|
|
|
| Author: | A.D. Mowat |
Abstract:
Seasonal fruit compositional data was collected at six intervals between anthesis and harvest, over two seasons, from 24 New Zealand persimmon orchards.
The fruit compositional dataset, described by 42 compositional attributes (fresh and dry fruit weight, peel hue, soluble solids concentration and content and soluble tannin concentration and content, uniquely identified by six sampling dates), was processed by an unsupervised neural network, a self organising map (SOM), that clustered the orchard samples.
For data combined from both seasons, the SOM found that the orchard samples could be grouped into four clusters, where cluster one and cluster four contained samples collected in the 1991-92 growing season and cluster two and cluster three contained samples from 1990-91 growing season.
The salient characteristics that distinguished each group were determined by analysis of variance (ANOVA) of the clusters in relation to fruit compositional attributes.
For orchard samples collected in the 1991-92 season, fruit weight, soluble solids and mean air temperatures were low in comparison to the 1990-91 season.
Differences between orchard samples within each season reflected differences in fruit maturity.
For example, at 25 weeks after full-bloom, cluster two fruit were more mature than cluster three fruit, and cluster one fruit more mature than cluster four fruit.
|
Download Adobe Acrobat Reader (free software to read PDF files) |
|