Performance indices for tests of soil nutrient status: Extractable phosphorus

Abstract

Soil phosphorus (P) extractants are often selected according to the correlation or regression between test values and crop performance (e.g., P uptake and/ or yield). Although this criterion is an essential determinant of extractant performance, it is often inadequate for evaluating whether extractants accurately discriminate between P‐deficient and P‐sufficient soils, or whether they produce reliable critical level estimates or repeatable soil P measurements. Four supplementary indices were evaluated that may provide a more direct assessment of extractant performance. The potential use and reliability of the indices were investigated in an evaluation of four soil P extractants, Modified Truog, Mehlich 3, Olsen, and ion‐exchange resin, using data from a greenhouse experiment. Coefficients of determination between relative dry matter yield and extractable P failed to identify differences among the extractants, ranging from 0.95 to 0.97. Coefficients of determination between extractable P and P added ranged from 0.96 to 0.97 except for one method at 0.83. The proposed indices, however, produced a ranking of the extractants related to their performance. The Kappa efficiency (K EFF) index indicated that Mehlich 3 provided better detection of P‐sufficient and P‐deficient soils than either Olsen or Truog (K FFF values of 0.92, 0.83, and 0.68, respectively). These index values reflect that the extradants correctly detected P deficiency in 17 of 18, 17 of 18, and 15 of 18 soils. The slight superiority of Mehlich 3 over Olsen was due to its correct detection of 9 of 9 P‐sufficient soils while the Olsen and Truog extradants correctly detected 8 of 9 P‐sufficient soils. Further studies are needed, especially field studies, to determine whether these indices accurately reflect the reliability of the extradants for use in diagnosis and recommendation. Because these indices directly assess success in identifying deficient and sufficient conditions, their use in extractant evaluations should provide more specific, purposeful evaluations than methods based solely on correlation and regression.