Amplitude spectrum of the last days
A Fourier transform of a short record shows bands; a long one shows lines. Step through the record lengths, or press play, and watch neighbouring constituents separate once the record is longer than one over their frequency difference.
Amplitudes are Hann-windowed and scaled so a pure sinusoid of amplitude A peaks at A ft. Zero-padding draws the curve smoothly between frequency bins, but it does not add resolution. Only a longer record does that.
Spectrogram: a 5-day window sliding along the record
Harmonic analysis: Budd Inlet’s constituents
↗ See these constituents turn the cranks of a Kelvin tide machine · or spin as wooden arms
Test: our constants vs NOAA on a week they haven’t seen
What’s left: hunting for a seiche
Power spectrum of the residual after subtracting our harmonic fit (Welch’s method, 3-day segments). Weather makes the slope at low frequencies. A basin resonance (a seiche, water sloshing along the inlet) would show as a bump at a fixed period, usually well under 6 hours. Look for one; there may not be one.
From a time series to named constituents
Numbers in teal come from the current analysis.
Use the data
/api/tides/spectrum — everything on this page: spectra, spectrogram, constituent table, holdout test (JSON).
/api/tides/harmonics — just the fitted constants (amplitude, phase, SNR), mean level and UTide’s constituent definitions (JSON, a few KB).
/api/tides/licor/download — the raw 5-minute record, MLLW-referenced (CSV; add ?format=json). Try it in Python with utide.solve, or in MATLAB with UTide or T_Tide.