Sampling Errors in Satellite-Derived Total Ozone Trends

Abstract

In the study or ozone trends, the importance of adequate sampling of large-scale space and time variability is well recognized. However, sampling errors also result from existence of time and space variability on scales much shorter than typical averaging scales. These shorter scale sampling errors are functions of space and time variances and autocorrelations of the residuals, i.e., of total ozone data which has had the known, longer scale variations removed. This paper examines these variances and autocorrelations as well as implied trend uncertainties for a satellite observing system. The data used are from the first two years of Nimbus 4 BUV. It is estimated that the slope of a trend line using 10 annual, global averages is subject to a standard error of ±0.057 percent per decade due to sampling error. Also, it is found that Southern Hemisphere trends are subject to a larger sampling error than Northern Hemisphere trends. The smallness of these sampling errors indicates that it would be far more beneficial to increase the accuracy of satellite measurements than to increase their density, for example, by adding another satellite.