Effects of soil sample pretreatments and standardised rewetting as interacted with sand classes on Vis-NIR predictions of clay and soil organic carbon

Abstract

Numerous studies have examined the soil analytical potential of diffuse reflectance spectroscopy in the near infrared range, alone or combined with the visible range (Vis-NIR). Soil organic matter (SOM), soil organic carbon (SOC) and clay content are the most commonly and successfully predicted parameters, but predictions are quite variable due e.g. to the range of soil types covered by the calibrations. Especially organic matter predictions are also suggested to be influenced by for example soil moisture content and inclusion of the visible range in the calibration. Excess quartz sand is also suggested to have a negative influence. This study was undertaken to examine the effect of a selection of standardised sample pretreatment procedures, including rewetting, on predictions of clay and SOC content. A subset of 400 samples was selected from a dataset of 3000 Swedish agricultural soils to cover clay and organic matter contents without co-variation. The selected samples were analysed by NIR and Vis-NIR on air dry samples, either carefully mixed to avoid stratification of particle size classes or shaken to promote separation, resulting in predominantly larger particles being analysed. Unshaken samples were also analysed immediately after standardised additional drying at 35 °C for 12 h and stepwise volumetric rewetting up to 30%. Shaking and additional drying had small negative effects on clay predictions, while drying only had small positive effects on SOC predictions. Volumetric rewetting to 20 or 30% before scanning reduced clay prediction errors by up to 15%, RMSEP reduced from 5.4% clay to 4.5% clay, and SOC prediction errors by up to 30%, from 0.9% SOC to 0.6% SOC, indicating that standardised rewetting should be considered. The mechanisms concerned could not be specifically identified, but known bands for water, hydroxyl and clay mineral-dependent absorption near 1400, 1900 and 2200 nm were involved in the improved clay calibrations and bands near 1700, 2000, 2300 and 2350 nm in the improved SOC calibrations. The SOC predictions were most inaccurate for soils with a high sand content. For these samples the average prediction error was more than twice as high as those for less sandy samples. Rewetting eliminated this bias, largely explaining the positive effects of rewetting.