ChescoWx · East Nantmeal, Chester County PA · 2021–2026

Equipment & Siting: How Much Do They Affect Accuracy?

Fan-aspirated vs passive radiation shield · Rooftop vs ground-level placement

Three stations have run side by side at East Nantmeal: a WeatherFlow Tempest on the roof, a Davis Vantage Pro 2 with its sensor suite in the yard the way the NWS recommends, and an Ambient Weather WS-2902 on a 5-foot pole nearby. A manual CoCoRaHS rain gauge sits alongside them. Years of paired readings show where the equipment matters, where the siting matters, and where neither matters much.

WeatherFlow Tempest
  • All-in-one unit on the house roof
  • Passive (unventilated) radiation shield
  • Haptic rain sensor: no bucket, it senses drop impacts
  • Sonic anemometer, no moving parts
Davis Vantage Pro 2
  • Sensor suite on a post in the open yard
  • Fan-aspirated radiation shield (air pulled across the sensor)
  • Tipping-bucket rain gauge
  • Wind from a Vantage Vue anemometer, also on the roof
Ambient WS-2902
  • All-in-one array on a 5 ft pole near the Davis
  • Passive (unventilated) radiation shield, like the Tempest
  • Temperature only is usable; its rain gauge failed in late 2020
  • History from Weather Underground monthly summaries
References
  • Rain: CoCoRaHS 4-inch manual gauge, read daily
  • Wind: Chester County Airport (KMQS), open exposure
  • Temperature: the aspirated Davis, the better-sited of the two
MeasurementTempest (roof)DavisWhat drives the difference
Daily high, sunny & calm days+3.3°F warmreferenceRoof + sun
Daily high, breezy days (2–3 mph)+0.0 to +0.3°FreferenceAgree
Hottest reading each summer month (2021–23)+2.8°FreferenceRoof, not shield
Daily low+0.5°F on averagereferenceRoof height
90°F days, 2021–2026458Roof + sun
Rain (liquid), share of manual gauge100%86–88%Roof no penalty
Snow and ice, share of manual gauge58%44%Both fail
Average wind, share of airport~30%~6–10%Trees + roof
Temperature · 1,488 paired days

The Tempest runs warm in the afternoon sun, and only then

Both stations were reduced the same way, from one-minute samples. On an average day the Tempest's high is 1.2°F warmer than the Davis and its low 0.5°F warmer. That average hides the real story: the gap depends almost entirely on the season and the weather.

Tempest minus Davis, by month

Daily highDaily low

Highs run +2 to +3°F warm from May to October, when the sun is strong and the air is calm. In March and April, the windiest months, the Tempest high is actually slightly cooler.

Daily-high difference by sun and wind

Rows: the day's peak sunshine at the Tempest (W/m²). Columns: the Tempest's mean wind (mph). Each cell is the average high difference; small type is the number of days. Cells with fewer than 15 days are left out.

Why: sun, still air, and a hot roof

The warm bias appears only when the sun is strong and the air is still. That pattern fits two causes. A passive shield heats up in the sun with no fan to carry the heat away, and a roof heats far more than a lawn and warms the air just above it. On the Tempest both happen at once, so on its own this comparison cannot say which matters more. The third station below answers that.

A third station · 45 complete months, 2021–2026

The deciding test: a passive shield on the ground

The Ambient WS-2902 uses a passive shield much like the Tempest's, but it sits on a 5-foot pole near the Davis instead of on the roof. If the shield were the main problem, the Ambient would run warm in summer too. It did not. Across nine complete summer months from 2021 to 2023, its hottest reading of each month matched the Davis within 1°F every time, averaging +0.0°F. The rooftop Tempest averaged +2.8°F in the same months.

Ambient (ground, passive)Tempest (roof, passive)

Hottest reading of each month minus the Davis, °F. Complete summer months only.

Ambient minus Davis, average of the monthly hottest reading, May–September, °F. From March 2024 its summer highs jumped 10–12°F too warm and stayed there. Its daily records from the past year show why. On cloudy days it still matched the Davis (+0.6°F), so the sensor itself is fine. On sunny spring and summer days it ran +6.8°F warm, and on about a third of them its "high" came between 8 and 11 AM, when air temperature is still rising. That points to morning sun reaching the sensor through a damaged shield. A passive shield at ground level is only as good as its condition.

These are monthly summaries from Weather Underground, not day-by-day pairs, and months Weather Underground marks as incomplete are left out. Monthly averages agreed as closely: Ambient −0.2°F against the Davis in 2021–23 summers, Tempest +1.0°F.

45days at 90°F or hotter on the Tempest
8on the Davis, same days, same yard

This is how a warm-biased station inflates a local climate record. A single fair-weather bias shows up most on exactly the days people remember and compare.

