Global LiDAR land elevation data reveal greatest sea-level rise vulnerability in the tropics

Accurate digital terrain models (DTMs) are fundamental for assessing exposure to coastal flooding and sea-level rise (SLR). Historically, suitable high-accuracy elevation data have been available only in limited regions, creating substantial uncertainty in global estimates of land and populations at risk. This study addresses that gap by applying a first global lowland DTM derived from satellite LiDAR (ICESat-2) to quantify the extent of land below key elevation thresholds relative to mean sea level (MSL), identify regional hotspots of exposure, and project changes in population exposure under a plausible scenario of relative sea-level rise (RSLR) by 2100. The core question is how improved LiDAR-based elevations alter global and regional estimates of SLR vulnerability and which regions are most affected. The work has high relevance for planning and policy because it indicates where adaptation and risk reduction are most urgently needed.

Prior global and regional assessments have shown that results of flood and SLR exposure analyses are highly sensitive to the accuracy of underlying elevation models, with widely used GDEMs such as SRTM prone to vertical biases that can underpredict vulnerability (e.g., Kulp & Strauss 2016; van de Sande et al. 2012; Griffin et al. 2015). Alternative global products (MERIT, TanDEM-X, CoastalDEM) improve some aspects but remain limited for low-relief coastal zones. Numerous studies highlight that many large deltas are already experiencing frequent inundation and are subsiding due to human activities (Syvitski et al. 2009; Nicholls et al. 2021), with the Ganges–Brahmaputra–Meghna (GBM) delta, for example, seeing 20–60% of land flooded annually and substantial cyclone-related mortality (Becker et al. 2020). SLR impacts on wetlands, coasts, and infrastructure are expected to intensify (Schuerch et al. 2018; Nicholls & Cazenave 2010). These findings collectively motivate the need for globally consistent, higher-accuracy elevation data to refine exposure estimates, particularly in tropical lowlands where subsidence and rapid population growth coincide.

Elevation dataset: The Global LiDAR Lowland DTM (GLL_DTM_v1) was produced at 0.05° (~5 × 5 km) resolution from ICESat-2 satellite LiDAR data collected between 14 October 2018 and 13 May 2020. Coverage: Global calculations used the full GLL_DTM_v1 extent (88°N–88°S). Tropical statistics were computed for 23.5°N–23.5°S. Definition of high-risk coastal lowland: Following Syvitski et al., coastal land below 2 m above mean sea level (+MSL) is considered most susceptible to storm surges and river floods. Population data and growth: Population distributions for 2000 and 2020 were taken from the UN-adjusted GPWv4 database. Population growth rates were computed for grid cells below 0 and 2 m +MSL based on 2000–2020 trends. Scenario of relative sea-level rise (RSLR): Projections to 2100 considered 1 m RSLR composed of approximately equal contributions from absolute SLR and land surface subsidence (LSS). SLR: A middle value of 0.5 m rise since 2020 was adopted within the IPCC RCP2.6/RCP8.5 2100 SLR range (0.29–0.59/0.61–1.1 m since 1986–2005). LSS: A middle value of 0.5 m cumulative subsidence (2020–2100) was applied based on rural-area rates (2.5–10 mm yr⁻¹) reported in the literature; higher urban rates (>20 mm yr⁻¹) were not used to maintain global uniformity. Subsidence drivers include deforestation, drainage, groundwater abstraction, and hydrocarbon extraction. Uncertainty assessment: Exposure estimates below 0 and 2 m +MSL and under 1 m RSLR were calculated using (i) a deterministic method and (ii) a modified deterministic method that incorporates vertical uncertainty. For GLL_DTM_v1 with RMSE=0.5 m below 2 m +MSL, areas under 1 m RSLR are reported with 68% confidence: lower/upper bounds correspond to 0.5 m and 1.5 m thresholds (1.0 ± 0.5 m). Results at 68% and 95% confidence (±1.96×RMSE) were also computed for comparisons with other GDEMs (SRTM, MERIT, TanDEM-X, CoastalDEM) using their reported RMSEs, within SRTM’s spatial coverage (60°N–56°S) for consistency.

