Models and sufficiency interpretation for estimating critical soil test values for the Fertilizer Recommendation Support Tool

Abstract

AbstractSoil test correlation determines whether a soil test can be used to predict the need for fertilization based on the critical soil test value (CSTV). Our objectives were to compare the CSTV estimated from five combinations of correlation models and yield sufficiency interpretations and to select one method for soil test correlation performed with the Fertilizer Recommendation Support Tool (FRST). Four models were fit to three datasets with strong (Mehlich‐1 K), moderate (Mehlich‐3 K), or weak (Olsen P) correlations between soil test P or K and crop relative yield. We tested the arcsine‐log calibration curve (ALCC), exponential (EXP), linear plateau (LP), and quadratic plateau (QP) models. The CSTV was defined as 95% of the maximum predicted yield for the ALCC and EXP methods, the join point for LP, and both the join point and 95% of the maximum for the QP providing five CSTV predictions. The five CSTVs ranged from 46 to 66 mg kg−1 for the Mehlich‐1 K dataset, 115 to 165 mg kg−1 for the Mehlich‐3 K dataset, and 7 to 16 mg kg−1 for the Olsen P dataset. Ten pairwise comparisons showed the estimated CSTV was numerically and sometimes statistically influenced by the model and sufficiency level interpretation. Despite differences among CSTVs, the frequency of significant yield responses above and below the predicted CSTV was generally comparable among the methods, with false‐negative errors occurring at 0%–18% of sites for a given dataset. The QP model with a CSTV at 95% of the predicted maximum was selected as the modeling approach for FRST.