Detection bias in monitoring birds by camera trapping in Chinese forests: An integrated analysis of functional traits and phylogeny

Infrared camera trapping technology has been widely employed to monitor bird diversity. However, little is currently known about the extent to which this technique contributes to regional bird distribution and abundance, the nature of the sampling biases inherent in such monitoring, and the functional traits that give rise to these biases. We utilised monitoring data from 91 nature reserves across China and integrated avian taxonomy, functional traits, and phylogeny. Bayesian and random forest models were employed to investigate detection bias and identify the key traits influencing the detection of birds by camera trapping. The results show that: (1) camera trapping recorded approximately 42% of China’s forest bird species, with variations in detection frequency across different orders and families; specifically, the families Phasianidae, Muscicapidae, Corvidae and Leiothrichidae were detected more frequently than all other families; (2) the avian community detected by camera traps showed clear functional trait divergence, yet remained predominantly assembled by stochastic processes; (3) Bayesian and random forest models identified tarsometatarsal length and feeding behaviour were key predictors of the probability of being detected by the camera, and there was a significant interaction between the them. In particular, tarsometatarsal length ( β = 0.04, 95% CI [0.02, 0.06]) and herbivory behaviour ( β = 0.03, 95% CI [0.01, 0.04]) showed a positive effect, whereas vertebrate behaviour ( β = −0.04, 95% CI [−0.07, −0.02]) had a negative effect, while other traits played only a minor role. These findings highlight the inherent detection biases associated with birds monitoring using camera trapping, providing a useful theoretical foundation and methodological guidance for targeted bias correction and the optimized application of camera trapping technology in bird diversity assessment.