We investigated the seed traits of Juglans mandshurica from 14 provenances in Changbai Mountains, northeastern China. Principal component analysis (PCA) and analysis of variance were applied to analyze the variations of seed traits of J. mandshurica from different provenances, in order to reveal the effects of spatial geography and topography on the seed traits. The results showed that there were significant differences among provenances in the yield per plant, weight, length and width of dried fruit. The variation coefficient of yield per plant among provenances reached 36.12%, while the variations in the weight, length and width of dried fruit were relatively smaller, with the variation coefficient ranging from 2.25% to 6.67%. PCA analysis showed that PC1 reflected the fruit yield, dried fruit weight and characteristics of tree morphological, while PC2 expressed the fruit phenotypic traits, such as the length and width of dried fruit and the diameter at breast height. The variation partitioning analysis showed that spatial variables could explain 13%-69% of the variations in the yield per plant, weight, length, width of dried fruit, tree height and diameter at breast height (DBH). Topographic variables could explain 19%-24% of the variations in the yield per plant, width of dried fruit, tree height and DBH. The spatial variables explained 33% and 24% of the variations of PC1 and PC2, respectively, and topographic variables explained 52% and 66% of the variations of PC1 and PC2, respectively. The correlation analysis represented that the yield per plant had significantly negative correlations with slope (P0.05) and width of dried fruit significantly positive correlations with elevation (P0.05). The cluster analysis showed that group II with the highest yield was the ideal group of fruit, while group III with the lowest yield but the highest DBH was the superior group of wood. Therefore, suitable provenances should be chosen according to the breeding goal in forestry practices.