Answers are grounded in this paper and cite their sources.
Remote sensing is a powerful technology for monitoring land surface materials, and an effective tool for managing soil carbon and agricultural systems. The complexity of soil with varying surface covers and soil materials makes it difficult to identify the unique remotely sensed spectral response for specific soil constituents and model the changes in these constituents from mixed pixels. The objective of the current study was in proof-of-concept to use a general strategy on diverse soils and farm settings to achieve significant predictive capability. The current study applies varimax-rotated principal component analysis (VPCA) of visible and near infrared spectral reflectance obtained from hyperspectral and multispectral datasets followed by stepwise multiple linear regression (VPCA-SMLR) to estimate the percentage of total combustible carbon (%TC), and total combustible nitrogen (%TN), in shallow soils (<15 cm). The VPCA-SMLR was calculated on the Malvern/Panalytical ASD FieldSpec 4SR hyperspectral reflectance of surface soil obtained in the field (VPCAF) and shallow soil measured under controlled laboratory lighting conditions (VPCASS). The VPCA-SMLR was applied on multispectral harmonized Sentinel 2 imagery (VPCAMSI) through a semi-automated averaging code satellite imagery processing in Google Earth Engine Python API. Model calibration-validation analysis on a test size of 20% of VPCAF generated a 77% explanation of variance in % TC, 66% explanation of variance in % TN, and 78% of variance in %SOM. The VPCASS produced 40% explanation of variance in shallow soil %TC, 66% explanation of variance in %TN, and 42% explanation of variance in %SOM. Field scale estimates (VPCAMSI) accounted for 46% of variance for %TC and 42% of variance for %TN. The variance inflation factor analysis (VIF = 1.0) indicates that the VPCA-SMLR reduced the impacts of heterogeneity and nonlinearity in complex soil matrix. The VPCAF /VPCASS /VPCAMSI detected goethite, hematite, epidote, gypsum, and quartz in the soil with a variance of up to 98%. The results suggested that the VPCA-SMLR can independently predict %TC and %TN for the field-obtained spectral data. The result of this work produces a broadly applicable remote sensing-based workflow, with limitations, for wide variety of land managers to have a proxy for soil nutrients or minerals in lieu of physical soil samples that would facilitate consistent monitoring over a time scale of years.
No takes yet. Share an insight, caveat, or question.
Braimah et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: