Validation of Soil‐Test‐Based Phosphorus and Potassium Fertilizer Recommendations for Flood‐Irrigated Rice

Abstract

Soil testing is a widely accepted practice for evaluating soil‐P and soil‐K availability but can be inconsistent for making accurate fertilizer‐P and fertilizer‐K recommendations. Our research objectives were to assess the accuracy of established soil‐ and tissue‐P and ‐K concentration interpretations for predicting flood‐irrigated rice (Oryza sativa L.) yield response to fertilization at three levels of significance (p ≤ 0.05, 0.10, and 0.25). Six treatments combining two fertilizer‐P rates (0 or 32 kg P ha−1) and four fertilizer‐K rates (0, 56, 84, 112 kg K ha−1) were applied at 24 sites. Soil‐test P (STP) interpretations were 40% accurate, regardless of the significance level, and whole‐plant‐P concentrations at the V6‐V7 stage were 50% accurate. Soil‐test K (STK) interpretations were 40 (p ≤ 0.05 and 0.10) and 36% (p ≤ 0.25) accurate, and whole‐plant‐K concentrations at the R2‐R3 stage were 61% (p ≤ 0.25) or 62% (p ≤ 0.05 and 0.10) accurate in predicting the yield response to K fertilization. Nearly all of the error in both soil‐test‐ and whole‐plant‐P and ‐K concentration interpretations occurred in the suboptimal categories. The accuracy (p ≤ 0.05) of yield response predictions for levels where no fertilizer was recommended was 100% for STP and 86% for STK and 82 to 100% for whole‐plant‐P and ‐K concentrations. The false‐positive error was the most common soil‐test recommendation inaccuracy and suggests that fertilizer recommendations are skewed to reduce the risk of yield loss from insufficient fertilization.Core Ideas Tissue analysis is more accurate at predicting flood‐irrigated rice response to P and K fertilization than Mehlich‐3 extractable P and K. False positive errors are the most common soil‐test‐based interpretation errors for flood‐irrigated rice. Mehlich‐3 extractable P is not a good predictor of rice yield response to P fertilization. The most accurate fertilizer P and K recommendations were interpreted at the 0.05 and 0.10 significance levels.