A research by Dhani Irwanto, 29 September 2025

Abstract

Relative sea-level (RSL) reconstructions provide critical baselines for understanding shelf flooding, paleo-river evolution, and human dispersal across Southeast Asia. Sundaland, a vast continental shelf now submerged, lacks dense offshore sea-level indicators compared to other regions. Here we compile and standardize RSL data from Singapore (Chua, 2021), the SEAMIS database (Mann et al., 2019), and Sunda-specific sites reported in Lambeck et al. (2014). These datasets were compared against the Lambeck et al. (2014) global mean sea-level (GMSL) reconstruction. Residual analysis shows Sundaland RSL generally tracked the global mean within ±5 m over the last 20 ka, with the largest departure (~−4 m) occurring around 11.6 ka. Anchored residual modelling, constrained to 0 m at present, was applied to correct for minor regional bias. The quadratic anchored model provided the best fit (RMSE = 1.82 m), ensuring physical consistency and modestly improving agreement with observed indicators. The final Sundaland curve, extended to 22.5 ka by splicing to the global mean, offers a robust working dataset for paleogeographic and paleoenvironmental reconstructions across Southeast Asia.

Keywords: Sundaland; relative sea level; Holocene transgression; Southeast Asia; sea-level curve; Lambeck et al. (2014); SEAMIS; Chua (2021)

1. Introduction

Southeast Asia hosts one of the most extensive continental shelves on Earth: Sundaland, a landmass that was periodically exposed during the Late Pleistocene and early Holocene. During glacial maxima, lowered sea levels exposed vast tracts of land that today lie beneath the Java Sea, Makassar Strait, South China Sea, Gulf of Thailand, and Malacca Strait. These transient landscapes played a crucial role in shaping the region’s ecosystems, hydrology, and pathways for human and faunal dispersal. Understanding the timing and magnitude of relative sea-level (RSL) changes in Sundaland is therefore central to paleogeographic reconstruction.

Reconstructing sea level over the last 22,500 years has been the focus of numerous global and regional studies. Early syntheses by Fleming et al. (1998, 2000) provided foundational reconstructions of global mean sea level (GMSL) through the Last Glacial Maximum (LGM) and Holocene. Milne et al. (2005) further refined these curves using glacial isostatic adjustment (GIA) models to account for Earth’s viscoelastic response to deglaciation. More recently, Lambeck et al. (2014) compiled a global database of RSL indicators and produced ensemble-model means that remain a benchmark for global GMSL reconstructions. At the regional scale, Mann et al. (2019) developed the SEAMIS database, collating thousands of Southeast Asian sea-level indicators, while Chua et al. (2021) provided a high-resolution Holocene RSL record from Singapore.

Despite this progress, Sundaland remains data-poor compared to other continental shelf regions. Many RSL indicators in Southeast Asia derive from estuarine, lacustrine, or mangrove settings that are highly local in character and sensitive to non-eustatic processes. Offshore indicators suitable for robust eustatic interpretation are sparse. Consequently, researchers often rely on global GMSL curves when reconstructing Sundaland paleogeography, though regional adjustments are desirable to account for local geoid and isostatic effects.

In this study, we compile offshore RSL data from three primary sources: the Singapore record of Chua (2021), the SEAMIS database of Mann et al. (2019) filtered for offshore indicators, and Sunda-specific sites reported by Lambeck et al. (2014). We evaluate these against the global mean of Lambeck et al. (2014), calculate residuals, and test adjustment functions constrained to present sea level. The aim is to produce a regionally adjusted Sundaland sea-level curve spanning 0–22.5 ka BP, suitable as a working reference for paleogeographic and paleoenvironmental modeling in Southeast Asia.

2. Methods

2.1 Data Sources

We compiled relative sea-level (RSL) data for Sundaland from three principal sources. First, the high-resolution Holocene sea-level reconstruction from Singapore reported by Chua et al. (2021) provided a standardized dataset of mangrove peat and coral-based indicators spanning the last ~8 ka. Second, the SEAMIS database (Mann et al., 2019) was filtered to include only offshore indicators within the Sundaland boundary box, excluding lacustrine, estuarine, and mangrove-derived data prone to significant local effects. Third, a subset of offshore Sunda sites reported by Lambeck et al. (2014) were extracted from their global compilation. Together, these sources supplied a coherent set of offshore RSL observations for the region.

