GroundTruth

Inspiration

Crop stress becomes visible to farmers 1-2 weeks after it starts. By then, yield loss is already happening. Satellites can detect stress earlier through changes in soil moisture, temperature, and light conditions, but this data is locked away in complex NASA databases that farmers can't access or interpret. We were inspired by the "Bridging Worlds" theme to build a system that bridges the gap between space-based Earth observation and ground-level agricultural decisions.

Our goal: to bridge the gap between farmers and the huge quantity of raw data available from NASA's databases.

What it does

GroundTruth is an agricultural early warning system that monitors crop health using NASA satellite data. Farmers input their field location and crop type. It pulls real-time data from NASA POWER satellites (soil moisture, temperature, precipitation, light levels) and analyzes trends over the past 30 days using statistical models to detect declining conditions. It then predicts future stress by projecting current trends forward and generating specific recommendations with urgency levels and actionable steps.

How we built it

We use an API call to the NASA Power database based on the most relevant parameters for farming. We use a provided post code which we translate to latitude and longitude. We then use some statistical modeling to make predictions into the future. We finally hand this to Claude LLM API to break down the implication in an easily understandable way.

Challenges we ran into

One challenge was not having direct communication with the end user as to what specifically farmer's would find most helpful in terms of parameters from the NASA DB. We had to make this interpretation ourselves.

Accomplishments that we're proud of

We are proud of how we divided work according to strengths and tackled our challenges in the most efficient way possible.

What we learned

There is a huge amount of data out there that could be useful for various industries, yet not everyone knows it is available or has access to it. LLMs can bridge that gap.