Decision support systems using models of fruit quality variability

Decision support systems to assist the management of fruit through harvest, storage and distribution often rely on models of key fruit property dynamics. Because market and harvest specifications generally describe fractions of a fruit population that must meet a specification, the underpinning models must also describe that variability. Traditional models of fruit properties over time describe the behaviour of fruit sample means and do not address sample variability directly. This paper describes an improved method of modelling the dynamic aspects of fruit behaviour, using statistical descriptors of biological age and aging rate, measurement uncertainty and 'other' sources of variability. For example, instruments with significant measurement uncertainties affect the observed variability within a sample. The modelling approach uses quantile functions (not necessarily of a Normal distribution) to describe the variability components within a fruit population. This allows a decision support system to directly estimate the date when a specified fraction of the fruit will meet specification. These techniques have been proven in decision support systems operating commercially since 2007.