Nutritional status in citrus plants, which is used as a guide for fertilisation, is normally determined by chemical analysis of leaves. According to standardised procedures, this is a destructive method. Leaf analysis detects symptomless detrimental conditions or confirms the nature of visible toxicity. This study proposes the use of a rapid, non- destructive, cost-effective technique to predict orange leaves nutritional status utilising a Vis–NIR (visible–near infrared) portable spectrophotometer and compares its results with standard chemical analyses. Tree nutritional status was evaluated by foliar analysis per- formed on 50 leaves. Chemical determinations on leaves detected N, P, K, Ca, Mg, Fe, Zn, Mn. For spectral acquisition, a ‘pen probe’ was used to measure the spectral reflectance response on each leaf. Mean reflectance values of all leaves for each treatment were compared by chemometric multivariate methods (PLS, partial least square) to both: a single reference chemical value and to all chemical parameters used together. The best model for single reference chemicals (coefficient of correlation r 1⁄4 0.995) and the tests (r 1⁄4 0.991) was obtained for potassium. Results also showed a high efficiency in the determination of nitrogen. For all chemical parameters used together, the analysed elements gave correla- tions in a range from r 1⁄4 0.883 for Mg to r 1⁄4 0.481 for P with standard error of prevision ranging from 0.01 for P to 12.418 for Fe.