The statistical effects of incomplete sampling of coherent data series

Abstract An expression is derived for the equivalent number of independent values obtained by partial sampling of a coherent, i.e. autocorrelated, time‐series or spatial series of data. The expression is a function of both the coherence of the data values and the coherence of the sampling, i.e. the non‐randomness of the distribution of available data. For a data series with high autocorrelation the distribution of the sampling in time or space is more important than for a relatively incoherent data series. Meteorological data are used as illustrative examples. The derived expression can be used in estimating the variance of sample means.