An essential ecosystem service is the dilution effect of biodiversity on disease severity, yet we do not fully understand how this relationship might change with continued climate warming and ecosystem degradation. We designed removal experiments in natural assemblages of Tibetan alpine meadow vegetation by manipulating plot-level plant diversity to investigate the relationship between different plant biodiversity indices and foliar fungal pathogen infection, and how artificial fertilization and warming affect this relationship. Although pathogen group diversity increased with host species richness, disease severity decreased as host diversity rose (dilution effect). The dilution effect of phylogenetic diversity on disease held across different levels of host species richness (and equal abundances), meaning that the effect arises mainly in association with enhanced diversity itself rather than from shifting abundances. However, the dilution effect was weakened by fertilization. Among indices, phylogenetic diversity was the most parsimonious predictor of infection severity. Experimental warming and fertilization shifted species richness to the most supported predictor. Compared to planting experiments where artificial communities are constructed from scratch, our removal experiment in natural communities more realistically demonstrate that increasing perturbation adjusts natural community resistance to disease severity.