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

Authors: SLATON, NATHAN - University Of Arkansas; PEARCE, AUSTIN - Field To Market: The Alliance For Sustainable Agriculture; GATIBONI, LUKE - North Carolina State University; OSMOND, DEANNA - North Carolina State University; Bolster, Carl; MIQUEZ, FERNANDA - Iowa State University; CLARK, JASON - South Dakota State University; DHILLON, JAGMANDEEP - Mississippi State University; FARMAHA, BHUPINDER - Clemson University; KAISER, DANIEL - University Of Minnesota; LYONS, SARAH - North Carolina State University; MARGENOT, ANDREW - University Of Illinois; MOORE, AMBER - Oregon State University; RUIZ DIAZ, DORIVAR - Kansas State University; SOTOMAYOR, DAVID - University Of Puerto Rico; SPACKMAN, JARED - University Of Idaho; SPARGO, JOHN - Pennsylvania State University; YOST, MATT - Utah State University

Submitted to: Soil Science Society of America Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/3/2024
Publication Date: 6/8/2024
Citation: Slaton, N.A., Pearce, A., Gatiboni, L., Osmond, D.L., Bolster, C.H., Miquez, F., Clark, J., Dhillon, J., Farmaha, B., Kaiser, D., Lyons, S., Margenot, A., Moore, A., Ruiz Diaz, D., Sotomayor, D., Spackman, J., Spargo, J., Yost, M. 2024. Models and sufficiency interpretation for estimating critical soil test values for the Fertilizer Recommendation Support Tool. Soil Science Society of America Journal. 88(4):1419–1437. https://doi.org/10.1002/saj2.20704.
DOI: https://doi.org/10.1002/saj2.20704

Interpretive Summary: Soil test correlation determines whether a soil test can be used to predict the need for fertilization based on the critical soil test value (CSTV). Soil test correlation supports the long-term goal of soil fertility research to understand how deficient a soil is in a given plant-essential nutrient and the extent to which the crop might respond to fertilization with that nutrient. The Fertilizer Recommendation Support Tool (FRST) initiative is a national project in the United States (US) that aims to provide soil test correlation information for user-selected datasets of P and K field trials for a variety of crops, soils, and climates. One challenge in developing the FRST is to choose a single soil test correlation model and interpretation for a variety of datasets in the FRST database. This study aimed to identify a soil test correlation model and sufficiency interpretation best suited for the FRST decision tool using soil test correlation. Our specific objectives were to gauge the effectiveness of five combinations of correlation models and relative yield sufficiency level to estimate the CSTV and uncertainty, review the strengths and weaknesses of the examined models, and incorporate metrics describing the probability of response into the soil test correlation process.

Technical Abstract: Soil test correlation determines whether a soil test can be used to predict the need forfertilization based on the critical soil test value (CSTV). Our objectives were to com-pare the CSTV estimated from five combinations of correlation models and yield suf-ficiency interpretations and to select one method for soil test correlation performedwith the Fertilizer Recommendation Support Tool (FRST). Four models were fit to three datasets with strong (Mehlich-1 K), moderate (Mehlich-3 K), or weak (OlsenP) 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 quadraticplateau (QP) models. The CSTV was defined as 95% of the maximum predicted yieldfor the ALCC and EXP methods, the join point for LP, and both the join point and95% of the maximum for the QP providing five CSTV predictions. The five CSTVsranged from 46 to 66 mg kg-1 for the Mehlich-1 K dataset, 115 to 165 mg kg-1 forthe Mehlich-3 K dataset, and 7 to 16 mg kg-1 for the Olsen P dataset. Ten pairwisecomparisons showed the estimated CSTV was numerically and sometimes statisti-cally influenced by the model and sufficiency level interpretation. Despite differencesamong CSTVs, the frequency of significant yield responses above and below thepredicted CSTV was generally comparable among the methods, with false-negativeerrors occurring at 0%–18% of sites for a given dataset. The QP model with a CSTVat 95% of the predicted maximum was selected as the modeling approach for FRST.