Ecology. 75(5), 1994, pp. 1512-1515 CO 1994 by the Ecological Society of America stable in old than in young stands. squirrels were neg- atively affected by forest fragmentation. and they were isolated by intensively managed timber lands and con- version of forest to nonforest habitats. The concluded NORTHERN SPOTTED OWLS: that woodrats showed similar patterns and that the home range size of Spotted Owls was related to prey INFLUENCE OF PREY BASE— abundance. These conclusions were based on compar- A COMMENT isons of flying squirrel and woodrat abundances among forest types, seral stages, and geographic regions. These Daniel K. Rosenberg. Cynthia J. Zabel, Barry R. issues are critical to resource land managers who are Noon, and E. Charles Meslow interested in silvicultural prescriptions that might en- hance prey densities for Spotted Owls and thereby in- Conflicts between the needs of the Northern Spotted crease the owls reproductive success and reduce the Owl (Stria occidentalis caurina) and the maintenance amount of foraging habitat used by the owl. In our of a timber-based economy in the Pacific Northwest comment, we question whether the results and meth- have motivated research on the habitat and area re- ods used by Carey et al. (1992) substantiate their con- quirements of the owl. The Northern Spotted Owl has clusions, and cite previous studies that are not consis- been studied extensively throughout its range and con- tent with their conclusions. sistently has been found to select older forest stands for foraging, roosting, and nesting (reviewed in Thomas Concerns Regarding Methods Used to et al. [1990]). At least six hypotheses have been pro- Estimate Prey Densities posed to account for selection of older forests by Spot- Flying squirrels were live-trapped in approximately ted Owls (Forsman et al. 1977, 1982, 1984, Carey et 10 x 10 grids, with trap stations spaced 40 m apart al. 1990, 1992. Rosenberg and Anthony 1992). Two within each grid. Trapping grids were set in 17 young of these relate to their prey. The prey hypotheses sug- and 17 old stands. for 8-10 nights. Trapping grids were gest that prey are more available to Spotted Owls by placed in three forest types: western hemlock (Tsuga either or both of two mechanisms: (1) prey are simply izeterophylla), Douglas-fir (Pseudotsuga menziesii), and more abundant in older forests (prey abundance hy- mixed conifer. Although 17 stands of each age class pothesis). and (2) prey are more efficiently hunted in were sampled, stands were not sampled consistently older forests because of an increase in the ability of (Carey et al. 1992: Fig. 4). The total number of samples owls to forage through more open understory structure (i.e., number of stands x number of times sampled) is (prey availability hypothesis). clear from only one region (southwestern Oregon: Ca- Carey et al. (1992) recently reported on associations rey et al.: Fig. 4): 43 for old stands and 18 for young between Spotted Owls and their prey in Oregon and stands. However, the total number of different stands Washington. They concluded that the abundance of sampled from this region was not mentioned. The prey across a landscape determined the carrying ca- number of stands sampled each year in this region pacity for Spotted Owls and that within-landscape fea- ranged from 1 to 8 (.t ± I SE. = 4.9 ± 0.7 stands) for tures such as stand age and forest fragmentation af- old stands, and from 0 to 5 (2.0 ± 1.3 stands) for young fected the abundance and demographic characteristics stands. Two or more stands of each age class were of prey. Northern flying squirrels (Glaucomys sabrinus) sampled in only five of nine (season—year) sampling and dusky-footed and bushy-tailed woodrats (Neotoma periods. However, six sampling periods were analyzed %uscipes and N. cinerea, respectively) are the most com- using a paired t test to compare flying squirrel density mon prey of the Northern Spotted Owl throughout its estimates between stand age classes. Each observation range, constituting the majority of the biomass con- in the paired comparison was apparently the difference sumed by owls (Forsman et al. 1984). Carey et al. of the mean of density estimates for a given year and (1992) concluded that northern flying squirrels were sampling period (spring, fall) between stand age classes more abundant in older forests, populations were more within a region. The analysis (paired I test using stand means) seems inappropriate because it ignores vari- U.S. Forest Service. Redwood Sciences Laboratory. Ar- ability within a stand age class for each sampling pe- cata. California 95521 USA. riod, ignores the difference in sampling intensity (i.e., = Address to which correspondence should be directed: Co- number of stands) between stand age classes. and does operative Wildlife Research Unit. Department of