A-47 
Figure 5.6. Confidence bounds computed based on quantiles of fraction of space 
computed on conditionally simulated surface estimates using variogram estimates 
from data. The base simulation has spatial correlation and no spatial or temporal trend. 
Sample size is 40. 
the kriging estimator. Additional analyses test whether this property might hold in 
general or is an artifact of the simple conditions (no spatial or temporal trend) under 
which this experiment was performed. 
In prior discussions we have noted that the CFD is the inverse of the CDF of the 
population of p's where p is fraction of space out of compliance with the criterion 
threshold. It is the variance of the p's that determines the steepness of the CFD: the 
smaller the variance, the steeper the CFD. In real applications, estimates of the p's 
have two important variance components. One variance component comes from true 
variance over time in the parameter being assessed. Another variance component 
comes from imperfect estimates due to sampling variability. In the base simulation 
with no spatial or temporal trend in the data, it is this second source of variance that 
controls the shape of the CFD. 
appendix a 
The Cumulative Frequency Diagram Method for Determining Water Quality Attainment 
