PRINCIPAL VARIATIONS IN THE CHEMICAL-COMPOSITION OF PEAT - PREDICTIVE PEAT SCALES BASED ON MULTIVARIATE STRATEGIES

The principal properties, here called the p-scales, of peat have been calculated on the basis of chemical analysis. The scales were derived from quantitative contents of carbohydrates, Klason lignin, amino acids, amino sugars and conventional chemical peat measurements. The variation in the chemical parameters was compressed using principal component analysis (PCA). Partial least squares (PLS) regression was used for prediction of botanical, microbial, physical and dewatering data. A rapid estimation of the scales has been made from near-infrared (NIR) spectroscopy and offers, indirectly, rapidly obtainable, chemically interpretable, biological information. A reduced scale based on carbohydrate data was also tested. The p-scales offer an interface between different areas of peat research. Strategies are outlined for the selection of a subset of chemical measurements among the variables used for characterization. A multivariate strategy based on these ideas is discussed.