A palaeo Tibet–Myanmar connection? Reconstructing the Late Eocene drainage system of central Myanmar using a multi-proxy approach

Strain resulting from the collision of India with Asia has caused fundamental changes to Asian drainage patterns, but the timing and nature of these changes are poorly understood. One frequently proposed hypothesis involves the connection of the palaeo Tsangpo drainage to a precursor to the Irrawaddy River of central Myanmar in the Palaeogene. To test this hypothesis, we studied the provenance of Palaeogene fluvio-clastic sedimentary rocks that crop out in central Myanmar, namely the Late Middle Eocene–Early Oligocene Pondaung and Yaw Formations. Isotopic analysis on bulk-rock and petrographic data indicate a primary magmatic arc source, and a secondary source composed of recycled, metamorphosed basement material. Although the exact location of both sources is hardly distinguishable because Burmese and Tibetan provinces share common lithological features, the presence of low-grade metamorphic fragments, the heterogeneity in Sr–Nd isotopic values of bulk sediments and westward-directed palaeoflow orientations indicate a proximal source area located on the eastern Asian margin. Central Myanmar was the locus of westward-prograding deltas opening into the Indian Ocean, supplied by the unroofing of an Andean-type cordillera that extended along the Burmese margin. We found no evidence to support a palaeo Tsangpo–Irrawaddy River, at least during the Late Eocene. Supplementary material: Data locations, and isotopic and petrographic results are available at www.geolsoc.org.uk/SUP18655.

The major East Asian rivers have elongated and irregularly shaped drainage basins that have been explained as the consequence of drainage reorganization associated with uplift of eastern Tibet, most probably in the Miocene (Brookfield 1998;Clark et al. 2004;Clift et al. 2008). The unique geometry of the Tsangpo River, with its abrupt and tight loop in the deeply incised gorges of the Namche Barwa massif, and the close proximity of several major rivers in the eastern Himalayan syntaxis region suggest that drainage patterns in the past were very different (see Fig. 1a; Brookfield 1998). These changed when the region was subjected to rapid uplift and erosion that caused a series of drainage reversal and capture events (Clark et al. 2004). One of the most popular models proposes a previous connection between the Yarlung-Tsangpo and the nearby Irrawaddy River via the Parlung and Lohit Rivers ( Fig. 1a; Stamp 1940). However, this commonly held hypothesis has not been extensively studied, and little attempt has been made to better constrain the timing of the proposed Irrawaddy-Tsangpo diversion. Nevertheless, recent work based on phylogenetic studies of freshwater fish has suggested that if any connection existed it would have ceased by the Early Miocene (Ruber et al. 2004).
The Central Myanmar Basin contains an almost continuous succession of Cenozoic sedimentary rocks most of which are associated with fluvio-deltaic depositional environments (see Fig. 1c). During the Palaeogene, rates of deposition were high (Metivier et al. 1999), and the large thicknesses of the Late Middle Eocene to Early Oligocene Pondaung and Yaw Formations represent a significant proportion of the Palaeogene sediments deposited in central Myanmar (United Nations 1978). Rocks of this age are good candidates for a provenance study to test whether palaeo-rivers of Myanmar were once connected to Tibet.
To define the sediment provenance, this study adopted a multiproxy approach, combining sedimentological, petrographic, and isotopic analysis. Rocks of Asian and Indian affinity are distinct in terms of their isotopic, geochemical and petrographic characteristics (DeCelles et al. 2001;Kapp et al. 2003;Chung et al. 2005). Because the Yarlung-Tsangpo River flows through the region of the Indus-Tsangpo Suture Zone, the boundary between Indian and Asian crust, any previous connection to the Irrawaddy would be reflected by the presence of sediments with features diagnostic of these regions (i.e. Transhimalayan material), thus provenance studies may be used to reconstruct palaeo sediment routing systems. Tools for tracing provenance are numerous, and have been widely used to help reconstruct the evolution of the Indo-Asian collision (see, e.g. Najman 2006, for a comprehensive review). Methods used include palaeocurrent analysis, quartzfeldspar-lithic fragment (QFL) point-counting, heavy mineral identification, and Nd and Sr isotopic analysis of bulk sediments (France-Lanord et al. 1993;Uddin & Lundberg 1998a;Colin et al. 1999;Pierson-Wickmann et al. 2001;Clift et al. 2001Clift et al. , 2006Najman et al. 2008).
