In this paper, correlation between modern leaf physiognomy and climate in China are examined, to optimize the use of leaf characters as a palaeoclimate proxy. A large dataset was compiled, recording the distribution of leaf physiognomic characters among 3166 native dicot trees species across 732 calibration grids on a county level. Grids span a range of ecological environments (tropical rainforests to alpine shrubs) across humid areas. Thirteen climatic parameters were included for each grid and 22 leaf physiognomic characters were scored for each tree species. The correlation between leaf physiognomic characters and climatic parameters were calculated based on single linear regressions (SLR) and multiple linear regressions (MLR). Results indicate clear spatial distribution patterns, linked to latitude, exist for all leaf characters, with temperature (Coldest Month Mean Temperature, CMMT) and precipitation (Growing Season Precipitation, GSP) being the main climate controls. Moreover, because leaf characters are more closely correlated with Precipitation during the Three Wettest Consecutive Months (P3WET), rather than with Precipitation during the Three Driest Consecutive Months (P3DRY), seasonal variations in rainfall associated with the Asian Monsoon might especially influence leaf physiognomic characters. Closer correlations between leaf physiognomy and climate are seen using MLR compared with SLR; therefore Mean Annual Temperature (MAT) and Mean Annual Precipitation (MAP) based on MLR equations provide the most promising basis for palaeoclimate reconstructions in China.