Macadamia kernel quality estimates are of fundamental importance to underst
anding tree responses to many experimental treatments and orchard managemen
t protocols. Experimental measures of macadamia kernel quality, collected u
nder field conditions, traditionally rely on the average of 100 fruit, samp
led from the estimated peak in fruit drop. To detect changes in kernel qual
ity over a single season, we measured variation in fruit quality of macadam
ia cv. 344. To sample this variation we measured 10 fruit from 6 blocks of
3 trees at each of 7 sites, over 4 harvests made at monthly intervals. For
all fruit collected we determined: husk, shell and kernel dry weights; kern
el recovery (the percentage of kernel to kernel and shell weight); and kern
el specific gravity from which oil content was estimated. A split-plot anal
ysis of variance model was used to determine variance estimates for each of
the fruit quality parameters measured. The percentages of partitioned tota
l variance of the quality parameters were lowest for sites (3.6-6.7%), inte
rmediate for harvests (3.3-41.1%) and highest for fruit (32.9-71.2%). Using
these estimates, we investigated the influence of varying the number of re
plicates per site and fruits per replicate on kernel quality estimates. The
analyses indicated that samples of 5 fruit, from each of 6 randomly locate
d blocks within a site, represent the minimum replication required to detec
t commercially relevant changes in the kernel quality parameters measured.
Larger sample sizes and increased replication did not significantly increas
e the precision of estimates.