Understanding the distribution of plant functional traits and functional entities in forest communities can contribute to exploring how species loss impact on forest ecosystem function, redundancy, and resilience. To figure out the ecosystem performance of fragile karst hills forests in response to biodiversity loss, this study analyzed woody plant data from two 1hm² deciduous broad-leaved mixed forests which is located in the karst hills of Guilin. The analysis included calculating functional diversity based on functional traits, constructing functional entities for computing functional redundancy, and applying Pearson analysis and Mantel test to examine the maintenance mechanism of species diversity indicators in the ecosystems. The results showed that: (1) All the functional diversity indicator of Quercus glauca+Zelkova schneideriana community were lower than those of Psydrax dicocca+Quercus glauca+Triadica rotundifolia community, and the functional evenness was not correlation between two communities (P > 0.05). Additionally, the functional richness, functional dispersion and Rao's quadratic entropy of the two communities showed highly significantly correlated (P < 0.001), and the functional divergence indicates showed significant correlation (P < 0.05). (2) The species richness and functional redundancy indices of the two communities exhibited similar linear relationship. Specifically, species richness was positively correlated with functional entities rank, functional redundancy, and functional over-redundancy, while species richness was negatively correlated with functional vulnerability. (3) There were strong correlations among different plant functional traits, different functional diversity indicators and different functional redundancy indicators, and no significant correlations between functional diversity indicators and functional redundancy indicators, but there were different degree of correlations between functional traits and functional diversity indicators, and between functional traits and functional redundancy indicators. It should not be ignored that the correlations between functional entities and species diversity indicators showed similar significance to species richness and species diversity indicators. In addition, species abundance was significantly correlated with species richness, functional divergence, functional dispersion, Rao's quadratic entropy and functional vulnerability. In conclusion, in evergreen deciduous broad-leaved mixed forest of karst hills, communities with high functional diversity demonstrated highly functional redundancy, while functional diversity and functional redundancy were independent factors. Furthermore, it is noteworthy that the insurance effect provided by communities with high species richness could not fully offset the vulnerability of the ecosystem. To realize the conservation of forest ecosystems in karst hills more effectively, not only maintain the unique function of the ecosystem only by protecting the species richness, but also should fully consider the contribution of species abundance to the ecosystem function.