Field-level agricultural frost alerts

A 2 m minimum temperature is not enough

Agricultural frost damage forms near the canopy, while a standard forecast at two metres may still read above freezing. On clear, calm nights the surface radiates heat quickly and cooled air drains into lower ground. Several degrees of difference can appear within the same field.

Meteon interprets air temperature together with dew point, cloud cover, wind, slope, exposure and hollows prone to cold-air pooling. The field-level output shows where and when risk rises, and which zones need closer checks.

Using frost alerts before you decide

1. Record an accurate field boundary. Terrain correction only helps when the drawn boundary matches the real area.
2. Judge timing and uncertainty together. Alongside the minimum, the cooling onset, expected duration and model agreement matter.
3. Check the riskiest zone. On-site measurement is most informative in the lower ground or where damage occurred before.
4. Account for phenology. The same temperature means different risk in dormancy, budburst and bloom.

Measured accuracy

We validated temperature estimates against ten Hungarian stations over 82,046 hours: mean error 1.04 °C, with 87.4% of values within ±2 °C.

Frost signals were checked on 14,952 station-days at 36 Hungarian stations, by season. In the spring bloom window the system flagged 92% of 635 frost days, with 17% of alerts false. Autumn is weaker: the 9 km grid misses many first frosts, so fewer frost days get a signal.

These are station-level results against 2 m air temperature with reanalysis inputs — they do not include live forecast error. Grass-level and within-field maps are estimates from a physical terrain and surface model, offered as decision support; they do not replace on-site measurement. Measurement and data-source background is on the data sources page.