Sbir phase i: development of AI and software platform for cost-effective soil carbon measurement and assessment

This $255,858 Project Grant from the National Science Foundation will support the development of machine learning models and a software and data platform for cost-effective soil carbon measurement and assessment. Funded under the NSF's Engineering program (CFDA 47.041), the award recipient Yard Stick PBC will develop novel digital pedology techniques to enable accurate in-situ soil carbon predictions. Specifically, the company will create new machine learning algorithms for soil carbon analysis, along with an integrated software and probe system and cloud-based data management platform. This will allow for automated, scalable, and authoritative soil sampling plans tailored for landowner needs and parcel characteristics. The outputs are intended to facilitate regional soil carbon modeling and adaptive sampling methodologies. The period of performance is from April 1, 2022 to December 31, 2022.

Abstract

The broader impact/ commercial potential of this small business innovation research (sbir) phase i project is to enable accurate soil health measurements. The proposed project will develop new machine learning (ML) models for accurate soil carbon predictions. The system will generate a statistically- and agronomically-significant soil carbon sampling design customizable for landowner preferences and the parcel?s characteristics. The envisioned system will integrate a handheld measurement probe and a cloud-based data analysis and management platform for automated, scalable, cost-effective, and authoritative soil carbon sampling. The proposed project will develop a novel machine learning (ML) model for the in-situ prediction and analysis of soil carbon and development of automated sampling plans. The proposed system integrates a soil stratification and sampling plan design algorithm; software integrated with a handheld probe; and an integrated cloud-based analysis and data management platform. The research will address technical challenges including: (1) detecting complex covariates; (2) aligning stratification with soil carbon verification protocols; (3) developing novel digital pedology techniques; and (4) developing machine learning tools able to iterate measurement plans in near real-time. The system's outputs will enable parallel regional models and adaptive sampling plans. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.