The Sundaland Paleo-River System: Reconstructing the Submerged Drainage Networks of the Last Deglaciation

A research by Dhani Irwanto, 13 October 2025

Abstract

This study reconstructs the paleo-drainage systems of the Sunda Shelf—now largely submerged beneath the Java, Karimata, and South China Seas—using high-resolution bathymetric and topographic datasets. Integration of GEBCO 2025 (15 arc-second), SRTM v3 (1 arc-second), and a deglacial inundation model reveals six major paleo-river systems and a large paleo-lake in the Gulf of Thailand. Watershed modeling was performed under consistent geomorphic thresholds (minimum watershed ≥ 1 000 km²; river length ≥ 10 km) after correcting for ship-passage artifacts in GEBCO data. The resulting networks portray an interconnected fluvial landscape that once linked the emergent landmasses of Sumatra, Java, Borneo, and the Malay Peninsula. These reconstructions illuminate the paleohydrological architecture that structured ecological corridors, sediment transport, and early human movement across Late Pleistocene Sundaland.

Keywords: Sundaland, paleo-river, deglaciation, GEBCO 2025, watershed modeling, Gulf of Thailand paleo-lake, Molengraaff River, Pleistocene hydrology

1. Introduction

During the Last Glacial Maximum (LGM) and the ensuing deglaciation, the Sunda Shelf constituted one of the world’s largest emergent plains, uniting the islands of Sumatra, Java, Borneo, and the Malay Peninsula. Rising sea level of more than 120 m since ~21 ka BP progressively drowned this continental platform, fragmenting it into the present Indonesian and Malaysian archipelagos. Reconstructing its paleo-river systems is essential for understanding patterns of freshwater and sediment routing, ecological and biogeographical connectivity prior to isolation, and the response of tropical fluvial systems to rapid post-glacial transgression. Earlier works (Molengraaff 1921; Voris 2000; Sathiamurthy & Voris 2006) outlined generalized drainage maps of Sundaland, but relied on coarse bathymetric data. With recent improvements in digital elevation models, it is now possible to delineate channels and basins at continental scale with greater realism. This paper extends previous reconstructions of relative sea-level change (Irwanto 2025a), sea-surface temperature evolution (Irwanto 2025b), and deglacial inundation rates (Irwanto 2025c) by mapping the paleo-hydrological network that organized the former Sunda landmass.

2. Data and Methods

2.1 Data Sources

To achieve a realistic reconstruction of the paleo-drainage framework across Sundaland, this study integrates the highest-resolution publicly available global terrain datasets. The bathymetric and topographic data were selected for their complementary spatial coverage—underwater and terrestrial—and for compatibility within a uniform geodetic framework. The sources are summarized below.

1. GEBCO 2025 Bathymetry (15-arc-second grid), representing global ocean depth data compiled from multibeam surveys and satellite altimetry.
2. SRTM Version 3 (1-arc-second grid), providing high-accuracy land elevations derived from radar interferometry.
3. Sundaland Deglacial Inundation Dataset (Irwanto 2025), previously produced from sea-level modeling, supplying shoreline reference surfaces for paleo-hydrological interpretation.

All datasets were resampled and mosaicked into a continuous elevation model referenced to WGS 84 geographic coordinates to ensure consistent vertical and horizontal alignment.

2.2 Methodology

The reconstruction of the Sundaland paleo-river systems followed a sequence of geomorphometric and hydrological analyses within a GIS environment. A composite digital elevation model (DEM) was produced by mosaicking and resampling the GEBCO 2025 bathymetry (15 arc-second) and SRTM v3 topography (1 arc-second) into a uniform WGS 84 geographic grid. The surface was hydro-flattened to remove discontinuities along modern coastlines and ensure consistent flow routing across subaerial–submarine interfaces.

2.2.1 Artifact Minimization

Bathymetric trench-like anomalies—known as ship-passage artifacts—were visually identified as linear depressions aligned with survey tracks in the GEBCO grid. These were locally corrected through neighborhood median filtering and manual editing of aberrant grid nodes using bilinear interpolation. The objective was to suppress artificial cross-flow pathways that could distort hydrological connectivity while retaining genuine topographic variability. Complete removal of such artifacts was not feasible without the original multibeam soundings; however, their hydrological influence was minimized to a negligible level.

