Speaker
Description
Continuous water-level monitoring in shallow coastal lagoons and wetlands is essential for ecological protection and water resource management. In extensive wetland systems such as the Danube Delta, ground gauges are sparse and often difficult to install and maintain. While satellite altimetry provides only limited coverage because its measurements are restricted to narrow ground tracks, SAR provides much denser spatial coverage, but retrieving centimetre-scale water-level variations from SAR is challenging in these environments. Small changes in water level produce only small horizontal shifts in the shoreline, which can be difficult to detect at the spatial resolution of freely available SAR data.
We introduce a method for estimating water-level changes by measuring sub-pixel shoreline movements along transects perpendicular to the shoreline. The approach uses multi-temporal ALOS-2 PALSAR-2 L-band quad-polarimetric SAR data and projects intensity profiles onto a fixed shoreline reference frame. The method was evaluated against continuous in-situ pressure loggers in two independent lagoons of the Danube Delta: Lake Sinoie at Histria and Lake Razim at Dolosman, using fixed transects per site.
Across a 34 cm seasonal drawdown, the in-sample retrieval shows strong agreement with observed water levels (r = −0.88, R² = 0.78, RMSE ≈ 5.85 cm). A linear regression between the observed water-level changes and the corresponding SAR-derived sub-pixel shoreline displacements was used to estimate the local shoreline slope, yielding 0.76 m/cm for Dolosman and 1.10 m/cm for Histria. Under leave-one-date-out cross-validation, the method achieves an RMSE of 6.53 cm, compared with 14.11 cm for a mean-level baseline, corresponding to a 53.7% reduction in prediction error.
These results demonstrate that sub-pixel L-band SAR shoreline tracking can provide a practical approach for monitoring centimetre-scale water-level dynamics in wetlands.