Methodology
Every primary reading on this site comes from a physical sensor buried in the ground at a federal monitoring station. This page explains where the data comes from, what it means when a reading is measured rather than modeled, and how the numbers are processed between the sensor and your screen.
How We Measure
Readings come from three federal sensor networks, with soil temperature measured by in-ground probes at five standardized depths:
| Network | Operator | Stations | Depths |
|---|---|---|---|
| SCAN (Soil Climate Analysis Network) | USDA NRCS | ~210 | 2", 4", 8", 20", 40" |
| CSCAN (Cooperative SCAN) | USDA NRCS | ~21 | 2", 4", 8", 20", 40" |
| USCRN (US Climate Reference Network) | NOAA | ~150 | 5, 10, 20, 50, 100 cm |
USCRN metric depths map to the same 2/4/8/20/40-inch reporting scale. New observations are pulled nightly from USDA AWDB and NOAA NCEI feeds, stored, and served alongside each station’s history. Every city page names its source station and the distance to it, so you always know where the number was measured.
Measured vs Modeled
There are two ways a website can tell you a soil temperature. It can report what a sensor in the ground measured, or it can estimate the value from a weather model: a reanalysis grid that infers soil conditions from air temperature, radiation, and precipitation over cells several miles across. Model estimates are useful, and we use them too, but they are estimates. They smooth over soil type, drainage, shade, and microclimate, the factors that make your yard differ from a grid-cell average, and the popular reanalysis layers blend the top several centimeters rather than reading a true 2-inch depth.
Our rule: stations first, models only as a labeled fallback. When your location is within range of a station (most of the country), the reading is a measurement. When the nearest station is more than about 75 miles away, we fall back to Open-Meteo modeled data, say so on the page, and mark the reading with low confidence. The confidence tier shown with every reading reflects exactly this distinction.
Historical Averages
Long-term comparisons (typical planting windows, seasonal averages, the average line on charts) come from per-station climatology: every recorded reading at a station is aggregated by calendar day and depth, producing a day-of-year average built from that station’s own measured history rather than a modeled reconstruction. Planting windows additionally use NOAA 1991–2020 frost normals.
Update Cadence
- Station observations refresh nightly (USDA publishes with about a one-day lag; NOAA USCRN similarly).
- Readings older than three days are flagged stale and the page says so rather than presenting an old number as current.
- The national heatmap image regenerates daily from the latest station readings.
Limitations
Honest data requires honest caveats. A station miles away is still not your backyard: full-sun sandy soil runs warmer than shaded clay in the same town. Sensor outages happen, and dead stations are retired from the site rather than padded over. For decisions at the margin, a $15 soil thermometer in your own lawn is the final word; this site tells you when it is worth walking outside to check.
Sources
- USDA NRCS Soil Climate Analysis Network
- NOAA US Climate Reference Network
- Open-Meteo (modeled fallback only)
Questions about the data? See about this site or browse the full station index.