Noktura

Metadata Reference

Understand AgContext metadata for consistent agricultural data annotation. This reference covers BBCH growth stages, crop types, soil conditions, and more.

Last updated: February 2026

What is AgContext?

AgContext is a standardized metadata schema for agricultural imagery. By using consistent metadata fields across datasets, researchers can:

  • Compare data across different studies and locations
  • Train machine learning models on diverse, well-labeled data
  • Search and filter datasets by specific agricultural conditions
  • Ensure reproducibility in research

Tip

You don't need to fill in every field. Add what you know – partial metadata is better than none!

BBCH Growth Stages

The BBCH scale is a universal system for coding growth stages of plants. It uses a two-digit code from 00 to 99, where the first digit represents the principal growth stage and the second digit represents the secondary stage.

Principal Growth Stages

CodeStageDescription
0GerminationSeed to emergence
1Leaf developmentFirst leaves unfolding
2TilleringSide shoot formation (cereals)
3Stem elongationMain shoot development
4BootingFlag leaf to head emergence
5Heading/FloweringInflorescence emergence
6FloweringAnthesis
7Fruit developmentGrain filling
8RipeningMaturation
9SenescencePlant death/harvest

Example BBCH Codes

  • BBCH 00 – Dry seed
  • BBCH 09 – Emergence (cotyledon breaks soil)
  • BBCH 12 – 2 leaves unfolded
  • BBCH 31 – Beginning of stem elongation
  • BBCH 65 – Full flowering (50% of flowers open)
  • BBCH 89 – Fully ripe

Note

If you're unsure of the exact BBCH code, select the nearest principal stage (first digit). Noktura's interface helps you narrow down the secondary stage.

Crop Types

Select the primary crop in your images. Noktura uses GBIF taxonomy for species identification, providing scientific names and common names.

Common Crop Categories

  • Cereals – Wheat, barley, oats, rice, maize, sorghum
  • Legumes – Soybean, peas, lentils, beans, chickpeas
  • Oilseeds – Canola/rapeseed, sunflower, flax
  • Root crops – Potato, sugar beet, carrot
  • Vegetables – Lettuce, tomato, onion, cabbage
  • Fruits – Apple, grape, citrus, berries
  • Pasture/Forage – Alfalfa, clover, ryegrass

Start typing to search for your crop. The search includes both common names and scientific names (e.g., “wheat” or “Triticum aestivum”).

Soil Conditions

Document the soil visible in your images for context.

Soil Color

Select the predominant soil color in your images:

  • Dark brown/black – High organic matter
  • Brown – Typical agricultural soil
  • Red/orange – High iron content
  • Yellow – Sandy or iron-depleted
  • Grey – Waterlogged or reduced conditions
  • White/light – High calcium or salt content

Soil Texture

  • Sandy – Coarse, gritty, drains quickly
  • Loamy – Balanced mixture, ideal for agriculture
  • Clay – Fine particles, holds water
  • Silty – Smooth feel, moderate drainage

Surface Cover

Describe what covers the ground surface in your images:

  • Bare soil – No cover, tilled or fallow
  • Crop residue – Stubble, straw, or plant debris
  • Mulch – Applied organic or synthetic cover
  • Living cover – Cover crops or weeds between rows
  • Full canopy – Crop completely covers soil

Weather Data

When your images include GPS coordinates and timestamps, Noktura automatically fetches historical weather data from Open-Meteo:

  • Temperature (min, max, mean)
  • Precipitation
  • Humidity
  • Wind speed and direction
  • Cloud cover

This data is attached to your dataset automatically – no manual entry required.

Camera Metadata (EXIF)

Noktura extracts technical metadata from your images automatically:

  • Camera make and model
  • Lens information
  • Focal length
  • Aperture (f-stop)
  • Shutter speed
  • ISO sensitivity
  • Image dimensions

Tip

EXIF data is preserved from your original files. If this data is missing, check if your image editing software removes EXIF during export.

Best Practices

  • Be consistent within a dataset – use the same metadata conventions
  • Add metadata before making datasets public – it helps discoverability
  • Use BBCH codes rather than general descriptions (“flowering” is ambiguous)
  • Include multiple crop types if your field is mixed
  • Update metadata if conditions change during a time series

Next Steps