The field is now being asked to report its own harvest
the field is now being asked to report its own harvest: a nasa and ibm earth model fine tuned on canola imagery predicts bushels per acre at 92 percent accuracy, a new benchmark pairs harvester records with sentinel 2 and soil data across four countries and nine years, and a soybean model needs nothing but routine weather.
the field now files its own yield forecast.
Context
IBM Research says that IBM and NASA's Prithvi-EO-2.0 models, fine-tuned on high-resolution images of canola crop growth, predicted with 92% accuracy how many bushels per acre the mid-summer crop would yield. The page read does not define the accuracy measure.
YieldSAT's project page describes a multimodal dataset for crop yield prediction at field and subfield (pixel) levels that combines combine harvester yield data, Sentinel-2 time series, weather, soil and topography across 4 countries, 4 crop types and 9 years (2016 to 2024). A September 29, 2026 ScienceOpen paper presents a deep learning framework that uses routine weather as the only time-varying input, plus two static inputs (crop year and an agro-environmental label), tested on 20 Brazilian soybean seasons.
The 92% belongs to the IBM and NASA canola work, and the page read gives no definition of accuracy, error bars or a test split, so it cannot be compared with the other two studies. The YieldSAT and soybean items are separate projects with different crops and scales.
The four countries and nine years match the YieldSAT page. The soybean model's claim of using only routine weather needs a qualifier: the paper says routine weather is the only time-varying input, with two static context inputs added. It is a national forecast for Brazil, and its best model, a Transformer, reports RMSE 149 kg per hectare and R squared 0.784, which is not a field-level result.
'The field now files its own yield forecast' is the author's line. The three sources are a model, a dataset and a benchmarked forecast study, and none of them describes a farm filing a forecast on its own.
Related work
- IBM wins a NASA Group Achievement Award (IBM Research, February 9, 2026) ↗Source for the canola result, the Prithvi-EO-2.0 models and the 92% figure.
- YieldSAT: A Multimodal Benchmark Dataset for High-Resolution Crop Yield Prediction (CVPR 2026) ↗Source for the dataset's data types, countries, crops and years.
- An Input-Frugal Deep Learning Framework for Weather-Driven National Crop-Yield Forecasting: A Case Study of Brazilian Soybean (ScienceOpen, September 29, 2026) ↗Source for the weather-only inputs and the Brazil soybean results.
Watch next
- Find the full canola study behind the IBM post and its accuracy definition. Compare YieldSAT's field-level errors with the national soybean errors.
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 20:32 IST.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →