Skip to main content
. 2020 Apr 14;10:6393. doi: 10.1038/s41598-020-63367-z

Table 3.

Comparison of the structure of the capture histories (CH) derived from classification of snow leopard images by eight observers (Obs 1–4 were non-experts, Obs 5–8 were experts) with the true capture history (TRUE).

CH structure Population estimate Bias in population estimate
5 4 3 2 1 True Remaining
TRUE 1 2 5 4 4 16.6 ± 0.9 +3.7% +3.7%
Obs1 0 2 3 6 7 20.3 ± 2.0 +27% +35%
Obs 2 1 1 1 4 8 23.9 ± 7.2 +49% +84%
Obs 3 1 2 3 6 6 21.0 ± 2.8 +31% +31%
Obs 4 0 2 2 9 7 22.8 ± 2.3 +42% +42%
Obs 5 1 1 3 5 8 23.5 ± 4.5 +47% +56%
Obs 6 0 3 2 6 9 22.8 ± 2.3 +42% +42%
Obs 7 2 0 5 5 3 16.1 ± 1.3 +0.6% +7%
Obs 8 3 0 2 5 8 23.2 ± 3.9 +45% +45%

The CHs were based on 5 sampling occasions; CH structure shows how many times each snow leopard individual was seen (where the ‘5’ column indicates an individual was recorded in 5 capture events, and the ‘1’ column indicates an individual was only identified by the observer once). Based on each observer’s CH a population abundance estimate was derived using a closed capture-recapture model (mean ± SD; see methods). The bias in the mean estimate for the population is shown relative to the true population size (n = 16) and also relative to the number of unique individuals remaining in each observers’ CH after accounting for animals removed from consideration because of capture event exclusion (for observers 1, 2, 5 & 7 the number of unique individuals assessed was n = 15, 13, 15 & 15 respectively; see Table 1).