How the model works

Last updated: 3 October 2026

Meteon works out the effect of the weather for your field's location, terrain, soil and crop. This page shows which data it uses, what it takes into account for each suggestion, and how accurate it is.

In three steps

Data for your field. Weather forecast, past weather, satellite images, terrain and soil — all tied to the boundary you drew.
Calculation for your field's conditions. Physical models and models trained on measured data work together to estimate what happens on your field.
Suggestion. We show what is worth checking or doing today, and why. The decision is yours.

Which data it uses

Weather forecast: hourly data from several international forecast models — temperature, rain, wind, humidity, sunshine, cloud cover.
Past weather: several years of historical weather data. The models learn from it, and we check them against it.
Weather stations: recent measurements from nearby stations pull the estimate towards measured values.
Terrain: elevation, slope, aspect and hollows. From this we work out where cold air collects and which way water runs.
Soil: texture, organic matter and water-holding capacity from a soil map.
Satellite images: Sentinel-2 images of vegetation and moisture (10–20 metre pixels), and Sentinel-1 radar images of surface wetting.
Your own data: crop, sowing date, irrigation, your own measurements and your log.

Data sources and licences →

What each calculation takes into account

Frost

What it takes into account: Night-time radiative cooling (cloud, wind, humidity), the surface energy balance, how cold air drains and settles on the terrain, and the crop's sensitivity at its growth stage.
What you get: An estimate of the grass-level minimum, the chance of frost, and the most exposed parts of the field.

Water and irrigation

What it takes into account: Evaporation (FAO-56 Penman–Monteith method), the crop's water demand at its growth stage, the soil's water-holding capacity in the root zone, fallen and expected rain, and logged irrigation.
What you get: Root-zone water, when water stress is expected, and the suggested dose and time window.

Crop condition

What it takes into account: Satellite vegetation and moisture indices (NDVI, NDRE, NDMI), compared with the field's own average.
What you get: Where the crop is weaker or stronger. The image does not show the cause — a field visit decides that.

Growth stage

What it takes into account: Daily temperature since sowing (heat units) and the crop type.
What you get: An estimate of the growth stage and when the next one is expected.

Soil temperature

What it takes into account: Surface temperature and how the soil conducts heat.
What you get: Temperature at several depths, for example in the seedbed.

Leaf wetness and spraying window

What it takes into account: Dew, rain, humidity, wind and temperature, hour by hour.
What you get: When the canopy is dry and when the weather is suitable. We do not recommend products or doses.

Zones within the field

What it takes into account: Terrain, soil and satellite imagery together.
What you get: An estimate of the differences between parts of the field.

Physics and trained models

The calculations are built on physics: energy balance, water balance, heat conduction, and how air and water move across the terrain.

We also use physics-informed neural networks, trained on our own multi-year weather archive and on measured data. A trained model is only used when it demonstrably improves the result on independent measurements.

How accurate it is

We compare the estimates with measurements from Hungarian weather stations.

92%of spring frost days were flagged (635 frost days, 36 stations)

83%of frost alerts turned out to be real frost

±1.0 °Cmean difference from measured air temperature (82,046 hours, 10 stations)

96%of temperature estimates stayed within 3 °C

The frost figures come from a retrospective test of the spring season, for air temperature at 2 metres. The uncertainty of a live forecast can be larger.

What we have checked against measurements

For each property: where the value comes from, and whether it has been compared with real measurements.

Air temperature, frost

Where it comes from: Forecast models, adjusted with nearby station measurements
Checked against measurements: Yes — against Hungarian station measurements (see the figures above)

Rainfall

Where it comes from: In Hungary, rain that has already fallen is measured data from nearby rain gauges (HungaroMet); expected rain comes from forecast models
Checked against measurements: Yes — the 7-day error of rain taken from nearby gauges is 2.6 mm (6.1 mm for the weather model), measured at 23 stations; outside Hungary rain is a model estimate

Soil moisture

Where it comes from: A water-balance calculation from rain, evaporation and the soil’s water-holding capacity
Checked against measurements: Not yet with on-site probes — you can enter your own measurement and the estimate follows it

Soil: texture, pH, organic matter

Where it comes from: An international soil map (about 250 metre resolution)
Checked against measurements: Not with on-site lab tests — indicative values; your own soil test is more accurate

Crop condition

Where it comes from: Sentinel-2 satellite imagery (the most recent cloud-free image)
Checked against measurements: A satellite measurement — the cause of a weaker patch is decided by a field visit

Estimates and measurements

Next to each value the interface shows where it comes from:

Measured: a weather station, a sensor or your own measurement.
Satellite: comes from an image, with the image date.
Model: the result of the calculation for the current state.
Forecast: applies to the coming hours and days.

Differences within the field are estimates: we work them out from terrain, the soil map and satellite imagery, and they have not yet been checked with on-site sensors. Values for the field as a whole are more reliable than the differences between its parts.

The “Model confidence” indicator shows how recent and complete the data behind the calculation is.

How it gets more accurate on your field

Set the crop and the sowing date — growth and water demand are calculated from these. Without a crop we give no irrigation suggestion, because water demand depends on the crop.
Log your irrigation — the water balance follows it.
Record your own measurements (soil moisture, temperature, rain) — the estimate follows the measured value.

Limits

The estimate does not replace a field visit or on-site measurement.
In cloudy weather there is no recent satellite image; the latest clear image is shown, with its date.
The location of local showers and thunderstorms is the hardest thing to forecast; rain on your field can differ from the estimate.
The grass-level temperature is a calculated value, without on-site measurement.

Liability in the Terms →