- Of global land below 2 m +MSL, 649,000 km² (62%) lies in the tropics, indicating a strong tropical concentration of exposure. - Assuming 1 m RSLR by 2100 and stable lowland population distribution, population on land below 2 m +MSL rises from 267 million (2020) to at least 410 million by 2100; 72% of this exposed population would be in the tropics, with 59% in tropical Asia alone. - Bangladesh: Land below 2 m +MSL equals ~22,000 km² (16% of 2020 land area) with 18.1 million people. SRTM yields only ~1,300 km² below 2 m +MSL, and no other GDEM exceeds ~13,900 km², indicating substantial underestimation by non-LiDAR models. Under 1 m RSLR, ~6,000 km² and 4.9 million people would be below MSL. - Indonesia: Has the largest national extent of land below 2 m +MSL globally at ~118,200 km² (6.3% of national area; 11.3% of global total), about 14× more than the ~8,100 km² inferred from SRTM. - Comparative analyses (Fig. 3) show that GLL_DTM_v1 yields larger and more plausible estimates of lowland area and exposed population than other GDEMs, especially in tropical regions, with quantified 68% and 95% confidence ranges based on elevation RMSE. - Many tropical deltas (GBM, Chao Phraya, Mekong, Pearl, Amazon, Niger) host dense populations in the 0–5 m +MSL zone, underscoring high exposure now and after 1 m RSLR.

By leveraging ICESat-2 LiDAR, the study provides more accurate global elevation estimates in low-relief coastal zones, directly addressing the longstanding uncertainty in SLR exposure assessments caused by GDEM vertical errors. The results reveal that previous models substantially undercounted low-lying land and population exposure, particularly across the tropics. The finding that nearly two-thirds of land below 2 m +MSL is in the tropics, and that future exposed population will be overwhelmingly tropical and concentrated in Asia, reframes global adaptation priorities. Case studies such as Bangladesh and Indonesia illustrate the magnitude of underestimation when relying on legacy elevation data. Quantified uncertainty bounds tied to elevation RMSE strengthen the robustness and usability of the estimates. Overall, the work underscores the need to integrate improved elevation data into coastal planning, flood risk management, and policy, with emphasis on rapidly urbanizing and subsiding tropical deltas.

This study introduces and applies the first global lowland DTM derived from satellite LiDAR (GLL_DTM_v1) to reassess land and population exposure to coastal flooding and SLR. The analysis shows that exposure is concentrated in the tropics, especially tropical Asia, and that prior assessments based on older GDEMs substantially underpredicted the extent of vulnerable land. Under a plausible 1 m RSLR by 2100, exposed population on land below 2 m +MSL could rise from 267 million to at least 410 million. The study advances global risk assessment by providing improved elevation data, explicit uncertainty quantification, and consistent global coverage. It calls for urgent adaptation planning in tropical lowlands and continued enhancements of LiDAR-based DTMs, including further uncertainty reduction and higher spatial resolution as satellite data collection progresses. Future research should refine subsidence estimates, incorporate spatial heterogeneity (including urban subsidence), and integrate elevation improvements with socioeconomic and protective infrastructure data in dynamic flood risk models.

- Elevation uncertainty: Although improved, GLL_DTM_v1 has an RMSE of ~0.5 m in lowlands; exposure ranges are reported at 68% and 95% confidence, but residual vertical errors remain. - Scenario assumptions: The 1 m RSLR scenario assumes equal contributions from SLR and LSS and applies uniform rural subsidence rates; higher and spatially variable urban subsidence rates were not incorporated to maintain global uniformity. - Resolution: The 0.05° (~5 km) grid may miss fine-scale topographic variability relevant for local planning. - Population data and constancy: Exposure projections assume stable 2020 lowland population distribution and use GPWv4 gridded data, which carry their own uncertainties. - Comparative extent: Some cross-GDEM comparisons were constrained to the SRTM coverage (60°N–56°S), not the full global extent.