Precision: ODFW Habitat Resurvey and Analysis
An analysis of survey precision was completed for the 2000 dataset and for the combined 19982000 data. The precision of an individual survey metrics was calculated using the mean variance of the resurveyed stream’s “Noise” and the overall variance encountered in the habitat survey’s “Signal”. Three measures of precision were calculated: the standard deviation of the repeat surveys (SD_{rep}), the coefficient of variation of the repeat surveys (CV_{rep}), and the Signal to Noise ratio (S:N). S:N ratios of less than 2 can lead to distorted estimates of distributions and limit regression and correlation analysis. S:N ratios between 2 and 10 are useful for analysis, but caution must be exercised due to the larger variances associated with each variable. S:N ratios greater than 10 are very good and indicate that variables have insignificant error caused by field measurements and short term habitat fluctuations (Kauffman et al. 1999).
An assessment of variable precision incorporated annual and composite variability for 19982000 (Table 9). The complete dataset encompassed 724 habitat survey sites and 85 sites where a repeat survey was conducted. The signal to noise ratio for channel length, gradient, percent dammed pools, deep pools per km, and 0.3 m conifers per 305 m of stream length varied widely between years, while the precision (reported as the standard deviation and coefficient of variation) remained consistently low. Measurements of signal to noise between years were consistent for percent secondary channels, percent fines, percent gravel, percent fines in riffles and percent gravel in riffles. Signal to noise ratios for woody debris were low but precision remained high. As in previous years, it appeared that a few sites with large amounts of wood proved difficult to count consistently thereby reducing the overall accuracy of the wood data collected.
All signal to noise ratios for variables in the combined 19982000 dataset exceeded a value of 2 with several dependent and independent variables with high ratios. Reliable independent variables included channel length, channel width, floodprone width and gradient. Reliable dependent variables were percent secondary channels, percent pools, percent dammed pools, deep pools per km, percent fines, percent bedrock and percent fines in riffles.
Table 9: Signal to Noise Ratios for 2000 and combined 19982000
Independent Variables:
Variable 
Year 
S.D. (repeats) 
CV 
S:N* 
Channel Length 
1998 1999 2000 19982000 
47.8 26.7 24.9 34.8 
6.6 3.5 3.5 4.8 
29.8 93.8 104.2 55.2 
Channel Width 
1998 1999 2000 19982000 
1.3 1.7 1.5 1.6 
18.1 19.5 17.4 18.4 
13.7 29.8 26.4 27.8 
Floodprone Width 
1998 1999 2000 19982000 
3.7 3.4 3.3 4.4 
25.9 27.6 22.0 32.0 
10.0 11.2 36.1 10.4 
Gradient 
1998 1999 2000 19982000 
0.5 1.8 0.8 1.1 
8.9 31.6 16.2 20.7 
172.9 11.8 51.4 30.9 
*S:N = Variance of all surveys/variance between repeat surveys
Dependent Variables:
Variable 
Year 
S.D. (repeats) 
CV 
S:N* 
% Secondary Channels 
1998 1999 2000 19982000 
3.0 3.1 3.0 2.9 
70.0 66.2 74.2 68.9 

