Surface pressure reflects the deep structure of the overlying atmosphere, and is recognized as an indicator of climate change. In this study, observed surface pressure at 71 stations over the Tibetan Plateau (TP) during 1979–2013 is analyzed and compared with monthly means from multiple reanalyses (NCEP1, NCEP2, ERA-Interim, MERRA and JRA55). During the studied period, surface pressure from both observations and the reanalyses increases slowly up until the mid-2000s but shows a decrease afterwards, leading to a recent fall in pressure. However, the surface pressure over the TP in spring has increased, probably explained by the thermal condition such as diabatic heating change. Observations and the multiple reanalyses are positively correlated at most locations indicating that reanalyses reproduce the interannual variation and long-term trend of observed surface pressure fairly well. Despite high inter-annual correlation, trend magnitudes over 1979–2013 are varied, with observations showing decreased pressure at most stations, but reanalyses showing increases in many cases. Compared with observations however, surface pressures from all reanalyses are underestimated usually by about 3–6%. There are significant positive correlations between surface pressure bias and elevation bias, suggesting that overestimation of elevation partially explains the surface pressure bias. A topographical correction method using the hydrostatic equation is therefore conducted and more than 90% of the biases of the reanalyses can be eliminated. Overall, this study points to the importance of better analyzing the importance of topography in the western TP to enhance understanding of reanalysis uncertainties in this region.