In the era of unprecedented biodiversity loss and a subsequent loss of ecosystem function, a core goal of restoration ecology is to identify planting strategies to restore ecosystem functions like soil carbon sequestration. Plant diversity, functional groups and functional traits can all be leveraged to improve ecosystem function, but there are major gaps in our understanding of which strategies are most impactful for increasing soil carbon (C). In this study, we seeded common California grassland species in 210 experimental plots and examined how plant functional groups, functional traits and diversity metrics influenced five unique soil carbon pools. We also evaluated effects of precipitation and land use history treatments on soil C pools across two years to examine how environmental conditions mediate plant-soil interactions, and if there are time lags in those effects. Land use history was the biggest single predictor of soil C, with the duration of shrubs on the site prior to our experiment outweighing the effects of all other variables. In linear mixed-effects model comparisons, environmental variables (precipitation, land use and year) alone best explained particulate organic carbon (POC), while adding species diversity metrics improved predictions for total carbon (TC), microbial biomass carbon (MBC) and mineral-associated organic carbon (MAOC). However, functional and species diversity correlated negatively with soil C pools in some environmental conditions, showing that diversity itself may not always boost soil C levels. Water-extractable organic carbon (WEOC) did not vary with any plant or environmental model. Analyzing effects of individual variables, we found that diversity, functional trait, and functional group correlations varied across soil C pools, environmental context, and the year of vegetation measured. These results suggest that different plant strategies contribute to soil C pools through disparate pathways on various timescales. Our study highlights how dynamic soil C is in grassland systems and that additional work is needed before restoration managers can use trait-based approaches to sequester soil C in various site contexts.