Lhasa terrane, which occupies the southern part of the Tibetan Plateau and which collided with Asia during the Late Jurassic-Early Cretaceous (Mitchell 1993). The Shan-Thai Plateau forms a large and elevated area in eastern Myanmar, and constitutes a local unit of the Sibumasu block (Bender 1983;Searle et al. 2012). At the transition between the Central Myanmar Basin and the Shan-Thai block, the deep basement of the Burma terrane is partially exhumed and comprises a succession of metamorphic belts ( Mitchell et al. (2004Mitchell et al. ( , 2012 and Pubellier et al. (2008); boundaries between Tibetan and SE Asian terranes are not precisely drawn because they are poorly known. The location of the Palaeogene Burmese samples of Allen et al. (2008) Mitchell et al. 2007Mitchell et al. , 2012. On its western margin, the Central Myanmar Basin is separated from the Assam Basin by the Indo-Burman Ranges, which constitute the Cenozoic accretionary complex produced by subduction of the Indian Plate under the Burma terrane (Brunnschweiler 1966;Maurin & Rangin 2009).
Following the Cenozoic deformation of Asia in response to India's penetration, central Myanmar primarily followed the northward motion of the Indian Plate and the rotational motion of the Indochina Plate, through intense strike-slip deformation and c. 20-30° clockwise rotation relative to South China (Tapponnier et al. 1986;Richter & Fuller 1996;Benammi et al. 2002;Hall 2002;Morley 2004). Until the Middle Miocene, strike-slip deformation was partially accommodated by the deep pull-apart formation of the Central Myanmar Basin, which led to exhumation of high-grade metamorphic rocks that partially form the Burmese basement (Rangin et al. 1999;Bertrand et al. 2001;Barley et al. 2003;Bertrand & Rangin. 2003;Searle et al. 2007). Middle Miocene spreading of the Andaman Sea and the emplacement along the eastern highlands of the Sagaing Fault accommodated the strike-slip motion (Khan & Chakraborty 2005). The total dextral displacement of central Myanmar has been variously estimated at 300-400 to about 1100 km (Maung 1987;Mitchell 1993;Curray 2005;Morley 2009).
The palaeogeographical reconstructions of Hall (2002) and Morley (2002Morley ( , 2004Morley ( , 2009) summarize and represent a compromise between the various palaeogeographical models. Late Middle Eocene simplified palaeogeography of the region is shown in Figure 2a, and a schematic structural map of central Tibet and Myanmar is proposed in Figure 2b. In the Eocene collision zone, central Tibet constitutes an area of old, inherited and young, collision-related highlands (Rowley & Currie 2006;Dupont-Nivet et al. 2008;Wang et al. 2008). North of the Indus-Tsangpo Suture Zone, the Asian basement is composed of granitic and sedimentary terrains of the Lhasa terrane (Kapp et al. 2003). The Transhimalayan arc represents the original Andean-type continental arc, formed by the subduction of Tethyan oceanic crust beneath the Asian active margin (Chung et al. 2005;Ji et al. 2009). In Eocene times, the Tibetan Sedimentary Series, a Palaeozoic to Eocene sedimentary succession deposited on the northern passive margin of India, may have been uplifted early and exposed south of the Indus-Tsangpo Suture Zone (Najman 2006;Najman et al. 2008). Further east, the Burmese margin was occupied by the Burmese magmatic arc, considered as the eastward continuation of the Transhimalayan arc and today represented by isolated granitoid exposures in the Burma basement terrane (Zaw 1990;Mitchell 1993;Mitchell et al. 2012). Further inland, the Shan and Yunnan highlands have been emergent since the Late Cretaceous period (Bender 1983), and have undergone minor uplift during the Palaeogene (Morley 2004). The Central Myanmar Basin constituted a NW-SE-oriented ribbon of pull-apart sub-basins, situated in a typical fore-arc position between the Burmese magmatic arc and the Indo-Burmese trench, where the Indo-Burmese rise is currently located.
The time at which the Central Myanmar Basin became isolated from the proto-Bengal fan and the Assam Basin on its western margin by the uplift of the Indo-Burman Ranges remains unknown. Maurin & Rangin (2009) demonstrated a Late Miocene uplift of the ranges, but the Central Myanmar Basin has been enclosed since at least the Middle Miocene, as it has been the locus of north-south prograding deltaic formations, recorded in the Pegu Group series (Khin & Myitta 1999). Poorly dated Eocene flysch-type sediments extend over the entire length of the inner wedge of the ranges, from the Arakan Yoma (southern edge of the ranges) to the Naga Hills (northern edge of the ranges; see Brunnschweiler 1966;Bannert et al. 2011), and indicate that the Indo-Burman Ranges were not emergent in the Middle Palaeogene. A combined petrographic and isotopic study of flysch sandstone samples from the Arakan Yoma identified a significant component of arc-derived material coming from the Asian active margin, and a persistent youngest zircon fission-track population at 37 Ma in the sandstones (Allen et al. 2008). Therefore, the Indo-Burman Ranges must have experienced a first uplift episode sometime between 37 Ma (Late Middle Eocene) and the Middle Miocene (Mitchell 1993;Allen et al. 2008).
