Mathematical models constitute a simple and accurate tool for non-destructive determination of leaf area. They have proven to be important in experimental situations where leaf area meters are unaffordable and unavailable. The purpose of this study was to develop a model for predicting leaf area of Gmelina arborea from leaf linear dimensions. Intact leaves were collected from trees in the Tubah Upland Forest of the Bamenda Highlands forest in northwest Cameroon. The length (L) and width (W) of each leaf were measured. Leaf area (LA) was determined by counting the number of cells and squares that fell within the boundary of the leaf margin traced on a graph paper. The measurements were subjected to regression analysis with LA as dependent variable and L, W, L×W as independent variable, while adjusting for linear, quadratic, and cubic equations. Goodness-of-fit was evaluated from the coefficient of determination (R2) and model accuracy was assessed from Relative Root Mean Square Error (RRMSE). Of the three variables and types of polynomial tested, the quadratic regression with W resulted in the highest R2, indicating that 96% of variation in leaf area is explained by this function. Model accuracy was generally good as demonstrated by values of RRMSE that ranged from 10.22 to 12.40. A plot between predicted and observed leaf areas resulted in a strong significant highly positive fit (R2 = 0.92, p = 0.00). It is therefore concluded that the model LA = -498.47+56.33(W)+1.53(W)2 can be applied for an estimation of leaf area in Gmelina arborea.