Estimating the accuracy of the NIRS prediction model based on soil t

Near infrared reflectance spectroscopy (NIRS) proved to be a rapid, reliable and cost-effective method for soil analysis, but its prediction accuracy relies on the proper calibration model. In this paper, NIR spectroscopy in combination with partial least-squares regression (PLSR) was used to predict soil chemical properties on regional and local scales. The results obtained show a higher potential of NIRS to assess soil fertility on local scale, compared to a regional scale, with excellent prediction for organic C, carbonates and total N content (R2 ≥ 0.98 and RPD > 6), and good prediction for soil pH (R2 =0.88 and RPD > 3).