2.2.2 Hydrological Modeling

Watershed delineation and stream extraction were performed using standard flow-accumulation algorithms in the GIS software. Flow direction was derived from the corrected DEM using the D8 algorithm[1], followed by computation of flow accumulation and stream order. A minimum contributing area threshold of 1 000 km² was imposed for first-order streams, and a minimum channel length of 10 km was adopted to exclude spurious or ephemeral drainages. The resulting stream networks were then vectorized and topologically validated to ensure connectivity and realistic drainage hierarchy.

2.2.3 Integration with Modern Drainage

Modeled paleo-channels were spatially aligned with the outlets of modern rivers to maintain genetic continuity between subaerial and submarine catchments. The procedure involved adjusting terminal flow paths toward existing estuaries and delta fronts, based on hydrological gradients and sediment-transport direction inferred from slope and curvature analyses. This ensured that modeled paleo-drainage systems remained compatible with present river mouths and physiographic boundaries.

2.2.4 Hydrological Synthesis and Visualization

The final drainage mosaics were categorized by basin identity and exported as vector shapefiles for cartographic visualization. Six regional-scale systems were defined through hierarchical clustering of flow accumulation zones, corresponding to the Java Sea, Eastern Java Sea, Karimata Strait, Gulf of Thailand, Mekong Extension, and Strait of Malaka systems. The outputs were compared against regional bathymetric contours and deglacial shoreline reconstructions (Irwanto 2025a; 2025c) to validate drainage coherence under the −122 m sea-level surface (~22.5 ka BP), corresponding to the Last Glacial Maximum (LGM).

2.3 Limitations

The reconstructed paleo-river systems represent a first-order geomorphometric model constrained primarily by topography and bathymetry. Several geomorphic and dynamic processes were not explicitly incorporated because of the scarcity and inconsistency of regional datasets. Consequently, the results should be interpreted as generalized hydrological frameworks rather than exact paleochannel geometries.

1. Sedimentary dynamics — Processes such as delta progradation, channel avulsion, floodplain aggradation, and littoral drift were not modeled. These factors can substantially modify valley morphology and estuarine geometry through time, especially during late-stage transgression.
2. Subsurface and tectonic influences — Localized tectonic subsidence, fault reactivation, and differential uplift may have altered drainage gradients and basin shapes after initial channel formation. These effects remain unquantified at the shelf scale.
3. Karst and dissolutional terrain — In regions underlain by carbonate lithologies (e.g., northern Java, western Borneo, and the Thai–Malay margin), subsurface drainage and sinkhole development may have influenced catchment connectivity in ways not captured by surface-based flow models.
4. Sediment compaction and isostatic adjustment — Post-depositional subsidence and isostatic rebound following deglaciation were not integrated into the DEM corrections, introducing minor uncertainty in absolute elevation and relative base level.
5. Bathymetric data quality — Despite artifact minimization, residual ship-passage anomalies in the GEBCO grid may still influence local flow routing. These artifacts—linear trench-like depressions created during data gridding—were reduced but cannot be wholly eliminated without access to original sounding lines. Their impact is minimized at the regional scale but may persist locally.
6. Temporal simplification — The modeling assumes a quasi-static topography corresponding to a single reference sea-level surface (−122 m RSL, ~22.5 ka BP). Progressive shoreline migration, sediment redistribution, and hydrological reorganization through subsequent millennia are therefore beyond the scope of this reconstruction.

Overall, the uncertainties above are unlikely to alter the broad configuration of the six major drainage systems identified in this study, but they may affect local channel positions and tributary details. Future work integrating seismic stratigraphy, sediment cores, and higher-resolution bathymetry could further refine the paleo-hydrological realism of the Sundaland reconstruction.

3. Results

3.1 Major Paleo-River Systems

The hydrological modeling reveals a coherent network of six principal drainage systems that occupied the Sunda Shelf before Holocene flooding. Each system integrates numerous tributaries draining from the emergent landmasses of Sumatra, Java, Borneo, and the Malay Peninsula. Their relative magnitudes and contributing regions are summarized in Table 1, while their spatial configuration is illustrated in Figure 1.

Figure 1. Reconstructed paleo-river systems and major drainage basins across the Sunda Shelf. Pale-blue shading indicates the extent of deglacial inundation; darker blue marks the paleo-lake in the Gulf of Thailand. Black lines show modeled paleo-rivers, gray lines depict modern rivers. © 2025 Dhani Irwanto.

