Hierarchical modelling partitions variation in vineyard fruit maturity for optimal sampling

Abstract

Background and Aims Sampling is problematic in perennial tree and vine crops because organ numbers are vast, budgets are low and maturation of fruit is asynchronous within and between individuals. Typically, the precision and the repeatability of sampling are low and methodologies used in the field vary widely within the industry. We define a combined statistical and cost model for fruit maturity sampling that could be applied widely and would enable precision and cost to be pre-defined. Methods and Results A two-stage statistical model, incorporating a ‘cost’ element was designed. The model was evaluated using an intensive set of data acquired from Cabernet Sauvignon vines. This was decomposed into between- and within-vine sources of variance for sugar measured as TSS, anthocyanin and phenolic substances. Conclusions Within a management unit, the between-vine variation may equal or exceed the within-vine variation. The two-stage protocol, for unit cost, gave a more precise estimate of TSS than other protocols. On-site costs were the main element in apportioning vines per plot and bunches per vine. The variance for anthocyanin and phenolic substances was greater than that of TSS and thus precision was lower for these at a given sample size. Significance of the Study A simple, two-stage sampling protocol resolves uncertainties inherent in existing protocols. It may be applied widely to grapevine and other tree crops.