See more from this Session: Sensor-Driven Digital Soil Mapping: I
Here we investigated the potential of mid-infrared spectroscopy combined with partial least squares regression (MIRS-PLSR) for rapid assessment of the spatial heterogeneity of SOC, black carbon (BC), and particulate organic matter (POM) of three size classes (POM1: 2000–250 µm; POM2: 250–53 µm; POM3: 53–20 µm) on a 1.3 ha test site. The stone content, texture of the fine earth, pedogenic oxides, relief, erosion (137Cs-activity), as well as the soil moisture were considered as effective parameters regulating SOC dynamics. Spatial patterns were analyzed by multidimensional scaling of a fuzzy-kappa similarity matrix, principal component analysis, correlation analysis, multiple regression, and semivariance analyses.
All SOC pools were successfully predicted by applying local calibrations for MIRS-PLSR (R² = 0.77–0.96). The PLSR model for BC predictions was characterized by aromatic absorptions; estimations of POM1 chiefly relied on specific signals of lignin and cellulose. The contents of POM2 were estimated by spectral bands assigned to degradation products as aliphatic C–H groups and aromatic moieties, while carboxylic groups essentially contributed to the prediction of POM3.
With exception of POM3 (R² = 0.20), multiple regression models employing the stone content, contents of pedogenic oxides, as well as the relief (hillslope), successfully predicted the spatial distribution of all SOC fractions (R² = 0.68–0.82). The observed variability was chiefly deterministic and could be attributed to saturation processes caused by disproportionately high input of plant debris as amounts of fine earth were reduced by increasing stone contents (4–60 %). The spatial distribution of BC was additionally affected by erosion.