Assessment of Lowest Astronomical Tide at Belgian North Sea Tide Gauges Using Long-Term Harmonic Analysis and Residual-Spectrum-guided Constituent Selection

This study presents an updated station-based assessment of Lowest Astronomical Tide (LAT) in the Belgian part of the North Sea using 23 years (2001–2023) of high-resolution water-level observations. Data from four tide gauges—Nieuwpoort (NPT), Oostende (OST), Zeebrugge (ZLD), and the offshore Westhinder station (MP7)—were analysed using the UTide harmonic-analysis framework. The duration and quality of the dataset enable, for the first time in Belgian waters, a systematic comparison of four analysis-window and reconstruction strategies and three tidal-constituent selection strategies. First, sensitivity to the analysis period and reconstruction strategy was assessed by comparing a long-record harmonic analysis, separate annual analyses, extended predictions based on annual coefficients, and complex averaging of annual harmonic coefficients. Second, sensitivity to constituent selection was evaluated using the UTide default set, a fixed Dutch constituent set, and an iterative residual-spectrum-guided procedure implemented with UTide routines. The results show substantial spatial and methodological variability in the estimated LAT values, highlighting the importance of station-specific analysis and careful constituent selection. The contribution of this study is therefore a systematic and reproducible assessment for Belgian waters rather than the development of a new theory of harmonic analysis. Existing station values from the Agency for Maritime and Coastal Services (MDK) and the LAT conversion grid are used as operational consistency benchmarks rather than as independent ground truth. Harmonic-parameter uncertainty was propagated using 1000 Gaussian Monte Carlo realisations. The reported spread is conditional on the selected harmonic model and does not represent a complete LAT uncertainty budget or a measure of absolute accuracy. Comparison of otherwise identical (N = 500) and (N = 1000) simulations showed maximum absolute differences of 0.007 m in the Monte Carlo mean, 0.009 m in the conditional parameter standard deviation, and 0.020 m in one percentile-interval bound across Methods M1–M4. For the preferred residual-spectrum-guided configuration, the corresponding maximum differences were 0.003, 0.004, and 0.006 m, respectively. The deterministic LAT is defined as the minimum obtained from the unperturbed harmonic reconstruction, whereas the Monte Carlo mean is the arithmetic mean of the minimum LAT values obtained from the perturbed realisations; these quantities are therefore reported separately. For the preferred 2001–2019 configuration, the deterministic LAT values were − 0.648, − 0.499, − 0.251, and − 0.300 m TAW/DNG at NPT, OST, ZLD, and MP7, respectively, with corresponding conditional parameter standard deviations ranging from 0.0031 to 0.0034 m. Sensitivity to the selected 19-year input window is reported separately.