The 3D distribution of phyllosilicate clay minerals in western Victoria

- MIR spectroscopy and 3-D models have been used to predict clay mineral distribution.
- Spectroscopic models were defined using legacy quantitative XRD measurements.
- Strong models were achieved for kaolinite, illite and smectite using MIR with the RMSECV all < 12 wt.%.
- Model trees account for 26 to 77% of total variation in map predictions.

The mineralogy of the clay fraction of soils is a major determinant of the behavior of soil. Conventionally these clay minerals have been determined using techniques such as X-ray Diffraction (XRD), but new complementary methods such as infrared spectroscopy can be used to rapidly and economically predict these minerals. This paper presents a methodology to predict these clay minerals at high-resolution that adhere to GlobalSoilMap (GSM) standards. Mid-infrared (MIR) spectroscopic models were formulated for clay minerals kaolinite, illite and smectite using partial least squares regression (PLSR) and legacy quantitative XRD determinations. Very strong models were achieved for kaolinite, illite and smectite and the root mean square error of cross validation (RMSECV) were all < 12 wt.%. Spectroscopic models were applied to 11,500 samples from western Victoria and harmonized to the GSM specified depth intervals using equal area splines. Clay minerals were then mapped using data mining model trees with a 10-fold cross validation to derive a mean prediction estimate and 90% prediction interval. Spatial models accounted for 26 to 77% of the total variation with kaolinite predictions for all 6 GSM depths ≥ 65%. Kaolinite is the dominant soil clay mineral of western Victoria for uplands and weathered volcanic terrains. Illite concentrations are higher where associated with weathered granitic plutons and in aeolian deposits of semi-arid environments. Smectite tends to occur associated with depressions of plains (volcanic and sedimentary). Further supplementation of additional sites and samples for landscapes with relatively sparse observations should contribute to refined and improved maps of these clay minerals.