Longitudinal variability in the Stratospheric Ozone distribution

Stratospheric ozone distribution over the Arctic does not exhibit the same simmetric pattern as in the Antarctic. This and other observations led to speculate that there might be a longitudinal effect. ANOVA was used to test this hypotesis No cogent reasons appeared leading to accept it.


Introduction

Ozone geographical distribution is the result of complex interactions involving its formation, destruction and transport as well as the meteorology of the stratosphere. For background information about ozone chemistry and distribution follow the link to the stratospheric ozone webpage.

As can be seen in the example pictures, during the anctartic spring when a "hole" is present, ozone concentartion are fairly simmetric over Antarctica due to the stability and persistance of the anctartic vortex. This stabilty may be partly expalined by the central simmetry of the continent, hogeneously sorrounded by the ocean. In the Arctic, on the other hand, the ocean is sorrounded by land, with the exception ot the notable exception of the North Atlantic-Norwegian sea. Ozone concentration over the Arctic, as a matter of fact, does not in general show a symmetric distribution.

The hypotesis formulated for this study is that this may produce a longitudinal distribution, so that areas at different longitudes may experience, over the long term, different average concentrations. In particular, it was speculated that an effect should show up in or around the Norwegian Sea, the area of asimmetry. There is also some evidence in the literature of a possible longitudinal effect [3] since some ground-based stations in the northern hemisphere (e.g. Iceland) have exhibited a tendency to an increased total ozone concentration, contrary to the general trend.


Methodology

Daily data of total ozone concentration have been provided by the Goddard Space Flight Center and distributed as a CD-ROM. They were collected by the Meteor 3, a polar orbitig satellite with an inclination of 82.5 degrees. The instument makes 35 omeasurements every 8 seconds, covering an area 50 to 200 km wide on the ground.

The technique chosen to test the hypotesis was the analysis of variance (ANOVA). This is a very powerful statistical tool which enables to discriminate the effects of each factor affecting the parameter of interest with great accuracy from a relatively small sample. It then allows to determine which factors are significant and which have no influence on the response. In this case it was decided that the factors that may affect the ozone concentration, besides longitude of course, are the latitude, the month and possibly the year. Due to the high latitude of the area of interest and the fact that the TOMS instrument has to rely on sunlight, data were available only from April through September. In order to have a balanced design, only the years 1992 and 1993 were used, other years lacking some months of data. Two days per each months were selected randomly.

The ANOVA design was 3-factorial, with the year as replicate. Month and Longitude were introduced as fixed effects. Actual analysis was carried out using the SAS statistical package v.6.

The locations were selected, among the available grid points, to cover the area of interest as well as two control continental regions. Values of latitude are 60.5, 65.5, 70.5, 75.5 and 80.5 degrees North. Values of Longitude are 120.625 W (western Canada), 44.375 W (Greenland), 6.875 W (Norwegian Sea), 15.625 E (Scandinavia) and 105.625 E (central Siberia). The sites are marked in the picture below (click to enlarge).


Results

Regression analysis indicates that the model used for the ANOVA seems to describe most of the observed variability, with an R2 value of 79.2%. No second- or higher-order interactions proved significant and were therefore discarded from the model.

Although this has to be considered only a preliminar study, no indications have emerged leading to accept the hypotesis. The Table below summarizes the results of the ANOVA.

FactorDFMean SquareF valueP valueSignificant
Longitude42313.634.360.0018YES
Latitude41024.491.930.1041NO
Lat.-Long. interaction16718.591.350.1599NO
Month5182072.48342.950.0001YES
Month-Long. interaction202092.463.940.0001YES
Month-Lat. interaction203052.075.750.0001YES
Year13021756.920.0001YES
Error529280844.99

There is indeed a weak evidence of a longitudinal effect, but this changes with the month. To check for statistical significance between pairs, because of the significance of the interactions, the Student-Newman-Keuls test (STK) has been used. This revealed no interesting pattern in the longitudinal variability. In particular, no statistically significant differences were detected in the area around the Norwegian sea at any time of the period considered.

Another relatively unexpected result is that, at least between 60 and 80 degrees North, latitude is not significant. This finding is however confirmed by looking at the graph of the zonally averaged ozone distribution (Seinfeld & Pandis). It is apparent that, with the exception of April, locations between 60 and 90 N do not differ. This also reveals the power of ANOVA, which in fact detected a significant month-latitude interaction but no significant latitude effect.

The fact that the year has proved an important variable affecting the ozone concentration while it had been considered only a replicate has interesting consequences. Clearly which year is chosen has no relevance in the chemistry of stratospheric ozone. Years, however, differ from one another in the meteorology which in turn affects ozone transport within the stratosphere. Inclusion of the year and its first-order interactions (all significant) in the model does not change in a substantial way any of the results and will not be presented here.

By far, the effect of the month is the most important and explaines almost all of the observed variability in ozone concentrations; monthly average variability at the 25 locations is shown in the graph below. All monthly averages are satistically different from one another in the SNK test.

The reason why the month is the variable dominating the stratospheric ozone concentration should be clear from the understanding of ozone chemistry. As the season changes, so does (and dramatically) the amount of solar radiation reaching the polar regions. Since most of the ozone there has been in fact produced in the tropics, any increase in solar radiation will enhance the catalytic cycles consuming ozone, thereby depleting its amount by the end of the summer.


Conclusions and future work


Send your comments or report improper links to Vito Ilacqua

Longitudinal Variability of Stratospheric Ozone/ Last updated April 29 1998/ Rutgers University - New Brusnwick, NJ