Predicting cadmium fractions in agricultural soils using proximal sensing techniques.

Cadmium (Cd) accumulation in agricultural systems has caused global environmental and health concerns. Application of phosphate fertiliser to sustain plant production unintentionally accumulated Cd in agricultural soils over time. Rapid and cost-effective Cd monitoring in these soils will help to inform Cd management practices. Compared to total Cd analysis, examining chemical fractions by sequential extraction methods can provide information on the origin, availability, and mobility of soil Cd, and to assess the potential plant Cd uptake. A total of 87 air-dried topsoil (0-15 cm) samples from pastoral farms with a history of long-term application of phosphate fertiliser were analysed using wet chemistry methods for total Cd and Cd forms in exchangeable, acid soluble, metal oxides bound, organic matter bound, and residual fractions. The data acquired using three proximal sensing techniques, visible-near-infrared (vis-NIR), mid-infrared (MIR), and portable X-ray fluorescence (pXRF) spectroscopy were used as input for partial least squares regression to develop models predicting total Cd and Cd fractions. The average total Cd concentration was 0.58 mg Cd/kg soil. For total Cd, cross-validation (cv) results of models using individual vis-NIR, MIR, and pXRF data performed with normalised root mean squared error (nRMSE) of 26%, 30%, and 31% and concordance correlation coefficient (CCC) of 0.85, 0.77, and 0.75, respectively. For exchangeable Cd, model using MIR data performed with nRMSE of 40% and CCC of 0.57. For acid soluble and organic matter bound Cd, models using vis-NIR data performed with nRMSE of 11% and 33% and CCC of 0.97 and 0.84, respectively. Reflectance spectroscopy techniques could potentially be applied as complementary tools to estimate total Cd and plant available and potentially available Cd fractions for effective implementation of Cd monitoring programmes.