Wet-Bulb Calculator

Real-time, pressure-corrected wet-bulb temperature by ZIP, city, or GPS — with a cooling-tower approach reference, a shade-WBGT estimate, a 24-hour trend plus next-24h forecast peak, and a fully offline manual psychrometric calculator.

▶ Live demo: https://jamesccupps.github.io/Wetbulb/

Built for HVAC / facilities work: the wet-bulb temperature is the floor a cooling tower or evaporative process can chase, and it drives worker heat-stress screening. This tool computes it the way a psychrometer would, corrected for the actual station pressure — not a sea-level approximation.

Features

• Live conditions by US ZIP, city (worldwide), or one-tap GPS.
• Pressure-corrected psychrometric wet-bulb — solved iteratively, not a fixed-pressure fit, and cross-checked against the Stull (2011) approximation.
• Cooling-tower reference — theoretical minimum cold-water temperature (= ambient wet-bulb), achievable leaving-water temperature at your design approach, and margin-vs-site design wet-bulb (persisted per location).
• Shade-WBGT estimate with an honest note that it is not a measured WBGT (see Accuracy & limitations).
• 24-hour trend + next-24h outlook — wet-bulb, dry-bulb and dew-point plotted from archived + forecast hourly data, with a forecast-peak callout to stage tower capacity ahead of the afternoon peak.
• Multiple saved sites — pin the buildings you manage and glance between them.
• Barometric trend (rising / falling over 3 h) and a WB-depression readout.
• Offline manual calculator — enter air temp, RH and pressure to get wet-bulb, dew point, WB depression and cold-water temp with no network at all. Pressure follows the unit toggle (inHg under °F, hPa under °C) and is range-checked, so a mixed-up unit is caught rather than silently returning wet-bulb ≈ dry-bulb.
• Share & export — copy a shareable deep link, copy a one-line summary for a work order/ticket, or download the 24 h series as CSV.
• °F / °C toggle, light / dark / auto theme, auto-refresh, installable PWA with offline shell caching.
• Zero dependencies, no build step. Plain HTML/CSS/vanilla JS.

How the wet-bulb is computed

The wet-bulb temperature Tw is the root of the psychrometric equation

e(T, RH) = es(Tw) − γ · (T − Tw)

where

• es(·) is the saturation vapour pressure over water, using the Alduchov & Eskridge (1996) coefficients: es(T) = 6.1094 · exp(17.625·T / (243.04 + T)) hPa;
• e = es(T) · RH/100 is the actual vapour pressure;
• γ = 6.65 × 10⁻⁴ · P (hPa/°C) is the psychrometer constant at the actual station pressure P — this is the pressure correction, equivalent to the standard 0.665 × 10⁻³ · P[kPa].

Because Tw always lies between the dew point and the dry-bulb temperature, the root is bracketed and solved by bisection (60 iterations → sub-millidegree convergence). Dew point uses the exact Magnus inverse of the same es coefficients, so esat, dew point and the wet-bulb solve are all on one water phase (correct for a supercooled-wick wet-bulb below 0 °C, and consistent with Open-Meteo's over-water dew point).

The Stull (2011) empirical formula is shown as an independent cross-check. It is a sea-level fit valid for RH 5–99 % and T −20…50 °C; outside that envelope the app labels the cross-check as not applicable rather than reporting a misleading Δ.

Accuracy & limitations

• The psychrometric solve was validated against an independent ASHRAE thermodynamic wet-bulb (adiabatic-saturation) solver over T −10…45 °C × RH 10–99 % × P {1013, 850} hPa: worst deviation 0.24 °C, typically under 0.10 °C. The bias is consistently positive, as expected for a psychrometric rather than thermodynamic wet-bulb.
• Psychrometric and Stull results agree to ≈ 0.3 °C typical and ≤ 0.85 °C over ordinary conditions (T 0–45 °C, RH 20–95 %). They diverge much further in cold, dry air: at the −20 °C / 5 % RH corner of Stull's own published envelope the gap reaches ≈ 3.95 °C (7.1 °F). That is a limitation of the Stull fit, not of the value the app reports — when the Δ exceeds 1 °C the method note says so explicitly. Both bounds are pinned by the test suite.
• Results are the ambient (shade) wet-bulb from a modelled 2 m observation — not a substitute for a sling/aspirated psychrometer reading at the equipment, nor for an ASHRAE 0.4 % / 1 % / 2 % design wet-bulb when sizing a tower. Enter your site's design wet-bulb in the tower card to compare today's reading against it.
• The heat-stress badge and the "shade WBGT" figure are wet-bulb-based estimates, not a measured WBGT. OSHA/ACGIH heat-stress limits are stated in WBGT, which adds a globe-temperature (solar) load and runs several degrees higher in direct sun. Treat the shade WBGT (0.7·Tw + 0.3·Tdb) as an indoor/no-solar screening estimate only; use a measured WBGT meter for compliance decisions.
• The shade-WBGT estimate is biased low, which is the unsafe direction. The WBGT definition uses the natural wet bulb Tnwb (an unaspirated wick in ambient air); this substitutes the psychrometric wet bulb, which is what a well-aspirated sling reads. In still air Tnwb runs above Tpwb, so in a low-air-movement space — a mechanical room, an attic, an unventilated penthouse — the real WBGT is higher than the figure shown. It also substitutes dry-bulb for globe temperature, which holds only with no radiant load. Do not use it to clear someone for work near hot equipment.
• The heat-stress band edges (26 / 29 / 31 °C wet-bulb) are a screening convention, not a published standard. Only the 35 °C survivability figure is sourced (Sherwood & Huber 2010).
• Weather values are model output (Open-Meteo), not direct station observations, and update on the provider's cadence.

