d of 
Olsen, 
data 
e to 
uds 
cquired 
nel 
0 a.m. 
al 
utility 
ists 
S 
form of 
stems 
remote 
ons. 
me ECS 
  
With these maps as guides, the Landsat image was displayed on a color video 
console and a cursor used to identify the image coordinates of the sample sites. 
Twenty of the 25 sites could be reliably identified in the Landsat imagery. The 
mean brightness counts for 3 by 3 pixel blocks at these locations were duly 
recorded for bands 4, 5, and 6. Using calibration constants tabulated by 
Robinove, et al., (1981), the mean brightness counts were converted to radiance 
values (mW/cm2-sr). 
Measurements of water clarity, by Secchi disc, were available for all 20 sites. 
Chlorophyll a concentrations were obtained for 11 of these sites, and at 6 of 
them, detailed measurements of conductivity, dissolved oxygen and nutrient 
concentrations were made. Using the stepwise linear regression approach of 
Draper and Smith (1966), equations were developed with water quality parameters 
as dependent variables. Candidate independent variables were radiance values 
for bands 4, 5, and 6, band-to-band radiance ratios and chromaticity ratios. 
Useful equations were identified for Secchi depth, chlorophyll a concentration 
and total phosphorus concentration. The equations are summarized in table 3. 
Table 3. - Water quality predictor models based 
on Landsat radiance values 
  
  
Number of 
Dependent variable Independent variable(s) r? S sites 
In (Secchi depth) R4 0.93 0.05 -1.5m 20 
In (chlorophyll a) R6, CR5 0.94. ..1.3.=-59 mg/m3 11 
Total phosphorus R6/R4 0.89 8.7 mg/m3 6 
  
R4, R6 are radiances in Landsat bands 4 and 6. 
CR5 is the band 5 chromaticity ratio, that is, band 5 radiance divided by 
the sum of band 4, 5, and 6 radiances. 
  
Having developed the equations in table 3, water quality maps could be prepared 
by applying them to digital images of the entire reservoir. Before this could 
be done, however, the band 6 imagery had to be cleaned up to eliminate the 
numerous bad data lines present, as shown in figure 3. This was accomplished by 
using a linear equation in terms of bands 5 and 7 (r7 20.99, $7 2.1 counts) 
to generate synthetic band 6 data. The synthetic data were inserted in the 
original band 6 image to replace the bad data lines, the result presented in 
figure 4. Once band 6 was cleaned up, cosmetically acceptable color-coded maps 
of chlorophyll a and total phosphorus concentrations could be prepared. The 
chlorophyll a model was not applied to areas where Secchi depth was less than 
1 meter. These areas were reported to have very high suspended sediment concen- 
trations. Such conditions retard algae growth by limiting light available for 
photosynthesis. Were the regression model applied in these areas, the high 
radiance values in band 6 would produce erroneously high estimates of chloro- 
phyll a concentration. 
ANALYSIS OF AIRBORNE SCANNER DATA 
Airborne MSS (multispectral scanner) data was analyzed in much the same fashion 
as the Landsat imagery. Water quality sampling sites were located on the video 
display and mean brightness counts recorded for 5 by 5 pixel blocks. These mean 
counts were converted to radiances by calibration constants supplied with the 
digital imagery. Again, a stepwise linear regression procedure was employed to 
identify predictor equations for water quality variables. These are summarized 
605 
maf : 
yw ai Ro A UR € (—— €