Sign inRegisterCheck connection

Third-Party Radar Mosaics Are Not a WSR-88D Volume

Third-party radar mosaics from the Weather Underground era: provenance, latency, and why a stitched consumer map is not a WSR-88D volume.

Back to Historical weather observations and forecasts

A national radar picture on a consumer weather site is usually a mosaic: many WSR-88D (and sometimes TDWR) products, remapped, timed, filtered, and painted on one map. That object is useful. It is not a NEXRAD volume. The historical TNET Weather URL /radar/wunderground existed because station pages of that era often hot-linked or framed Weather Underground radar. This article is about provenance and latency in that class of third-party mosaic. It is not a Weather Underground product page, not a screenshot gallery, and not a current loop.

TNET does not rehost Weather Underground imagery or copy its branding. For official U.S. radar, use NWS Radar, which provides both single-site products and a national mosaic with NWS attribution. TNET is not a warning service.

Historical context

PWS templates in the 2000s needed a colorful map that non-specialists recognized. Weather Underground (WU) was a widely linked consumer network: station IDs, personal pages, and a radar map that looked like “the weather.” Operators pointed img tags at WU radar endpoints or told visitors to click through. The recovered TNET Doppler page from April 2005 even labeled Arizona loops as coming from Weather Underground, with separate “ground clutter removed” and “scientific” views. Those captures are historical. They are not today’s mosaic, not KIWA’s current volume, and not restored here.

The scientific remainder of /radar/wunderground is the mosaic problem that outlived any one brand: once you leave a single antenna, you must say who stitched the map, when each contributing volume was valid, and what processing sits between the pulse and the pixel.

What a mosaic is

A WSR-88D volume is polar. Gates along a radial, several elevation angles, a few minutes per volume, beam height rising with range. NCEI archives that Level-II base data and the Level-III products generated from it. A display client such as GRLevel3 typically shows one site’s Level III product at a time (Level III and GRLevel3).

A mosaic is a derived map:

  1. Select contributing radars (CONUS NEXRAD, sometimes adjacent Canadian radars, sometimes airport TDWRs).
  2. Convert polar products to a common grid.
  3. Choose a compositing rule where beams overlap (maximum reflectivity, closest radar, hybrid scan, quality-controlled merge).
  4. Apply clutter suppression, smoothing, or color tables chosen by the publisher.
  5. Publish a PNG, GIF loop, or map tile with a legend that may or may not name those choices.

NOAA and NWS produce official mosaics (including MRMS—Multi-Radar Multi-Sensor—as a separate, documented QPE and radar-mosaic family). Consumer sites produce their own redraws. Both are mosaics. Only the official chain has an NWS product identity you can cite without guessing. A WU-era map sat in the consumer class: convenient, recognizable, and one or more processing steps away from the volume.

Provenance: whose pixel is this?

On a PWS page the honest caption answers four identities:

  • Sensor. Which radars contributed? A CONUS mosaic is not “the Phoenix radar.” KIWA may be in the blend over Mesa; Flagstaff (KFSX), Tucson, and Yuma may own pixels a short drive away. At long range, the mosaic may be showing another site’s high beam, not KIWA’s lowest tilt.
  • Product. Base reflectivity versus composite versus a precipitation-estimate layer. Consumer palettes often resemble dBZ but may be rain-rate colors. If the legend does not say, do not invent dBZ values from the PNG.
  • Processor. NWS, NOAA MRMS, a university, or a commercial weather site. “Radar” is not a source. Weather Underground, in the era this URL names, was a redistributor and renderer, not the WSR-88D program.
  • Time. Mosaic valid time is often the time of the latest tile or a rounded clock. Individual radars in the blend can be one volume older. That is latency, not animation style.

Without those four, a visitor will read a smooth green blob as “rain at my gauge now.” The gauge is still a point observation. The mosaic is a processed field. Radar as remote sensing is the physics. This page is the chain of custody.

Latency is not the same as volume time

Several clocks hide in a third-party radar image:

  • Volume time — when the contributing WSR-88D finished the tilt or product.
  • Mosaic compose time — when the publisher merged grids.
  • CDN or page time — when your img tag last fetched a file, or when your browser showed a cached GIF.
  • Station publish time — when Cumulus or Weather Display uploaded the HTML.

A “live” WU-style loop on a PWS site in 2005 was already a delayed mosaic of delayed products, then delayed again by FTP. That does not make it useless. It makes a 25-minute GIF a historical sequence of derived maps. Refreshing the page does not create new pulses. If the publisher applied “ground clutter removed,” that filter is another derived step with its own false-negative risk.

Official NWS Radar states that the site displays radar with forecasts and alerts and that radar products are also available as services. Even there, a national mosaic view is not identical to the KIWA station view. Bookmark the view you mean.

Why a mosaic is not a WSR-88D volume

Keep this list next to any consumer radar thumbnail:

  • Geometry. A volume has beam height, beam width, and missing low-level rain at range. A mosaic hides those cones inside a flat map. Overlap regions can look more complete than any single radar.
  • Time. A volume is internally consistent to a few minutes. A mosaic can mix volumes that did not exist at the same second.
  • Moments. Level II includes velocity, spectrum width, and dual-pol fields. Most public mosaics show a single reflectivity-like layer.
  • Algorithms. QPE, storm tracks, and “future radar” on consumer maps are modelled or nowcast layers. They are not observed gates.
  • Clutter policy. One publisher’s “clean” mosaic is another’s missed AP; the 2005 TNET wording already split clutter-removed versus scientific views, which is the right instinct if both are labeled.

Use a mosaic to see whether precipitating echoes exist on a synoptic scale. Use a named site product to see how one antenna sampled the column. Use a rain gauge to see what reached the ground at one point. Three evidence types.

Practical rules if you still embed a third-party map

  1. Prefer linking radar.weather.gov over hot-linking a commercial tile you do not control.
  2. If a license still allows an embed, attribute the publisher and NOAA/NWS as the underlying radar program.
  3. Caption “mosaic,” not “Doppler from our station.”
  4. State that times may lag the latest volume.
  5. Do not crop the legend or the timestamp.
  6. Do not paste another company’s screenshots onto TNET. Describe the class of product; do not reproduce the UI.
  7. Do not treat the mosaic as satellite. GOES stays on GOES-West.

Local Phoenix geometry—why Flagstaff-to-Nogales on one GIF was already a beam-height cartoon—belongs on Phoenix-area NEXRAD.

Modern relevance

Connection and weather research depend on source identity: which dataset, which time, which processing. That is the same discipline as refusing to call a WU mosaic a WSR-88D volume. TNET’s public note on source identity is data sources, quality controls, and methodology. Atmospheric precipitation structure, as a family of weather evidence rather than as a branded map, is grouped on evidence families.

The Horizon study review is independent research context for weather and Starlink. It does not convert a consumer radar GIF into a speed forecast. How TNET labels observed versus derived information is how the service works.

This URL does not display Weather Underground radar. Historical observation pages: observations hub.

Sources