Predictive model for harvest readiness and storability of long day yellow onions

Overview

Onion growers face significant post-harvest losses as stored bulbs lose weight, texture, color, and nutritional value over time. This research delivers a data-driven predictive model that estimates the ideal harvest window and storage duration for long day yellow onions, helping producers minimize losses and deliver consistent quality to buyers. By integrating field-level growing conditions with post-harvest handling practices, the model provides lot-specific guidance that supports smarter harvest planning, storage management, and marketing decisions.

Technical specifications

Modeling approach:

• Cox regression framework adapted from survival analysis to predict the time until an onion lot loses marketability
• Marketability event defined by measurable thresholds for weight, texture, color, and nutritional value

Input features include:

• Preharvest conditions: irrigation practices, fertilizer regimes, genotype, chemical treatments, and harvest stage
• Postharvest conditions: bulb handling methods, curing protocols, storage environment, and processing treatments
• Feature selection applied to identify the most predictive variables before model training

Outputs:

• Estimated storage duration for held-out onion lots
• Risk profile indicating when a given lot is likely to fall below marketable quality thresholds

Technology readiness level

The project is in the model development and validation phase. The team will train the Cox regression model on existing datasets where marketability outcomes are known and then evaluate predictive performance on held-out samples. Future work will focus on expanding the training dataset across additional genotypes and growing regions, refining the feature set, and validating predictions in real storage environments with industry partners.