In the Minbu and Chindwin Sub-Basins, the locations of the present study, the Pondaung and Yaw Formations represent a significant proportion of the Palaeogene deposits that are locally up to 2000 m in thickness (United Nations 1978 T ra n s h im a la y a n a rc C M B S h o rt e n in g L h a s a T e r r a n e Indian Plate Q i a n g T a n g T e r r a n e s  Late Middle Eocene continental fauna, and fluvio-deltaic deposits constitute the main outcrops (Jaeger et al. 1999;Beard et al. 2009), characterized by a high density of palaeosols: the 'cherry-red, bright buff and cream-white earths' already observed by Pilgrim & Cotter (1916). Lateral variations in facies are common, and fine-grained floodplain sediments alternate with pond-like deposits, crevasse splays and channel bodies . The magnetostratigraphic study of Benammi et al. (2002) of the upper part of the formation yielded a constant normal polarity throughout 319 m of deposits, indicating that this decametre-thick series would have been deposited in less than 1 Ma (maximum length of Bartonian polar chrons potentially contemporaneous with these deposits; Benammi et al. 2002), owing to a high accumulation rate (>0.3 mm a −1 ). The Yaw Formation is the marine successor of the Pondaung Formation, and has been far less studied until now. In the Chindwin Sub-Basin, the Yaw Formation is especially thick and yielded Late Eocene to Early Oligocene foraminifera (Nagappa 1959).

Sampling sites and analytical methods
During the 2011 and 2012 winter seasons, we explored, described and sampled several Eocene exposure sites in both the Minbu and Chindwin Sub-Basins, paying special attention to the continental Pondaung and marine Yaw Formations. In the Minbu Sub-Basin, 11 sites were explored in the fossiliferous area of the Bahin township (site A, Fig. 1a); In the Chindwin Sub-Basin, most of the visited outcrops are situated in Kalewa township, along the Kalewa-Kalaymnyo road (site B, Fig. 1a), where the Yaw Formation is especially exposed. All studied sites are from the western margin of the Central Myanmar Basin, close to its boundary with the Indo-Burman Ranges (Fig. 1), and have been exhumed and exposed with a slight dip angle by successive Mio-Pliocene inversion episodes (Pivnik et al. 1998). Sites were logged, described and photographed; palaeoflow directions were measured on sandstone trough cross-bedding on 3D outcrops, according to standard field methods (Collinson & Thompson 1989).
Sandstone and mudstone samples of both the Yaw and Pondaung Formations, sampled in a 25 km wide area in Bahin township (site A, Fig. 1a), were selected to represent all of the various observed facies. To determine their contents in quartz, feldspar and lithic grains, five sandstone samples were selected and mounted in a synthetic resin from which thin sections were prepared at the Georessources laboratory (Vandoeuvre les Nancy); 400 points were then counted according to the Gazzi-Dickinson method (Dickinson 1985). We used the classification diagram of Garzanti & Vezzoli (2003) for metamorphic lithic grains, and calculated the Metamorphic Index, which expresses the average degree of metamorphism in rock fragments. Six sandstone samples were selected for heavy mineral identification and counting; dense minerals were concentrated with bromoform using the 63-250 μm fraction; resulting extracts were weighed, mounted in a synthetic resin and prepared as thin sections at the Hydrasa laboratory (Poitiers), whereas residues were crushed for mineralogical identification by X-ray diffraction (XRD) using a Philips Xpert diffractometer. Dense minerals were identified with the help of diffractograms, and 200 minerals were counted in thin section according to the 'ribbon counting' method (Mange & Maurer 1992). Six samples were prepared for clay mineralogy determination. Mineralogical identifications were made by XRD on the <2 μm size fraction, isolated by gravity settling. Oriented mounts were obtained using 10 mg of the clay fraction dried at room temperature; untreated and glycolated subsamples were analysed to identify swelling clay minerals (1 h saturation with ethylene glycol vapour). XRD analyses were performed on a Brucker D8 Advance diffractometer in AD state at the Hydrasa Laboratory (Poitiers).
