Review of Bull Trout Presence/Absence Protocol Development Including the Washington Validation Study

Over the past decade, the U.S. Fish and Wildlife Service has worked toward defining sampling protocols for detection of threatened juvenile bull trout (Salvelinus confluentus) populations in the Coastal-Puget Sound and Columbia River regions. Previous detection protocols (Bonar et al. 1997) have set sample size guidelines using an assumed fixed capture efficiency and an arbitrary threshold density of bull trout. Sample sizes were selected so that if the capture efficiency was correct and no bull trout were detected, one would conclude it was unlikely that bull trout existed at or above the threshold density. Because of the fixed nature of the capture efficiency and threshold density, the sample size recommendations were also fixed. Recent protocols by Peterson et al. (2002) have incorporated two fundamental changes: (1) capture efficiencies that are modeled as a function of habitat characteristics; and (2) use of observed bull trout densities rather than arbitrary threshold densities. With the recent protocols, the resulting sample size recommendations vary with both the habitat characteristics and with density of juvenile bull trout used. Our review of the recent protocols resulted in substantial concern with both fundamental changes.

To address the utility of the habitat modeled capture efficiencies, validation studies were conducted in Washington streams during the summer of 2003. The natural variation in observed capture efficiencies within habitat categories was large making prediction difficult. The observed variation ranged as much as 60 percentage points (e.g., ranging from 10% to 70%) within a habitat category with much overlap among habitat categories. With so much variation and overlap, we felt that the stratification by habitat type suggested by the model acted more to dilute than improve the quality of the capture efficiency estimators. In fact, we found there to be less prediction error for both night snorkeling and one-pass electrofishing when we used the standard capture efficiency of 0.25 (Rieman and McIntyre 1995) across all habitat types than with the predicted capture efficiencies that we calculated from the habitat model. Until the practical value of the habitat-based capture efficiency model is more fully addressed, we recommend the resources used to measure habitat characteristics would be better devoted to enhance the direct sampling for bull trout.

Regarding the use of observed bull trout densities, we find it illogical to substitute actual densities for hypothetical thresholds. If one had actual densities, an effort to detect presence would be unnecessary. On the other hand, if actual densities measured elsewhere were to be used in sample size determination, then the density used will essentially function as the threshold. For example, using Idaho density patterns to set sample sizes in Washington protocols essentially declares that locations with densities less than those observed in Idaho are not important to detect in Washington. Because the choice of threshold has policy implications, we feel the choice of threshold needs input from a broader audience that includes managers and policy makers.

Based on our review, we have five general recommendations for Washington:

• Revisit the threshold density concept by creating a process that includes policy input to determine minimum threshold density and power (acceptable risk) criteria.
• Reanalyze the habitat data for practical value. This might be achieved by quantifying the actual cost vs. benefit (in terms of power and sample size) of using habitat modeled capture efficiencies and should include the uncertainties including the prediction error, the sampling error and the natural variability in observed efficiencies.

• Continue research to improve actual capture efficiencies as methods with better capture efficiencies will reduce the necessary sample sizes.

• Consider habitat-based models to help predict bull trout presence (as opposed to capture efficiency) so that the initial choice of sampling sites based on judgment will yield power greater than a strictly random selection would.

• Until the above steps are taken, continue to use the procedures outlined Bonar et al. (1997) for presence/absence sampling using a global mean value for capture efficiency updated with the data from all the studies conducted in Washington.