Table IV-1. (continued). 
Procedure 
2007 Addendum 
2008 Addendum 
Statistical Modeling: Turbidity-to-Kj 
conversion 
U.S. EPA 2007. p79. Statistical 
Modeling - Model definition and 
regionally specific models. “A multiple 
regression model of K d vs. 1.5 root of 
turbidity [i.e., turbidity 171 - 5 ] x 
chlorophyll x salinity provides the best 
fit of the K d -to-turbidity relationship”. 
A multiple regression model of K d vs. 
1.5 root of turbidity [i.e., turbidity 1/1 5 ] 
+ chlorophyll a + salinity provides the 
best fit of the K d -to-turbidity 
relationship. The general form of the 
models then are K d =(x* turbidity 3 ) + 
(y*chlorophyll b ) + (z*salinity c ) + C 
where: 
• a.b and c are exponents on 
their respective water quality 
parameters and a=( 1/1.5). b=l 
and c = 1; 
• x, y and z are region-specific 
constant multipliers for the 
respective three water quality 
parameters defined in Table IV-2; 
• C is a region-specific constant; 
and 
• Turbidity is measured in NTUs, 
chlorophyll a is reported in ug/L 
and salinity measures are taken in 
parts per thousand (ppt). 
(K d ) using in situ K d calibration measurements and coincident continuous water 
quality monitoring data. A single equation for baywide application was not found to 
be appropriate (Appendix D). Rather, a series of regionally-specific multiple regres¬ 
sion models for determining light attenuation (K d ) from turbidity, chlorophyll and 
salinity data were developed (Table IV-2). Details of the regionally-specific regres¬ 
sion equation derivations supporting their application for turbidity conversion to K d 
throughout Chesapeake Bay and its tidal tributaries and embayments are docu¬ 
mented in Appendix D. 
Turbidity conversion to a K d measure is not a 1:1 unit conversion. On page 79, U.S. 
EPA (2007) specifically discussed the multiple regression model approach but 
initially provided a multiplicative form of a general equation where K d = 7.5 root of 
turbidity x chlorophyll a x salinity as providing the best fit to the K d -turbidity rela¬ 
tionship. Table IV-2 provides the updated additive form of the regression model and 
region-specific groupings of tributaries as defined through State-specific cluster 
analyses in Maryland and Virginia. Virginia-specific analyses were the first 
completed and published the use of the 1.5 root for turbidity conversion to K d (U.S. 
EPA 2007). Maryland-specific analyses showed that a 1.6 root yielded the lowest 
root mean square prediction error and highest r-square value. However, this differ¬ 
ence in root, the associated error and r-square for the 1.5 vs. 1.6 root associated with 
turbidity-K d conversion were so minor (i.e., thousandths-decimal-place differences) 
that it was decided for consistency across the jurisdictions to use the results for the 
chapter iv 
Refinements to Procedures for Assessing Chesapeake Bay Water Clarity and SAV Criteria 
