Effect of model choice on critical soil test value of phosphorus for corn in long-term trials in North Carolina

Abstract

The soil test correlation determines the critical soil test value (CSTV) of phosphorus (P) required to achieve 95%–100% of the maximum crop yield. However, CSTV predictions vary with the mathematical model used, which has implications for fertilizer recommendations. This study compared the P CSTVs for corn (Zea mays) estimated using four models, (1) modified arcsine-log calibration curve (ALCC), (2) linear plateau (LP) at the join point (JP), (3) quadratic plateau (QP) at the JP (QP-JP), and (4) QP at 95% of maximum yield (QP-95), and then calculated the frequency of crop response at different Mehlich-3 soil test phosphorus (STP) concentrations. Corn was grown in long-term trials in 2010, 2012, and 2014 in the Piedmont, Coastal Plain, and Tidewater regions of North Carolina. The P CSTVs obtained with ALCC, LP-JP, QP-JP, and QP-95 models were 42, 24, 31, and 26 mg kg−1, respectively, at the Coastal Plain site and 55, 43, 55, and 49 mg kg−1 at the Tidewater site, but these models could not calculate CSTVs at the Piedmont site. Nevertheless, the 95% confidence interval of CSTV did not differ for these models and sites analyzed. The frequency of corn response to STP declined with increasing STP, reaching 10% at 37.0 and 44.9 mg kg−1 at Coastal Plain and Tidewater sites, respectively, defining critical soil test range (CSTR) of 26–37 and 45–49 mg kg−1. Additional approaches combined with CSTV using broader datasets may help to refine the CSTR definition and improve fertilizer recommendations.