Sign inRegisterCheck connection

Publishing Weather Facts as a Small Knowledge Base

How a personal weather site can publish sourced weather facts: update discipline, rotation versus editorial selection, and avoiding unsourced folklore.

Back to Personal weather station notebooks

A station website can teach units, local climate, and phenomena without pretending to be a textbook and without becoming a rumor mill. The method is to treat each “weather fact” as a small knowledge-base record: a claim, a source, a date checked, and a rule for when it may appear. Random rotation is a presentation choice. Editorial selection is a quality choice. They are not the same operation.

Historical context

The historical TNET Weather notebooks described a flat file of short facts, split on a delimiter, shuffled, and included on page footers. The recovered PHP and the Arizona sample blurbs are evidence that the old site wanted a rotating teaching strip. They are not a fact database to copy. Some of those blurbs mixed useful definitions with climate extremes that would now need a current NCEI or NWS citation. This article rebuilds the workflow, not the old file.

Input

Each candidate fact needs a structured payload before it is eligible to print:

| Field | Role | |---|---| | Claim | One sentence the reader can use | | Topic | Unit, phenomenon, local climate, safety pointer, or instrument note | | Source | Primary URL or document, not “a forum” | | Source date | When the source said this | | Checked date | When you last opened that source | | Geography | Global, national, or a named station/climate division | | Status | active, seasonal, retired, disputed |

Reject the record if any of those is empty. A delimiter in a text file (:::) can store the fields. A delimiter cannot create authority.

Good inputs are definitions (what a knot is), instrument caveats (what a backyard rain gauge misses), and dated official extremes (NWS climate summaries). Bad inputs are unsourced “Arizona never…” claims, viral monsoon myths, and restated warnings that belong on weather.gov instead of a footer.

Process: sourcing and updating

Write the claim so it can be checked

Prefer a form that names the quantity, the place, and the authority:

A knot is one nautical mile per hour. The international nautical mile is 1852 metres.

That sentence can be verified. Contrast a folklore form:

The monsoon always starts on June 15.

The second sentence has no station, no definition of “start,” and no source. It should never enter the file.

Prefer primary sources

Use NWS, NCEI, NHC, and WMO for climate and hazards; use the instrument or software vendor for logger behavior; use peer-reviewed papers for research findings. Blogs and social posts can point you at a primary source. They cannot be the source. The same rule is expanded in the resource bibliography notebook.

Climate normals move. U.S. official normals are 30-year products; the set in force in 2026 is 1991–2020 (NCEI U.S. Climate Normals). A fact that says “normal July rainfall is…” without a normals period is unfinished.

Update on a schedule, not when you remember

Assign a review interval by volatility:

  • Stable metrology (units, compass rose, dew-point definition): review yearly.
  • Climate normals and station records: review when NCEI or the local NWS office publishes a new normals set or a new record event, and at least yearly.
  • Safety pointers: review each season; point to live NWS, never cache a warning.
  • Local station notes (siting changes, gauge moves): review when the metadata changes.

Store checked date in the record. A shuffled fact with a five-year-old check date is an unpublished erratum.

Random rotation versus editorial selection

Shuffling is acceptable for eligible facts that are all true at once. It is the wrong tool when:

  • The fact is seasonal (wind chill in July; monsoon onset in January).
  • The fact is a live hazard (heat advisory language).
  • The fact is a record that was broken last week.
  • You have mixed definitions with persuasion (“you should buy…”).

Editorial selection means a human or a dated rule chooses the pool before shuffle: for example, only status=active and topic=units on instrument pages, and only monsoon-tagged facts from June through September in the Southwest. Shuffle inside that pool. Do not shuffle the whole archive and hope.

If you display HTML inside a fact, sanitize it. A knowledge base that accepts raw markup will eventually print a broken tag or an injected link. Prefer plain text plus a source URL rendered by the template.

Output

The page should show:

  1. The claim, in language a non-specialist can use.
  2. A visible source (office or document name, not a raw tracking URL).
  3. Optionally the checked year, if the claim is a normal or a record.
  4. Nothing that looks like a watch, warning, or forecast.

Example of a publishable teaching fact (definition, not a climate record):

Dew point is the temperature to which air must be cooled at constant pressure to reach saturation. It is a moisture variable, not a comfort slogan. Source: NWS glossaries and psychrometric practice; see the apparent-temperature notebook.

Example of a record fact that is not publishable until you attach an official table:

“Phoenix’s greatest 24-hour rainfall is …” — withhold until you cite a current NWS Phoenix or NCEI station report with the date of the event and the station identifier. Do not recycle a 2006 footer.

The station site’s job is to teach the kind of quantity (what 24-hour rainfall means, midnight-to-midnight versus storm-total) and then point at the official extreme.

How a station site can teach without being a textbook

Stay inside three honest roles:

Units and encoding. Knots versus miles per hour, millimetres versus inches, 0–360° versus compass labels. Pair with the wind-encoding notebook.

Siting and instruments. What your sensor can and cannot see. A rooftop thermometer is a rooftop thermometer.

Pointers to official climate. Link NWS local climate pages and NCEI station normals. Do not retype their tables.

Leave textbooks to textbooks: atmospheric dynamics, climate-change attribution, and warning policy. Leave live hazards to NWS. A rotating fact is a glossary card, not an authority.

Troubleshooting

A reader treats the footer as a forecast. Add a one-line frame: “Background note, not an NWS product.” Remove any future-tense weather claims from the file.

Two facts contradict each other. That is a knowledge-base bug. Retire one, or qualify both with geography and date. Shuffle will otherwise teach both on consecutive page loads.

HTML in the data file breaks the layout. Stop storing presentation in the claim. Store title, body, source_url.

An extreme was beaten. Status becomes retired the day the office publishes the new record. Do not wait for the annual rewrite.

You cannot find a primary source. Do not publish. “Everyone knows” is the folklore filter failing.

The pool is all local color and no method. Add unit and instrument cards until at least half the active set would still be true if the station moved.

Modern relevance

TNET’s research library makes the same distinction this notebook asks of a hobby fact file: observed, modelled, and derived information are labeled, and folklore does not enter the evidence stack. See evidence families for how atmospheric and network inputs are grouped, and the notebooks hub for neighboring methods. Official climate and hazards remain on NWS and NCEI.

Sources