding the difference to the insect cuticle δ 13 C value. However, this parameter deviated by no more than 0.5‰ over the last 22,000 years, so we did not incorporate this change in our calculation. We take e values from insects cultured on C3 or C4 biomass 9 , where e3 = 0.8, and e4 = 0.

Changes in δ 13 C values of forest end-members.
The stable carbon isotope composition of forests varies with the type of forest present 10 . A more seasonal climate might have resulted in tropical seasonal forests or tropical deciduous forests replacing tropical rainforest. Baseline δ 13 C values are approximately 2 ‰ more positive for tropical seasonal forests relative to rainforest, and tropical deciduous forests are 3.8 ‰ higher than tropical rainforest 10 . Notably, δ 13 C values from Saleh cave are above baseline δ 13 C values for even tropical deciduous seasonal forests, unambiguously indicating that tropical grasses were a substantial part of the environments of southern Borneo during most of the past 40,000 years and that the hydroclimate was significantly drier and/or more seasonal (Supplementary Fig. 2).

High molecular weight n-alkanes in guano.
Guano deposits are composed of the undigested components of the bat communities diet, which is largely composed of insect cuticles 11 . Insects produce a variety of hydrocarbons, including linear n-alkanes, alkenes, methyl branched alkanes, and alkadienes 12 . Although high molecular weight n-alkanes with a strong odd over even preference are almost exclusively used as a molecular biomarker of vascular plants in sedimentary records 13 , insects also produce these compounds with a strong odd over even preference 14,15 . A detailed investigation of the hydrocarbons in bat guano determined that although insects almost exclusively produced branched alkanes, high-molecular weight n-alkanes (e.g., C27, C29), were directly assimilated into insect cuticles via plants and synthesized by the insects 15 . In fact, it was estimated that odd high-molecular weight n-alkanes were assimilated vs synthesized at a ratio of approximately 0.3, whereas even n-alkanes were mostly assimilated directly from plants at a ratio of ~0.7.
Nonetheless, δ 13 C values of high-molecular weight n-alkanes are a record of local vegetation, as both synthesized and assimilated compounds are directly reflective of the insects diet 14 . The strong covariation with guano δ 13 C values yields confidence that both proxies yield a reliable signal without diagenetic interference.
Supplementary References