Therefore, we additionally performed simulations for instant decline (in one generation), of the same degree as above, for example, 97.5% instant decline gives population sizes at each generation of: 2000, 50, 50, 50 … 50.
In marked contrast to results for the exponential decline, the temporal arrangement of samples following instant decline (clustered samples, before and after start of decline, etc).
Under instant decline, on the other hand, using K, decreased genetic diversity can be identified with good power within one or two generations after decline, for severe (99%) or moderate (97.5%) declines, and notably even while population size remains constant.
For most pairwise comparisons, especially for comparisons separated by two or more generations, using K, power for instant declines was >0.90.
Therefore, if historical samples are unavailable, for instant declines in particular, monitoring schemes should be implemented as soon as possible, before all variation disappears.
For instant declines, power is usually improved by approximately 0.20 if samples are available before the onset of the decline as compared to when sampling started multiple generations into the decline.
However, weak (90%) declines yielded lower power, in agreement with Pinsky and Palumbi (2014), who tested power to detect genetic erosion after 90% instant declines in large fish stocks.
Instant declines may occur from disease outbreak, natural or anthropogenic catastrophe (e.g., oil spill, volcano eruption), population collapse (e.g., fisheries), or sudden over-exploitation of wildlife resources (e.g., American bison, northern elephant seal, exotic pet trade).
The power to detect genetic erosion differed strikingly for exponential and instant declines, and among indicators, while the precise number and distribution of temporal samples available has less effect.
However, for instant declines, some later pairwise comparisons (e.g., comparing generations 8 and 10) show low or no power (Fig. 4), possibly because alleles at many loci become fixed, after which no significant differences could be further observed.
This is readily observed in the half-matrix of power for instant declines which shows power increasing steadily with time between samples, whether early or late in the bottleneck, making the matrix relatively symmetric (though with some loss of power in the latest generations, see Discussion).
