NASA Space Apps Challenge 2026 · Dominican Republic
AUTO-FARM
NASA data for the next planting decision
A farmer-first decision-support tool: NASA Earth observations + local soil +
crop traits + farmer priorities, to compare rotation strategies.
Team: Brian, Ruben, Yoel, Oscar, Marlon, Alison · UNIROMANA
Project demo1 / 7
The problem
Farmers decide with last year's weather
Same country, different reality
In one week Azua received +103 % of its normal rainfall while Constanza
received +3 %. A national average would hide both farms.
Data does not reach the field
Satellites produce global layers; the farmer needs an answer for his plot:
what to plant, when to irrigate, what risk he faces.
Rotation choices involve water, soil health, pests, markets and
risk — a crop that helps one farm can hurt another.
Source: Open-Meteo/ERA5 (non-NASA), own 1991–2020 normal2 / 7
What we built
One screen that opens on the farm
Real satellite map
Three demo plots over real imagery, with sensors, roof and camera labelled
as simulated.
NASA observations
NASA POWER climate and risks (heat, heavy rain, dry days, frost, root-zone
moisture, evapotranspiration), SRTM terrain, MODIS NDVI, GPM IMERG.
Local soil + crop traits
ISRIC SoilGrids texture, carbon and pH (non-NASA, labelled); crop water
demand and critical months.
Farmer priorities
Sliders for water saving and diversification, plus a rainfall scenario to
see which strategy holds up.
NASA POWER · SRTM · MODIS · IMERG · ISRIC3 / 7
The core deliverable
Three rotations, compared with evidence
76/100
Alternative 1: maize → bean
93/100
Alternative 2: cassava → peanut → sorghum
- Why it is suggested, and the NASA evidence behind it (dry months of the farm).
- Benefits: legumes fix nitrogen; more crop diversity spreads water risk.
- Information gaps and risks that need local review by an agronomist.
Scores are orientation only, not an agronomic recommendation.
Rotation engine: NASA POWER climatology + crop traits4 / 7
Farmer-first design
Plain language, a phone, and a real bot
No jargon
“127 L/m²” instead of “126.8 mm”; “green 65 out of 100” instead of NDVI 0.65;
a field glossary inside the app.
Real Telegram bot
@climaxeliobot answers /estado, /clima, /techo on and free
questions using only the farm's data.
Mobile and automatic
Pinch-zoom map, automatic irrigation below 25 % soil moisture and roof
closing with real ERA5 rain.
Honest labels
Sensors, roof and camera are simulated and said so; sources marked NASA or
non-NASA.
Works without internet: self-contained demo5 / 7
Validity and limits
What we know, and what we do not
- Uncertainty shown: FDR correction over 2,592 grid cells, bootstrap
confidence intervals, adjusted p-values.
- Satellite vs ground: NASA POWER is reanalysis; MODIS pixel ~289 m and
IMERG ~2.3 km, larger than a 1 ha plot — explained on screen.
- Declared gaps: bare-field calendar, pest cycles, local market and
agronomic validation are not covered yet.
- Roadmap: SMAP, ECOSTRESS, GEDI, GRACE and Landsat are next; they need
Earthdata access and pixel harmonisation.
Support tool: it does not replace farmer judgment or local
agronomic expertise.
Every limitation is stated in the product6 / 7
Thank you
Run it, ask it, break it
Demo:
kirixdj.github.io/tierra/web/
Code: github.com/KIRIXDJ/tierra
Bot: @climaxeliobot on Telegram
Team: Brian, Ruben, Yoel, Oscar, Marlon, Alison · UNIROMANA ·
NASA Space Apps Challenge 2026, Dominican Republic
Pitch: 7 slides, 30 seconds7 / 7