Nine samples of various grain sizes, chosen to represent both bed-load and suspended-load sediments, were also selected for Nd and Sr isotopic analysis of their bulk silicate fractions. Four additional samples were separated into clay fraction (<2 μm) and sand fraction (>63 μm), to test grain-size dependent effects. Nd and Sr chemical separation and isotopic analyses were performed at the CRPG, according to the standard procedures of the laboratory, adapted from those of Pin et al. (1994). Briefly, after decarbonation in hydrochloric acid, the silicate fractions were dissolved in HF-HNO 3 with a small amount of HClO 4 , and Sr and Nd were separated using Eichrom Sr-spec, TRU-spec and Ln-spec resins. Sr isotopic composition was measured by thermal ionization mass spectrometry, using a Triton Plus instrument operated in static mode. The 87 Sr/ 86 Sr ratios were corrected for mass fractionation assuming 86 Sr/ 88 Sr = 0.1194. During these measurements the 87 Sr/ 86 Sr value for the NBS 987 standard was 0.710263 ± 0.000023 (2σ). Nd isotopic compositions were measured using a Neptune Plus multicollector inductively coupled plasma mass spectrometry system. Nd isotopic ratios were normalized to 146 Nd/ 144 Nd = 0.7219. During the period of sample measurement the JNdi Nd standard gave a mean value of 143 Nd/ 144 Nd = 0.512077 ± 0.000017 (2σ). Nd and Sr procedural blanks were insignificant compared with the amounts of Sr and Nd measured in the samples.

Sedimentological observations and palaeocurrent analysis
Sandstone bodies of the Pondaung Formation are fairly homogeneous and show many common characteristics; logs and detailed information have been previously published by Soe et al. (2002). Bodies are lenticular, wing-shaped, and usually isolated in clayrich deposits, or stacked in thick agglomerations where each body can still be recognized. Their dimensions vary from 15 to 800 m in width, for 1-15 m of thickness (with width/thickness ratios between 8:1 and 100:1). The smaller sandstone bodies are single units whereas larger bodies form multi-storey stacked sandstone successions. Each sand body displays a characteristic inverse-grading, with coarse and erosive basal layers of low-angle planar stratifications, overlain by several tabular sets (20-80 cm thick) of trough cross-stratification, which are the dominant facies of the sandstone bodies. Basal lags of pebbles and cobbles are common. Among the clasts, volcanic rocks, composed of olivine and pyroxene-rich volcanic conglomerates, basalts, trachytes and rhyolites were notably recognized. Palaeoflow measurements on trough cross-bedding show an unequivocal unimodal westward direction before any rotational correction (see Fig. 3a). The array of measured directions extends from the NW to the SSW.
In the Minbu Sub-Basin, the Yaw Formation is poorly exposed. Outcrops in the Bahin area (area A in Fig. 1a) display horizontally laminated mudstones with sparse millimetre-scale silty and wavy laminae, and rare centimetre-thick carbonate layers. These deposits yielded several shark teeth and shellfish fragments; ichnotaxa are rare, but thalassinoides and Planolites have been locally recognized; metrethick tempestite sandy sets have also been observed and yielded shellfish and fossil wood fragments. In the Chindwin Sub-Basin, the Yaw Formation is far better exposed in the Kalewa township (area B in Fig.  1a), and thick outcrops, exposed during excavation for coal mining, could be logged (see Fig. 3b). The Yaw Formation comprises a succession of coarsening-upward, 10-30 m thick sequences. Basal deposits of the type-sequence are composed of grey to black laminated mudstones with a high content of organic matter (facies Fo), interrupted by decimetre-thick, laterally continuous carbonated layers (micritic wackestone, facies Cw). These mudstones are 2-15 m thick and are overlain by a coarsening-upward succession of small 3D ripples (2-3 cm) with organic matter laminae as drapes (facies Sp) and centimetre-thick sets of wavy stratifications (facies Sw). Set surfaces of both facies are extremely rich in plant debris; the sets can contain flasers of pure organic matter, and can be interrupted by several centimetres of laminated, sapropelic coal (Facies Cs). Grains are subangular, silty to fine grained, and sandstones can be micaceous. thalassinoides and teichichnus ichnotaxa have been observed (Cruziana ichnofacies), but ichnotaxa as fossil remains are rare. These layers may be abruptly overlain by stacked sets of trough cross-stratification (60-100 cm thick), with flaser bedding with organic matter and lignite drapes (facies St). This facies is composed of fine to coarse grains, is locally rich in mud clast breccia, and locally yielded Roselia ichnotaxa, freshwater gastropods and fish bones. These sandstone units can be up to 6 m thick, and the edges of the bodies are not well exposed. The sequence is usually capped by a laterally extensive, metre-thick lignite layer containing continental vertebrate fossils (facies Cv). These gradually evolve into Fo facies through a progressive increase in clastic content. Palaeocurrent analysis on trough cross-bedding (facies St) of the coarsening sequences displays similar patterns to the Pondaung Formation, with unimodal, west-directed results, from NW to SW, before any rotational correction (see Fig. 3a).

