The objective of this research was to determine the sample size (number of years) to estimate the ten-day average of daily sunshine duration in 30 locations of the Rio Grande do Sul State, Brazil. With sunshine data duration of the period from 1960 to 2007, 1,080 time series (30 locations x 36 ten-days) of ten-day of daily sunshine duration average were formed. The aleatory and normality, in each time series, was verified through the run test and Lilliefors test, respectively. For each ten-day and locality, it was determined the sample size using bootstrap resampling with replacement of 3,000 samples. Twenty-five years of data are enough to predict the ten-day average of daily sunshine duration, with amplitude of bootstrap confidence interval of 95%, equal to 2.00 hour day-1.
sample size; time series; bootstrap resampling; confidence interval
