Sign inRegisterCheck connection

WXSIM as a Local Station Model

WXSIM is a local mesoscale and statistical weather model run at a personal station. It is a scenario, not NWS and not TNET Connection Outlook.

Back to Historical weather observations and forecasts

WXSIM is a local weather-modeling program run on a station operator’s computer. It produces a site-specific scenario: expected temperature, humidity, sky, wind, and related fields for that customized location, from a mix of ingested data and the program’s own calculations. It is not a National Weather Service product. It is not TNET Connection Outlook. A page that once printed a WXSIM table under a Mesa heading was a model display, not an official forecast and not a live observation.

The historical URL /wxsim.php recovered a Mesa, Arizona WXSIM-formatted forecast on the original weather-station site. The snapshot identified a run updated Sunday, 18 April 2010, 6:55 a.m. Period temperatures from that run are not current and are not reprinted here. The page’s own historical note said forecasts were generated six times a day and uploaded at 2:30 a.m., 7:30 a.m., 11:30 a.m., 2:30 p.m., 4:30 p.m., and 10:30 p.m. That cadence is historical site practice. It is not a claim that a model is running here now.

Historical context

Personal weather sites in the 2000s often added a local-model tab beside the NWS tab. WXSIM, created by Thomas J. Ehrensperger and described on wxsim.com as an interactive local atmospheric modeling system, filled that tab for many operators. Companion tools ingested surface, upper-air, and larger-model fields and, optionally, the home station. Third-party PHP scripts formatted the output for the web; the recovered page credited a Saratoga-Weather.org formatting script. Those scripts displayed model text. They did not convert the model into NWS.

The recovered heading said the forecast was “generated” for Mesa using WXSIM. That wording is easy to misread as TNET issuing an official forecast. The producer of the scenario was the WXSIM run on the operator’s machine. TNET Weather today does not operate that model, does not endorse it, and does not treat a 2010 table as current Mesa weather.

What WXSIM is

Ehrensperger’s public description is specific enough to classify the product without copying his marketing claims. WXSIM is a Windows program that builds a forecast for a customized site. It combines:

  • astronomically calculated sun angle and radiation terms in its temperature logic;
  • ingested real-time surface and upper-air information;
  • larger numerical model fields (the public site names NOAA’s Global Forecast System among the external inputs);
  • optional personal weather-station samples via common station software;
  • operator interrupts, so a human can inject timing or other information before or during a run;
  • optional bias-correction that learns from past temperature and humidity errors at that site.

The author calls it an interactive local atmospheric model. That is the right category: scenario / local model output. It is mesoscale in intent (a point and its immediate surroundings, not a continental grid the operator built) and statistical in part (climatological files, bias learning, customized site parameters). It is not a Weather Forecast Office. It is not a WMO official forecast.

WXSIMATE, the companion downloader described in the project’s technical notes, is an ingest layer. Ingested GFS fields remain forecast input from a global model. Ingested station samples remain observations. WXSIM’s issued table is the local model’s output. Keep those labels if you display the stack.

This article does not document WXSIM’s internal equations, and it does not rank the program against NWS. Factual credit: WXSIM and wxsim.com are the origin of the software. No endorsement is implied. TNET is not affiliated with the author.

Scenario, not NWS, not TNET

Not NWS. Official U.S. public forecasts and warnings are issued by the National Weather Service. A WXSIM run may ingest NWS or NOAA model data and may look similar to an NWS point forecast in layout. Similarity is not issuance. If you need the official Phoenix-area forecast, use weather.gov or NWS Phoenix. If you need warnings, use those sites and local emergency authorities. TNET is not a warning service.

Not an observation. Even when the model ingests the last temperature from the logger, the issued high for this afternoon is a model value valid for a future interval. The logger’s 6:55 a.m. reading is an observation. Putting both on one graphic does not make the high a measurement.

Not TNET Connection Outlook. Connection Outlook is a derived TNET estimate of how a Starlink connection may perform for a selected city, date, and interval. It interprets atmospheric conditions among other evidence families. It does not run WXSIM, does not publish WXSIM tables, and does not promise a download speed. A local weather-model scenario and a connection outlook are different products with different questions.

Not a consensus. A station that also reprints NWS text has two forecasts. See station-site composites. Do not average them silently.

What a local-model page must disclose

A defensible WXSIM (or similar) display states:

  1. Software name and version if known — “WXSIM,” not “TNET forecast.”
  2. Site the model was customized for — the cty.fdt location, which may be the backyard and still not be the NWS zone.
  3. Run or issue time — when this table was produced.
  4. Valid periods — the model’s own day and night windows.
  5. Ingest summary in words, not weights — that the run used external model fields and/or station samples, if that is true for the setup. Do not invent which input dominated.
  6. A non-official line — this is a local model scenario, not NWS.

The historical six-per-day upload list is a publishing schedule. It is not a sampling theorem and not a verification score. A missed upload leaves the previous run on the server. That previous run is historical model output. It must keep the old issue time.

Graphs of a WXSIM run (the recovered page linked a wxsim/latest.png) are derived pictures of forecast fields. A PNG timestamped 5 August 2010 is a historical plot. Embedding it as “latest” without the run time repeats the live-needle error from old consoles.

How to read a local-model table

Work it as you would any model guidance.

  • Compare like periods: model tonight versus NWS tonight, not model tonight versus the 10-minute temperature.
  • Treat UV index, solar irradiance, and heat index from the model as calculated scenario fields. They are not a pyranometer and not an NWS heat product unless the page says they were copied from NWS.
  • If the model offers hypothetical or educational modes, those runs are scenarios by construction. Do not file them as yesterday’s weather.
  • Verification belongs in a dated table: forecast issued at T, observed maximum on the valid day, source of the observation. A single anecdotal “it did well last Tuesday” is not a skill score.

The forecast versus observation article is the vocabulary. The NWS metro product article is the official public baseline this model must not impersonate.

Practical checklist

  • Heading names WXSIM (or the actual engine), not NWS and not TNET as issuer.
  • Issue time is present; retrieval time is separate if the HTML is a copy.
  • No watch or warning is implied by a model icon. Hazards go to weather.gov.
  • No recovered 2010 period high is presented as today’s Mesa weather.
  • No claim is made that TNET uses WXSIM, and no TNET formula is described.
  • No claim is made that cloud cover in the model, by itself, sets Starlink throughput.

Modern relevance

Local models still exist because official grids are not a backyard energy budget. They remain scenarios. Official public forecasts remain with NWS. Connection planning remains a different question: given atmospheric and other evidence, how might a link behave in a chosen interval.

That last question is what Connection Outlook is for. It interprets atmospheric evidence. It does not sell a weather-model franchise, does not replace WXSIM, and does not promise a speed. Use it as a labeled derived estimate. Use weather.gov as the official forecast. Use this URL only to understand what a station-run local model was.

Related historical records, including weekly station reports, sit in the observations hub and on the weekly report page.

Sources