Leaf nitrogen content (LNC) and leaf mass per unit area (LMA) were assessed by near-infrared spectroscopy (NIRS) on fresh and dried plants of durum wheat (Triticum turgidum ssp). Individual leaves were scanned with a portable spectrometer and reference analyses of LNC and LMA were then carried out. Partial least squares (PLS) regression was used for calibration and cross-validation. INC was accurately predicted for both fresh and dry leaves whatever the phenologic stage (correlation coefficient of calibration R(2)cal ranging from 0.932 to 0.958, standard error of cross-validation SECV ranging from 0.215 to 0.320% (dry matter)). LMA was predicted with R(2)cal = 0.942 and SECV = 4.84g m(-2). The combination of these two calibrations made it possible to predict leaf nitrogen per unit area (R(2)cross-validation = 0.94, SECV = 0.248 gN m(-2)) and provides a relevant and non-destructive tool for following the dynamics of three major leaf parameters.