Frequently Asked Questions

To become a user, you need to send an email to info@cropt.ag and schedule a meeting to determine which usage options and functionality packages best suit your needs. Don’t worry, meetings are organized quickly and take very little time.

All data generated through the application of AI models is available both via the Space Garden platform and through an API. This means the data can be integrated into your own platform or existing information system.

Absolutely. Cropt is ready to meet all your requirements related to agricultural production monitoring, predictive algorithms, satellite detection, and more.

Soil data is based on the SoilGrids platform maintained by the World Soil Organization. Climate data comes from the EU Copernicus Climate Change Service (C3S), satellite data is sourced from the European Space Agency, and agricultural production data comes from internal databases.

We use AI algorithms trained on more than 13,000 field data points related to yields, varieties/hybrids, agronomic practices, etc. The models have an error margin below 10%, but larger deviations are possible due to hyperlocal effects such as plant diseases, waterlogging, variable productivity zones within a field, non-standard seeds or sowing dates, and specific agronomic operations.

The classification accuracy is around 94%, but this figure can vary from year to year, region to region, and crop
to crop.

The soil data is not generated by Cropt but comes from third-party sources. Based on experience, we estimate its accuracy at around 70–80%.

Initial predictions can be provided at the beginning of the season. These predictions are updated every two weeks, and the accuracy increases as the season progresses.

Winter crops can be identified in March. After the flowering of rapeseed in April, wheat and rapeseed can be distinguished. For spring crops (corn, soybeans, sunflower, sugar beet), more satellite images are required, so detection is possible from mid-July to the end of August.

For each historical yield data point, corresponding soil, elevation, and climate data are extracted. These inputs are used together with satellite data to train the prediction algorithm. When predicting yield for a new location or season, the trained algorithm makes predictions based on available soil, elevation, climate measurements, and satellite imagery.

Relevant spectral bands (near-infrared, shortwave, visible light…) are extracted, and cloud and shadow corrections are applied. Various vegetation indices such as NDVI, NDWI, and others are calculated to assess vegetation status and changes over time. AI models are then applied to these indices to generate desired outputs (e.g., crop types, detected damage, etc.).

Yield predictions are updated every two weeks. During this time, new climate data is collected. Based on this, predictions for the remainder of the season—and thus for harvest yield—are adjusted. Crop maps can also change during the season, as more satellite images help the AI algorithm to achieve a higher precision.

Artificial Intelligence is a field of computer science that develops systems capable of learning, reasoning, and problem-solving similar to human intelligence. In our system, AI connects input parameters such as soil, climate, and satellite data with outputs like yield.

Variance measures yield stability, indicating how much values fluctuate across seasons. Risk refers to the minimum expected yield, i.e., the value below which there is only a 5% chance the yield will fall.

Seed distribution maps are based on three criteria: yield (maximum potential), stability (minimal variation across climates), and performance (a combination of yield and stability). Based on these, three maps show which variety/hybrid performs best for each criterion, using color to indicate the optimal choice at each location. Additionally, for each variety/hybrid, three more maps highlight where it ranks among the top 3 for each criterion.

This is an assessment of a plot's expected contribution compared to others of the same crop. For example, a score of 10 is assigned to corn fields expected to produce the highest yields relative to other corn plots.