Iron is one of the indispensable nutrition elements to crop, which is a component of haemoglobin, cytochrome, oxidase of cytochrome, catalase, peroxide etc, and it is strongly relative to the formation of aroma precursor. The spatial variability in micro sampling scale (at the meter or submeter level) was more useful for precision farming than that in small or large sampling scale. So a micro sampling scale of 11m×10m was fixed in this paper to study the spatial variability of soil total iron before tobacco planting and after tobacco harvest. Later, the iron content in tobacco leaves sampled from the same scale was mensurated to study the relationship of iron content between soil and flue-cured tobacco leaves. Results showed that the mean of soil iron content reduced to 24.79 g kg-1 after tobacco harvest from 30.3 g kg-1 before tobacco planting. There was a medium spatial correlation (the nugget effect was 48.8%) and highly spatial heterogeneity of the soil total iron before tobacco planting, the optimum theoretical model of soil total iron variation function was a linear one (R2=0.999). And the ANOVA analysis also showed the difference of total iron content among soil samples was significant (p=0.00008). While the optimum theoretical model of soil total iron was a spherical one（R2= 0.956）after tobacco harvest. The maximum total iron content shift a little from north to south, the minimum content shift a little from south to north, and the spatial heterogeneity of soil total iron after tobacco harvest was higher than that before tobacco planting. The iron content of flue-cured tobacco leaves was significantly correlated to the total iron in soil before tobacco planting (X1) not to that after tobacco harvest (X2) and to the difference content between X1 and X2. Finally, a regression equation of soil total iron before tobacco planting（the independent variable）and the iron content of flue-cured tobacco leaves (the dependent variable）was constructed, and test result showed that it had a good expectant effect in the distributing interval of the independent variable.