Abundance Estimators and Truth: Accounting for Individual Heterogeneity in Wild House Mice

Abstract

We compared the actual abundance of 12 confined populations of wild house mice (Mus musculus) with closed population estimates based on mark-recapture data. Goodness-of-fit tests consistently detected individual heterogeneity in capture probability. Estimators designed to take such heterogeneity into account were expected to perform best, and generally did. However, of the 9 abundance estimators we considered, only Chao's modified moment estimator had no obvious bias and produced confidence intervals that always included the actual population size. The reliability of the 9 estimators and minimum number known alive (MNA), as indices of abundance, was quantified by (1) linear regression, and (2) calculating a Spearman rank correlation coefficient between the ranking obtained using the index and the true ranking. We found Petersen estimates to provide the most accurate ranking, and MNA also performed well. Since Chao's modified moment estimator was designed for populations with high levels of heterogeneity and low average capture probability, we concluded that this is an accurate description of trap response for confined mice populations. Body-length frequency distributions and sex ratios for mice caught once, more than once, and not at all revealed that body length and sex could explain some of the observed individual heterogeneity. Males were more likely to enter traps than females, and large mice were more likely to be trapped than small mice. Field populations of house mice are monitored regularly throughout southeastern Australia because high densities cause considerable damage to wheat crops and place substantial stress on farming communities. We found that field populations trapped during the breeding season exhibit heterogeneity in capture probability similar to that for confined populations, with the same relationships between capture probability, sex, and body length. However, these relationships did not hold for the same field population later in the season when breeding and recruitment had ceased. Chao's modified moment estimator was applied to field data and found to be stable when populations were trapped for ≥5 nights, but unreliable trapped for <5 nights.