Table 1. Major Paleo-River Systems of Sundaland

System	Principal Source Regions	Approx. Watershed Area (km²)
Java Sea	Southern Borneo, Northern Java, Southern Sumatra	≈ 570 000
Eastern Java Sea	Southern Borneo (Barito, Kapuas-Murung, Kahayan)	≈ 180 000
Karimata Strait (Molengraaff)	Eastern Sumatra, Western Borneo	≈ 630 000
Gulf of Thailand	Eastern Malay Peninsula, Chao Phraya Basin	≈ 1 020 000
Mekong Extension	Lower Mekong and adjacent South China Sea margin	≈ 690 000
Strait of Malaka	Eastern Sumatra, Western Malay Peninsula	≈ 260 000

Table 2. Connectivity between Modern Rivers and Paleo-River Systems

Paleo-River System	Modern River(s) — Borneo	Modern River(s) — Sumatra	Modern River(s) — Java	Modern River(s) — Malay Peninsula/ Mainland
Java Sea	Mendawai, Sampit, Pembuang	Tulang Bawang	Bengawan Solo, Serang, Cimanuk, Citarum	–
Eastern Java Sea	Barito, Kapuas-Murung, Kahayan	–	–	–
Karimata Strait (Molengraaff)	Kapuas	Musi, Batanghari, Indragiri	–	–
Gulf of Thailand	–	–	–	Johor, Rompin, Endau, Kuantan, Kelantan, Tapi, Mae Klong, Chao Phraya, Bang Pakong
Mekong Extension	–	–	–	Mekong main stem and tributaries
Strait of Malaka	–	Kampar, Rokan, Barumun, Belawan	–	Malacca, Perak

Note: Table 2 illustrates the continuity between present river mouths and modeled paleo-channels, supporting the inference that many modern estuaries originated as terminal segments of these ancient systems.

3.2 Paleo-Lake in the Gulf of Thailand

A closed depression of approximately 93,000 km² with an outlet sill near −55 m relative sea level indicates the existence of a vast paleo-lake in the central Gulf of Thailand. The basin morphology suggests prolonged freshwater retention during the early deglacial stages before eventual overtopping and breaching into the South China Sea. This lacustrine phase is congruent with transitional sedimentary records documented in regional core studies (Horton et al., 2005; Chabangborn et al., 2020; Zhang et al., 2022)[2] that show shifts from freshwater-dominated facies toward more estuarine or marine-influenced depositional environments during rising sea levels.

For example, sediment cores along the western Gulf of Thailand (e.g., CP3, CP4, CP5) record stratigraphic transitions consistent with increasing marine influence around 7.9 ka BP, as evidenced by grain-size ratios, microfossil assemblages, and mangrove pollen influx (Chabangborn et al., 2020; Horton et al., 2005). These findings provide independent support for a broad freshwater-to-estuarine transformation compatible with the modeled paleo-lake hydrology of this study.

3.3 Effect of Artifact Reduction

Accurate delineation of flow paths across the shallow continental shelf requires correction of artificial trench-like depressions generated by gridding along ship-track data. After iterative smoothing of the GEBCO 2025 bathymetry, several improvements were achieved in the modeled drainage topology. These corrections produced morphologically consistent valley alignments and more realistic connectivity among adjacent basins. The most notable adjustments are outlined below:

1. The Java Sea system expanded westward, integrating southern-Sumatran tributaries previously misrouted toward the Sunda Strait.
2. A distinct Eastern Java Sea system emerged, isolating the Barito, Kapuas-Murung, and Kahayan catchments from the Java Sea basin.
3. Linear transverse channels formerly produced by ship-track artifacts were removed, restoring natural curvilinear drainage.

4. Discussion

4.1 Paleogeographic Significance

The reconstructed networks demonstrate that the Sunda Shelf once functioned as a contiguous fluvial plain. Drainage convergence zones in the Java Sea, Karimata Strait, and Gulf of Thailand align with present depocenters identified in seismic surveys and sediment-core analyses (Horton et al., 2005; Chabangborn et al., 2020). These relationships clarify the shelf’s role as both a sediment sink and a corridor for freshwater discharge during deglaciation, providing the physical context for the rapid transgression and shoreline fragmentation patterns described in Irwanto (2025c). The Gulf of Thailand paleo-lake and its subsequent marine transgression exemplify this dynamic transition from terrestrial to marine environments, illustrating the sedimentary continuity between the ancient fluvial systems and the modern shelf basins.

