Advancing snow leopard density estimates to assess population dynamics over time in Afghanistan

Journal of Forestry · Springer Nature · 2026-11-12 · DOI: 10.1007/s10344-026-02176-7

Аннотация

Reliable population estimates for elusive species require accounting for uncertainty in individual identification. We combined AI-assisted image matching with spatial capture–recapture models to estimate changes in snow leopard ( Panthera uncia ) density in Wakhan National Park, Afghanistan, using camera-trap data from 2012–2013 and 2017–2019. We fitted models separately to each two-month sampling session and jointly across the six sessions within each survey period. Because unpaired cameras recorded asymmetric flank patterns, we compared datasets that naively pooled left- and right-flank detections, potentially assigning a single animal multiple identities, with datasets restricted to either right- or left-flank detections. In single-session models, pooled-data density (estimate ± SE) increased from 1.38 ± 0.39 to 3.61 ± 0.90 individuals per 100 km² between survey periods. Corresponding estimates increased from 1.44 ± 0.54 to 3.41 ± 1.04 for right-flank data and from 1.05 ± 0.36 to 2.24 ± 0.73 for left-flank data. Pooling left- and right-flank detections inflated density relative to the mean of the two flank-specific estimates by 10.8% in 2012–2013 and 27.8% in 2017–2019 in single-session models, and by 4.7% and 25.1% in multi-session models, with the largest inflation in the period with the highest detection rates. These findings indicate an apparent increase in snow leopard density in Wakhan while demonstrating that unresolved partial identities can substantially bias population estimates. Explicitly modeling identity uncertainty, or averaging flank-specific estimates when paired cameras are unavailable, can strengthen inference for snow leopards and other individually identifiable species. Continued monitoring is needed to determine whether the apparent increase has persisted.

Читать у издателя →

Похожие статьи