In a two-year experiment (2002-2003), five N application rates [0, 60, 120, 180, and 240 kg(N) ha-1, marked N0, N60, N120, N180, and N240, respectively] were applied to sugar beet cv. Rizor arranged in a Randomized Complete Block design with six replications. Leaf shape parameters [leaf area (LA), maximum length (L), maximum width (W), average radial (AR), elongation (EL), and shape factor (SF)] were determined using an image analysis system, and leaf area index (LAI) was non-destructively measured every two weeks, from early August till mid-September (four times). Years, samplings, and their interaction had significant effects on the determined parameters. Fertilization at the highest dose (N240) increased L and sampling×fertilization interaction had significant effects on LA, L, W, and SF. For this interaction, W was the best-correlated parameter with LA and LAI meaning that W is a good predictor of these parameters. Two proposed models for LA estimation were tested. The model based on both leaf dimensions [LA = 0.5083 (L×W) + 31.928] predicted LA better than that using only W (LA = 21.686 W - 112.88). Instrumentally measured LAI was highly correlated with predicted LAI values derived from a quadratic function [LAI = -0.00001 (LA)2 + 0.0327 LA - 2.0413]. Thus, both LA and LAI can be reliably predicted non-destructively by using easily applied functions based on leaf dimensions (L, W) and LA estimations, respectively. and J. T. Tsialtas, N. Maslaris.