Document Type

Thesis - Open Access

Award Date

2022

Degree Name

Master of Science (MS)

Department / School

Agronomy, Horticulture, and Plant Science

First Advisor

Jason Clark

Abstract

Researchers have pointed to changes in climate and land management practices to justify the need to reevaluate the accuracy of current South Dakota (SD) corn (Zea mays L.) P and K fertilizer recommendations. Also, an increase in soil health understanding has created the potential for soil health measurements to be used to improve the accuracy of these recommendations. The objectives for this study were to 1) evaluate the current P and K critical values and 2) determine the effect of including soil health indicators on improving fertilizer recommendation accuracy. This project was conducted throughout central and eastern SD from 2019-2021 at 97 experimental areas that varied in management, landform, and soil type. Fertilizer addition treatments of 112 kg P2O5 ha-1 and 112 kg K2O ha-1 were compared to a control with no P or K fertilizer. Soil health and fertility samples (0-15 cm) were collected before fertilization and analyzed for physical, chemical, and biological characteristics. A linear plateau model indicated the soil test P (STP) critical value needs to be increased from 16 to 20 mg kg-1 and soil test K (STK) needs to be decreased from 160 to 140 mg kg-1. However, both new critical values either 1) had low correlation values to yield response or 2) were not significantly better than the old critical values. Therefore, more sites and years of data are needed to confirm if a change in critical values is needed. Random forest variable importance methods found differences among variables, although differences were not substantial enough to clearly identify what variables were most important in predicting yield response to P and K fertilization. Decision tree analysis found several variables for P (STP, CEC, soil respiration, and clay content) and K (STK, tillage, and soil pH) that when split using a decision tree, improved prediction accuracy from 63% (STP or STK used alone) to 74% and 77%, respectively. These results demonstrate that soil health indicators along with soil fertility testing improves the accuracy of our yield response predictions to P and K fertilizer.

Number of Pages

173

Publisher

South Dakota State University

Share

COinS
 

Rights Statement

In Copyright