The statistical test is based only on the rankings of concentrations. For example, the rankings of cad- 

 mium concentrations at the site outside Hempstead Harbor River in Long Island Sound were: 



where the highest value, rank 1, occurred in 1987 and the lowest, rank 8, occurred in 1991. The corre- 

 lation coefficient between these ranks and years is -0.905. The coefficient is negative because concen- 

 trations are decreasing as years increase. The correlation is statistically significant because the coeffi- 

 cient is greater than 0.738 (or less than -0.738) meaning that there is less than a 5% chance that rankings 

 could fall out in that order purely by chance. The conclusion is that cadmium concentrations are de- 

 creasing at this site. If the ranking had been: 



the coefficient would have been - 1 .000. indicating a perfect correlation, but perfection is not necessary 

 for a determination that a trend exists. Note that the actual concentrations were not used. The actual 

 concentrations were (in ppm-dry): 



but even if they could had been: 



Year 86 87 88 89 90 91 92 93 



Cone 13.2 48.7 9.60 7.93 7.90 0.40 0.50 0.45 



the rank order would have been the same, the Spearman correlation would have been the same, and the 

 same conclusion would have been drawn. 



Proceeding with site-by-site and chemical-by-chemical non-parametric correlations, we have calcu- 

 lated correlations for 14 chemicals at the 154 sites with 6 or more years of data. Table 5 from O'Connor 

 (1995) summarizes the results by indicating chemicals for which there is a 90 or 95% probability of a 

 correlation between concentration and year, and whether that trend is an increase (I) or decrease (D). 

 For XPAH, the numbers of years are 4 to 6 rather than 6 to 8, and for butyltin die numbers of years are 

 5 or less. These fewer numbers of years have been taken into account when assigning 90 and 95% 

 levels of confidence. 



The 90% probability trends are included for two reasons. First, because there are fewer years for XPAH 

 and ZBT, relaxed statistical criteria are needed to identify trends. Secondly, inclusion of 90%-level 

 trends highlights similarities among groups of sites. For example, the decreasing concentrations of Cd 

 and Cu in Long Island Sound are emphasized by including the Huntington Harbor site where the 

 decrease found at three other sites with 95% certainty is also found but with 90% certainty. 



