SoilSight: A Systems-Based Approach to Soil Organic Carbon Estimation

Overview

Have you ever wondered what’s in your soil? Farmers do all the time! However, costly and time consuming measurement techniques make understanding soil and the effect of agricultural management practices difficult. To address this, I developed SoilSight, a machine learning model. It predicts Soil Organic Carbon (SOC), a key indicator of soil health and fertility, from the ratio of light reflected back from a sample. This ratio is known as reflectance. To build SoilSight, I collected 26 soil samples, obtained SOC values from combustion, and measured the associated reflectance using a low-cost reflectometer. I then trained, tested, and tuned 18 models evaluating their individual accuracy in predicting SOC. The SoilSight model is accurate (MAE = 1.40) and applicable (AUC = 0.92) at the field level. My next steps are developing an API of SoilSight, allowing for immediate soil analysis and informed sustainable agricultural management practices.

Awards (3)

• Challenge Award
• Bronze Medal
• Selected for CWSF 2026

Competition history

• CWSF 2026 Agriculture, Fisheries & Food

Related projects

Closest projects by meaning, across every fair and year in the corpus.