The role of temperature control in the kiwifruit cool chain revealed by artificial intelligence

Zespri, the world largest marketer of kiwifruit, exports to more than 50 countries, aiming to ensure that the product meets the different offshore market quality standards. After harvesting, many different stages in the cold chain, including temperature control during shipping, can affect the fruit quality in different ways. Despite the huge variety of factors there is limited sampling of fruit in pallets making it challenging for Zespri to identify the key stages that affect fruit quality and thus address them appropriately. This work is motivated by the need to develop a data-driven solution for this business challenge, which will allow Zespri to improve their decision process to improve the kiwifruit quality that reaches all offshore markets.

Previous attempts at understanding how the different stages in the cool chain can impact fruit quality had limited success. A full picture based on quantitative, data-driven analysis is yet to emerge. Here I present a novel approach to address this problem. Making use of data from the 2018, 2019 and 2020 seasons I incorporate state-of-the-art machine learning models, time series analysis, Bayesian Networks and a mechanistic fruit softening model to reveal how the different stages in the cool chain impact fruit quality.

I demonstrate that it is possible to link the fruit quality outcomes in offshore markets in terms of cold chain properties in its different stages, and also identify the key cold chain properties that are most responsible for affecting fruit quality metrics.

The outcomes of this research allow Zespri to plan to allocate resources efficiently in the cold chain stages that are most critical for the market/fruit variety/period of interest to address fruit quality outcomes. More importantly, it constitutes the backbone knowledge over which Zespri can develop a new decision process for its Vessel Assessment Committee, using a robust data-driven quantitative foundation.

Alvaro Orsi. Principal Research Scientist at PlantTech. More than 12 years of experience in Scientific computing, AI and Machine learning applied to address both scientific and data-driven industry challenges. The former by pursuing an academic career as a Computational Cosmologist in research centres in the UK, Chile and Spain. The latter since 2019 as a Principal Scientist at PlantTech in New Zealand, where I find solutions for the horticulture industry through artificial intelligence technology.

I have published over 50 articles in top scientific astrophysical journals, and mentored several postgraduate students (Masters and PhDs). I have also led scientific groups in large international collaborations and organised international scientific meetings.