u 



U 



a 

 <d 

 l. 

 cd 



■o 



CD 

 (_ 

 =5 

 ■*-> 

 TO 

 t_ 

 CD 

 O. 



0.6 



0.4 

 0.2 

 

 0.2 

 0.4 



cd -0.6 



-0.8 



840 



880 



1920 

 year 



1960 



2000 



Figure 5b. Difference between annual mean Northern Hemisphere air temperature and 

 20-year average of 1951 to 1970 (Jones etal., 1986). 



mean concentrations have to line up 

 almost as perfectly as the carbon- 

 dioxide concentrations in Figure 5a (i.e., 

 the same direction of change occurs 

 between each consecutive pair of years). 

 Five years simply do not provide a long 

 enough record to extract trends from 

 noisy data. 



The Spearman test was run on annual 

 mean concentrations of 14 chemicals at 

 the 141 sites with four or five years of 

 data. Site-by-site and chemical-by- 

 chemical results of this test are listed in 

 Appendix B. Among 1,974 chemical- 

 site combinations (14 chemicals x 141 

 sites), there are 239 cases ( 1 52 decreas- 

 ing and 87 increasing) with strong cor- 



relations between those concentrations 

 and time. Finding trends in only 13% of 

 the possible cases, reflects a lack of 

 trends so strong that, like atmospheric 

 carbon dioxide, they are evident with 

 only a few years of data. There may be 

 weaker trends that will be revealed with 

 more years of data. 



The statistical test allows for 5% of the 

 correlations to be random occurrences 

 rather than real connections between 

 concentration and time. So 99 (0.05 x 

 1 ,974) of them may not be real. Never- 

 theless, of the 239 trends, the ones de- 

 serving close attention are those found 

 among groups of sites. Among the nine 

 Long Island Sound sites in Appendix B, 



22 



