Alarm thresholds marked. Currently frozen -- see banner above.
0-360°. A jump between 0 and 360 in the trace is a real wraparound, not a sensor glitch.
Negative values mean the channel is running backwards.
Instantaneous volumetric flow rate.
Radar return strength. Velocity readings below the floor are unreliable.
Every plotted value in text form. Rows are bucket averages when the window is long enough to need down-sampling.
| Time | Level (m) | Velocity (m/s) | Flow (m³/s) | Wind dir (°) | Signal | Samples |
|---|
Predicted water level and wind direction, generated by
ml/train.py. Water forecast is currently trained on
synthetic data (see banner above) -- swap to real data once the
BGT-RD103 is replaced, no other change needed.
Forecast reports are generated offline; run
python3 -m ml.train --synthetic on the Pi and view the
result in the terminal, or --report forecast.json to
save it for later reference.