Performance of field‐scale lab vs in situ visible/near‐ and mid‐infrared spectroscopy for estimation of soil properties

Comparison of laboratory versus in situ visible/near- (visNIR) and mid-infrared (MIR) spectroscopy for prediction various soil properties is required to demonstrate trade-offs between accuracy efficiency. Field measurements were made on an arable field Germany (silt loam Haplic Luvisol) using visNIR (ASD FieldSpec 3 Hi-Res) MIR (Agilent Technologies 4300 Handheld FTIR) material was collected lab (Foss XDS Rapid Content Analyzer) (Bruker-TENSOR 27) dried ground determination total, labile (>63 μm light), stabilized heavy + <63 oxidizable) resistant organic carbon (OC) content, total nitrogen (Nt), pH texture. Partial least squares regression models calculated five repeated partitions the dataset (n = 238) into training (75%) test (25%) sets. Lab spectral outperformed OC (root mean squared error [RMSE] 0.24–1.0 g kg−1), Nt (RMSE 0.026–0.088 0.12–0.28) texture 0.53–1.5%). For both spectra, comparable or slightly better than sand, silt clay. Spectral estimations 0.34–0.47 kg−1) 0.41–0.85 (lab spectra) substantially (field inferior from multiple linear regressions OC, Nt, clay as predictors. Variable importance projection scores elucidated differences mechanisms by spectrometer fraction, found mineral signatures highly important. this field-scale study with 14% median gravimetric water content (GWC), loss measurement lower MIR. Analysis driest soils (<9% GWC) estimation vice versa wettest (>18%), demonstrating moisture dependence performance rankings. Highlights vs (visNIRS) (MIRS) require comparison C fractions, N, MIRS C, N superior, while visNIRS. At visNIRS MIRS. The ranking dependent.