light proportional to a known exposure. The transmitted light was measured 
through each step of the film wedge on each frame. A transmittance versus 
exposure curve was plotted for each frame and this was used to find the 
exposure related to the densitometric reading for each of the lakes. 
All the above calculations were done with computer programs developed 
for this project. The program includes a graphing subroutine for plotting 
exposure versus secchi depths. These plots are shown in Figures 1-6. The 
programs include provisions for inputting calculated lake exposures and 
secchi depths into a non-linear regression curve fitting subroutine for 
statistical analysis. 
Of the four ERTS bands, band 5 and, to a slightly lesser extent, band 4 
showed the best relation between lake image exposure and secchi depth. Band 
5 of the multispectral scanner senses red band wavelengths from .6-.7y. The 
plot of band 5 exposure versus secchi depth is shown in Figure 2. An expo 
nential model was used to calculate the least squares regression represented 
by the solid line. The following equation describes this line: 
EXPOSURE = .0543 + .148e* 073 ( secchl de P th ) 
The root mean square residual (standard deviation) about the regression line 
is .02524. The mean measurement error in band 5 of two replicate sets of 14 
lakes was 6%. Given this small measurement error, much of the scanner about 
the regression line can be assumed to be a function of the 1 to 25 day 
interval between the sampling date and ERTS overflight date. The root mean 
square residual is an indicator of how reliably the fitted curve predicts a 
secchi depth for a given exposure. Assuming a normal error distribution, an 
envelope of one standard deviation (.0254) on each side of the fitted curve 
can be expected to contain a given lake exposure 68% of the time (see Figure 
6). An envelope of two standard deviations (.0508) on each side of the curve 
will contain a given lake exposure 95% of the time. 
The plot of band 4 (green band, wavelength .5-.6y) lake exposure is 
shown in Figure 1. There is very little contrast in band 4 between the lake 
image and the land surrounding it. To reduce the chance of measurement error 
caused by the inability to distinguish between the two, 12 of the smaller 
lakes were not analyzed, leaving 25 lakes in the band 4 sample. The expo 
nential equation describing the least squares fit is: 
EXPOSURE = .144 ♦ .176e' 0979(seCChl de P th3 
The root mean square residual (standard deviation) is .0402. The mean 
measurement error in band 4, calculated from two replicate sets of 14 lakes, 
is 4%. 
The plots of the infrared wavelengths, band 6 (.7-.8y) and band 7 
(.8-l.ly) exposure versus secchi depths are shown in Figures 3 and 4 
respectively. There was no significant correlation between lake exposure 
and secchi depth in either of these two bands.