Engine 0.4.0 · 13 November 2026
Method
An assessment reports measured indicators, a feasibility verdict, a ranked crop screen and a written interpretation. This note describes where each comes from and where the method stops.
Sources
Soil properties are read from a global soil property model compiled from field samples and remote sensing. Climate figures are normals calculated from a reanalysis record of past daily weather. Elevation comes from a digital elevation model. Map imagery is supplied under licence by a commercial tile service, and place names by OpenStreetMap. Each indicator in a report carries its source.
The environmental sources are models. Soil is estimated at a resolution measured in hectares rather than metres, and describes the soil profile in general terms. It does not describe the condition of a particular field in a particular season.
Scoring
Each crop in the catalogue carries two ranges for the factors that determine where it can be grown: a tolerated range, outside which the crop fails, and an optimal range, within which it performs best. pH, altitude, annual rainfall and mean temperature are treated this way. Slope has a limit beyond which cultivation requires earthworks. Drainage is assessed against the crop's sensitivity to waterlogging. Annual crops also carry a seasonal water requirement, measured against the rainfall of the best matching period in the climate record.
A measured value is converted into a score for each factor according to its position relative to the range. The factor scores are combined with a limiting-factor weighting, in which the weakest factor contributes more to the result than the average. A site is therefore not made suitable by favourable readings elsewhere when one constraint rules the crop out. The result is a class, from highly suitable to not suitable, and the constraint responsible is named.
The ranges and the weighting are maintained by GEODESIA and are not published in full. Questions about a particular result can be raised through the contact form.
Safeguards
Before scoring, the boundary is resolved to a country and region so that the assessment names the actual location. A boundary over water is refused, with the reason. A boundary larger than a single farm or estate is refused. A feasibility check then considers rainfall, heat, frost, slope, soil reaction and area. Where it concludes that cultivation is not viable, no crop ranking is presented.
Interpretation
The written sections are produced by a language model from the measured values and the crop screen. The model is instructed to use the supplied figures exactly and not to introduce prices or yields. It remains a model and can misjudge emphasis. The measured values are the evidence; the text explains them.
Limits
An assessment does not establish:
- effective soil depth, stoniness, salinity or nutrient status;
- water rights, borehole yield or the reliability of an irrigation supply;
- microclimate, frost pockets, wind exposure or field-scale drainage;
- pest and disease history, or previous chemical use;
- market access, prices, labour or establishment costs;
- yield or profit.
Validation and version
The ranges are drawn from published crop requirement tables and reviewed against regions with well-documented agriculture. The engine is tested against contrasting environments: dry lowlands, humid highlands, coastal and arid sites. It has not been validated by field trials plot by plot. Where a result disagrees with observation, that is the most useful feedback we can receive.
Every assessment records the engine version and date, and every report prints them. When the method changes, the version changes. To report an error or question a result, use the contact form and quote the reference code at the top of the sheet.