Ship’s Barometer 0.3.0

An offline Signal K weather instrument with a responsive pressure chart. No QuestDB, InfluxDB, internet connection, API key or extra history plugin is required.

Open Ship’s Barometer from Signal K’s Webapps after enabling the plugin. The webapp shows current sensor pressure in hPa, a measured three-hour change, the existing tendency and three clearly labeled forecasts: wind-direction, pressure-only and seasonal outlook. Frontal outlook and expected wind behavior explain the existing front analysis beneath those forecasts. Measured conditions appear separately underneath when enabled. These are the same published Signal K values available to other displays. Select 6, 12, 24 or 48 hours. Day, Night and Red use the same controls and Auto rules as Endurance: the theme button cycles the manual modes, and Auto follows vessel environment.mode. Day selects Day; Night selects Red while navigation.state is sailing or motoring, and Night while anchored or moored. Missing day/night state falls back to Night. Your manual selection is stored in your browser.

History and data quality

• A separate recorder retains at most 72 hours (3 days) and 4,320 samples.
• At most one actual observation per 60 seconds; no timer-generated samples.
• Original Signal K timestamps are retained, including across restarts. Invalid, future, out-of-order and other-vessel observations are rejected. Source IDs are required; known external weather sources are excluded.
• The recorder locks to the first accepted pressure source. Use History pressure source to select an exact Signal K source ID when multiple sources exist. The chart’s source selection is independent of the legacy forecast engine.
• Lines stop at gaps longer than 90 seconds. Single observations remain visible as points. Missing observations are never filled. No history is imported from the forecast engine or an external database.
• The three-hour delta compares actual readings within one minute of the three-hour baseline; it is unavailable without that baseline or with stale current data. It describes the change between those observations, even if there are gaps between them.
• Current pressure and analysis are hidden when pressure observations are over five minutes old. Historical points remain available, with their actual times.
• pressure-history.json in the plugin data directory is atomically replaced every three minutes when changed, and on orderly shutdown. Expired observations are pruned during ingestion, reads, loading and saves. An abrupt power loss may lose the latest three minutes. History storage errors appear in the webapp and server log.
• The chart displays raw sensor pressure. The existing analysis retains its own altitude correction, sampling and source handling, so its reference can differ.

The existing pressure/tendency/forecast calculation files are unchanged. The optional local weather assessment remains opt-in. Its existing explanatory reasons appear when enabled and fresh. External weather forecast comparison is not included; the instrument uses onboard observations and works entirely offline.

Install, update and rollback

For the local release tarball, in the Signal K installation directory (commonly ~/.signalk), run npm install /absolute/path/signalk-ships-barometer-0.3.0.tgz. Restart Signal K, enable Ship’s Barometer in Plugin Config, then open its Webapp. For containers, use the installation directory inside the Signal K container and retain the persistent Signal K data volume. This is local installation guidance; initial onboard review on TBD confirms live observations and historical collection. Forecast accuracy has not been field-calibrated.

Before updating, back up the plugin configuration and its data directory. To roll back, reinstall [email protected] (or your saved previous tarball) and restart Signal K. The new history file is independent of legacy offline.json and local-assessment.json. Keep the backup until your aboard checks pass.

Development

Run npm ci then npm test. Tests cover retention, minute bounds, original timestamps, source isolation, persistence/restart, corrupt storage, write failures, chart gaps, endpoint validation and plugin lifecycle, alongside the inherited forecast tests. The read-only /plugins/signalk-ships-barometer/history?hours=24 endpoint accepts only 6, 12, 24 and 48 hours. Browser assets ship locally with the package.

Ship's Barometer

Offline weather interpretation from the boat's own sensors. Ship's Barometer combines traditional barometric forecasts with an optional assessment of measured pressure changes, wind strengthening/easing, moisture conditions and water-temperature context.