Petrographic data
QFL and lithic grain plots of the point-counting results are presented in Figure 4. All samples show a significant proportion of lithic detritus and plot in the magmatic arc province on the QFL plot of Dickinson (1985). Volcanic lithic grains are dominant, whereas sedimentary and metamorphic lithic grains are less common, and are found in similar proportions. The Metamorphic Index of Garzanti & Vezzoli (2003) shows very low values, between 100 and 150 (on a scale varying from 100 to 500).
Heavy mineral fractions show a strongly depleted assemblage (see Table 1); in other words they display very low contents of transparent heavy minerals and a scarcity of pyroxene and amphibole, indicating extensive diagenetic dissolution. Opaque phases (rutile, ilmenite, Cr-spinel) represent a significant proportion of the observed grains (from 17 to 45%). Epidote, ilmenite and rutile are the most frequently detected minerals, with minor amounts of Cr-spinel, alkali amphibole (glaucophane), clinopyroxene, tourmaline and anatase. Titanite, pumpellyite, hornblende and chlorite were occasionally detected. We note the lack of zircon in the ultra-stable phase.
XRD analyses of <2 μm fractions of six samples show a prominent contribution of smectite with subordinate amounts of kaolinite and illite. Chlorite has also been detected in the analysed sample of the Yaw Formation, as well as smectite-illite mixed layer clays in one pedogenized sample of the Pondaung Formation.

Nd and Sr isotopic results
Results for both the Yaw and Pondaung Formations are plotted in Figure 5a. Pondaung isotopic results show an unusually wide range, from +0.3 to −7.8 for εNd (with an average of −4.3) and from 0.705 to 0.718 for 87 Sr/ 86 Sr (with an average of 0.7119). Among these results, two distinct families can be distinguished: sediments from the Paukkaung, Nyuangpinle and Pangan localities yielded very high εNd (+0.3 to −3.8) and low to moderate 87 Sr/ 86 Sr ratios (0.705-0.711) and constitute one source family; sediments from the Yarshe, Thaminchauk and Ganle localities yielded substantially lower εNd (−3.9 to −7.8) and higher 87 Sr/ 86 Sr ratios (>0.714), and form a second family. Neither family displays a stratigraphic trend or any geographical zonation (see Fig. 5b and c). The <2 μm and >63 μm fractions of the same samples show significant differences in εNd (up to four) and 87 Sr/ 86 Sr (up to 0.003), but stay within the range of values of the corresponding family.
Yaw isotopic results also display a wide range of values, with εNd from −2.5 to −7.5 (with an average of −5.3) and 87 Sr/ 86 Sr from 0.708 to 0.714 (with an average of 0.7115). Nevertheless, these results form a unique group, which occupies the space between the two Pondaung families on the εNd v. 87 Sr/ 86 Sr diagram. We note that three εNd measurements on Palaeogene flysch of the Indo-Burman Ranges yielded similar values (between −4 and −7.4; see Allen et al. 2008).

Sedimentological interpretation and depositional patterns
Wing-shaped sandstone bodies of the Pondaung Formation display a standard fining-upward sequence dominated by trough crossbeds; their sedimentary profile and their characteristic wing shape clearly evoke channel bodies (Gibling 2006). Soe et al. (2002)  interpreted the Pondaung Formation as representing a fluvio-deltaic environment, where crevasse splay, swale-fill, overbank, fluvial and distributary channel, marsh and prodelta deposits were identified. The coarsening-upward sequences of the Yaw Formation, with their gradual transition from laminated mudstones to cross-laminated ripples with shallow marine ichnotaxa, sandstone bodies and continental fossil-bearing coal, show a continuous shift from shallow marine to continental facies. They are interpreted as deltaic deposits, with successive shifts from the distal prodelta (facies Fo/Sw), to the delta front (facies Sp, Sw and Cs), and finally to the delta plain, with distributary channel deposits (facies St) and swamp coals (facies Cv; see, e.g. Styan & Bustin 1984;Postma 1990). The depositional environments of the Pondaung and Yaw Formations display a gradual shift from onshore fluvial or deltaic deposits to purely deltaic deposits, and reflect an overall transgression, which could be caused either by increased subsidence or by eustatic variations. In the present-day reference frame, mean channel palaeoflow directions of both formations are unimodal and (westward-) directed towards the Indo-Burman Ranges (see Fig. 1). These results show that during the time of interest the Central Myanmar Basin was not topographically enclosed on its western border, as deltaic formations prograded through its western margin, and that the Indo-Burman Ranges were not emerged. This provides a younger time constraint on the early emergence of the Indo-Burman Ranges than previous estimates (post 37 Ma, Allen et al. 2008), and shows that central Myanmar was open to the Indian Ocean until at least Early Oligocene times. Taking into account a 20-30° clockwise rotation for Indochina since the Eocene, and assuming the absence of significant rotation between the Burma block and Indochina, these deltas would have been SW-directed in an Eocene geographical reference frame (Richter & Fuller 1996;Benammi et al. 2002;Morley 2009). The palaeo-shoreline drawn by these deltas would have been oriented NW-SE, corresponding to the general Eocene orientation of Indochina (see Fig. 2a). These observations suggest that regardless of the extent of strike-slip motion of the Burma block along Indochina since the Eocene, Pondaung-Yaw upstream river systems had to flow through the eastern margin of the Burma block.

