84 
MODELING APPROACH 
Continuous monitoring data for Maryland and Virginia were analyzed to determine 
a method of post-calibrating fluorescence/chlorophyll to match extractive chloro¬ 
phyll more precisely. Because the instruments are identical, it was assumed that the 
relationships between the Dataflow fluorescence and chlorophyll a would show 
similar patterns. Maryland data were available for 2003 through 2005 for approxi¬ 
mately 21 tidal tributaries (not all tributaries were sampled in all three years). 
Virginia data came from the York River. Initial tests indicated that no more variation 
occurred between Maryland and Virginia data than among the tidal tributaries in 
Maryland. This finding simplified the post-calibration model geographically by 
allowing combination of data from both states. 
A second test of the data evaluated potential differences among years. This test also 
proved negative, which signified that all three years of data could be combined when 
developing the post-calibration model. Tests of season and tributary differences 
suggested that the final model would need to account for temporal and spatial differ¬ 
ences. Further analyses indicated the need for two tributary groups and two season 
groups, meaning that four calibration curves will be required. Significant variables 
in the model also included water temperature, turbidity, and pH. Significance is 
defined here as a p-value of less than 0.05. 
Initial results indicate that four calibration curves would be needed, two for season 
and two for tributary. All four models contain fluorescence, water temperature, 
turbidity, and pH. 
ANALYSIS ISSUES 
Several issues were addressed in conducting the analysis to formulate the decision 
rules and calibration curves. Similar to the turbidity/K d relationship, many of the 
issues related directly to the decision to lump or divide the data when computing 
calibration curves and decision rules. The argument in favor of lumping (to perform 
the analysis on a data aggregate) reasons that better estimates result when large 
numbers of observations are averaged. On the other hand, the in situ to in vitro rela¬ 
tionship may not be consistent across all subsets of the data (i.e., between different 
tidal tributaries and embayments). If so, dividing the data and developing algorithms 
for each set may lead to better overall precision. 
Seasonal Patterns 
Because species composition can affect the relationship of in situ to in vitro chloro¬ 
phyll measurements, this relationship may change with the seasons. Thus, one 
aggregate-or-divide issue requiring resolution is the effect of seasons. 
The in situ/in vitro difference generally follows a seasonal pattern consistent with 
known species composition patterns for Chesapeake Bay and its tidal tributaries. In 
situ chlorophyll measurements have a negative bias when phytoplankton populations 
chapter vii 
Shallow-water Monitoring and Application for Criteria Assessment 