This is an independent continuation of signalk-barometer-trend v2.3.4, created by oyve. The underlying barometer-trend library is also by oyve. Their work supplies the original pressure analysis and forecast models; this project adds integration corrections, local observations, quality checks and independent packaging. It is maintained by SCarns and is not endorsed by the original author.

What it does

• Converts Signal K true-wind direction from radians to the degrees required by the forecasting library, including a one-time conversion of legacy saved history.
• Retains all 21 pressure/forecast paths used by the original 2.3.4 plugin and existing dashboards.
• Optionally publishes environment.outside.weather.assessment: measured changes, short reasons, SI units, source IDs, original timestamps, missing inputs and quality flags.
• Uses local humidity/temperature or a supplied Derived Data dew point, with checks for disagreement and observation-time mismatch.
• Compares air and dew point with local water temperature while explicitly distinguishing transducer-depth water from a measured surface temperature.
• Preserves separate pressure and assessment history across restarts. Runs without any weather API or internet connection after installation.

The assessment is descriptive and experimental. It does not provide a calibrated rain probability, future wind speed, confirmed fog, gust forecast or new automatic alarm. Heat index and wind chill remain with Derived Data. The optional assessment is off by default.

Install

Available on npm. Refresh the Signal K store and search for Ship's Barometer, or install from your Signal K user directory:

cd ~/.signalk
npm install --save signalk-ships-barometer

Download signalk-ships-barometer-0.3.0.tgz from GitHub Releases and copy it into your Signal K user directory, normally ~/.signalk:

cd ~/.signalk
npm install --save ./signalk-ships-barometer-0.3.0.tgz

Disable Barometer Trend before enabling Ship's Barometer. The plugins share the original forecast output paths; running both would create competing publishers. You can keep the old package installed for rollback.

Restart Signal K and enable Ship's Barometer in plugin settings. To use the extra assessment, enable Local weather assessment. Check installation and migration for source/depth settings and history handling.

Releases are published on npm under scarns, with the Signal K discovery keyword and app icon metadata. Store indexing/cache refresh may take time. See publishing to the Signal K store for future releases.

Local inputs and output

| Input | Signal K path | Units | |---|---|---| | Pressure | environment.outside.pressure | Pa | | Air temperature | environment.outside.temperature | K | | True wind direction | environment.wind.directionTrue | rad | | True wind speed | environment.wind.speedTrue | m/s | | Humidity | environment.outside.humidity or environment.outside.relativeHumidity | ratio | | Optional supplied dew point | environment.outside.dewPointTemperature | K | | Optional water temperature | environment.water.temperature | K |

Set the assessment source IDs explicitly when multiple local sources share a path. Blank source fields retain the first valid local source. Sources identified as Open-Meteo, NOAA or WeatherAPI are excluded from the assessment. Only this vessel's timestamped observations are accepted. The source IDs and timestamps used remain visible in the assessment.

The new assessment contains pressureState, windState, moistureState, waterContext, fogContext, qualityFlags, measurements, units, inputs, missing and reasons. windOutlook is explicitly uncalibrated.

Evidence and development

The first release comes from a local archive study and installation testing. The wind-angle fix corrects a unit-contract error; it does not establish improved rain accuracy. The pressure-to-future-wind association was inconsistent across periods and decision-window timings, so the new assessment reports observed changes rather than an automatic wind warning. Water context has software tests, but no boat-fog accuracy score.

The release is based on the working 2.3.4 lineage. Upstream's newer main branch uses a different library interface and path layout; this project intentionally retains the installed 2.3.4 output contract. See validation and limits.

Requires Node.js 18 or newer. For development:

npm ci
npm test

Runtime dependencies are bundled in the release tarball; development dependencies are not. Contributions should include meaningful tests and retain the distinction between measured conditions, hypotheses and verified forecast outcomes.

License and credits

Apache License 2.0. Original license text is retained; modified-file attribution and additions are recorded in NOTICE and CHANGELOG.md. Runtime dependencies retain their own license notices. The nautical artwork was generated with OpenAI image generation at SCarns' direction.