in table 4. Figure 5 is a contrast enhanced band 6 image of the Halfway Hollow 
reach of the Blacks Fork Arm. It illustrates the strong role of radiance in the 
red portion of the spectrum for mapping chlorophyll a concentration, as indi- 
cated by the second regression model of table 4. T 
Table 4. - Water quality predictor models based 
on airborne MSS radiance values. 
  
  
Number of 
Dependent variable Independent variable(s) r? S sites 
In (Secchi depth) R4 0.94 0.04 - 1.8 m 19 
In (chlorophyll a) R6, R3 0.91 ^ 19+ 8. 2ng/mB 10 
  
R3, R4, R6 are radiances in bands 3, 4, and 6 of table 1 
  
Other investigators (e.g., Mace, 1982), have successfully found predictor 
equations for variables such as total phosphorus, nitrate-nitrogen, and con- 
ductivity using airborne MSS data. For the study on Flaming Gorge, there were 
not enough measurements of these variables at sites appearing in the imagery to 
be able to establish statistical relationships with the remote sensing data. It 
is felt that the reason only two water quality parameters could be modeled from 
the MSS data is due to the shortage of ground truth and not due to any defi- 
ciency.in the remote sensing data. 
It was noted when inspecting the MSS imagery of the Blacks Fork Arm of the 
reservoir, which experienced dramatic blooms of the bluegreen alga Aphani- 
. zomenon on September 9, that thermal infrared data seemed to discriminate 
between areas of organic and inorganic turbidity. To investigate this further, 
an image of the reach was prepared for multispectral classification using bands 
2, 5,—,--and-10. — (Bands 2,-5,and7:correspond: Lo those recommended by Johnson 
and Bahn (1977) for preselection for mapping chlorophyll a and suspended sedi- 
ment.) Training statistics were derived for areas of high suspended sediment, 
algae mats and swirls, and relatively clear water. Figure 6 shows a plot of the 
training sets' mean values for band 5 versus band 10. It can be seen that high 
inorganic turbidity is characterized by high radiance in the red band with low 
temperature, whereas high organic turbidity exhibits a relatively high tempera- 
ture (8 °C higher) and lower red band radiance. (Each count in band 10 is 
roughly equivalent to 0.5 °C.) A maximum likelihood classification using these 
statistics produced a qualitative mapping of the Blacks Fork Arm which seemed to 
agree with the map obtained by applying the chlorophyll a model to the reach 
(again, excepting waters with Secchi depth < 1 meter). Apparently, the sun 
heats the algae at the surface, and since these algae contain gas vacuoles, 
there is significant contrast in temperature with the water. It appears that 
further investigations would be justified to see if these results are peculiar 
to this image or whether they point to a new tool for discriminating between 
organic and inorganic turbidity in multispectral imagery. 
SUMMARY 
Numerous studies have demonstrated the utility of multispectral imagery for 
mapping surface water quality variables. This paper reports some of the first 
efforts by the Bureau of Reclamation to use this tool. The results have been 
well received by Bureau water quality specialists. In fact, the methodology may 
be considered "operational" in that the Landsat calendar is now routinely 
606 
ES 
ses em te 
EM tape ee AR EE NAS 
= ne a 
te " T 
CE M... rama dmm pida 
RSS 4 
Ic mi 
con 
sec 
In 
an 
fer 
COS 
Sup 
air 
stu 
pro 
pre a AA ON 
Wit 
B.