Provenance significance of petrographic observations
Point-counting analysis of sandstone samples yielded fairly uniform results, with similar QFL percentages and lithic contents. All sandstone samples plot within the magmatic arc provenance on the QFL plot. A sample of Palaeogene flysch from the Indo-Burman Ranges analysed by Allen et al. (2008) plots close to the sandstone samples on the QFL plot, but is more enriched in volcanic lithic material.
Heavy mineral fractions are highly depleted in the more alterable fractions and must be interpreted with caution. The presence of epidote, glaucophane and pumpellyite suggests a low-grade metamorphic source contribution (Mange & Maurer 1992). The high abundance of Cr-spinel may indicate an ultramafic source, which could have been partially removed during later sample dissolution. Concerning the clays, high smectite content in the mudstones suggests a mafic character for the drained area, as smectite commonly results from the weathering of basic substratum (Wilson 2004).
Interestingly, the combination of recycled sediments, low-grade metamorphic rocks and hypothetical ultramafic detritus is common in foreland basins associated with Andean-type margins, and constitutes an 'axial belt source' resulting from the erosion of the local basement . Based on our different petrographic approaches, two main sources can be distinguished, which are consistent with the standard description of common orogenic sediment provinces of : a 'magmatic arc source', which contributed to the numerous volcanic lithic clasts, and an 'axial belt source', contributing metamorphic and sedimentary lithic clasts and low-grade metamorphic and probable ultramafic heavy minerals.

Explaining the dual isotopic signal of the Pondaung Formation
Two distinct isotopic source contributions can be distinguished for the Pondaung Formation. The supply of the Pondaung river system appears to be heterogeneous and shifts from one source to the other, both geographically (Fig. 5b) and stratigraphically (on decametrescale, see Fig. 5c). At least three stratigraphic shifts in sourcing would have occurred in less than 50 ka, according to the estimated sedimentation rate of the Pondaung Formation (Benammi et al. 2002). The high isotopic variability of the deposits in the Pondaung palaeo-floodplain points to a rapidly evolving river system with an unbuffered, heterogeneous load.
Yaw marine sediments are more homogeneous, and possess isotopic ratios that are situated exactly between the two identified source contributions of the Pondaung river system. The similarity of the mean isotopic ratios and the homogeneous petrographic results of the two formations suggest that that Yaw and Pondaung Formations were fed by the same sources. At least two distinct sources were isotopically identified, but the two formations reflect different extents of homogenization of material from these sources. Whereas the Pondaung deposits retain the distinct isotopic signatures of these  Allen et al. (2008) and main provenance provinces, following Dickinson (1985) and Najman (2006). Q, quartz; F, feldspar; L, lithic fragments (Lm, metamorphic; Ls, sedimentary; Lv, volcanic). two provinces and thus reflect inefficient mixing, the Yaw deposits display a relatively well-mixed signal of the two sources. The differences in the isotopic patterns of the two formations can be easily explained by their respective continental and marine characters. The Pondaung and Yaw Formations can be seen respectively as the onshore and offshore parts of the same unbuffered deltaic system, which was supplied by an isotopically heterogeneous drainage basin. Onshore Pondaung deposits, which were located landward, represent the successive ephemeral contributions of distinct areas of the drainage basin, whereas marine or deltaic Yaw deposits, located seaward of the coast, reflect efficient mixing of approximately equal contributions from the two sources, producing a smoothed and fairly homogeneous isotopic signal. As the Yaw Formation overlies the Pondaung deposits without unconformity, the changing degree of isotopic homogenization reflects a transgression-related shift towards marine conditions favouring more efficient mixing. The heterogeneous mixing in the Pondaung deposits is inconsistent with a long-distance, stable, sediment supply from two distal catchment areas, as extensive transport would tend to homogenize the isotopic signal along the river course (e.g. Singh & France-Lanord 2002). These results suggest a fast-evolving river system with local catchments, rather than a long transport distance for the Pondaung-Yaw deposits.

