Quality Control of Tide Gauge Observations and Gap Filling Using Data Fusion

Ensuring the accuracy and continuity of sea level observations is crucial for coastal monitoring, operational forecasting, and oceanographic modelling. This study evaluates the quality control and comparability of radar tide gauge measurements at two Portuguese ports—Leixões and Sines—using five consecutive days of high-resolution (1-minute) sea level records. Several statistical approaches were applied to assess instrument agreement and reconstruct missing data segments, including process capability analysis, Kalman filtering, and ARIMAX time series modelling with external regressors. Artificial gaps representing 20% of the time series were introduced and uniformly randomly distributed to test each method’s reconstruction accuracy. The analysis revealed strong linear relationships between co-located sensors, despite the presence of systematic biases and moderate phase-dependent deviations. Ebb–flood stratification was used as a diagnostic tool to identify phase-dependent discrepancies, particularly at Sines, highlighting potential hydrodynamic and geometric influences. Among the tested approaches, ARIMAX models provided the highest reconstruction accuracy (RMSE < 5 mm) and more effectively captured temporal autocorrelation than linear regression or Kalman filtering. These findings demonstrate the potential of sensor redundancy and statistical data fusion to ensure short-term data continuity in tide gauge networks, particularly in response to operational disruptions. The proposed methodology supports real-time quality control workflows and gap filling without relying on long-term climatological records.