Data sources & attribution

• Weather (current + hourly archive/forecast) © Open-Meteo — free, no API key.
• City geocoding via the Open-Meteo Geocoding API.
• US ZIP geocoding via Zippopotam.us — free, no API key.

No API keys are used or committed; all requests are client-side over HTTPS.

Running locally

Just open index.html in a browser — everything except the installable service worker works from file://, including the live API lookups and the offline manual calculator.

To exercise the PWA / service worker (which requires an http(s) origin), serve the folder:

python -m http.server 8000
# then visit http://localhost:8000

Tests & CI

The psychrometric core (js/psychro.js) is a UMD module: the same file powers the browser and is imported by Node's built-in test runner — no bundler, no dev dependencies.

node --test          # or: npm test

CI runs the suite on Node 18/20/22 via GitHub Actions (.github/workflows/ci.yml).

Deploy to GitHub Pages

1. Push to main.
2. Settings → Pages → Source: Deploy from a branch → main / root.
3. The site is served at https://jamesccupps.github.io/Wetbulb/.

For rich link unfurls, upload assets/social-card.svg (or a PNG export of it) as the repo's social preview under Settings → General → Social preview.

Project structure

index.html               markup only
css/styles.css           styles + light/dark theming
js/psychro.js            pure psychrometric core (browser + Node, UMD)
js/app.js                UI: fetch, render, chart, persistence, events
test/psychro.test.js     node:test suite for the core
sw.js                    service worker (offline shell; APIs stay network-only)
manifest.webmanifest     PWA manifest
assets/                  icon + social card
.github/workflows/ci.yml CI

Changelog

1.1.0 — Audit fixes.

Correctness

• A corrupted or stale wb_last in localStorage no longer dead-ends the app. isFinite() coerces strings, so a stringified latitude passed the old guard and then threw in placeKey(); boot now validates lat/lon as real, in-range numbers and falls through to the default instead of stranding the user on a raw TypeError with no recovery path. Saved-site entries are filtered the same way.
• Clearing the site design wet-bulb now actually clears it. There was no delete path, so the next computeAndRender — i.e. within one auto-refresh — restored the old value and its margin callout.
• Manual-calculator pressure follows the unit toggle (inHg / hPa) and is range-checked against a plausible station-pressure envelope. Typing 1010 into the inHg field previously produced a wet bulb ≈ dry bulb with no warning.
• hourlyNowIndex scans the full hourly array instead of breaking on the first later stamp; local wall-clock stamps are not strictly increasing across the DST fall-back hour.
• Chart summary no longer prints Infinity–-Infinity when no observed hours are in range.

Accessibility

• Heat-stress badge is legible in light theme. heatStress() hardcoded two hex colors that were never theme-aware; "Caution" measured 1.29:1 and "Extreme" 1.93:1 against the badge background, and the three theme-aware bands only reached ~4.2:1. All five bands in both themes now measure ≥ 4.7:1 (WCAG AA).
• The location field's label used display:none, which drops it from the accessibility tree and leaves the field named by its placeholder. Replaced with a clip-based .sr-only.

API

• Breaking: heatStress() returns { t, level } instead of { t, c }. The core no longer emits CSS colors — callers map the level slug to a theme-aware custom property, which is what fixes the contrast bug at its root.
• Breaking: removed windChillF(); it was exported and tested but never used by the app.

Service worker

• respondWith() always settles with a real Response. A cold cache plus offline previously resolved it with undefined, surfacing as an opaque network error.
• The stale-while-revalidate refresh is registered with waitUntil() while the event is still active, so the worker is not killed before a redeploy self-heals.
• Cache version bumped to wetbulb-v2 — this release changes the psychro.js API and the CSS tokens together, so a mixed old/new shell must not be served.

Privacy

• A GPS fix is written to the URL at 2 dp (~1.1 km) instead of 4 dp (~11 m). That URL lands in the address bar, in browser history, and in whatever "Copy link" is pasted into; 2 dp is coarser than Open-Meteo's own grid, so the reading is unchanged.

Docs & tests

• Corrected the Stull accuracy claim. The documented "≈ 0.85 °C at the corners of Stull's valid range" was the maximum over the test grid (T 0–45 °C, RH 20–95 %), not over the envelope stullValid() actually admits — at −20 °C / 5 % RH the gap is ≈ 3.95 °C. Added tests that pin both the ordinary-conditions bound and the envelope-wide maximum, and the method note now flags a Δ over 1 °C in the UI.
• Documented that the shade-WBGT estimate is biased low (psychrometric wet bulb substituted for natural wet bulb), and that the heat-stress band edges below 35 °C are a screening convention rather than a published standard.
• Added the ASHRAE thermodynamic-wet-bulb validation result (worst deviation 0.24 °C).
• CSV export carries timezone, tz_abbr and utc_offset_seconds alongside the local timestamp.

1.0.0 — First public release. Extracted a testable psychrometric core with a CI'd node:test suite; added multi-site pins, next-24h forecast peak, shareable links, copy/CSV export, dew-point chart series, barometric trend, shade-WBGT clarification, light/dark theming, PWA offline, and accessibility fixes (aria-live status, accessible unit toggle & chart, focus-visible, reduced-motion). Fixed a status-bar XSS, added null-field validation, and corrected approach-input reconversion across unit changes.

License

MIT © 2026 James Cupps