K-4 
Assessments were produced for each of 85 Chesapeake Bay Program segments and 
sub-segments containing benthic data. Segments (TMWA 1999) are Chesapeake Bay 
regions having similar salinity and hydrographic characteristics. In Virginia, 
segments were sub-divided into smaller units by the Virginia Department of Envi¬ 
ronmental Quality. Sub-segments were produced for each of the mainstems of rivers 
and bays (e.g., James River mesohaline) and for some of the smaller systems 
opening into the mainstem (e.g.. Pagan River). Assessment samples were assigned to 
segments and sub-segments using GIS software. Hydrographic data collected synop- 
tically with the benthic data were used to assign each sample to one of seven habitat 
classes used in the calculation of the B-IBI. These are the same habitat classes used 
in the reference data set. 
3.2. DEGRADED AREA 
The new method developed for the 2006 assessment was based on the confidence 
limit and bootstrap simulation concepts described in Alden et al. (2002). Specifi¬ 
cally, bootstrap simulation (Efron and Tibshirani 1998) was applied to incorporate 
uncertainty in reference conditions. Bootstrap simulation is used to assess the accu¬ 
racy of an estimate by randomly sampling n times, with replacement, from an 
original data set. In our case, we wished to estimate the score corresponding to the 
5 th percentile of the B-IBI reference distributions for the good sites (by habitat). 
Because the reference distributions were based on small sample sizes, the percentiles 
were not well defined and would likely vary if different sets of reference sites were 
sampled. Thus the need to estimate this parameter more accurately with bootstrap 
simulations. Bootstrap simulations make no assumptions, except that the reference 
data are a representative sample from a “super population” of reference sites. 
For each habitat, a threshold based on the 5th percentile B-IBI score of the reference 
data set for the good sites (or the maximum B-IBI score observed for the degraded 
sites, see below), was determined. This threshold was not intended to serve as a crite¬ 
rion for classifying individual B-IBI scores, rather it was used to categorize the 
segment as impaired or not based on the proportion of sites below the threshold (i.e., 
degraded area) and the variance associated with this estimate. The variance in the 
estimates of proportions for each segment was estimated by the simulations. 
The B-IBI scores for the reference good and degraded sites had degrees of overlap 
that ranged from quite high in the tidal freshwater and oligohaline habitats to moder¬ 
ately low in the mesohaline and polyhaline habitats. An assessment sample is more 
likely to come from an impaired benthic community if the B-IBI score for this 
sample is within the range of scores observed for sites known to be degraded. There¬ 
fore, two criteria were established for determining the threshold: its score had to be 
within the lower bound of the good reference distribution (i.e., 5th percentile), and 
it had to be within the upper range of observed scores for known degraded sites (i.e., 
the reference degraded sites). If the 5th percentile score for a simulation run was not 
within the range of scores for the reference degraded sites, then the maximum B-IBI 
score for the reference degraded sites was selected as the threshold. Thus, in this 
study, sites with low B-IBI scores below thresholds were likely to be impaired and 
unlikely to come from good reference areas. 
appendix k 
2006 303(d) Assessment Methods for Chesapeake Bay Benthos 
