Soil Science Society of America Journal · 2023 · 15 citations · 38 references
AgricultureCrop SciencesSoil SciencesNutrient ManagementSoil FertilitySoil ScienceCrop Nutrient RequirementsFertilizer ManagementPhosphorus ManagementPhosphorus NutritionSoil Fertility EvaluationSoil TestingControlled Field TrialsCrop Response PredictionFertility Response TrialsPhosphorus Response ModelingPotassium Fertilizer ManagementPotassium Fertilizer RatesSoil Potassium TestingSoil-test Based Fertilizer Response
Abstract Soil test phosphorus (P) and potassium (K) and the associated soil test critical values (CSTVs) are widely used as the primary diagnostic tool to determine if a crop needs P and K fertilizer. However, identifying robust and predictive CSTVs requires a large number of field trials in representative soils and environments to assess crop fertilizer responses. From 1976 to 2021, a total of 457 P trials and 458 K trials were conducted in corn ( Zea mays ), soybean ( Glycine max ), and wheat ( Triticum aestivum ) across 40 counties in Ohio. The trials consisted of both multi‐rate, multi‐year trials and single‐year, single‐rate and large strip trials. Across all crops, grain yield responded to P and K fertilizer in 23.4% and 25.3% of trials, with average yield increases of 8.8% and 11.0%, respectively. Corn yield was more responsive to P and K fertilizer (∼30% of P and K trials) than soybean yield (14% of P trials, 20% of K trials). Several models were run to identify CSTVs, but models imparted systematic bias in CSTVs, and a superior model was not identified due to poor model fit and indiscernible criteria to guide model selection. Instead of a single CSTV, data were classified into five discrete categories based on relative yields and Mehlich‐3 soil test values. These classifications provided a useful and consistent framework for assessing the probability of crop response to P and K fertilizer. Ohio soils with Mehlich‐3 values greater than 20 mg P kg −1 and 130 mg K kg −1 have a low probability of responding to P or K fertilizer application.
38
R: A Language and Environment for Statistical Computing
R Core Team · 2000 · 353K citations · Full text
Natural Sciences, Computer Science, Statistical Computing +10