Statistically correct determination of long-period trends in geophysical data series

Abstract

Linear trend analysis of geophysical time series is considered in connection with the problem of determining long-period variations (possibly of man-made origin) in the presence of short period noise of larger amplitude. Criteria for testing hypotheses about linear trends are presented for the cases of independent observations and of the observations whose correlation function is known. The possibility of increasing the precision of the results using area-averaged values is investigated. Examples are given of the long time series analysis of air temperature, carbon dioxide and water vapor content. Time series of total atmospheric ozone content at some stations are also considered, and conditions for correct statistical analysis of such data are given.