Using the Oklahoma Mesonetwork to develop and test a practical sampling error statistic for meteorological time series

Abstract

The error variance based upon sampling in time which accounts for the missing values in meteorological and climatological averages in a time series is developed. This statistic can be substituted for those already developed which assume a random or uniform distribution of discrete measurements in time. The statistic is unique from other error statistics developments published in the meteorollogical literature in that it relates the error variance of local time averages (eg, monthly) around the local population averages rather tha the long-term mean and allows for missing data or unevenly sampled data. A comparison of theoretical values with trhose derived from an autoregressive time series and a nonstationary reference data set indicate that the statistic can be appropriately used to provide researchers with a quantitative measure of the usefulness of their data.