With rapid population growth and industrialization, increased accumulation of soil Cadmium (Cd) becomes a severe threat to soil quality. Conventional soil Cd measurements in the laboratory are expensive and time-consuming, involving complex processes of sample preparation and chemical analysis. Previous studies found that a combination of visible near-infrared reflectance (Vis-NIR) spectroscopy and some soil properties (soil organic matter (SOM), iron, and pH) as auxiliary information has the potential to predict soil Cd concentration effectively through statistics models. This study aimed to identify the feasibility of using sensor data of Vis-NIR and portable X-ray fluorescence spectrometry (PXRF) to replace soil auxiliary information in improving the prediction of soil Cd concentration. The sensor data of Vis-NIR and PXRF, and Cd concentrations of 128 surface soils from Yunnan Province, China, were measured. Outer-product analysis (OPA) was used for synthesizing the sensor data and Granger-Ramanathan averaging (GRA) was applied to fuse the model results. Artificial neural network (ANN) models were built using Vis-NIR data, PXRF data, and OPA data, respectively. Results showed that: (1) ANN model based on PXRF data performed better than that based on Vis-NIR data for soil Cd estimation; (2) Fusion methods of both OPA and GRA had higher predictive power (coefficients of determination (R²) = 0.89, root mean squared error (RMSE) = 0.06, and ratios of performance to interquartile range (RPIQ) = 4.14 in ANN model based on OPA fusion; R² = 0.88, RMSE = 0.06, and RPIQ = 3.53 in GRA model) than those based on either Vis-NIR data or PXRF data. In conclusion, there exists a great potential for the combination of OPA fusion and ANN to estimate soil Cd concentration rapidly and accurately.

Artificial neural network, Outer-product analysis, Granger-Ramanathan averaging, Soil Cd concentration, Fusion method.