Root traits are fundamental characteristics of belowground ecosystems that regulate plant growth and drive ecosystem functioning. Nevertheless, the way root traits respond to environmental factors and consequently influence productivity remains unexplored on large geographic scales. We examined the root traits of exotic Spartina alterniflora and native Phragmites australis across China’s coastal salt marshes. Using structural equation models (SEMs), we quantified the direct and indirect effects of mean annual temperature, soil nutrients (e.g., soil dissolved inorganic nitrogen and phosphorus), and root traits on aboveground net primary productivity. Our results showed that root traits of S. alterniflora were more sensitive to changing soil nutrient availability than those of P. australis. The SEMs indicated that soil nutrient availability increased S. alterniflora productivity by increasing root nitrogen concentration and root length density. In P. australis, temperature could increase productivity by both increasing root length density and soil nutrient-mediated root nitrogen concentration. The studied root architectural trait (root length density) and nutrient trait (root nitrogen concentration) were effective in predicting productivity, whereas none of the root morphological traits (i.e., specific root length, root tissue density, and root diameter) significantly affected productivity. We provide the first empirical evidence that root trait-based responses modulate the effects of climate and soil nutrients on geographic variation in vegetation productivity, but these effects are species specific.