Table VI-5. Sample attainment results when assessing with varying time windows 
Dissolved Oxygen Criterion 
Time Windows 
Meeting Criterion 
Percent of 
Observations 
at or above Criterion 
30-day Mean (3 mg liter' 1 ): 
Sequential 
2 of 2 
100% 
Rolling window 
31 of 31 
100% 
7-day Mean (2.5 mg liter' 1 ): 
Sequential 
7 of 8 
87.5% 
Rolling window 
46 of 55 
83.6% 
Instantaneous Minimum (1.7 mg liter' 1 ) 
Pool of hourly measurements 
1,250 of 1,484 
84.2% 
uses. Nevertheless, the number of such data sets on hand is substantial and growing, 
relative to the number and location of fixed monitoring stations. 
Developing and verifying the method will be an ongoing process. Short-term fore¬ 
casts based on synthetic data are created and compared to actual semicontinuous 
records not used in the original forecasting process. There are some, but not many, 
instances in which semicontinuous data are available at the same site in different 
years. Also, in some instances, multiple semicontinuous records are available for the 
same region. In these cases, one record is used in the spectral analysis and equation 
development and the other is used to verify the results. With data recorders deployed 
for the specific purpose of validating and refining the forecasting models, better veri¬ 
fication will be available in the future. 
Even with these issues resolved, there are still questions concerning how synthetic 
time-series data sets should be adapted to enable an assessment of spatial extent and 
frequency of attainment in a manner consistent with criteria assessed by other analyt¬ 
ical methods. 
Logistic Regression Approach. This method modifies and significantly updates 
a method developed originally to measure attainment of the 1992 Chesapeake Bay 
dissolved oxygen restoration goal (Jordan et al. 1992). The early work demonstrated 
predictable relationships, on a segment-by-segment basis, between seasonal mean 
dissolved oxygen concentrations and the percent of observations above a target 
concentration. The relationships proved to be strong and applicable in areas where 
dissolved oxygen concentrations ranged above and below the goal target concentra¬ 
tions. Given the tidal water quality monitoring data record that spans more than 17 
years with the measurements from multiple depths (the vertical dissolved oxygen 
profile is collected at 1- to 2-meter intervals), the regression models are now month- 
and depth-specific in many segments. Based on the monthly mean dissolved oxygen 
concentration measured at a specified depth, the models predict the percent of time 
chapter vi • Recommended Implementation Procedures 
c 