4.2 Biogeographic Evidence

The modern distribution of freshwater and estuarine taxa implies historical continuity through these ancient waterways. The river threadfin (Polydactylus macrophthalmus), today restricted to the Kapuas (Borneo) and Musi–Batanghari (Sumatra) rivers, exemplifies vicariant separation of populations once joined by the Molengraaff River system (Motomura et al., 2001). Comparable disjunctions among mangrove species and aquatic mollusks reinforce the paleohydrological connections inferred from this model.

4.3 Comparison with Previous Models

Relative to Voris (2000) and Sathiamurthy & Voris (2006), the present reconstruction offers higher spatial fidelity and improved hydrological realism. GEBCO 2025’s finer resolution delineates meanders and tributary curvature previously unresolved, while artifact correction enhances drainage continuity. The resulting systems exhibit asymmetric basins and multi-branch deltas more consistent with tropical alluvial morphodynamics.

4.4 Hydrological Implications

The extensive low-gradient plains inferred from the model suggest slowly meandering rivers traversing broad floodplains, capable of sustaining vast wetlands and delta complexes. These channels likely transported large sediment loads toward the shelf edge, influencing near-shore nutrient dynamics and the eventual formation of submerged ridge sequences visible in present bathymetry.

4.5 Broader Implications for Human and Biotic History

During lowered sea levels, the integrated river corridors of Sundaland provided continuous freshwater, fertile soils, and navigable routes across the emergent shelf. Such corridors would have facilitated the dispersal of human groups, enabling occupation of interior basins and coastal margins long before marine transgression. The interconnected fluvial plains may have served as arteries for cultural and genetic exchange across what is now Island Southeast Asia.

Large alluvial tracts along the paleo-Kapuas–Musi–Batanghari (Molengraaff) and Gulf of Thailand systems possessed the ecological capacity to sustain proto-agricultural communities. These environments echo the environmental settings of later riverine civilizations elsewhere, suggesting that the Sunda Shelf offered similar opportunities for early food-producing and settlement behaviors, as discussed in the Riverine Civilizations section of Irwanto (2015).

Following progressive inundation, former trunk rivers evolved into coastal estuaries and deltaic plains, maintaining their roles as communication axes. The transformation from fluvial to estuarine transport networks likely fostered the emergence of hydraulic and navigational knowledge, promoting the transition from inland cultivation to maritime resource exploitation.

As shelf flooding severed continental routes, human communities adapted to rising waters by shifting toward littoral livelihoods. Former river valleys became sheltered bays and straits—natural conduits for early seafaring. This environmental forcing may have seeded the maritime orientation that later characterized Austronesian and other early Southeast Asian cultures.

Submergence of the Sunda Shelf fragmented once-continuous habitats, isolating freshwater and terrestrial species. The split distribution of Polydactylus macrophthalmus across Sumatra and Borneo (Motomura et al., 2001) typifies post-inundation vicariance. Similar processes likely affected elephants in Kalimantan (Fernando et al., 2003; Sharma et al., 2018), freshwater turtles, and riverine vegetation, producing the biogeographical mosaics evident today.

4.6 Empirical Evidence for Lacustrine-to-Estuarine Transition in the Gulf of Thailand

Multiple sediment-core studies from the Gulf of Thailand and adjacent coastal plains substantiate the interpretation of a paleo-lacustrine stage followed by progressive marine influence during the early to mid-Holocene. Cores from the western Gulf of Thailand (CP3, CP4, CP5; Chabangborn et al., 2020; Jiwarungrueangkul et al., 2022) reveal stratigraphic successions where fine-grained lacustrine and deltaic units are overlain by brackish to marine estuarine facies, accompanied by increases in mangrove pollen, foraminiferal abundance, and marine microfossils.

Similarly, Horton et al. (2005) documented analogous palaeoenvironmental transitions in coastal cores from the Malay–Thai Peninsula, with a clear evolution from freshwater swamp and fluvial deposits to tidal-flat and estuarine sediments synchronous with the mid-Holocene sea-level rise. Regional syntheses of shelf sedimentation (Zhang et al., 2022) further demonstrate that the Sunda Shelf experienced a widespread hydrological reorganization, wherein formerly subaerial basins became drowned estuaries and shallow marine embayments as sea levels rose rapidly between ca. 10 and 7 ka BP.

