Improving models to predict holocellulose and Klason lignin contents for peat soil organic matter with mid-infrared spectra

Abstract. To understand global soil organic matter (SOM) chemistry and its dynamics, we need tools to efficiently quantify SOM properties, for example, prediction models using mid-infrared spectra. However, the advantages of such rely on their validity accuracy. Recently, Hodgkins et al. (2018) developed quantitatively predict peat holocellulose Klason lignin contents, two indicators stability major fractions matter. The may help large-scale gradients have been used in various studies. A research gap fill is that these not validated detail yet. What are limitations how can improve them? This study provides a validation with aim identify concrete steps models. As first step, provide several improvements original training data. limitation identified data representative range diverse samples. causes both biased estimates extrapolation uncertainty under In addition, practice produce unrealistic predictions (negative values or >100 mass-%). Our improved partly reduce observed bias, better predictive performance data, avoid predictions. Finally, proof concept contents also be predicted mineral-rich samples (e.g., mineral admixtures potentially soils). key step will collect formed conditions. opens directions develop operational from