2.2 Data Preprocessing

All datasets were standardized to a common reference frame (present mean sea level = 0 m). Indicators flagged as problematic or ambiguous in their source publications were excluded. Further, we manually inspected spatial distributions to eliminate points located on land, rivers, lakes, or swamps. Two extreme outliers that deviated significantly from regional and global trends (Strait of Malacca at 20.176 ka BP, −57.9 m; and 12.348 ka BP, −34.2 m) were removed.

The cleaned Sundaland dataset was then sorted by reference citation and age, and restricted to ages younger than 22.5 ka BP to match the interval of interest. For comparison with global curves, we used the refined global mean sea-level (GMSL) dataset of Lambeck et al. (2014), based on their ensemble means and smoothed at 1 ka resolution.

2.3 Binning and Mean Curves

To address the uneven temporal distribution of data, Sundaland RSL indicators were grouped into 1 ka bins. For each bin, we calculated the mean, spread (range), and number of contributing points. This produced a smoothed Sundaland mean curve covering 19.25–0 ka BP. The global Lambeck et al. (2014) dataset was likewise smoothed in 1 ka bins over 22.5–0 ka BP for direct comparison.

2.4 Residual Analysis and Model Fitting

Residuals were computed as the difference between Sundaland binned means and the Lambeck global means at corresponding ages. Two adjustment functions were tested:

1. Linear anchored model:

r(a) = b₁ ​· a

2. Quadratic anchored model:

r(a) = c1 ​∙ a + c2 ​∙ a²

where r(a) is the residual (m) at age a (ka BP), and b₁, c₁, c₂​ are fitted coefficients.

To ensure physical consistency, both models were constrained such that:

r(0) = 0

This guarantees that the adjusted curve matches present-day sea level. Model performance was evaluated using root mean square error (RMSE).

2.5 Curve Extension to 22.5 ka

Because the Sundaland dataset contains no indicators older than 19.25 ka BP, the adjusted curve was extended to 22.5 ka BP using a conservative splice. The Lambeck global mean was combined with a constant offset equal to the residual at 19.25 ka.

Adj(a) = GMSL_Lambeck​(a) + r(19.25),  19.25 < a ≤ 22.5

This procedure ensured a seamless join while avoiding instability from extrapolating the quadratic model beyond the fitted range.

2.6 Outputs

The final outputs include (i) a regionally adjusted Sundaland RSL curve for 0–22.5 ka BP, (ii) benchmark tables comparing Sundaland and Lambeck global means at key ages (20, 15, 11.6, 8, 6 ka), and (iii) shapefiles and CSVs suitable for direct integration into paleogeographic and GIS analyses.

3. Results

3.1 Data Coverage

The combined offshore Sundaland dataset, derived from Chua (2021), SEAMIS (Mann et al., 2019), and Lambeck et al. (2014), yielded 132 offshore RSL points between 19.25 and 0 ka BP after filtering and outlier removal. No valid offshore data older than 19.25 ka were identified. In contrast, the Lambeck et al. (2014) global database provides continuous coverage from 22.5 ka to present. This discrepancy necessitated splicing of the Sundaland curve to the global mean beyond 19.25 ka.

3.2 Regional vs. Global Comparisons

Figures 1 and 2 illustrates the distribution of Sundaland offshore RSL indicators by source, plotted alongside the Lambeck et al. (2014) global mean sea-level (GMSL) curve. The Sundaland observations generally track the global mean within ±5 m, though with greater scatter in the late Pleistocene bins.

Figure 1. Offshore relative sea-level (RSL) indicators from Sundaland, compiled from Chua (2021), SEAMIS (Mann et al., 2019), and Lambeck et al. (2014). Data points are plotted by reference source, alongside the Lambeck et al. (2014) global mean sea-level (GMSL) curve for comparison. The Sundaland dataset spans 19.25–0 ka BP; no offshore indicators older than 19.25 ka were identified.

