Your weather
5 Help from nature
6 Plan your crops
Watch it grow press Run it first
What should I plant?
Crops that feed the soil
How good is this AI?
Plant Doctor
Photo of a sick plant in, name and fix out.
No photo? Pick what you see
My Farms
Time-lapse
Watch your farms over 10, 20 or 30 years.
1 Pick farms
2 Pick the future
Missions
Fix a farm in trouble. Fewer hints, more stars.
The NASA data behind the simulator
Each satellite measures a different part of the farm's story. The simulator brings them to one monthly time step for one field, so they can be compared.
Use your own field's NASA POWER data
The six built-in regions use approximate long-term monthly values. For your real field, get the data from NASA POWER:
- Enter your field's latitude and longitude, then open the link. It asks NASA POWER for monthly climatology of
T2M_MAX, T2M_MIN, PRECTOTCORR, GWETROOT. - Save the page that opens as a
.jsonfile. A CSV from the POWER Data Access Viewer also works, including monthly time series. - Load the file below. Your field becomes a new region in the Farm Lab.
Open source
Field Shift is free under the MIT licence. Read every line, change the crop numbers below without writing code, or download the whole project and make it your own. It is one HTML file, so it runs in any browser with no install.
Crop book
Your local prices, yields and limits. Changes apply at once to the simulator, Missions and Time-lapse, and are kept in this browser. Edited numbers are highlighted.
Files
Common changes
- Add a crop: copy one line in
CROPSat the top ofsrc/engine.js, give it a new name and numbers, and it appears everywhere. Drawings follow itsspritetype and colours. - Add your region: add an entry to
REGIONSinsrc/engine.jswith 12 monthly values from NASA POWER, or simply load a POWER file in the NASA Data tab. - Change the look: colours and fonts are the variables at the top of the
<style>block inindex.html. - Retrain the AI: run
node training/gen.js 6000to make training data, thenpython training/train.pyto write a newsrc/model.json, thenpython build.py.