Evaluating camera survey methods for estimating Eld’s deer abundance in the Hainan Datian National Nature Reserve

Abstract Accurately estimating population abundance is crucial for conserving endangered ungulates like Eld’s deer ( Rucervus eldii ), yet methodological challenges persist in tropical ecosystems. To evaluate different approaches, we compared two camera-trap-based models – spatial mark-resight (SMR) and the random encounter model (REM) – for estimating Eld’s deer abundance in Hainan Datian National Nature Reserve, China, using the same dataset. Model estimates were also referenced against a preliminary belt transect survey count. From October 2021 to January 2022, we deployed 50 camera traps and utilized GPS data from eight collared deer under a closed-population assumption. The SMR model, incorporating spatial capture-recapture data, yielded a density of 18.45 ± 2.21 deer/km 2 . The REM model yielded a lower density estimate of 6.25 ± 0.11 deer/km 2 . Between the two models, the SMR estimates were similar in magnitude to the belt transect survey results (19.78 ± 0.76 deer/km 2 ). Our analysis indicates that SMR provides more robust estimates when marked individuals are available, whereas REM offers a practical solution for unmarked populations but depends critically on variation in encounter rates among camera stations. This comparison highlights the performance characteristics and parametric sensitivities of both models in a tropical context, highlighting the need for careful method selection based on individual identifiability and parameter availability in practice.