Low-Elevation DEM Sensitivity in Landsat-Based Reconstruction of Long-Term Relative Lake Volume Anomalies

Abstract

Long-term lake storage monitoring remains challenging in data-scarce high-elevation regions because continuous water-level records and bathymetric data are often unavailable. Annual Landsat-derived areas of Xiao Qaidam Lake from 1996 to 2025 were combined with DEM-based height–area–volume (H–A–V) relationships to reconstruct DEM-referenced relative water-level anomalies and DEM scenario-based relative volume anomalies. Sentinel-2, three DEMs, ICESat-2 ATL13, uncertainty propagation, and buffer tests were used for evaluation. All three DEM scenarios showed long-term increases but different magnitudes. FABDEM contained a 91.04 km2 low-elevation platform, 99.9% of which overlapped the 2011–2015 water-mask union, consistent with source-period water-body flattening or gap filling. Lake area increased from 78.42 to 126.63 km2. Across the three DEM scenarios, the 2025 DEM-referenced relative water-level anomalies ranged from 3.29 to 3.99 m, and the 2025 DEM scenario-based relative volume anomalies ranged from 3.46 to 4.14 × 108 m3. Without bathymetric validation, the magnitude of ΔV remains DEM scenario-dependent. During 2019–2025, DEM-referenced relative water-level anomalies were consistent with ATL13 relative water-level anomalies (r = 0.885; RMSE = 0.195 m). Annual relative volume increments were positively associated with climatic water balance. The framework provides a DEM-aware approach for monitoring relative water-level and volume anomalies in data-scarce closed basins.