ESTIMATION OF ANIMAL ABUNDANCE WHEN CAPTURE PROBABILITIES ARE LOW ANDHETEROGENEOUS

Obtaining reliable estimates of abundance or relative abundance under conditions of low numbers of captures and recaptures is crucial to pro perly assess population status of species that are of management conce rn; however, these characteristics make estimation difficult. We appli ed the commonly used jackknife (Burnham and Overton 1978, 1979) and mo ment (Chao 1988) estimators of abundance to capture-recapture data fro m northern flying squirrel (Glaucomys sabrinus) populations that had l ow (p congruent-to 0.10), heterogeneous, capture probabilities and low densities (approx 2 squirrels/ha). The jackknife estimator selection procedure, higher-order jackknife estimators, and moment estimator wer e sensitive to the number of trapping occasions. These estimators tend ed to have low precision. Comparisons of estimators suggested specific , lower-order jackknife estimators performed well. Monte Carlo simulat ions corroborated results from field data. The moment estimator tended to have low bias, but the high root mean square error made the estima tor less reliable than lower-order jackknife estimators. First- and se cond-order jackknife estimators tended to be the most reliable (low bi as and precise) estimators when the number of trapping occasions (t) w as greater-than-or-equal-to 12. However, confidence interval coverage (% replications in which the constructed confidence interval included true N) was low with the first-order jackknife estimator, reflecting t he negative bias of the variance estimator. We improved confidence int erval coverage by an ad hoc adjustment to the variance estimator; cove rage with the adjusted estimator approached the nominal 90% level at t greater-than-or-equal-to 12. Reliable estimates of abundance can be a chieved under conditions often encountered in field studies (small N a nd low, heterogeneous, capture probabilities) with lower-order jackkni fe estimators, a modification of the variance estimator for the first- order jackknife estimator, and greater-than-or-equal-to 12 trapping oc casions.