Commodities

Practices

Potato virus Y (PVY) is a major threat to seed potato production because even low infection levels can jeopardize certification, increase roguing labor, and reduce seed market value. Current scouting relies on manual, whole-field inspections that are time-intensive and can miss early or spatially uneven infections. Although UAV-based sensing and machine learning have shown promise for detecting PVY, producers lack decision-ready tools that are reliable, and interpretable.

This project asks:

1. Can early-season PVY detection and management usefulness be improved by shifting from pixel-level predictions to patch-based field maps with calibrated uncertainty?
2. Can active or semi-supervised learning improve model robustness while reducing labeling costs?
3. How can map design and risk thresholds best support producer actions while minimizing false positives and wasted labor?

Research will be conducted with a cooperating Montana seed potato producer using UAV imagery, targeted ground validation, and machine-learning models that generate uncertainty-aware PVY risk maps. Outputs will be translated into a producer-ready workflow including flight timing guidelines, map interpretation tools, and scouting prioritization protocols.

Results will be disseminated through an on-farm demonstration/field day, a hands-on workshop for producers and agricultural professionals, and practical extension materials including factsheets and short instructional videos. Digital resources will supplement, not replace, in-person outreach.

Expected outcomes include improved early PVY detection, reduced scouting labor per acre, better protection of certification status, and greater confidence in management decisions. Environmental benefits include fewer unnecessary field passes and more targeted interventions, while social benefits include reduced stress and improved decision efficiency during critical scouting periods.

PROJECT OBJECTIVES (Research + Education)

Research Objectives

R1. By the end of Year 1, develop an operational UAV-to-map pipeline that produces georeferenced PVY risk maps and calibrated uncertainty layers for cooperating seed potato fields. (Deliverable: map products + documentation.)

R2. By the end of Year 2, improve PVY detection reliability by implementing patch-based models and active/semi-supervised learning to reduce labeling burden while improving robustness across collection dates and field conditions.

R3. By the end of Year 3, evaluate decision utility with producers by comparing (a) standard scouting vs. (b) map-guided targeted scouting: measuring time spent scouting/roguing, number of confirmed PVY detections found per hour, and producer confidence in decisions.

Education Objectives

E1. Deliver 1 on-farm demonstration/field day and 1 hands-on workshop by Year 3 to train seed potato producers/ag professionals to interpret PVY risk + uncertainty maps and integrate them into scouting/roguing plans.

E2. Produce and distribute at least 3 educational products by Year 3: (i) a 2-3 page factsheet, (ii) a flight + data collection checklist, and (iii) a "map interpretation and action guide" (plus a short recorded walkthrough).

E3. Evaluate learning and intended practice change at each outreach event using the Western SARE Survey and Evaluation Tool plus a brief pre/post knowledge check tailored to PVY decision-making.