Early estimation of expected fruit tree yield is important for the market planning and for
growers and exporters to plan for labour and boxes. Large variations in tree yield may be found,
posing a challenge for accurate yield estimation. We evaluated a multilevel systematic sampling
procedure for fruit yield estimation. In the Spring of 2009 we estimated the total number of fruit in
several rows in each of 14 commercial fruit orchards growing apple, kiwi, and table grapes in central
Chile. Survey times were 10-100 minutes for apples, 85 minutes for table grapes, and up to 150
minutes for kiwis. At harvest in the Fall, the fruit were counted to obtain the true yield. Yields ranged
from lows of several thousand (grape bunches), to highs of more than 40 thousand fruit (apples,
kiwis). In 11 orchards, true errors less than 10% were obtained. In two highly variable orchards we
obtained absolute true errors of about 20%. An analysis based on systematic sub-sampling of
sample data across each sampling stage was used to determine how to distribute sampling effort to
acheive the desired precision.