Demographic compensation maintains shrub population growth in a grazed desert steppe

Abstract Background and Aims Grazing is a pervasive disturbance in dryland grasslands, yet its effects on shrub populations remain difficult to predict because it can influence different demographic processes in various ways. Understanding whether shrub populations decline or persist under grazing requires demographic approaches that integrate vital rates across the full life cycle in temporally varying environments. Methods We monitored populations of Caragana microphylla, a long-lived leguminous shrub widely distributed across the dryland grasslands of northern China, from 2022 to 2024 in both grazed and grazing-excluded plots in the desert steppe of Inner Mongolia. We modelled size-dependent survival, growth, reproduction probability and reproductive output, and used stochastic integral projection models to estimate stochastic population growth rate (λs). We further used stochastic elasticity analyses and stochastic life table response experiments (LTREs) to identify the vital rates and temporal components underlying treatment differences in λs. Key Results Grazing increased survival and the probability of reproduction, but reduced growth and reproductive output. Despite these divergent vital-rate responses, λs remained close to one in both treatments and was higher in grazed populations. Stochastic LTREs revealed that the higher λs in grazed populations arose primarily from positive contributions of survival across all sizes and of fecundity in small-to-medium-sized individuals. These positive contributions offset negative contributions from growth and shrinkage across all sizes, as well as negative fecundity contributions in larger individuals. Temporal variation in fecundity made a negative contribution to λs under grazing, but this effect did not outweigh the positive contribution of mean vital-rate compensation. Conclusions Overall, C. microphylla populations maintained near-stable stochastic population growth under grazing through compensatory changes among vital rates, even though plant growth and reproductive output declined. These findings demonstrate that shrub population responses to grazing cannot be inferred from changes in individual vital rates alone. Full life-cycle stochastic demographic models provide a useful framework for understanding population dynamics in temporally variable environments.