Functional and phylogenetic clustering shape rodent co-occurrence via multi-scale mechanisms in a subtropical forest

Species co-occurrence has been widely investigated in various animal species, yet the roles of trait similarity, scatter-hoarding behaviour and habitat factors remain incompletely understood. We examined the effects of three groups of factors-functional similarity (i.e. phylogenetic, body size, diet), scatter-hoarding behaviour (i.e. scatter-hoarding intensity) and habitat factors (i.e. patch size, seed richness, seed abundance and stand age)-on species co-occurrence probability among sympatric rodent species with 7 years of field data using live and camera traps in a subtropical forest. At the community level, we assessed phylogenetic and functional structure using null model approaches and developed an abundance-weighted average species co-occurrence probability (ASCp) to evaluate assembly mechanisms. We found the species co-occurrence probability of rodents was significantly and positively associated with phylogenetic relatedness, body size similarity and diet similarity of pairs of rodent species using live traps at the species-pair level (such correlations were significant only for phylogenetic relatedness and diet similarity using camera traps). Species co-occurrence probability of rodents was significantly and positively associated with scatter-hoarding intensity using both methods. Community-level analyses revealed significant phylogenetic clustering and functional clustering in scatter-hoarding intensity and diet breadth, supporting a dominant role of environmental filtering. At the community level, ASCp was significantly and positively correlated with patch size when assessed using camera traps, and significantly and negatively correlated with stand age using live traps. Notably, ASCp positively correlates with community-weighted scatter-hoarding intensity, though this effect weakens in larger patches. Our results suggest that species traits, scatter-hoarding behaviour and habitat factors independently shape species co-occurrence patterns of rodents in forest ecosystems across multiple spatial scales, but the underlying mechanisms are distinct, which is explained by environmental filtering (traits), reciprocal pilferage (scatter-hoarding) and habitat heterogeneity (patch habitat factors), respectively. The species co-occurrence probability patterns showed some differences between the two methods, probably because animals may exhibit different responses or sensitivities to various survey approaches.