Articles
Mango shelf-life modelling
Article number
1364_19
Pages
151 – 156
Language
English
Abstract
Short shelf-life of fresh produce can disappoint wholesalers, retailers, and consumers, damage reputations, and lessen repeat purchases.
With contemporary technologies, monitoring supply chain conditions from farm to retail is accessible to supply chain stakeholders.
Thereby, agile decision-making in a first expired first out context is enabled by real time data collection.
Remaining shelf-life (RSL) models were developed for mango fruit using data collected in laboratory trials simulating real-world export air and sea freight conditions.
Storage duration and temperature matrices covering a range of potential export scenarios and conditions were applied for R2E2 and Kensington Pride (KP) mango cultivars.
Harvest time and a phytosanitary treatment were also considered.
Models were trained and validated with split data set (70 and 30%, respectively) and verified with real world shipment monitoring data.
Simulation experiments showed that specific regression models were required to account for differences between cultivars and between harvest times (early or late). Shelf-life prediction intervals (PI) at the 90% confidence level were ±3.2 days for R2E2 and ±2.8 days for KP, respectively.
Root mean square errors (RMSE) at 90% were 4.1 days for R2E2′ and 4.8 days for KP. Dry matter content at harvest as a co-variable inconsistently improved RSL predictions.
A web-based dashboard and a mobile phone application were developed to demonstrate RSL modelling to industry.
With contemporary technologies, monitoring supply chain conditions from farm to retail is accessible to supply chain stakeholders.
Thereby, agile decision-making in a first expired first out context is enabled by real time data collection.
Remaining shelf-life (RSL) models were developed for mango fruit using data collected in laboratory trials simulating real-world export air and sea freight conditions.
Storage duration and temperature matrices covering a range of potential export scenarios and conditions were applied for R2E2 and Kensington Pride (KP) mango cultivars.
Harvest time and a phytosanitary treatment were also considered.
Models were trained and validated with split data set (70 and 30%, respectively) and verified with real world shipment monitoring data.
Simulation experiments showed that specific regression models were required to account for differences between cultivars and between harvest times (early or late). Shelf-life prediction intervals (PI) at the 90% confidence level were ±3.2 days for R2E2 and ±2.8 days for KP, respectively.
Root mean square errors (RMSE) at 90% were 4.1 days for R2E2′ and 4.8 days for KP. Dry matter content at harvest as a co-variable inconsistently improved RSL predictions.
A web-based dashboard and a mobile phone application were developed to demonstrate RSL modelling to industry.
Authors
Y. Chen, D.C. Joyce, N.A. White, P. Bryant, A. Valida, H.V. Duong, A.J. Macnish, L. Smith
Keywords
dry matter, harvest time, postharvest, prediction, supply chain
Online Articles (60)