Comparison of infrared spectroscopy and laser granulometry as alternative methods to estimate soil aggregate stability in Mediterranean badlands

• IR spectra outperformed LG-data to predict aggregate stability values and classes.
• NIR-DR and MIR-ATR data provided best prediction (RPD = 3.55 & 2.88; R2 = 0.93 & 0.90).
• Spectral prediction relied on a tight link between SOC and aggregate stability.
• The combination of IR and LG data did not yield a better prediction model.

Soil aggregate stability is a key indicator of soil resistance to erosion, but its measurement remains fastidious for large scale uses. Alternative time and cost-effective methods are thus needed. Our objective was to assess and compare the efficiency of laser granulometry (LG) and soil mid- and near-infrared spectroscopy (MIR/NIR) as alternative methods to assess soil aggregate stability in Mediterranean badland soils. A collection of 75 badland soil samples was used, showing wide variations in soil aggregate stability. Three different categories of measurements were performed: (i) the aggregate breakdown kinetics of the [< 1 mm] size fraction under stirring and sonication, tracked by repeated particle size distribution measurements, using LG, (ii) mid-(diffuse-MIR-DR and attenuate transmitted reflectance — MIR-ATR) and near-(NIR-DR) infrared spectra of the fine soil fraction [< 2 mm] and (iii) the soil aggregate [3–5 mm] stability, using the standardized method (ISO/FDIS 10930, 2012). Partial least squares regression models were used to predict soil aggregate stability using LG data and infrared spectra. Results showed that NIR-DR and MIR-ATR data provided the best prediction model for soil aggregate stability values (RPD = 2.61 & 2.74; R2 = 0.85 & 0.87), followed by MIR-DR data (RPD = 2.24; R2 = 0.89) and finally LG data (RPD = 2.12; R2 = 0.80). For a quantitative use of the models to assign soil samples to standardized soil aggregate stability classes (ISO/FDIS 10930, 2012), infrared spectra also provided the best accuracy, with a misclassification rate below 30% for NIR-DR and MIR-ATR models, while it reached 43% with the LG-based model. The combination of IR and LG data did not yield a better prediction model for soil aggregate stability values and classes. Infrared-based method also provided best results in terms of time-saving strategy, reducing the measurement time to 8 min only. To conclude, infrared spectra (NIR-DR and MIR-ATR) outperformed LG-data to predict soil aggregate stability. Further development of this technique would require calibrating a set of soil-type specific prediction models for a wide range of soil types.