The relationship between leaf physiognomy and precipitation has been explored worldwide in regions under different climate conditions. Unlike the linear relationship established between the percentage of woody dicot species with entire margins and mean annual temperature, precipitation has been reported to correlate to different leaf physiognomic characters depending on the region where the correlation is studied. To investigate if precipitation can be calculated from leaf physiognomic characters on the basis of regional sample sites, data from 50 mesic to humid forests in China were analyzed in this study. With data from Chinese forests, the leaf-area analysis based on linear regression between natural logarithms of leaf size and mean annual precipitation (MAP) shows no significant correlation. Both single and multiple linear regression analyses fail to confirm the correlation between leaf physiognomy and precipitation, which may result from the similarity of modern spatial distribution of temperature and precipitation in China. Our results show that, due to variations in climatic conditions among sampling regions, leaf physiognomic characters that correlate to precipitation are not consistent worldwide, and applications of models without considering regional constraints could mislead our understanding of palaeoclimate. Therefore, when choosing a leaf physiognomic model for palaeoclimate reconstructions, it is important to determine if the leaf physiognomy of the palaeoflora lies within the leaf physiognomic spectrum of the model used.

The relationship between leaf physiognomy and precipitation has been explored worldwide in regions under different climate conditions. Unlike the linear relationship established between the percentage of woody dicot species with entire margins and mean annual temperature, precipitation has been reported to correlate to different leaf physiognomic characters depending on the region where the correlation is studied. To investigate if precipitation can be calculated from leaf physiognomic characters on the basis of regional sample sites, data from 50 mesic to humid forests in China were analyzed in this study. With data from Chinese forests, the leaf-area analysis based on linear regression between natural logarithms of leaf size and mean annual precipitation (MAP) shows no significant correlation. Both single and multiple linear regression analyses fail to confirm the correlation between leaf physiognomy and precipitation, which may result from the similarity of modern spatial distribution of temperature and precipitation in China. Our results show that, due to variations in climatic conditions among sampling regions, leaf physiognomic characters that correlate to precipitation are not consistent worldwide, and applications of models without considering regional constraints could mislead our understanding of palaeoclimate. Therefore, when choosing a leaf physiognomic model for palaeoclimate reconstructions, it is important to determine if the leaf physiognomy of the palaeoflora lies within the leaf physiognomic spectrum of the model used.