54 
variables have no statistically significant interaction with tributary. Starting with the 
groups shown in Figure D-l and juggling a bit, we arrive at the following groups: 
Group 1: 
Bush River 
Gunpowder River 
St. Mary’s River 
Magothy River 
Middle River 
Group 2: 
Lower Patuxent 
Potomac River 
Eastern Bay 
West/Rhode River 
Group 3: 
Severn River 
South River 
Fishing Bay/Chicamacomico 
Group 4: 
Little Choptank 
Miles/Wye River 
Group 5: 
Upper Patuxent 
Group 6: 
Lower Chester River 
Upper Chester River 
These groupings reflect a strong geographical pattern which strengthens their 
validity. The Upper Patuxent River falls in a group alone because of the negative 
coefficient for chlorophyll. This coefficient seems quite unusual when juxtaposed 
with the positive coefficients for all other tributaries. This may be the result of some 
spatial pattern that is confounded with chlorophyll and warrants additional study. 
Shown below are the ANCOVA tables for each group illustrative that the interaction 
terms lack significance (p > 0.01) (Tables D-7 thru D-16). Based on these results, we 
infer that the primary independent variables of the calibration equation: turbidity, 
salinity, and chlorophyll, have a uniform effect for each tributary group. In some trib¬ 
utary groups, some independent variables, e.g. salinity for group 1, appear to be not 
important. The model could be reformulated to omit these variables in these tribu¬ 
tary groups. 
The calibration equations for each tributary group are: 
Group 1: 
K d = 0.5545 + 0.3172*(rl_5Turb) + 0.0160*(Chlorophyll a) - 0.0138*(Salinity) 
Group 2: 
K d = -0.1247 + 0.2820*(rl_5Turb) + 0.0207*(Chlorophyll a) + 0.0515*(Salinity) 
appendix d 
Derivation of Regressions 
