A hit counter on a weather site counts fetches. It does not count storms, sensor quality, or whether yesterday’s rainfall is fit to use. At /visitors.php the historical TNET Weather station site published visitor statistics: recovered headings promised a global map of recent hits and a continental U.S. map. Those maps are not rehosted. This article is about what such pages actually measured, why the totals are a poor scientific metric, and how access logs can still serve as provenance of who retrieved what, when—without turning vanity into evidence.
It is distinct from /stat/counter.php, which is the implementation slot for a PHP counter package. This URL is the public meaning of the number. That URL is the gadget.
Historical context
Personal weather templates of the 2000s treated traffic as proof of life. A sidebar said “you are visitor 77,164.” A dedicated visitors page drew dots on a world map from geolocated IPs. Operators compared totals the way they compared rain totals. The analogy is false. Rain is an observation of the atmosphere (with its own errors). A hit is an observation of the web server.
Recovered TNET Weather chrome advertised Mesa-area weather, lightning, webcam, maps, and history, and used the same boilerplate description on the visitors page as on unrelated URLs. That reuse is a clue: the visitors page was site furniture, not a climate product. PHP Web Stat and similar packages wrote cookies, stored IP-derived locations, and printed Today / Yesterday / Total. Start dates on those widgets are historical configuration, not a scientific epoch. They are not reprinted here as current analytics.
Modern TNET Weather does not claim those counts, does not operate that counter, and does not treat restored maps as a live audience. Domain restoration is not a continuation of the original site’s traffic, and it is not a continuation of the original team.
What a visitor number is
Strip the map art and the number is a count of events the software chose to call a visit.
Page views. Each request that the script saw. Bots, prefetch, and a user hitting refresh are views. A thunderstorm does not create views. A Reddit thread does.
Sessions or “uniques.” A cookie or an IP-plus-agent heuristic. Carrier NAT makes many people look like one. Mobile networks and crawlers make one actor look like many. Uniques are a derived statistic. They are not identities.
Mapped hits. An IP is passed through a geolocation table and plotted. The table is a commercial or public guess. It is not GPS. A pin in “Phoenix” is not a verified observer in Phoenix, and it is not a meteorological station.
Referrers. Sometimes logged, often stripped. A referrer can explain why a ClientRaw URL was hammered. It still is not weather.
None of these series can be quality-controlled like a temperature. There is no calibration bath for a hit. There is no WMO siting classification for a bot.
Why counters fail as science
They measure popularity of a URL, not quality of a sensor. A well-sited thermometer in a quiet suburb can have a tiny counter and a usable series. A webcam pointed at a highway can have a huge counter and no climate value. Using hits to rank stations is how entertainment beats metadata.
They mix humans, bots, and other stations. Parsers that fetch clientraw.txt every minute are “visitors.” Search crawlers are visitors. The operator loading their own dashboard is a visitor. Inflating the total does not thicken the atmosphere.
They are easy to game and easy to break. Reloading, hosting the counter on a hot-linked image, or losing the database resets the epoch. A “start 2008” widget that was wiped in 2011 is a broken instrument pretending to be a long record.
They leak people while claiming to measure weather. IP logs and hit maps are the privacy surface described on the privacy article. Publishing a world map of “who looked at my thermometer” is a choice about visitors, including neighbors. It is not a required part of a climate record.
They tempt operators to optimize the wrong thing. A news post written to attract hits is the opposite of an operations log. Sensor swaps do not trend. Outages should be recorded even when nobody is watching.
Official observation programs do not use website hits as an ingest criterion. NOAA’s MADIS/CWOP path cares about the packet, the metadata, and quality flags—not about whether the operator’s homepage was busy.
Log provenance versus vanity hits
Access logs still have a legitimate scientific adjacent use. They are not climate. They are provenance of access.
A later analyst may need to know:
- whether a public file was actually being fetched at the times a third party claims to have used it;
- whether a gap in an external archive matches a period when the server was down (see the operations log);
- whether a sudden change in a copied series matches a scraper that started or stopped.
That use requires raw enough logs to be checked, with time, path, and status, and a clear statement that the log is not the weather file. It does not require a public running total. It does not require a map.
Vanity hits are the opposite presentation: a single integer, no path, no distinction between /visitors.php and clientraw.txt, no bot filter that you can audit, often a cookie whose only job is to make the integer climb. Treat that integer as decoration. If you must keep it, keep it off observation pages so a screenshot of “current conditions” does not include a crowd.
Implementation details of a named PHP counter belong on /stat/counter.php. Do not read this article as a setup guide, and do not read that page as a claim that TNET still runs the package.
How to read an old visitors page
- Do not cite the total as evidence that the station was important or accurate.
- Do not treat a geolocation pin as an observer’s location. Station siting is metadata on the About and configuration pages, not a hit map.
- Do not assume the counter survived domain lapse or restoration. A new publisher starts a new log.
- If you need climate, go to the observation files and to official archives (NCEI, NWS), not to
/visitors.php. - If you need to know how a specific counter script stored Today/Yesterday/Total, use the counter implementation page, not this essay.
Recovered map headings are evidence that this URL was a traffic display. They are not a reconstructed audience, and they are not current analytics for tnetweather.com.
Practical guidance for operators
- Put last-updated and station ID on the observation page. Put hits, if you insist, on a separate URL like this one.
- Do not use the counter as an uptime monitor. A silent logger can still serve a popular stale page; a live logger can have twelve visitors.
- If you geolocate IPs, treat the result as personal-adjacent data, not as a mesomap.
- When you retire the station, freeze or remove the counter. A climbing total on a dead instrument is a frozen page pretending to be alive. Label it on the archived terms.
- Do not copy another site’s counter ID into your skin. You will inflate their vanity and empty yours, and you will send your visitors’ IPs to a third party you did not mean to involve.
Modern relevance
Analytics suites replaced many PHP widgets. The scientific mistake migrated with them: dashboard owners still screenshot session counts next to a performance chart and call it validation. Connection tests, weather evidence, and visitor traffic remain different families. Mixing them is how a popular page gets mistaken for a good measurement.
TNET’s public research language keeps observed, modelled, and derived values apart. A hit count is not in those families. It is operational telemetry about the website. Useful in logs. Misleading on a climate page.
TNET research bridge
If the question is “what counts as a source, and how is quality talked about in public,” the answer is data sources, quality controls, and methodology. That page is the primary next step from a visitors URL: it is the discipline that refuses to treat a fetch count as an atmospheric record.
How the service works describes the modern research product. The legacy hub and site-history hub collect identity, privacy, operations, and archive-label pages that sit beside this one.