Figure 2. Offshore relative sea-level (RSL) indicator points compiled for Sundaland, plotted against the modern shoreline base. Colors indicate data source: Chua (2021), SEAMIS (Mann et al., 2019), and Lambeck et al. (2014). Points represent offshore-only records retained after filtering, with outliers and inland indicators excluded. The map highlights the spatial distribution of sea-level observations across the Sunda Shelf, Makassar Strait, South China Sea, and Malacca Strait.

Benchmark comparisons (Table 1) demonstrate the close correspondence between Sundaland and the global mean. At 20 ka BP, Sundaland RSL averaged −113.4 m compared to −118.1 m in Lambeck’s GMSL (residual +4.6 m). At 15 ka BP, the difference narrowed to +1.5 m. The largest departure occurred at 11.6 ka BP, when Sundaland RSL was ~3.9 m lower than the global mean. By 8 ka BP and younger, residuals converged to within ±2 m.

Table 1. Benchmark comparisons between Sundaland mean RSL and Lambeck (2014) global GMSL.

Age (ka BP)	Sundaland RSL (m)	Lambeck GMSL (m)	Residual (m)
20.0	−113.4	−118.1	+4.6
15.0	−92.6	−94.0	+1.5
11.6	−54.6	−50.7	−3.9
8.0	−5.8	−8.1	+2.3
6.0	+0.7	+0.2	+0.5

3.3 Residual Modeling

Residuals between Sundaland and Lambeck means were modeled using anchored linear and quadratic functions constrained to 0 m at present (Figure 3). Both models reproduced the observed pattern, but the quadratic anchored fit achieved a slightly lower root mean square error (RMSE = 1.82 m) than the linear anchored fit (RMSE = 1.95 m). The quadratic model was therefore selected as the best-fit adjustment function.

Figure 3. Residuals between Sundaland offshore mean RSL (1 ka bins) and Lambeck et al. (2014) global mean sea level. Anchored linear and quadratic fits are shown, both constrained to pass through 0 m residual at present. The quadratic anchored model achieved a slightly lower RMSE and was selected as the preferred adjustment function.

3.4 Final Adjusted Sundaland Curve

The resulting anchored quadratic adjusted Sundaland curve is shown in Figure 4. The curve closely follows the Lambeck global mean, with a slight downward offset (~4 m) during the terminal Pleistocene–early Holocene transition (~12–11 ka BP). From ~8 ka BP onward, the Sundaland curve is virtually indistinguishable from the global mean, with present-day sea level set to 0 m by design.

Figure 4. Final adjusted Sundaland mean sea-level curve (anchored quadratic model, 0–19.25 ka BP), plotted against the Lambeck et al. (2014) global mean. The Sundaland curve shows a modest offset (~4 m lower) around 11.6 ka BP, but otherwise tracks the global mean within ±2 m during the Holocene.

3.5 Extension to 22.5 ka

Because no Sundaland offshore indicators are available prior to 19.25 ka, the curve was extended to 22.5 ka by applying a constant residual offset equal to the fitted residual at 19.25 ka. This approach provided a seamless join to the Lambeck global mean while avoiding instability from extrapolation of the quadratic function. The final curve therefore spans the full deglacial interval (22.5–0 ka BP) with regional fidelity (Figure 5).

Figure 5. Extended Sundaland adjusted sea-level curve (0–22.5 ka BP). Beyond 19.25 ka, the curve was spliced to the Lambeck et al. (2014) global mean using a constant residual equal to the fitted offset at 19.25 ka, ensuring a seamless join without extrapolation.

4. Discussion

4.1 Agreement between Sundaland and Global Mean

The Sundaland offshore dataset demonstrates close agreement with the global mean sea-level (GMSL) curve of Lambeck et al. (2014). Across the deglacial interval (20–0 ka BP), residuals rarely exceed ±5 m, with most falling within ±2 m during the Holocene. The modest offset of ~−4 m around 11.6 ka BP suggests that regional processes, including glacio-isostatic adjustment (GIA) and sediment loading, may have exerted localized influence. Nevertheless, the small magnitude of the deviations indicates that Sundaland sea level largely followed the global trajectory of deglaciation.