Collectively, these datasets reinforce the interpretation that the Gulf of Thailand depression functioned initially as a large freshwater basin and subsequently transitioned to a semi-enclosed marine embayment—a sequence consistent with the modeled topography and hydrological reconstruction presented in this study.

5. Conclusion

The integrated analysis identifies six major paleo-river systems and a large Gulf of Thailand paleo-lake that together shaped the hydrological framework of emergent Sundaland. By combining GEBCO 2025 bathymetry, SRTM v3 topography, and hydrological modeling, this study refines previous reconstructions and establishes a physically consistent depiction of the shelf’s drainage architecture. Beyond geomorphology, the findings elucidate how these fluvial networks structured ecological corridors and human pathways before Holocene transgression, laying groundwork for future interdisciplinary research.

References

Molengraaff, G.A.F. (1921). Modern Deep-Sea Research in the East Indian Archipelago.

Voris, H.K. (2000). Maps of Pleistocene Sea Levels in Southeast Asia. Journal of Biogeography, 27, 1153–1167.

Sathiamurthy, E., & Voris, H.K. (2006). Maps of Holocene Transgression and Pleistocene Coastlines, Southeast Asia.

Chabangborn, A., Phantuwongraj, S., Sinsakul, S., Choowong, M., & Nakagawa, T. (2020). Environmental changes on the west coast of the Gulf of Thailand during the Holocene. Quaternary International, 555, 3–16. https://doi.org/10.1016/j.quaint.2019.12.020

Horton, B. P., et al. (2005). Holocene sea levels and palaeoenvironments, Malay–Thai Peninsula. The Holocene, 15(8), 1189–1203. https://doi.org/10.1191/0959683605hl887rp

Zhang, H., Liu, S., Wu, K., Cao, P., Pan, H-J., Wang, H., … Shi, X. (2022). Evolution of sedimentary environment in the Gulf of Thailand since the last deglaciation. Quaternary International, 629, 36–43. https://doi.org/10.1016/j.quaint.2021.02.018

Jiwarungrueangkul, T., Jirapinyakul, A., Sompongchaiyakul, P., & Rattanakom, R. (2022). Response of sediment grain size to sea-level rise during the middle Holocene on the west coast of the Gulf of Thailand. Arabian Journal of Geosciences, 15, 167. https://doi.org/10.1007/s12517-022-09450-3

Motomura, H., et al. (2001). Redescription of a rare threadfin (Perciformes: Polynemidae), Polydactylus macrophthalmus (Bleeker, 1858), with designation of a lectotype and notes on distributional implications. Ichthyological Research, 48, 289–294.

Fernando, P., Vidya, T.N.C., Payne, J., Stuewe, M., Davison, G., Alfred, R.J., Andau, P., Bosi, E., Kilbourn, A., & Melnick, D.J. (2003). DNA analysis indicates that Asian elephants are native to Borneo and are therefore a high priority for conservation. PLoS Biology, 1(1), 110–115.

Sharma, R., Goossens, B., Heller, R., Rasteiro, R., Othman, N., Bruford, M.W., & Chikhi, L. (2018). Genetic analyses favour an ancient and natural origin of elephants on Borneo. Scientific Reports, 8, 880.

World Wildlife Fund (WWF). (n.d.). Borneo Pygmy Elephant. Retrieved from http://www.worldwildlife.org/species/borneo-pygmy-elephant

Irwanto, D. (2015). Sundaland: Tracing the Cradle of Civilizations. Sections “Riverine Civilizations” and “Kalimantan Elephants.”

Irwanto, D. (2025a). A Refined Relative Sea-Level Curve for Sundaland.

Irwanto, D. (2025b). Holocene and Deglacial Sea-Surface Temperatures in Sundaland.

Irwanto, D. (2025c). Deglacial Rapid Inundation and Land-Loss Rates of Sundaland.

Footnotes

[1] The D8 (Deterministic Eight-node) algorithm is a standard flow-direction model in hydrological GIS analysis. It assigns each raster cell a single downslope direction toward one of its eight neighboring cells—north, northeast, east, southeast, south, southwest, west, or northwest—based on the steepest descent gradient. This approach enables efficient computation of flow accumulation and watershed delineation across large terrain datasets.

[2] Direct core evidence for a continuous freshwater “lake” spanning the entire Gulf basin is limited; thus, the paleo-lake interpretation should be regarded as a working geomorphological hypothesis derived from modeled topography and hydrological potential, constrained by regional sedimentary analogs.