A-13 
Defining the CFD Ideal 
As defined above, the CFD is a data driven formulation. But the data used to formu¬ 
late the CFD are a sample of points taken from a population. Defining the CFD 
becomes complex when one considers the many different levels for which it might 
be defined. At one level, the CFD might be defined based on the true state of a 
segment. Imagine that the state of a segment could be frozen for sufficient time to 
permit deployment of an analog sampler (that is one that measures water quality 
continuously rather than in discrete samples) to assess the percent of area out of 
compliance at that instant. Now stretch that imagination one step further to relax the 
condition that the segment be frozen and allow that these analog measurements of 
percent of area out of compliance be determined continuously in time. With this 
information, a determination of the CFD for the true state of the segment is possible. 
While the information needed to construct the ideal CFD is not obtainable, it is 
important to ask how well the CFD based on obtainable data represents this ideal 
(see also Section 5). Is a data driven CFD consistent for the ideal CFD in the statis¬ 
tical sense? Loosely speaking, consistency implies that the data driven CFD should 
get closer to the ideal CFD as more data are used. Is the data driven CFD unbiased 
for the ideal CFD? Unbiasedness implies that even with small amounts of data, the 
data driven CFD on average covers the ideal CFD. 
One might argue that if both the assessment CFD and the reference CFD are data 
driven, then it is not important for the CFD to approximate the ideal. Even so, it is 
important to understand the behavior of the CFD as a function of samples size and 
the relative temporal and spatial contributions to the variance in the water quality 
parameter. If the curve changes shape as a more data are used, this could result in 
unfair comparisons between assessment and reference regions. In Section 4, statis¬ 
tical properties for both types of reference curves are evaluated further. 
Defining Reference Curves 
Two approaches to defining the reference curve are being considered. One is a 
biologically based definition. The idea is to identify appropriate reference regions 
with healthy biological indicators and compute the reference CFD for these regions. 
For example, healthy benthic IBI scores might be used as indicators of adequate 
bottom dissolved oxygen. Thus after stratifying by salinity zone and perhaps other 
factors, a series of dissolved oxygen reference CDF curves could be computed from 
the existing 20+ year monitoring data base. When it is not possible to establish a 
reference condition some more arbitrary device must be employed. Alternatives are 
discussed in Section 4.0. 
Discussion of Each Step 
Step 1 - Data Collection. One of the advantages of the CFD approach is that it will 
accommodate a variety of input data and still arrive at the same assessment endpoint. 
Data collection methods currently in place include: fix station data, cruise track data, 
continuous monitor data, aircraft flight path data, and satellite imagery data. Because 
of the interpolation step, all of these data can be used (and potentially combined) 
appendix a 
The Cumulative Frequency Diagram Method for Determining Water Quality Attainment 
