Knowledge of the leaf characteristics of the coffee tree, such as leaf dimensions, is of great importance for management of this crop, since it directly impacts on plant development. We evaluated the genetic diversity of 43 Coffea canephora genotypes and developed and compared mathematical models for estimating the leaf area of distinct genotypes using linear characteristics. Leaves from 2½ year old trees were collected from the upper middle third of the plant and the length of the central vein and maximum width of the leaf were measured; the leaf area was subsequently measured to determine real leaf area (RLA). The variables leaf length (L), leaf width (W), RLA and length x width (LW) were subjected to Pearson correlation analysis and grouped by the Tocher optimization method. All combinations were tested by linear models according to the measured parameters, and for each model R 2 was adjusted and Bayesian information criterion tested. After choosing the variable, equations were defined considering two parameters, which were ©FUNPEC-RP www.funpecrp.com.br Genetics and Molecular Research 18 (4): gmr18486 D. Dubberstein et al. 2 subjected to cross-validation by comparing between observed x predicted areas. The 43 genotypes formed three groups according to the Tocher procedure, wherein one group was comprised of 41 genotypes. High Pearson linear correlations were found between LW x RLA (0.99), followed by W x RLA (0.95), and as such, LW best estimated the coffee leaf area; but the variable width can also be adopted, with greater ease of field measurement. The equations designed including both variables were significant at 1% and 0.1% according to the F test, and cross-validation analysis confirmed the adjustment of the equations, with equal or very similar values.