Fisheries not allow for lack of independence of the repeated sam- and Wildlife. Oregon State University, Corvallis, Oregon 97331 USA. pling strategy that was used (i.e.. it artificially reduces U.S. Fish and Wildlife Service. Cooperative Wildlife Re- variance). The rationale for using a paired test is that search Unit. Oregon State University. Corvallis, Oregon 97331 each estimate in the first sample is somehow related USA. with a specific estimate from the second sample (Sne- July 1994 NOTES AND COMMENTS 1513 decor and Cochran 1967:97), and this does not seem very low, and a 10 individuals were captured in only to be the case with the samples presented in Carey et 2 of 79 samples (area—season combinations) in stands al. (1992). An analysis of variance (ANOVA) would sampled three or more seasons. Woodrats were not be more appropriate to use, but must account for the captured during most (72%, 57 of 79) samples (Carey repeated sampling (e.g.. repeated-measures or split-plot et al. 1992: Table 5). These data make unbiased esti- ANOVA) and unequal sample sizes between stand age mation of population size difficult. It is not clear which classes in each season and year. data were used to calculate woodrat densities among Woodrats were live-trapped in the same 10 x 10 forest stand types. It appears that data were combined grids where flying squirrels were trapped. In addition, from grids and line transects. The number of stands in 27 stands were sampled using two parallel line tran- Table 4 (Carey et al. 1992). used to calculate woodrat sects, with 20 stations per transect spaced at 20-m densities by stand type. exceeds the number of stands intervals and two traps per station. Traps were oper- in Table 5 (Carey et al. 1992). where the number of ated for 3-4 nights. individuals captured in trapping grids was presented. Two types ofestimators were used to assess the abun- Long belt transects can sample within-stand variation dance of flying squirrels and woodrats: an enumeration better than trapping grids for woodrat populations (Sa- method (minimum number known to be alive, MKA) kai and Noon 1993). Therefore, biases are likely when and Chapmans modified Lincoln—Peterson (LP) esti- combining population size estimates from these two mator, a mark—recapture method. The estimator used trapping techniques, and the comparison of results was based on the number of individuals captured: MKA among forest stand types is suspect. was used if <10 individuals were captured, and LP if a 10 individuals were caught. Density was calculated Problems with Interpretation of Data as D = N/A, where N was computed as stated above Flying squirrels. Carey et al. (1992:233) stated and A was calculated using area of grid plus one-half ". densities in old forest were consistently higher the mean maximum distance moved within the trap- than densities in managed, young forest . .". Data ping grid. presented by Carey et al. (1992) do not support this The use of two different estimators by Carey et al. conclusion. During three of nine sampling periods (sea- (1992) was inappropriate. The MKA method yields son—year combinations) only old-growth stands were negatively biased estimates whenever capture proba- sampled, and during one of nine sampling periods only bilities are <1.0 (Hilborn et al. 1976, Nichols and Pol- one young stand was sampled (Carey et al. 1992: Fig. lock 1983, Nichols 1986). MKA performed poorly as 4). Thus, a 2 stands of each age class were sampled a relative measure of abundance in simulated data sets only during five of nine sampling periods. Among these that were modeled using flying squirrel population five sampling periods when comparisons between age characteristics (Rosenberg 1991). Flying squirrel cap- classes were possible, the ranges overlapped between ture probabilities vary temporally and spatially (Carey young and old stands in four sampling periods (spring et al. 1991, Rosenberg 1991, Witt 1991), thus making 1986-1989), and mean densities were similar (within comparisons among stands and regions using the MKA 0.5 squirrels/ha, based on Carey et al. [19921: Fig. 4) method inappropriate (Nichols 1986). Carey et al. in at least two sampling periods. For the Olympic Pen- (1992) cited a previous report (Carey et al. 1991) as an insula, no data were presented except the overall (3- apparent justification for using the MKA method. yr) mean and SE of each stand age class and the results However, in that report the number of different indi- of a paired t test. Again, the appropriateness of the viduals captured was compared to results from a va- paired t test is unclear. The coefficients of variation (cv riety of population estimators, an inappropriate way = sE//) of the means were extremely high: 200% for to determine reliability of estimators (White 1992). second-growth and 40% for old-growth.