Issues and potentials of multi-sensor water level monitoring: lesson learned at Recentino Lake, Italy

Abstract. Monitoring surface water levels in reservoirs is often hindered by the sparse distribution of in situ gauges and inconsistencies among height reference frames. This study compares water levels obtained from the SurfaceWater and Ocean Topography (SWOT) satellite altimeter, an in situ gauge station, and a Unmanned Aerial Vehicle (UAV) photogrammetric survey, at Recentino Lake, a small artificial reservoir in central Italy, to quantify discrepancies arising from unknown or inconsistent height reference frames. SWOT data were processed using a two-step outlier removal procedure to derive a reliable water level time series, while the UAV survey provided a high-resolution Digital Elevation Model (DEM) (GSD 1.6 cm/pixel) in a certain epoch from which water level was extracted at the water-dam interface. All datasets were transformed to a common height reference frame for direct comparison. Assuming the gauge time series as reference, at the epoch of UAV survey, the UAV-derived water level differs from the gauge one by -0.17 m, while the SWOT and gauge time series show moderate agreement (Pearson correlation of 0.69) and a mean/median difference of - 0.08 m. Also, differences between ascending and descending SWOT passes (Pearson correlation of 0.65 and 0.76, respectively) indicate orbit-dependent effects on SWOT water level measurements. These findings emphasise the relevant contribution of multi-source datasets for water reservoir level monitoring, mainly for the detection and correction of height reference frame inconsistencies, including the bias between SWOT and the national height reference frame.