Knowledge of quantitative palaeoclimates is a crucial for the evaluation of climate changes for the earth system. In order to improve the reliability of climate reconstruction,especially the climatologies during the glacial periods outside the modern observed climate space, an improved inverse vegetation model has been designed to quantitatively reconstruct past climates, based on pollen biome scores from the BIOME 6000 project. The method has been validated with surface pollen spectra from China by reconstructing the modern climate at each site and comparing it with the observed values. There are no systematically regional errors between pollen biomes and reconstructed biomes by inverse vegetation method. The high correlation coefficients (R) between the actual and reconstructed climate for the present-day pollen sites are 0.95,0.89, 0.82, 0.89, and 0.94 for the mean temperature of the coldest month, the growing degree-days above 5℃,the ratio of actual to equilibrium evaportranspiration, the annual mean temperature and the annual precipitation, respectively. It demonstrated that the inversion method worked well for most climate variables in China. This new approach can improve our understanding on the climate changes of East Asian monsoon evolution.