Earlier work on food plants showed that NIRS applications appeared to be effective for authenticating varieties. In our work, we confirmed that result for coffee varieties, but we also showed that inter-variety relations are not stable from one harvest to the next. We put forward the hypothesis that the spectral signature is affected by environmental factors. The purpose of this work was to find a way of reducing environmental variance in order to increase measurement reliability and enable practical application in breeding. Spectral collections were obtained on ground green coffee samples from agronomy trials. Two harvests of bean samples from 11 homozygous introgressed lines, and the cv ¿Caturra¿ as the control, supplied from 3 different sites, were compared. For each site, squared Mahalanobis distances between the 12 varieties were estimated based on NIR spectra. Matrix correlation coefficients between matrices were assessed by the Mantel test. The higher the coefficients, the more we concluded that inter-variety relations were identical from one environment to another. We showed that very good stability was obtained for inter-variety relations across the sites when using the two harvests data. At the same time, the spectrum was regarded as a succession of phenotypic variables, each resulting from an environmental and genetic effect. Heritabilities were calculated with confidence intervals. The stability of inter-variety relations was largely increased if only the most heritable zones of the spectrum were kept. It could therefore be considered that a coffee variety can be characterized by a NIRS signature acquired over a set of harvests. We indicated how this typical signature can be used in breeding to assist in selection.