Homogenisation and trend detection analysis of broken series of solar UV-B data

Summary

Statistical techniques have been developed to homogenise a broken series of clear-sky solar UV-B radiation, measured by a Robertson Berger (RB) meter over the period 1981–90 at Invercargill, New Zealand, and to analyse the series for long term trend. Statistical modelling of the quasi-linear UV-B/ozone relationship evident in the departures of daily clear-sky UV-B data and coincident satellite ozone data from their respective mean references has been used to provide a self-consistent de-seasonalised data set of UV-B and ozone departures, and to bridge a major gap in calibration that separates the data set into two periods, 1981–86 and 1988–90. The choice of UV-B reference is important to the quality of the results and particular attention was given to the methodology for defining it. Four alternative objective adjustment procedures for calibrating the 1988–90 period against the 1981–86 period were examined. Because our interest lies primarily in the higher values of summer, a UV-B-weighted procedure was chosen. The modelling and homogenisation techniques developed may have application in related analysis problems. Analysis of the data for the independent 1981–86 period showed large trends in ozone and UV-B, but this was mostly due to a period of very low ozone values during 1985. Over the whole period, 1981–90, the ozone trend was −4.7%/decade. The corresponding UV-B trend was +5.8%/decade, but this result is not independent of the corresponding ozone trend because the homogenisation procedure imposes the assumption of the derived UV-B/ozone relationship on the 1988–90 UV-B data. However, the evidence suggests there is little reason to doubt that solar UV-B radiation has increased at the site by about +6% per decade.