How the App Measures Soil Moisture

Soil moisture tells you whether a crop has water to draw on, whether the ground will carry machinery, and when irrigation would pay. The app gives every field a soil-moisture figure for today: the percentage of the top 10–20 cm of soil that is water, with a plain label underneath — Wet, Moist, Drying or Dry.

1. A daily water budget for each field

Think of the topsoil as a tank that holds about 100 mm of water. Every day:

• Rain on the field tops the tank up — and so does any irrigation you log in the Management Hub.
• Evaporation (from the soil and through the crop) draws it down. The app uses the international-standard FAO-56 reference evapotranspiration for the field's own location. While the soil is moist it evaporates at the full rate; as it dries, the soil gives up water more slowly, just as real soil does.
• Anything over full drains away.

The app runs this budget day by day over the last 92 days of the field's weather (rain and evaporation from Open-Meteo), ending today. How full the tank is becomes a moisture percentage using the field's soil type: sandy soil holds less water than loam, clay holds more. If no soil type is set, the app uses a loam range fitted across UK farmland.

2. How we know it works

We tested it against COSMOS-UK, the UK Centre for Ecology & Hydrology's network of field-scale soil-moisture stations. Each station measures the water in roughly the top 10–30 cm across a circle a few hundred metres wide — about the size of a field. We used 18 stations on arable land and improved grassland across Great Britain, over two years, and ran the model exactly as the app does, from the same weather feed.

• Correlation 0.90 with the stations' daily readings (1.0 would be perfect tracking of every rise and fall).
• Typical error about 5.5 percentage points, with almost no overall bias — inside the 6-point target we set for the app.
• One probe reading logged in the app's calibration panel tunes a field to its own soil (most of the remaining error is the soil's own wet/dry level), and three readings tighten it further.
• Only 92 days of weather are needed: starting the budget from any assumption 92 days back changes today's figure by under 2 points.

3. Why not satellite radar?

Until November 2026 the app estimated soil moisture from Sentinel-1 radar using a Water Cloud Model. Tested against the same COSMOS-UK stations, it read about 27 points too dry on average and tracked real changes poorly (correlation 0.43). We then tried the best-regarded radar method — change detection, comparing each field with its own driest and wettest radar returns, kept separate for each satellite viewing angle — and smoothed it over time. That improved the correlation to 0.69, still well short of the water budget. Feeding radar into the water budget made the budget worse, so we left it out.

The reasons are well known in the research literature: C-band radar sees only the top few centimetres, and its signal changes with crop growth, cultivation and the satellite's viewing angle as well as with moisture. For UK farmland, the field's own rain and evaporation predict soil moisture more reliably. The radar background articles in this Knowledge Hub remain as general science.

4. What it does and doesn't capture

• Irrigation only counts if it's logged. On an irrigated field, log each watering (in mm, or flow rate and duration) so the figure includes it.
• Local downpours that the weather grid misses, field drains, springs and waterlogged hollows aren't seen. The figure is for the field as a whole.
• Deeper soil responds more slowly than the topsoil. The irrigation advice converts today's figure into a root-zone water deficit using the crop's rooting depth and the soil type.
• Absolute level varies with soil: a sandy field reads a few points high and a heavy clay a few points low until its soil type is set or a probe reading is logged.