Biodiversity loss poses a significant threat to ecosystem functioning. However, much of the empirical evidence for these effects is based on artificial experiments that often fail to simulate the structure of natural communities. Hence, it is still unclear whether natural diversity losses would significantly affect the functioning of "real-world" ecosystems. As subordinate and rare species constitute most of the diversity in natural communities and are often more vulnerable to local extinction, we evaluated their contribution to ecosystem functioning in a naturally species-rich grassland. We focused on two mechanisms by which they can support ecosystem functions: redundancy and complementarity. We conducted two long-term field experiments (>6 years) simulating contrasting biodiversity loss scenarios through the manual removal of plant species and measured the consequences of species loss on various ecosystem functions related to carbon dynamics. The latter were examined seasonally to explore diversity effects outside the typical peak of vegetation. We found that dominant removal led to substantial reductions in aboveground phytomass and litter production and altered the annual carbon fixation capacity of the vegetation, highlighting the pivotal role of dominant species in driving ecosystem functioning. Despite high species diversity, other species could not fully compensate for the loss of a single dominant even after more than 25 years, challenging assumptions about redundancy. Complementarity effects were not detected at the peak of vegetation but were evident in early spring and autumn when subordinate and rare species enhanced ecosystem functions. Surprisingly, belowground phytomass, soil organic carbon content, and litter decomposition were unaffected by species removal, suggesting complex interactions in belowground processes. These findings underscore the importance of dominant species in maintaining ecosystem functioning and emphasize the need for nuanced approaches to studying biodiversity loss in real-world communities. Comprehensive seasonal measurements are essential