Calyces of Hibiscus sabdariffa are traditionally used to prepare a red beverage rich in anthocyanins (delphinidinand cyanidin-3-O-sambubioside) by decoction in water. These molecules are subjected to degradation during storage leading to a undesirable color intensity decay and change from red to brown in few months. The chemical origins of this quality loss have been already identified as scission and mainly condensation reactions leading to uncolored or brown polymers of anthocyanins (Sinela et al., 2016). Scission and condensed products are known to be instable and/or of various molecular weight. Their concentration is therefore very difficult to get unless numerous time- and money-consuming analysis. An alternative of getting mecanistic and kinetic insigth into anthocyanin degradation during processing or storage is the UV-VIS spectra analysis with chemometric tools. Indeed, UV-VIS spectra are easy to get, but may be complex to interpret as they result from the contribution of all molecules that absorb in this spectral region, which is the case of most of the polyphenols. The objective of this study is to use spectral deconvolution by mean of Multivariate Curve Resolution- Alternating Least Squares (MCR-ALS) on spectral evolution of Hibiscus sabdariffa extract obtained by soaking calyx powder in water at a ratio 1/9 (g/g) during one month of storage at 37°C. In this work, we propose a twosteps approach: a first deconvolution on two fractions of the extract, low molecular and high molecular weigth fractions obtained by size-exclusion chromatography and a MCR done on the spectrum of the whole extract. Absorbance from 250 to 700 nm (each 0.5 nm) was recorded on 10 samples at different storage times. MCRALS was carried out on the 10 spectra using a toolbox developped by Jaumot et al. (2005) on Matlab (The MathWorks, Inc., Natick, MA, USA) with a non negativity constraint algorithm. To improve the mathematical resolution and the chemical validity, the data set was aug