USGS Home
Patuxent Wildlife Research Center

NAAMP III Archive - aquatic surveys
Home | Archive by Alphabetical Order | Archive by Category

The Power to Detect Trends in Amphibian Populations Using Aquatic Funnel Traps

Klaus O. Richter and Michael J. Adams

King County Natural Resources Division
506 Second Avenue, Suite 720
Seattle, WA 98104
email: klaus.richter@metrokc.gov

University of Washington
College of Forest Resources
Box 352100
Seattle, Washington 98105-2100
madams@u.washington.edu


INTRODUCTION

Funnel trapping is one of several methods used to determine the distribution and relative abundance of amphibians in lentic habitats. In the Pacific Northwest it may be the simplest, most repeatable method for surveying wetland-breeding amphibian communities since many species are difficult to observe and few have loud calls. Funnel trapping minimizes the effect of individual bias and can be consistently replicated over time (Adams et al. in rev.). Consequently, aquatic funnel trapping shows promise as a method for detecting population trends in amphibians.

In Washington State (USA), aquatic funnel traps have been used to census amphibian populations, identify habitat use and assess biological interactions between species (Richter 1995, Adams 1996, West and Adams 1996, Leonard pers. comm.). Aquatic funnel traps have not been used for long-term monitoring of amphibians and their efficacy as a monitoring technique requires exploration.

The purpose of this paper is to provide estimates of trap rates and variance for Pacific Northwest amphibians and to explore power under a variety of trapping scenarios using those estimates.

METHODS

Estimates of the coefficient of variation (CV ) and trap rates were obtained for Rana aurora, Hyla regilla and Ambystoma macrodactylum using funnel trap data from Gee's Wire Minnow Trap captures. Trapping occurred in the Puget Sound Lowlands, Washington. Two homogenous habitats from two wetlands were repeatedly trapped from 1994-1996 (Table 1). Estimates of CV for within and among years were calculated. We ran simulations using a broad range of CV's that encompassed these values.

In practice, CV is best estimated using a sampling regime that is similar to the monitoring design. Thus, if sites will be monitored once each year (one count/plot/year), CV should be calculated from a series of mean trap rates measured once each year (Gibbs pers. comm.)

We used the program 'Monitor' (Gibbs 1995) to estimate the power to detect a trend using different combinations of CV, mean initial value (trap rate), number of wetlands surveyed, and number of years surveyed. Parameterization is described below.

Plots

Number Monitored: We define plots as wetlands and simulate a range from 3 to 21.

Counts/Plot/Survey: We define a count as some number of traps set in a wetland overnight. One count per year was used in all simulations reported.

Initial Values: This section allows an initial count (in this case, the mean number of animals caught per trap) to be entered for each wetland as well as an estimate of the standard deviation expected with repeated counts over time. A weighting factor that determines the importance of that wetland to trend calculations can also be added. We simulated mean initial counts of 1, 5, and 10 animals per trap and CV's of 50%, 100%, and 150%. We weighted using the initial count since trends are likely to be more evident in larger populations.

Since the wetlands monitored will exhibit a range of population sizes, we entered a range of initial values for the wetlands in each scenario. We kept the mean value of this range constant when varying the number of wetlands surveyed so that the effect of mean initial value and number of wetlands surveyed could be separated. We assumed that CV does not vary with count (not tested) and applied the CV to these initial counts to produce an appropriate range of standard deviations.

Surveys

Number Conducted: We varied the number of years surveyed from 3 to 21.

Occasions: We used one survey per year in all simulations.

Trends

Type: Exponential trends were simulated since population effects tend to be multiplicative.

Significance Level: We used alpha = 0.10 since amphibian populations tend to be highly variable and it is often more desirable to minimize the likelihood of committing a type II error (failure to detect a significant trend) rather than a type I error.

Number of Tails: We used a two-tailed test since both declines and increases are possible.

Constant Added: We added a constant of 0.5 to allow log-transformation.

Trend Variation: The variation in trends among wetlands was held at 0.05 in all simulations. This is a moderate value (Gibbs 1995). We have not measured this parameter.

