Quantifying the Importance of Soil‐Forming Factors Using Multivariate Soil Data at Landscape Scale

• Published in: Journal of geophysical research. Earth surface 2021-08, Vol.126 (8), p.n/a
• Main Authors: Eger, A., Koele, N., Caspari, T., Poggio, M., Kumar, K., Burge, O. R.
• Format: Article
• Language: English
• Subjects: Bayesian analysis; Bayesian mixing model; Biota; Climate; Data; Datasets; Digital mapping; Greywacke; hillslope soils; Infrared spectroscopy; Loess; Multivariate analysis; Multivariate statistical analysis; multivariate statistics; Oxides; Probability theory; proximal soil sensing; Rocks; Soil; Soil analysis; Soil characteristics; Soil chemistry; Soil formation; Soil mapping; soil parent material; Soil sciences; Soil types; soil‐forming factors; Statistical analysis; Statistical methods; Topography; Tracers; Variation
• ISSN: 2169-9003; 2169-9011
• DOI: 10.1029/2021JF006198

The role of soil‐forming factors (time, parent material, climate, biota, and topography) on soil processes has commonly been studied using soil sequences where only one factor varies between sites. However, when multiple factors change, it becomes difficult to partition the importance of different soil‐forming factors for soil formation. We show for an altitudinal gradient how proximal sensing (portable XRF, Fourier‐Transform Infrared [FTIR]), multivariate statistics, and Bayesian mixing modeling can help to quantify the importance of two soil‐forming factors. First, we confirmed the existing qualitative soil‐landscape model of concomitant shifts in parent material (greywacke loess to mafic volcanics) and climate (higher precipitation) with altitude, leading to increases in pedogenic oxides, soil carbon, and soil Fe, while Si concentrations and pH declined. Second, we applied a mixing model using immobile elements as parent material tracers to quantify the parent material contribution in soils across our gradient. Third, we conducted a variation analysis to determine how much variation in the soil FTIR spectra could be explained by parent material and climate. We found that parent material explained a larger proportion of the variation than climate. However, because of the strong spatial correlation between climate and parent material, a large proportion of the variation in the FTIR data could be explained by either parent material or climate, potentially concealing the leading role of parent material. Given that parent material is often omitted in modern digital soil mapping, our results emphasize the importance of parent material as a predictor of spatial soil distribution. Plain Language Summary The type of soil that exists in a landscape depends on the multiple factors like climate, rock type, or topography. Teasing these factors apart in their contribution to shape the characteristics of the soil is difficult. However, this knowledge is important to understand the processes in soil and make predictions about where soils occur in a landscape. In our study, we analyzed soils that differed by climate and rock type to determine which of the two factors was more important for the soil characteristics. We obtained soil chemical data of thousands of variables, and used statistical methods that were specifically developed to understand such large data sets. We found that rock type could explain the differences between our soils better than climate. Our work i