Understanding long-term herbivore impacts on plant communities is essential for ecosystem management. However, most studies examine short-term responses or treat plant communities as homogeneous vegetation, potentially missing life-form-specific trajectories of plants and their interactions with environmental change. We hypothesized that (1) deer impacts would vary among plant functional groups based on different palatability and browsing tolerance, and (2) these associations would strengthen over time with cumulative browsing pressure. Our objective was to test whether effects of deer on plants vary by their dominant life-form (trees, shrubs, graminoids, forbs) and whether these associations strengthen or weaken over time after accounting for concurrent environmental change. To achieve this objective, we examined associations between Sika deer (Cervus nippon) and multi-decadal shifts in plant community structure across the Chugoku region, Japan, using Japanese national survey data on natural environment by Ministry of Environment for both sika deer and plant communities from 1978, 2003, and 2020 (3,808 site-year plant community observations from 1,270 sites), with deer presence/absence linked to each observation.
We fitted life-form-specific binomial generalized linear mixed models (logit link) to estimate occurrence probability for each functional group while accounting for spatial clustering via Mesh7 random effects (5 km² grid cells). Fixed effects included deer presence, deer duration index (0–3 years cumulative presence), year (reference: 1978), deer presence × year interactions, and standardized covariates representing climate (temperature, precipitation, snow depth), topography (elevation, slope, aspect), and land-use/land-cover composition (K-1 approach with paddy field as reference).
After controlling environmental covariates, deer associations were strongly life-form-specific and temporally dynamic. Shrubs exhibited increasingly negative associations, strongest in 2020 (deer presence×2020: β = -0.73, p<0.001; net odds ratio ≈ 0.55), indicating sustained browsing pressure. Trees showed unexpected temporal reversal: initially negative in 1978 (β = -0.12, p=0.044) but shifting positive through significant interactions with later years (presence×2003 and ×2020 both p<0.001; net 2020 odds ratio ≈ 1.31), suggesting competitive release or browse escape. Forbs displayed negative deer associations in 1978 (β = -0.32, p=0.012) without evidence of intensification, while graminoids showed weak negative duration effects (β = -0.10, p=0.015) with minimal support for deer presence terms.
Our results demonstrate that long-term plant community responses to deer are life-form-dependent, temporally dynamic, and occur alongside broader climate and land use change. Findings support different management strategies prioritizing vulnerable shrub communities.