published in: Meteorologicke Zpravy, vol 48 (1995), No.1, p.9-17
language: Czech + English abstract; includes 14 figures and 1 table with English captions. 

Assessment of the limits between which daily average values of total ozone can normally vary

Jaroslava Kalvova Mathematical and Physical Faculty, Charles University, Prague, Czech Republic. Martin Dubrovsky Institute of Atmospheric Physics, Hradec Kralove‚ Czech Republic

visit my home page  

ABSTRACT:
An increasing depletion of the ozone layer in the recent years has awaked a great interest of general public to be regularly informed about the state of the ozone layer. Rather than the actual values of the total ozone concentration, it is wanted to state if the deviation from some reference standard is within the frame of the common variability or it may be considered significant, either below-normal or above-normal. This type of information assumes the knowledge of the "typical" value -- e.g. mean or median -- of the total ozone concentration as well as of the variability (standard deviation or interquartile range) for given day of the year. The present paper aims to derive the annual courses of the relevant characteristics from the 29 years series (1962-1990) of daily total ozone measured in Hradec Kr lov‚, Czech Republic. Three approaches are considered to provide estimates of the annual courses of the respective characteristics: (1) unsmoothed daily values, (2) pentad statistics and (3) daily values smoothed by robust locally weighted regression (RLWR). It is argued that only the third method provides reasonably smooth curves applicable as a reference standard. The statistical tests has proved that even the severely smoothed curves satisfactorily portray the annual course of relevant characteristics. Although several brief singularities may be detected (the best marked one being in February) during the year, more detailed analysis should be yet performed to confirm their statistical significance and possibly to explain their origin. The relationships between the mean and standard deviation on the one hand and quantile characteristics on the other hand indicate that application of either approach will be similarly efficient for classifying the below-normal and above-normal values of the total ozone. 

Figures:

Annual course of median and upper and lower quartiles smoothed by Robust Locally weighted regression with halfspan = 50 (green lines) and 15 (red lines). The degree of the smoothing polynom is 2 in both cases.