CSU's TerraScope turns complex soil data into decisions farmers can use
Colorado State University is building an AI-driven platform to convert fragmented soil and field data into clear, decision-ready guidance for producers. The goal: faster, cheaper, and more confident calls on irrigation, nutrient management, and long-term soil stewardship amid drought and volatile weather.
The $1 million TerraScope project is funded by the U.S. Department of Agriculture and brings together computer science, soil science, statistics, and outreach experts across CSU.
Why it matters
Soil health anchors water retention, root growth, and nutrient cycling. Yet monitoring change over time-and tying it back to specific management choices-is still hard and time-consuming.
Producers need signal, not more dashboards. TerraScope is built to turn heterogeneous inputs into a small set of trustworthy, actionable outputs.
How TerraScope works
The team will fuse on-the-ground measurements with remote sensing and couple them with advanced simulations to surface patterns that single data streams miss. Think spatiotemporal alignment of field observations, farm records, and satellite-derived indicators with process-based models.
- Inputs: field observations and farmer records, soil sampling, satellite data, weather, and topography.
- Core engine: AI models informed by simulations to detect relationships, fill gaps, and forecast trajectories.
- Outputs: decision-support prompts tied to thresholds (e.g., moisture stress risk, timing for field operations) with uncertainty estimates.
A major challenge is harmonizing formats, resolutions, and terminology-especially for variables like soil moisture. The team is building shared schemas, spatiotemporal up/downscaling, and quality-control pipelines to make the data usable at the field level.
Built with producer feedback
Usability is a first-class requirement, not an afterthought. The project includes continuous feedback loops with agricultural producers to validate data priorities, interface design, and the clarity of recommendations.
This co-development approach aims to deliver tools that fit real workflows across Colorado's diverse climates, soils, and topography-while remaining portable to other regions.
Who's involved
- Soil and Crop Sciences: Eugene Kelly, Keith Paustian
- Computer Science: Shrideep Pallickara, Sangmi Pallickara
- Statistics: Jay Breidt
- Sociology (community food systems and social sustainability): James Hale
- Integrated Rocky Mountain-region Innovation Center for Healthy Soils: research and outreach leadership by Megan Machmuller
Technical notes for researchers
- Data harmonization: ontology crosswalks for soil variables, unit normalization, provenance tracking, and versioning.
- Spatial-temporal fusion: gap-filling and resolution alignment for field observations and satellite products; handling irregular sampling intervals.
- Modeling approach: hybrid AI + simulation to keep physical constraints in the loop and improve out-of-sample generalization.
- Interpretability: feature attributions tied to agronomic factors and management actions.
- Validation: cross-site evaluation, uncertainty calibration, and field-level ground truthing with producer partners.
- Data governance: privacy for producer-provided records and transparent model updates.
What success looks like
- Higher predictive skill for soil moisture and condition indices at the field scale.
- Clear, timely recommendations that reduce unnecessary field passes and sampling.
- Improved yield stability and water-use efficiency under weather variability.
- Adoption by producers due to ease of use and trust in outputs.
Where this can go next
Beyond soil moisture and management timing, TerraScope could inform carbon accounting, drought planning, and nutrient optimization. As the platform matures, expect portable workflows that adapt to new regions and cropping systems, and tighter integration with satellite missions like NASA SMAP.
For methods, tools, and case studies at the intersection of AI and scientific work, see AI for Science & Research.
Your membership also unlocks:
