Assessing the influence of climate, soil fertility, and species identity on leaf trait relationships is crucial for understanding the adaptations of plants to their environment and for interpreting leaf trait relationships across spatial scales. In a comparative field study of 171 plant species in 174 grassland sites across China, we examined the trade-offs, defined as negative covariance between two traits, between leaf persistence (leaf mass per area, LMA) and leaf productivity (mass-based photosynthetic rate, Amass, N and P content, and photosynthetic N use efficiency, PNUE). We asked to which extent these trade-offs were influenced by: (1) variation among sites within species, decomposed into variation due to climatic and soil variables; (2) variation among species within sites, decomposed into variation among taxonomic, functional, or phylogenetic groups; and (3) the joint contribution of variation among species and sites. We used mixed-model analysis of covariance to partition bivariate relationships between leaf traits into trade-off components. We found significant mass-based persistence-productivity trade-offs of LMA-Amass, LMA-N, LMA-P, and LMA-PNUE consistent with previous broadscale findings. Overall, (1) variation among sites within species explained 14-23%, (2) variation among species within sites explained 20-34%, and (3) the two together explained 42-63% of the total covariance between leaf traits. Interspecific trade-offs of LMA-Amass, LMA-N, and LMA-P were stronger than inter-site ones. A relatively low amount of covariance was explained by climatic and soil variables. However, we found the trade-offs were stronger for LMA-N and LMA-P at higher precipitation and for LMA-PNUE at greater soil fertility, if displayed by major axis regression, which combined both intra- and interspecific variation. Residual trade-offs within species and sites were weak, suggesting that intraspecific, intra-site variation in physiology was less important than variation imposed by species identity or environmental differences among sites. Our results from grassland biomes add evidence for the fundamental nature of productivity-persistence trade-offs in plants. No individual factor emerged as the single major cause for these trade-offs. Rather, the total covariance between leaf traits was explained by a combination of factors, each contributing a range of explanatory power.