Prediction of β-glucosidase and β-glucosaminidase activities, soil organic C, and amino sugar N in a diverse population of soils using near infrared reflectance spectroscopy

There is a need for methods that can rapidly measure multiple biological properties simultaneously. A near infrared spectroscopy (NIRS) method was used to predict β-glucosidase (EC 3.2.1.21) and β-glucosaminidase (NAGase, EC 3.2.1.52) activities, and soil organic C and amino sugar N concentrations in 184 diverse soils of Ohio. The laboratory-measured values of the variables were calibrated against NIR spectral data with partial least squares regression analysis. Statistical analysis of the spectral data was done using the multivariate analysis software Unscrambler 8.0 (CAMO Inc). The first differential transformation of the spectral data in the NIR region (1100–2498 nm) generally yielded best results for developing multivariate calibration models. The multivariate models developed were validated using the full cross validation method and the test set method with a test set size of approximately 45 samples. The R2 values, testing variation between concentrations as measured by the NIR method and chemical methods, were 0.91 for organic carbon (OC), 0.92 for amino sugar N, and 0.82 for both soil β-glucosidase and β-glucosaminidase enzyme activities. Our study showed that the NIRS method has the potential to simultaneously, rapidly and accurately predict values of multiple related variables. The equipment needed for the NIRS method is not expensive and can be used where very large numbers of samples need to be rapidly analyzed. Indeed, the prediction equations can be constantly improved as more data points are entered into the correlations between laboratory-measured values and NIRS values.

► NIRS can predict soil variables simultaneously, rapidly, inexpensively and accurately.
► β-glucosidase, β-glucosaminidase, organic C and amino sugar N were assessed by NIRS.
► R2 values relating NIRS to chemical methods were 0.82–0.92 for these four variables.
► NIRS prediction equations can be improved as more data are entered into correlations.