Publication Date:
2018-10-01
Description:
In paleoenvironmental studies, the mineralogical composition of sediments is an important indicator. In combination
with other indicators, they contribute to the understanding of changes in sediment sourcing as well as in
weathering and depositional processes. Fourier transforminfrared spectroscopy (FTIRS) spectra contain information
on mineralogical composition because eachmineral has a unique absorption pattern in the mid-IR range. Although
easily obtained, FTIR spectra are often too complex to infermineral concentrations directly. In this study,
we use a calibration set of ca. 200 sediment samples conventionally measured using X-ray diffraction (XRD) in
order to developmultivariate, partial least squares (PLS) regressionmodels relatingmineral contents to sediment
spectra. Good correlations were obtained for the most common minerals (e.g. quartz, K-feldspar, illite, plagioclase,
smectite, calcite). Correlation coefficients ranged from 0.85 to 0.92, coefficients for the validation varied
from 0.64 to 0.80, the number of latent variables (PLS regression components) in the models ranged between
3 and 7 and the range of variation of the RMSEcv gradient was from 15.28 to 5.7.
Repository Name:
EPIC Alfred Wegener Institut
Type:
Article
,
isiRev
