Resumo Executivo
A per-field record built entirely from open satellite archives for the people who advise, insure and lend on US row crops: one cloud-screened look per field per week, the field’s history back to 1984 to read it against, and a plain-word status. Crop type, land cover, field boundaries, yield and management zones sit on that record, each scored against a public reference built outside Axine.
Principais Descobertas Metodológicas
A 47 ha corn pivot at Seward, Nebraska, is read across 43 of 43 seasons of open imagery.
Corn: 0.96 of held-out pixels agree with the USDA Cropland Data Layer, Iowa 2021.
Soybean: 0.95 of held-out pixels agree with the USDA Cropland Data Layer, Iowa 2021.
Every figure in a field document is traced to its source scene and to a stats file that can be checked without re-running anything.
Física e Metodologia de Inversão
Landsat and Sentinel-2 scenes back to 1984 are cloud screened and read once a week per field. Each product follows a named published method and is scored, as a count over a denominator, against a public reference independent of the method. Results are labelled by what they have earned: a record, a screening layer or a validated reading.
Citação e Identificador de Objeto Digital (DOI)
@techreport{axine_field_record_2026,
author = {Axine Labs},
title = {A weekly field record from forty-three years of open imagery: crop type, land cover, zones, yield and boundaries, measured against public truth},
institution = {Axine Labs Technical Whitepaper Series},
number = {10},
year = {2026},
doi = {10.5281/zenodo.23191338},
url = {https://axinelabs.com/papers/field-record}
}