Relative pollen productivity (RPP) serves as an indispensable parameter for supporting quantitative palaeovegetation reconstruction based on fossil pollen, making the accurate determination of RPP critically significant. Focusing on the Qaidam Basin, northern Tibetan Plateau, this study conducted systematic field vegetation investigation and pollen analysis. To obtain the relevant source area of pollen (RSAP) and RPP values for eight key plant taxa, the Extended R-Value (ERV) model was employed. Additionally, we explored the primary factors influencing the differences in RSAP and RPP values across desert steppe in China. Our results showed that the combination of ERV sub-model 2 and Prentice model for estimation yielded optimal results (RSAP = 1710 m), indicating that pollen has the optimal relationship with maternal plants at this distance. Taking Amaranthaceae as the reference taxon, Artemisia exhibits a higher RPP value, while those of Asteraceae, Nitraria, Ephedra, Tamaricaceae, Poaceae, and Calligonum are relatively low. Notably, the RPP value of Calligonum represents the first documented record in China. Variations in RSAP were mainly influenced by sedimentary basin size, vegetation spatial structure, and topography. For the same plant taxon, variations in RPP value were primarily attributed to species composition, fall speed of pollen (FSP), and reference taxon. The desert steppe RPP dataset based on pollen data can provide critical boundary conditions for the quantitative reconstruction of palaeovegetation and further offers essential foundational data to support vegetation-climate coupling simulations.