Predicting Soil Phosphorus-Related Properties Using Near-Infrared Reflectance Spectroscopy

Near-infrared reflectance spectroscopy (NIRS) is a rapid, inexpensive, and accurate analysis technique for a wide variety of materials, and it is increasingly used in soil science. The objectives of our study were to examine the potential of NIRS to predict (i) soil P extracted by two methods [Mehlich 3 (M3P) and water (Cp)1, soil total P (TP), annual crop P-uptake, and annual P-budget, and (ii) other soil chemical properties [total C (TC), total N (TN), pH, and K, Al, Fe, Ca, Mg, Mn, Cu, and Zn extracted by Mehlich 3]. Soil samples (n = 448) were taken over a 7-yr period from an experimental site in Levis (Quebec, Canada) where timothy (Phleum pratense L.) was grown under four combinations of P and N fertilizer. The NIRS equations were developed using 80% of the samples for calibration and 20% for validation. The predictive ability of NIRS was evaluated using the coefficient of determination of validation (12,2) and the ratio of standard error of prediction to standard deviation (RPD). Results show that M3P, Cp, crop annual P-uptake, and annual P-budget were not accurately predicted by NIRS (R-v(2) < 0.70 and RPD < 1.75). Similar results were found for K and Cu. However, NIRS predictions were moderately useful for TP, TN, Fe, and Zn (0.70 <= R-v(2) < 0.80 and 1.75 <= RPD < 2.25), moderately successful for TC and Al (0.80 <= R-v(2) < 0.90 and 2.25 <= RPD < 3.00), successful for pH and Mg (0.90 <= R-v(2) <=. 0.95 and 3.00 <= RPD <= 4.00), and excellent for Ca and Mn (R-v(2) > 0.95 and RPD > 4.00). The NIRS predictive ability of several soil properties appears to be related to their relationship with soil organic C. Although NIRS can predict several soil properties, prediction of total P was the only soil P-related property, correlated to soil C, that was moderately useful.