Table 4-15. Principal variance components for proposed monitoring framework 



Among-site variance: Variation due to differences in the indicator value among a sample of stream 

 sites. This component represents the environmental "signal" to be detected and interpreted with 

 respect to an ecological condition. 



• Function of number of sites sampled (inclusion probability) 



• Calculated based on routines in Diaz-Ramos et al. (1996) 



• Performance Objective: Target subpopulation sample size of 50, with minimum of 30 



Extraneous variance: Remaining temporal, spatial and measurement-related variation. Collectively, 

 these components represent "noise" that inhibit the ability to detect and interpret the environmental 

 "signal." Extraneous variance is characterized using a randomly-selected subset of the probability 

 sample sites. These sites are revisited across and within years. 



• Components of extraneous variance: 



• Coherent variance across years: Amount that all sites in a region vary in 

 common due to regional-scale effects (e.g., climate, hydrology); important 

 component in ability to detect trends in regional population of sites 



• Among-year variance: Interaction of site and annual variability; amount 

 an individual site varies among years 



• Within-year variance: Temporal variance at a site within the defined 

 index period. Also contains measurement-related error, crew errors, etc. 

 Important component in determining status of resource population 



Approach: If temporal variance is significantly less than total observed 

 variance, then measurement-related components are not important. If 

 temporal variance contributes substantially to total variance, then examine 

 measurement-related components for possible sources. 



• Spatial variance: Variance among different ecological subregions 

 (see Guideline 11) 



• Calculated based on 2-factor analysis of variance model (sites, coherent variance 

 across years), with an interaction term representing among-year variability at an 

 individual site 



• Performance Objectives: 



• Variance within the index period should be approximately 10% of total variance 

 to minimize effect on status estimation and maximize discriminatory ability. 



• Variance between years must be minimal relative to total; target capability 



is to detect a 2% change per year in a regional population mean with a Type 1^ 

 error of 0.1 and a Type IP error of 0.2. 



^ Type I error (false positive) is the probability of concluding that a trend is present when in truth it is not. 

 "Type II error (false negative) is the probability of concluding that a trend is absent when in truth it is present. 



of interest, and are based on the probability that an individual site will be selected as part of the survey sample 

 from the universe of potential target population sites. The relative magnitudes of different components of 

 variability for the indicator and each candidate metric were identified (Fig. 4-6), following Urquhardt et al. 

 (1998). With respect to estimating the status of resource populations using the proposed indicator, within- 

 year (index period) variability should comprise 10 percent or less of the total extraneous variance (Larsen et 

 al. 1 995, Larsen 1 997). The indicator itself achieved this target, with index variability contributing approximately 

 5% to the total extraneous variance. However, performance of individual metrics was mixed: Six metrics 



4-28 