Identification of the provenance areas
In the direct vicinity of the SE Asian margin, five main geological provinces are believed to contribute to the Eocene Burmese supply: the Transhimalayan magmatic arc, the Lhasa terrane and the Tethyan Sedimentary Series in central Tibet, the Burmese magmatic arc and the Burma terrane in Myanmar (see Table 2). In central Tibet, the northern part of the Lhasa terrane is dominated by Palaeozoic to Mesozoic sedimentary rocks intruded by S-type granitic batholiths of the Northern Plutonic Belt (Kapp et al. 2003;Chiu et al. 2009). At the southern margin of the Lhasa terrane, the Transhimalayan arc area is dominated by Andean-type volcanism (the massive Linzizong volcanic rocks), I-type plutonic intrusions (the Gangdese batholith) and ophiolite emplacement at the Indus-Tsangpo suture (Chung et al. 2005;Ji et al. 2009). South of the suture, the Tethyan Sedimentary Series consists of Cambrian to Eocene sedimentary and very low-grade metasedimentary rocks, mainly composed of phyllites, limestones and quartzose sandstones (DeCelles et al. 2001;Najman et al. 2008). Lhasa basement rocks, Transhimalayan volcanic and ultramafic rocks, and Tethyan Sedimentary Series display different εNd and 87 Sr/ 86 Sr signals, which can be easily distinguished (Chu et al. 2006;Clift et al. 2006;Najman 2006;Najman et al. 2008).
In Myanmar, the Burma terrane basement is variously represented. The Gaoligong Belt, Slate Belt and Mogok Metamorphic Belt form a continuous strip of Palaeozoic sediments and low-to high-grade metamorphic rocks with large granitic intrusions, from the eastern Himalayan syntaxis to the Tenasserim highlands (see Fig. 1b; Mitchell et al. 2004;Searle et al. 2007;Xu et al. 2008). The Burma terrane basement also crops out in central Myanmar (see e.g. Mitchell et al. 2007;Shi et al. 2009;Bannert et al. 2011), and in the core of the Indo-Burman Ranges (Maurin & Rangin 2009). The Burmese extension of the Transhimalayan arc is notably represented by the Baunmauk and Salingyi batholiths, and Mokpalin-Sit-Kisin diorites, which have positive εNd and low 87 Sr/ 86 Sr ratios (Zaw 1990;Mitchell et al. 2012). Compilation of isotopic data on other Burma terrane basement rocks shows a range in values that is similar to those of the SE Asian continental blocks, with low εNd (between −3 and −15) and high 87 Sr/ 86 Sr ratios that may exceed values of 0.740 (Zaw 1990;Chen et al. 2002;Mitchell et al. 2007;Xu et al. 2008;Liu et al. 2009;Mitchell et al. 2012).
Other structural units are considered not have made significant contributions to the isotopic signal of the sediment sources, either because their exposure is insignificant in the direct drainage area or because their contribution to the solid load of the drainage system is not important. Further inland, the contribution of the Quiangtang terranes of central Tibet to the sedimentary supply of the Pondaung-Yaw river system seems improbable, owing to their distal character and the presence of an early orographic barrier, formed by a proto-Tibetan Plateau (Wang et al. 2008). A contribution from the Shan-Thai block, linking the southwestern region of   Chen et al. 2002;Chu et al. 2006;Xu et al. 2008;Liu et al. 2009;Mitchell et al. 2012 Burmese arc Volcanic rocks, I-type granitoids 0 to +8 <0. 708 Zaw 1990;Mitchell et al. 2007, 2012 Pondaung and   present-day Yunnan province in China and the Shan-Thai Plateau, is plausible (see Fig. 1a). Nevertheless, Cambrian to Early Cretaceous carbonates are the dominant constituent of the incised rocks of the Shan-Thai block (Chhibber 1934;Mitchell et al. 2007). Because carbonate clasts have not been recorded in the studied series, we argue that the contribution of the Shan-Thai area to the solid load was probably insignificant. The first source contribution recorded in the Pondaung and Yaw sediment (source 1 in Table 2) displays εNd values close to zero and low 87 Sr/ 86 Sr; these results suggest a contribution of Burmese or Transhimalayan arc material and are consistent with the presence of magmatic arc clasts in these samples. Unfortunately, Table 2 shows clearly that the Burmese and Tibetan extensions of the Asian Andean arc are isotopically almost identical; it is therefore hard to distinguish between a Tibetan or a Burmese origin for the first source contribution, based on petrographic and isotopic features alone. The lack of positive values for εNd suggests minor mixing with a secondary source, probably the surrounding terrane basement. The second source contribution (source 2 in Table 2) has the isotopic characteristics of the Lhasa and Burma terrane basements, and is consistent with the metamorphic and sedimentary clasts recorded by sample petrographic data. The evidence for low-grade metamorphism suggests either a contribution from the Burmese basement, where ultramafic and low-grade metamorphic rocks are prominent (Mitchell et al. 2004;Shi et al. 2009;Bannert et al. 2011), or a contribution of the Tethyan Sedimentary Series. The latter source seems improbable or minor, as detritus sourced from the Tethyan Sedimentary Series is characterized by very low εNd values, which are not consistent with our results (White et al. 2002). The presence of low-grade metamorphic material is therefore better explained by an origin in the Burmese basement for the second source, even though we cannot definitively reject a Tibetan provenance, owing to the limited lithological information available for several Tibetan areas such as the eastern Himalayan syntaxis.

