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:

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 with a current reading is more than about 99 miles away, or its reading is unsuitable, we use modeled data and say so on the page. Each reading shows its station, distance, date, and modeled status directly so you can judge whether it fits your local conditions.

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: the median (P50) date of the first 32°F freeze from the nearest station within 40 km. Beyond that distance no frost date is shown, because frost turns on elevation and cold-air drainage that a distant station cannot represent.

Soil Moisture: Compared Against Each Station’s Own History

Soil moisture is reported as a comparison, not as a raw number. A volumetric water content of 25 percent is wet for a sandy soil and dry for a clay, so the same figure is not comparable between two stations. For each station, the current reading is placed against that station’s own moisture readings for the same time of year, and the page reports where it falls: much drier, drier, normal, wetter, or much wetter than normal for the date, with the percentile rank alongside.

The baseline is built per station, per depth, and per calendar day from readings within 15 days on either side of that day, pooled across every year the station has reported (the most recent 5 years for USDA SCAN stations, whose probes drift over longer spans). Days when the soil was at or below 32°F are left out, because a frozen sensor reads ice rather than liquid water. A comparison is shown only when the baseline holds at least 30 readings from at least 2 distinct years; otherwise the page shows the measured reading and says the comparison is held back. The baseline is recomputed periodically, and one older than 120 days is treated as stale and withheld. There is no national average of soil moisture on this site, because averaging raw water content across different soils would not mean anything; the national map and state pages count stations by band instead.

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
• Soil Temps location estimates, clearly labeled, where a current suitable station reading is unavailable

Questions about the data? See about this site or browse the full station index.