Comparison of laser-induced breakdown spectroscopy, colour sensing, as well as visible, near-infrared, and mid-infrared spectroscopy to predict soil properties

• Seven different instruments to predict soil properties using the same samples were compared. • Soil properties included P, K, Ca, Mg, Al, buffer pH, pH, SOM, and CEC. • LIBS and Vis-NIR sensor systems produced results with relatively lower uncertainties. • Several prediction models, all for CEC, yielded relatively accurate predictions. • Neither instrument was able to overperform commercial labs for P, pH, Mg, and SOM. Seven spectroscopic instruments’ ability to predict nine soil chemical properties [ i.e. , extractable phosphorus (P), extractable potassium (K), calcium (Ca), magnesium (Mg), aluminum (Al), soil pH (pH), buffer pH (BpH), soil organic matter (SOM) and cation exchange capacity (CEC)] were assessed using a diverse set of 798 air-dried and compressed soil samples from Southern Quebec. The instruments varied in terms of spectral range and resolution [ e.g ., colour/visible (Vis), visible and near-infrared (Vis-NIR), mid-infrared (MIR), and laser-induced breakdown spectroscopy (LIBS)], complexity, and interaction with the target. The performance of these instruments was compared using relevant test statistics derived from partial least squares regression (PLSR) models, such as the root mean squared error of prediction (RMSE P ) and the coefficient of determination (R 2 ) for simple linear regressions between sensor measurements and soil test reference values. Analysis of the LIBS spectra produced the lowest RMSE P values for P, K, Mg, Ca, pH, BpH, and SOM, while two Vis-NIR instruments yielded prediction models with the lowest RMSE P for Al and CEC. Overall, prediction models were relatively accurate for Ca, Mg, Al, SOM, and CEC (R 2 = 0.70–0.81), and less certain for P, pH, BpH, and K (R 2 = 0.53–0.65).