Integrating Systematic and Opportunistic Camera Trap Data to Monitor Eurasian Lynx Demography

ABSTRACT

Monitoring large carnivores is essential for understanding population dynamics and conservation outcomes, yet it is often constrained by limited resources and imperfect detection. Camera trapping is a cornerstone of such monitoring, and citizen science offers opportunities to extend temporal and spatial coverage. Here, we assess the potential and limitations of integrating systematic and opportunistic camera trap data to monitor the demography of the Eurasian lynx ( Lynx lynx ) in Croatia. We analyzed 4201 lynx photographs collected over 14 monitoring seasons (2010–2024), combining systematic camera trap surveys conducted between 2018/2019 and 2023/2024 with opportunistic records gathered throughout the entire study period. Opportunistic camera trapping contributed 41% of all records and increased sampling effort, while the proportion of photographs suitable for individual identification remained stable across seasons. In total, 225 individual lynx were identified, with individuals recorded on average 17.4 times. Most individuals were redetected within 2 months, whereas the longest interval between successive detections exceeded 4 years, indicating pronounced heterogeneity in detectability. Between 2018 and 2023, 201 individuals were identified, and minimum population size increased from 64 to 89 individuals. The proportion of individuals of unknown sex remained high across seasons (mean 33%) and showed a moderate positive correlation with the number of opportunistic photographs, suggesting limitations in sex assignment from opportunistic data. Mean individual turnover was estimated at 31%, with interannual variation (25%–39.2%) and mean age of 2.59 (±1.94) years, while the oldest identified individual had 9.41 years. Our findings show that integrating opportunistic records with systematic camera trapping collected year‐round enhances demographic monitoring of Eurasian lynx, likely increasing detection probability while highlighting the need to account for detection heterogeneity and data‐quality limitations.