The summer gap is growing

June to September, sunny (900+ W/m²) and light-wind days only. The Tempest's warm bias grew from about +1.8°F in its first summer to over +3°F by 2025. Weathering or dust on the shield, or a change on the roof around it, would do this. Cleaning the shield would test the first.

Overnight lows: the roof's own effect

Lows show a smaller, different pattern. On breezy nights, especially in the cold season, the rooftop Tempest runs about 0.9°F warmer: it sits above the shallow layer of cold air that settles over the ground on the Davis. On calm winter nights the roof reverses that, and the Tempest runs 0.9°F colder, because a roof radiates heat to a clear sky faster than the ground does. Net result: the Tempest recorded 364 nights at or below freezing to the Davis's 395.

Rain · 1,629 matched days vs the manual gauge

For rain, the rooftop Tempest beat the ground station

The manual CoCoRaHS gauge is the reference. On days with liquid rain the Tempest's haptic sensor caught 100% of the gauge total over five years, year after year between 95% and 107%. The Davis tipping bucket, properly sited on the ground, caught 86–88%. Tipping buckets are known to under-read in heavy rain, and they need regular cleaning.

Share of the manual gauge, by daily rain amount

Tempest (roof)Davis bucket (ground)

Days with 0.10" or more on the manual gauge, excluding summer 2025 when the Davis bucket was clogged. 100% means a perfect match.

Where both fail: snow and ice

58%Tempest, winter precipitation days
44%Davis bucket, same days

Neither sensor can measure snow. The Tempest needs drops to strike it, and snowflakes barely register. An unheated bucket fills with snow and reports it days later as it melts, or never. On January 25, 2026, the manual gauge melted down 1.40" of water from 12.8" of snow. The Tempest recorded 0.01".

Day to day

On any single rainy day the Tempest landed within ±20% of the gauge 66% of the time, and within 0.05" half the time. It missed 7 of 351 real rain days of 0.10" or more. Light drizzle under a tenth of an inch is unreliable on every automated gauge.

Wind · 1,800 paired days

Wind is a siting problem, and both anemometers have it

Both wind sensors here are on the roof, so this comparison is about exposure. The standard is 10 meters (33 feet) above open ground. Chester County Airport meets it. A house roof among mature trees does not.

Average daily wind, mph

TempestDavis Vue

The Tempest reads about 3½ times the Davis. The Davis logged a daily average under 0.5 mph on 63% of days. The Tempest did so on 1%.

Compared with the airport

Share of Chester County Airport (KMQS) average wind. Even the better sensor reads less than a third of an open-field site, and less again in summer when the trees are in leaf. The Davis started at about 20% and has fallen to under 10% as its bearings wear. That part is equipment.

Gusts tell the same story

Peak gusts are less affected by sheltering than the average. The two stations' typical daily peak gust is almost the same, about 10–12 mph. But the Tempest has recorded gusts to 51 mph, and the Davis never above 32 mph in the same period. The worn cup anemometer cannot spin up fast enough to catch the strongest bursts.

For anyone choosing or reading a home station

What five years at one site says about accuracy

Temperature: location first, then upkeep

Placement mattered more than the shield. An inexpensive passive shield on a 5-foot pole over grass matched the fan-aspirated Davis in summer, while the same kind of shield on a roof read about 3°F too warm on sunny, calm days. A shield must also be kept in good condition: when the Ambient's failed, its highs ran 10°F or more too warm. Night lows are much less sensitive to either.

Rain: equipment and upkeep beat placement

The rooftop haptic sensor matched a manual gauge closely for liquid rain. The ground-level bucket read low and failed outright when it clogged. A manual gauge is still the only reliable measure of snow's water content. Automated totals in winter should be treated with caution.

Wind: exposure beats everything

No sensor fixes a sheltered location. Trees and buildings cut these readings to between 6% and 30% of the airport's, and the loss changes with the season and wind direction, so no single correction factor can repair it. Wind needs height and clear surroundings.

Limits of this comparison

  • This is one site with one of each instrument. Another Tempest, or a Tempest on a pole over grass, may behave differently.
  • For temperature the Davis is the reference, not certified truth. Its aspirated shield is the recognized standard, and the Ambient's independent agreement with it in 2021–2023 supports its readings.
  • The Ambient comparison uses Weather Underground monthly summaries (each month's hottest, coldest and average reading), not paired daily values. It separates roof from shield only for 2021–2023, while its shield was sound, and it assumes the two passive shields perform alike.
  • After August 2021 the airport wind reference is estimated from Philadelphia's record, scaled by month to match Chester County Airport. That estimate matches the real airport record closely where they overlap.
  • Days when either station was frozen, offline or under 20 hours of data were excluded.