Discussion and conclusion: a former Tsangpo-Irrawaddy river connection?
Petrographic and isotopic results suggest a dual source for the Pondaung-Yaw drainage system; namely, the Asian magmatic arc and the Asian basement terranes. Distal (Tibetan) or proximal (Burmese) sources are isotopically and petrographically difficult to distinguish, because the two regions share similar features.
However, palaeoflow measurements indicate an eastern provenance for the Pondaung-Yaw river system. The disorganized stratigraphic shifts in sediment provenance and the isotopic discrepancies between marine and onshore fluvio-deltaic sediments suggest a quickly evolving river system with local catchments rather than a long transport distance. A local provenance is supported by the presence of numerous pebble lags in the Pondaung successions. In addition, the presence of low-grade metamorphic rock fragments is consistent with a local source, because such fragments are found in abundance in the Burma terrane basement, but crop out only rarely in the Tibetan Lhasa terrane. The assemblage of volcanic arc and axial belt detritus is in accord with the local unroofing of an Andean-type cordillera , which extended along the Burmese margin during the Palaeogene and experienced an early episode of uplift (Morley 2004). The Pondaung and Yaw deposits are therefore better explained as resulting from the progressive exhumation and erosion of the local Burmese Andeantype margin, rather than from the distal Tibetan area.
The isotopic and petrographic results for the Pondaung and Yaw sediments, and the orientation of the two former delta systems, support a source in the Burmese margin for the Palaeogene flysch of the Indo-Burman Ranges, as proposed by Allen et al. (2008). Differences in lithic content between the flysch sample and the Pondaung-Yaw samples (see Fig. 4) suggest temporal and/or spatial variations in volcanic supply along the Burmese margin. Further west, in the proto-Bengal fan deposits, the arrival of small amounts of arcderived material occurred sometime between 50 and 38 Ma (Kopili Formation), but seismic evidence of a dominant input from the NW suggests that these deposits are more likely to be from the Transhimalayan rather than the Burman portion of the arc . The Burmese subduction trench might have acted as a trap for offshore Burmese-sourced detritus, preventing the contribution of any Burmese sediment to the proto-Bengal fan. Despite the subordinate Transhimalayan contribution, Bengal Palaeogene sediments display a predominantly Indian cratonic provenance, with a minor component of the Tethyan Sedimentary Series in the Late Eocene-Early Miocene Barail Formation . The paucity of Tibetan input in the proto-Bengal fan and central Myanmar Eocene deposits, although puzzling, is not necessarily suggestive of the absence of Tibetan supply into the Bay of Bengal, as such sediments may have been deposited in the remnant ocean basin trapped between the Indian craton and the Burmese block, and later subducted below the Indo-Burman Ranges (see Fig. 6; Uddin & Lundberg 1998b;Najman et al. 2008). Potential Tibetan contributions in the northern end of the Chindwin Sub-Basin, where we did not sample, cannot be completely excluded, but their contribution to the general sedimentary supply into central Myanmar would have been insignificant.
Could an ephemeral Burmese-Tibetan connection have existed, in the Oligocene, before the Early Miocene divergence of Tsangpo and Irrawaddy fish fauna (Ruber et al. 2004)? Oligocene and Early Miocene series of the Central Myanmar Basin are unfortunately either poorly studied or poorly exposed (Bender 1983). Sedimentation rates in central Myanmar reached their lowest level in the Oligocene (Metivier et al. 1999); this suggests that if a connection existed, it would not have supplied significant quantities of sediments.
Our integrated study of the Pondaung and Yaw Formations showed that during Late Middle Eocene to Early Oligocene times, central Myanmar was open to the Indian Ocean, and was the locus of SW-directed deltas, most probably sourced in the proximal highlands along the SE Asian shoreline. The Pondaung and Yaw deltaic systems are interpreted as supplied by the unroofing of an Andeantype cordillera that extended along the Burma terrane, representing the eastward continuation of the Transhimalayan arc. A potential Tibetan contribution to the sedimentary supply of central Myanmar appears unlikely. A connection between the Tibetan Plateau and central Myanmar before the Neogene shifts in SE Asian river networks, as suggested by former studies, seems not to have been recorded in the Pondaung-Yaw system.