Harmonic Tide Prediction

Demo only — not for navigation/operations.

This script Illustrates harmonic tide prediction with standard Python tools. It is a non-profissional harmonic prediction tool for sea level that generates a short animation. The script reads an hourly time series in CSV format, converts levels from >millimeters to meters, splits the record into two halves (training and test), fits a tidal harmonic model by least squares on the first half, and, on the second half, animates sliding 7-day windows comparing observed vs. predicted levels. It also computes and plots 4-week moving averages over the full period.

This software is not for profissional use. It was created using vibe programming to illustrate how harmonic prediction can be done using standard libraries in Python.

Do not use for navigation or operational decisions.

There is a PDF Manual available.

Examples

Tide Prediction vs Colleted Data for Salvador-Brazil

Average sea level along 100 years for Honolulu

What it does

• Reads hourly CSV year,month,day,hour,level_mm
• Converts mm → m, removes -32767 (missing)
• Splits data: train (1st half) / test (2nd half)
• Fits ~37 tidal constituents (NOS/CO-OPS) by least squares
• Animates a sliding 7-day window (observed × predicted)
• Plots 4-week averages + linear trend (m/yr)
• Saves MP4 (ffmpeg) and PNG; logs to tide.log

Requirements

• Python ≥ 3.8 · numpy, pandas, matplotlib
• ffmpeg on PATH (for MP4)

Data

Sample CSVs in ./data/: Honolulu, Fortaleza, Salvador
Source: University of Hawai‘i Sea Level Center (UHSLC) — the data keep UHSLC terms.
Cite if used:
Caldwell, Merrifield, Thompson (2015), Sea level measured by tide gauges… NOAA NCEI, doi:10.7289/V5V40S7W.

CSV format (no header): year,month,day,hour,level_mm (UTC).
Minimum 100 records after filtering.

Quick start

Interactive:

python tide.py

Example:

python tide.py --csv_file ./data/Honolulu.csv --initial_year 1993 --final_year 2020   --local "Honolulu" --animation_weeks 4 --animation_speed 1.0 --week_seed 42

Key options

• --csv_file PATH · --initial_year INT · --final_year INT
• --week_seed INT (start position reproducible)
• --animation_weeks INT (duration, in weeks)
• --local "Label" (titles/filenames)
• --animation_speed FLOAT (use 1.0)

Outputs

• MP4 animation: {local_lower}_weekly_average.mp4
  or {csv_basename}_weekly_average_{YYYYMMDD_HHMM}.mp4
• PNG averages: {csv_basename}_average_level_4_weeks_{YYYYMMDD_HHMM}.png
• Log: tide.log

Notes / limitations

• Educational code; no nodal corrections or QA beyond -32767
• Frequencies may be ill-conditioned on some datasets
• Model trained on first half only

License

Code: MIT.
Data: UHSLC terms (see site above).