While three hypotheses (niche dimension, light asymmetry, soil acidification) are proposed to explain the decline of plant diversity following nutrient enrichment, our knowledge of the underlying mechanisms remains vague. To reveal the cryptic mechanisms, we analysed the century-long Park Grass Experiment under modern coexistence theory by fitting Lotka-Volterra competition models with time-series data. Supportive of the last two hypotheses, intrinsic population growth rates (r) decreased and competitive differences or hierarchies reshuffled with nutrient addition and soil acidification remarkably, mostly favouring grasses over legumes or forbs. Moreover, the altered r (affecting species existence) and competitive differences (affecting species coexistence) effectively explained the diversity loss and recovery. However, intraspecific and interspecific competition decreased dramatically but niche differences increased and decreased respectively with nutrient addition and soil neutralisation, poorly supporting the first hypothesis. These findings greatly deepen our insights into fundamental mechanisms underlying the response of plant diversity to nutrient deposition in nature.