Rounding: Decimal rounding was used since trap-rates (as opposed to whole counts) were simulated.

Trend Coverage: We examined power to detect a -5% annual trend in all simulations.

Replications: We used 500 replications which gave estimates of power that seemed to vary ca. 5%.

Results are displayed as line plots using the 'connect the dots' format, but note that the power calculated for a monitoring scenario will vary randomly from one simulation to the next. The best representation of such data would be a smooth curve that summarizes the output.

RESULTS

As expected, power increased with more wetlands surveyed, more years surveyed, and with higher initial counts. Power decreased with higher CV's. The most dramatic increases in power were gained by monitoring more wetlands (Fig. 1). With more wetlands surveyed, power increased more rapidly over time. When only three wetlands were monitored, power remained low even after 21 years.

Monitoring wetlands with higher initial counts (note that CV was held constant but may be correlated with count) gave only slight increases in power (Fig. 2). Lower CV's yielded small gains in power (Fig. 3).

When more than 3 wetlands were surveyed, power increased rapidly over the first 10-15 years. After 20 years, good power (i.e., >80%) could be achieved under a variety of monitoring scenarios.

DISCUSSION

The effect on power by a broad range of initial counts (1-10 animals per trap) and CV's (50-150%) were upstaged by the effect of number of years and wetlands surveyed. These results suggest it will be difficult to devise a funnel trapping monitoring plan that will be effective in the short term or with only a few wetlands.

We noted large increases in power with an increasing number of wetlands monitored but held trend variation constant. In practice, trend variation may increase as more wetlands are added which could negate the increase in power (not tested).

CV can be minimized by trapping at the same point in the seasonal cycle of a wetland each year. This would help remove variation due to the annual decline of larval numbers in wetlands. We suggest trapping late in larval development after the mortality curve (which is extremely steep early in the larval period) has flattened out.

Trapping multiple times within a season would likely improve power but cannot be realistically simulated without an estimate of how such a regime will affect the CV. Trapping multiple times per year might provide more accurate estimates of annual population levels but could also increase CV (particularly if the annual mortality trend is not removed from the data (Gibbs, pers. comm.).

Our estimates of CV came from monitoring homogenous habitats rather than whole wetlands. Clearly, monitoring whole wetlands will increase within plot variability but this variability is not directly incorporated into the simulations. Increases in within plot variation may increase count variation but were not examined here.

Before funnel traps can receive wide use for assessing population trends, additional issues need to be examined. For example, we don't know how capture probability will be affected by changes in habitat structure, predator invasions, temperature, etc.

Our simulations suggest that achieving adequate power using funnel traps is possible for long term monitoring programs (>15 years) involving 12 or more wetlands.

REFERENCES

Adams, M. J. 1996. Amphibian distribution patterns on the Fort Lewis Military Reservation: Associations with impoundments and exotic vertebrates. Unpubl. Final Rep. to U. S. Army. 62pp.

Adams, M. J., K. O. Richter, and W. P. Leonard. in review. Surveying and monitoring amphibians using aquatic funnel traps. p. 52-61. In D. H. Olson, W. P. Leonard, and R. B. Bury, eds.). Sampling Amphibians in Lentic Habitats.

Gibbs, J. P. 1995. Monitor: Users Manual, Department of Biology, Yale University, New Haven, CN, USA.

Richter, K. O. 1995. A simple aquatic funnel trap and its application to wetland amphibian monitoring. Herpetological Review 26:90-91.

West, S. D., and M. J. Adams. 1996. A Survey of Amphibians on the Submarine Base Bangor, Kitsap County, Washington. Unpubl. Final Rep. to U. S. Navy. 29pp.

ACKNOWLEDGEMENTS

We thank J. Gibbs for advice on parameterizing the program 'Monitor'. J. Erickson provided helpful comments on this manuscript.

 

U.S. Department of the Interior
U.S. Geological Survey
Patuxent Wildlife Research Center
Laurel, MD, USA 20708-4038
http://www.pwrc.usgs.gov/naamp3/naamp3.html
Contact: Sam Droege, email: Sam_Droege@usgs.gov
Last Modified: June 2002