Reliable and realistic models for lignin content determination in poplar wood based on FT-Raman spectroscopy

Studying lignin content is critical for analyzing the inherent constitution, performance, and application of lignocellulosic materials. However, the traditional methods (wet chemical methods) for the determination of lignin content have several limitations such as labor intensive, time-consuming preparations, and the use of toxic reagents. To address these shortcomings, based on 140 groups of poplar samples, a strategy of 1064 nm FT-Raman spectroscopy combined with several algorithms was proposed in our study. Before modeling, smoothing algorithm (Savitzky-Golay), baseline correction algorithm (adaptive iteratively reweighted penalized least squares), and deconvolution algorithm (Gaussian and Lorentzian) were applied to extract the Raman information. Also, several peaks including 1095, 1378, and 2895 cm(-1) were selected as the candidates of internal standard peak for normalizing the lignin-related peaks. Subsequently, lignin content predictive models were established based on data extracted from FT-Raman spectra combined with regression algorithms, including principal components regression (PCR), partial least square regression (PLSR), ridge regression (RR), lasso regression (LR), and elastic net regressions (ENR). Consequently, credible models were obtained based on data normalized by peak of 2895 cm(-1) combined with RR, LR, and ENR algorithms (Pearson's R > 0.88, RMSE < 0.62, MAE < 0.46, and MAPE < 1.92%). Thus, the lignin quantitative models based on Raman spectroscopy developed here can be used to predict the lignin content in poplar clones accurately and efficiently.

Automated summary

In this study, a strategy of 1064 nm FT-Raman spectroscopy combined with several algorithms was proposed to establish lignin content predictive models with reliable results. Internal standard peaks were selected for data normalization, and credible models were obtained based on data normalized by a specific peak combined with regression algorithms. The developed models can accurately and efficiently predict the lignin content in poplar clones.