L-2 
1. INTRODUCTION 
Dauer et al. (2002) submitted a report to the US EPA Chesapeake Bay Program 
Office on the development of diagnostic approaches to determine sources of anthro¬ 
pogenic stress affecting benthic community condition in the Chesapeake Bay. The 
objective of the study was to develop analytical tools capable of classifying regions 
in Chesapeake Bay identified as having degraded benthic communities into cate¬ 
gories distinguished by the type of stress experienced by those communities. The 
tool was successful at identifying regions with high probabilities of sediment 
contamination. However, prior to implementation, it was recommended that the 
operational effectiveness of the diagnostic tool be further tested using additional 
validation data sets. 
In this Addendum the results of two additional tasks are presented. First, the linear 
discriminant function was independently derived to verify the accuracy of the devel¬ 
opment of the function. Second, two additional putative validation data sets were 
used to assess the validity of the linear discriminant function. 
2. LINEAR DISCRIMINANT FUNCTION 
In this task it was discovered that four samples from the original calibration data set 
were not included in the derivation of the final linear discriminant function originally 
reported in Dauer et al. 2002. The final validation of the linear discriminant func¬ 
tion with these additional four samples was identical to that reported in Table 21 for 
the Baywide scenario, i.e. using the All Province sediment contaminant classifica¬ 
tion, namely, with an overall percent correct classification of 75.14%. The new 
coefficients for this function are given in Table 1 of this Addendum (revised Table 
24 of Dauer et al. 2002). 
3. ADDITIONAL VALIDATION DATA SETS 
Two putative data sets were used for further validation of the Contaminant Discrim¬ 
inant Tool (CDT) as presented in Dauer et al. 2002. 
ELIZABETH RIVER WATERSHED 
The first putative data set consisted of 125 random samples collected in 1999 from 
the Elizabeth River watershed (Dauer and Llanso 2003). An additional 100 random 
samples collected 25 per year from 2000-2003 were also used (Dauer 2001, 2002, 
2003, 2004). All samples were analyzed using the CDT function and placed into 
categories based upon the posterior probability of inclusion into the Contaminant 
Group. Due to the high levels of contaminants recorded historically in the Elizabeth 
River watershed (Hall et al., 1992, 1997, 2002; Padma et al. 1998; Conrad et al. 
2004), the a priori expectation was that a high percentage of samples declared 
degraded by the Benthic Index of Biotic Integrity would be placed into the Contam¬ 
inant Group. The results from the Elizabeth River watershed are compared to results 
appendix 
Addendum to the Report 
