Is it meaningful to estimate a probability of extinction?

Abstract

I calculate probabilities of quasi-extinction of natural populations from time series of census data or estimated abundance. The results depend sensitively upon estimated parameters. Short time series or poor fits of the model to data lead to wide confidence intervals for the probability of quasi-extinction. In many cases there are substantial errors in estimated abundance. These errors further increase the size of the confidence interval, to the point where the estimates may become meaningless. A further complication is that natural populations are subject to occasional catastrophes, and these increase the probability of extinction. Unfortunately, evidence concerning the frequency of occurrence of catastrophes and the distribution of their sizes is generally lacking. Assessments that fail to compute confidence intervals or that disregard observation errors or possible catastrophes may be unduly optimistic about the viability of natural populations.