4.2 Value of Regional Adjustment

Although the residuals are minor, the anchored quadratic adjustment provides a useful refinement. By explicitly constraining the residual function to 0 m at present, the model corrects for late-Holocene drift observed in unconstrained fits and ensures physical consistency with present-day sea level. The quadratic fit modestly outperformed the linear model in reproducing observed residuals (RMSE 1.82 m vs. 1.95 m), supporting its adoption as the working adjustment function. The resulting curve thus reflects both global deglacial trends and the limited but consistent Sundaland-specific signal.

4.3 Implications for Sundaland Paleogeography

A robust RSL curve is essential for reconstructing the paleogeography of Sundaland. The final adjusted curve indicates that at the Last Glacial Maximum (~20 ka BP), relative sea level was approximately −113 m, rising rapidly during the terminal Pleistocene and early Holocene. By ~8 ka BP, sea level was within ~−6 m of present, reaching near-modern levels by ~6 ka BP. These thresholds correspond to the timing of progressive submergence of shelf plains and reorganization of drainage networks across the Java Sea, South China Sea, and Gulf of Thailand. The refined curve provides a more reliable baseline for GIS-based reconstructions of shelf flooding, paleo-river evolution, and ecological connectivity in the region.

To further illustrate the paleogeographic implications of the refined Sundaland sea-level curve, we generated a series of shelf exposure maps corresponding to the five benchmark relative sea-level positions reported in Table 1 (20 ka, 15 ka, 11.6 ka, 8 ka, and 6 ka BP). These maps (Figure 6) highlight the progressive submergence of the Sunda Shelf, including the loss of low-lying plains, fragmentation of drainage networks, and eventual formation of semi-enclosed seas such as the Java Sea and Gulf of Thailand. At 20 ka BP, much of the Sunda Shelf remained emergent, exposing broad riverine networks. By 15 ka BP, slight portions of the shelf were inundated. At 11.6 ka BP, sea-level rise sharply reduced terrestrial connectivity. By 8 ka BP, only remnant plains survived above sea level, and by 6 ka BP the shelf was largely flooded, leaving an archipelagic configuration close to the modern geography. The maps were generated using Digital Elevation Models (SRTM 30 and GEBCO 2020) combined with a paleo-stream model (Irwanto, 2020). Processes such as sedimentation, scouring, limestone solution, tectonic movement, littoral drift, delta formation, meandering, river regime change, and riverbed mobility were not considered, owing to limited availability of consistent regional data.

(a)

(b)

(c)

(d)

(e)

Figure 6. Paleogeographic reconstructions of Sundaland at five benchmark sea-level positions from Table 1: (a) 20 ka BP (−113 m), (b) 15 ka BP (−93 m), (c) 11.6 ka BP (−55 m), (d) 8 ka BP (−6 m), and (e) 6 ka BP (0 m). Maps were generated using Digital Elevation Models (SRTM 30 and GEBCO 2020) and a paleo-stream model (Irwanto, 2020). Geological and geomorphological processes such as sedimentation, scouring, limestone solution, tectonic displacement, littoral drift, delta formation, meandering, river regime change, and riverbed movement were not incorporated, as consistent regional datasets are unavailable.

4.4 Slowing of Sea-Level Rise during the Younger Dryas

Both the Sundaland adjusted curve and the Lambeck et al. (2014) global mean show evidence for a deceleration in relative sea-level (RSL) rise during the Younger Dryas interval (~12.9–11.6 ka BP). Prior to this interval, the Meltwater Pulse 1A event (~14.5–13.0 ka BP) drove rapid sea-level rise, with Sundaland recording an average rate of ~1.3 m per century. During the Younger Dryas, this rate slowed to ~1.1 m per century, before recovering to ~1.2 m per century in the early Holocene (~11.6–10.0 ka BP).

