Methods Data for 10 phenotypic traits of seeds and fruits including fruit diameter, peduncle length, thousand kernel weight, seed length, seed width, seed wing length, seed wing width, seed width ratio, seed wing width ratio, ratio of seed length and seed wing length, were sampled from 36 natural L. formosana populations across China. Analysis of variance (ANOVA), correlation analysis, and regression analysis were used to analyze the phenotypic variations among and within populations.

Important findings Results showed that the mean phenotypic differentiation coefficient was 57.55%, which indicated that the phenotypic variation among populations was the main source of the phenotypic variation in L. formosana. The mean coefficient of variation was 15.83%, ranging from 10.05% to 24.31%, suggesting that the phenotypic variation of seeds and fruits was very plentiful. The coefficient of variation was different among populations. The mean coefficient of variation in Anfu was the highest (16.73%), while that of Jianyang was the lowest (11.48%). The pattern of geographic variation of most seed and fruit phenotypic traits was random. There was no significant correlation between the adjacent populations in phenotypic traits. The mean temperature in January, altitude and latitude played important roles in affecting the phenotypic traits. The length-width ratio of seeds decreased with the increase of altitude, which can be used to infer the altitude of the distribution of a population. Our results provide useful information for the collection of wild resources and the breeding of this economically important species.