U.S. Department of Agriculture: Agricultural Research Service, Lincoln, Nebraska


Document Type


Date of this Version



Published in COMMUN. SOIL SCI. PLANT ANAL., 32(15&16), 2325–2345 (2001).


The widespread use of soil phosphorus (P) data, either in the context of agronomic or environmental management, requires an explicit understanding of potential errors related to soil P testing. This study compares a variety of soil P extraction methods, each performed by 9 separate laboratories on 24 soils from across the United States. Soil clay content ranged from 0 to 47%, pH from 4.2 to 8.6, and Mehlich-3 P concentration from 2 to 205 mg kg-1. Average interlaboratory coefficients of variation (CVs) ranged from 0.11 to 0.22 for solution extracts (Bray-1 P, Fe-strip P, Mehlich-3 P, and Olsen P) and from 0.11 to 0.70 for saturated paste extracts (resin capsules and resin membranes, incubated for 2, 4, and 7 days). For soil tests based upon solution extracts, Olsen P exhibited the greatest variability among laboratories (CV = 0:22); despite its reputed suitability for a wider range of soils than Bray-1 and Mehlich-3. Soil test data were highly correlated, with the lowest correlations occurring between Olsen and Bray-1 P or Olsen and Mehlich-3 P (r = 0:77 and 0.84, respectively) and the highest correlations occurring between Olsen P and Fe-strip P or Mehlich-3 and Bray-1 P (r = 0:94 for both correlations). Results indicate that some common soil test P protocols, when carefully conducted, yield data that may be reliably compared, such as in the compilation of regional and national soil databases.