Guideline 10: Temporal Variability - Across Years 



Indicator responses may change over time, even when ecological condition remains relatively stable. 

 Obsen/ed changes in this case may be attributable to weather, succession, population cycles or other 

 natural inter-annual variations. Estimates of variability across years should be examined to ensure that 

 the indicator reflects true trends in ecological condition for characteristics that are relevant to the 

 assessment question. To determine inter-annual stability of an indicator, monitoring must proceed for 

 several years at sites l(nown to have remained in the same ecological condition. 



Combined Performance Objectives for Guidelines 9 and 10 



1 . Identify important components of variance based on the proposed monitoring framework and 

 sampling design. 



2. Demonstrate that the magnitude of individual components is within performance criteria established 

 for the proposed monitoring program. 



The use of a probability-based survey design as the monitoring framework requires a modified approach to 

 defining and estimating important components of an indicator's spatial and temporal variance. In the case of 

 a multi-metric index, it is also necessary to determine variability in individual candidate metrics in order to 

 select the final suite. 



Important sources of variation for the indicator within the proposed monitoring framework and target 

 performance criteria for EMAP have been identified (Table 4-1 5). Note the inclusion of "population variance," 

 which is the variation due to the relationships between the probability sample and the survey design. It is 

 solely a function of the number of probability-based samples used to estimate the resource population of 

 interest. It has been determined that 50 samples provides population estimates with 90 percent confidence 

 bounds that are approximately ±10 percent of the proportion (Larsen etal. 1995, Larsen 1997). The survey 

 design is flexible in allowing one to define a posfer/or/ various resource subpopulations of interest within the 

 constraints of sample size. 



Sources of temporal variation in the indicator value (or metric score variable) are included in "extraneous 

 variance" (Table 4-15). These components and descriptions are based on Larsen et al. (1 995) and Urquhardt 

 et al. (1998), for indicators associated with monitoring frameworks similar to that of the proposed indicator. 

 Within-year variability (Guideline 9) is estimated as index period variance. Variability across years (Guideline 

 1 0) for this indicator is addressed by two separate components of variance: the coherent variability of all sites 

 across years, and the among-year variability of individual sites. Within-year variability also includes 

 "measurement-related" errors described under Guideline 8. Because the proposed monitoring framework 

 emphasizes regional scales and populations of sites, rather than patterns at individual sites, the importance 

 of measurement-related error is reduced. Measurement-related errors are considered in detail only when 

 within-year variability is unacceptably large relative to the total extraneous variance of the indicator. In such 

 cases, it must be determined whether the variability is due primarily to temporal variability or to measurement- 

 related errors. 



Variance components were estimated using all 298 sites in the mid-Atlantic highlands region (including repeat 

 visits within and across years), weighted by the appropriate population expansion factor. These expansion 

 factors are used to extrapolate the results from each site in the survey sample to the entire resource population 



4-27 



