When Laser-Induced Breakdown Spectroscopy (LIBS) spectra exhibit complex spectral features as the result of two or more overlapping emission lines, it is common to apply multi-peak fitting in order to extract the relevant information. In this paper, we propose an alternative method to multi-peak fitting based on advanced data processing. To illustrate this new strategy, we consider the quantification of lithium from a series of spectra related to samples being part of the calibration samples for the ChemCam instrument dedicated to the analysis of Mars geological matrices. In this case, one can observe a spectral overlap between the unresolved doublet of lithium at 670.78 and 670.79 nm and the line of calcium at 671.769 nm. A first PLS model built from 1713 variables related to the spectral range 492.7-856.8 nm revealed the strong influence of this doublet of lithium but also a significant overfitting and instabilities preventing from any exploitation of this model. Reducing the number of variables to 17 corresponding to the spectral window 669.07-672.49 nm then allowed to build a 3-component PLS model characterized by the apparently satisfying indicators, R-2 = 0.95 and RMSECV = 11 ppm. However, this model was not considered as satisfying due to the significant nonlinearity between LIBS signal and concentration values, which it was unable to describe, since it is linear by definition. Thus, our PLS model was exploited through the loading weights plot of the first component, explaining > 84% of the variance of the lithium concentration values, which allowed to select the 9 variables exhibiting the highest correlation with the lithium concentration values. Then, the sum of the related values of intensity was exploited to build a univariate quadratic model, characterized by R-2 = 0.93 and RMSECV = 17 ppm. Comparing this result with the previous results obtained after multi-peak fitting, namely R-2 = 0.92 and RMSECV = 18 ppm, we conclude that our alternative approach can be considered as a satisfactory option.< Leer menos