Although the reduction is modest in magnitude relative to MWP-1A, the pattern is consistent with global reconstructions (e.g., Lambeck et al., 2014) and reflects the temporary suppression of ice-sheet melting under cooler Younger Dryas conditions. In Sundaland, the slowdown delayed the inundation of low-lying shelf areas, temporarily stabilizing coastlines and prolonging terrestrial connectivity across parts of the shelf. This pause likely influenced ecological transitions, vegetation succession, and the persistence of land corridors available for human and faunal dispersal during the terminal Pleistocene.

4.5 Limitations and Uncertainties

Despite the refinement, the Sundaland curve is constrained by the sparse distribution of offshore indicators. Many available data points derive from coastal settings sensitive to local processes, necessitating strict filtering. The absence of offshore indicators older than 19.25 ka BP required splicing to the global mean, an approach that assumes Sundaland residuals remained constant during the LGM. While this is supported by the general agreement between regional and global curves, additional offshore indicators would strengthen confidence. Furthermore, glacio-isostatic adjustment models specific to Southeast Asia remain limited, and incorporation of such models could refine the regional residual correction beyond the empirical quadratic approach used here.

5. Conclusion

We present a refined relative sea-level (RSL) curve for Sundaland spanning 0–22.5 ka BP, derived from offshore indicators compiled from Chua (2021), SEAMIS (Mann et al., 2019), and Lambeck et al. (2014). Comparison with the Lambeck et al. (2014) global mean demonstrates close agreement, with residuals generally within ±5 m. Anchored residual fitting ensured consistency with present sea level, and the quadratic model provided the best representation of regional deviations.

The final Sundaland curve indicates that sea level at the Last Glacial Maximum (~20 ka BP) was ~−113 m, rising rapidly during the late Pleistocene and stabilizing near modern levels by ~6 ka BP. This trajectory is broadly consistent with global deglacial trends but incorporates a small regional correction specific to Sundaland.

The curve provides a robust baseline for paleogeographic reconstruction, including shelf flooding, paleo-river evolution, and ecological connectivity across Southeast Asia. Limitations remain due to the sparsity of offshore data, particularly before 19.25 ka BP, where the curve relies on splicing to the global mean. Future work should prioritize expanding offshore datasets and incorporating region-specific glacio-isostatic adjustment models to further improve precision.

References

Chua, D. K. H., Bird, M. I., Grice, K., Gadd, P. S., Trevathan-Tackett, S. M., Heijnis, H., … Hua, Q. (2021). A new Holocene sea-level record for Singapore. Quaternary Science Reviews, 270, 107152. https://doi.org/10.1016/j.quascirev.2021.107152

Fleming, K., Johnston, P., Zwartz, D., Yokoyama, Y., Lambeck, K., & Chappell, J. (1998). Refining the eustatic sea-level curve since the Last Glacial Maximum using far- and intermediate-field sites. Earth and Planetary Science Letters, 163(1–4), 327–342. https://doi.org/10.1016/S0012-821X(98)00198-8

Fleming, K., Johnston, P., Zwartz, D., Yokoyama, Y., Lambeck, K., & Chappell, J. (2000). Global sea level since the Last Glacial Maximum: A review of the records. Quaternary Science Reviews, 19(17–18), 1809–1826. https://doi.org/10.1016/S0277-3791(00)00021-1

Lambeck, K., Rouby, H., Purcell, A., Sun, Y., & Sambridge, M. (2014). Sea level and global ice volumes from the Last Glacial Maximum to the Holocene. Proceedings of the National Academy of Sciences, 111(43), 15296–15303. https://doi.org/10.1073/pnas.1411762111

Mann, T., Bender, M., Saito, Y., Hanebuth, T. J. J., & Stattegger, K. (2019). SEAMIS: Southeast Asian Sea-Level Database for the past 40 ka. Quaternary Science Reviews, 219, 112–125. https://doi.org/10.1016/j.quascirev.2019.07.037

Milne, G. A., Long, A. J., & Bassett, S. E. (2005). Modelling Holocene relative sea-level observations from the Caribbean and South America. Quaternary Science Reviews, 24(10–11), 1183–1202. https://doi.org/10.1016/j.quascirev.2004.10.005