Document Type

Thesis - Open Access

Award Date

2018

Degree Name

Master of Science (MS)

Department

Agronomy, Horticulture, and Plant Science

First Advisor

Doug D. Malo

Abstract

Soil organic matter (SOM) is composed of living biomass, dead plant and animal residues, and humus. Humus is a class of complex, organic molecules that are largely responsible for improving soil water holding capacity, nutrient mineralization, nutrient storage, and other critical soil functions. Soil organic carbon (SOC) accounts for approximately 60 percent of SOM and thus SOC is recognized as a strong indicator of soil health. Land use changes and intense cultivation of arable soils in the United States over the past century have led to large decreases in SOM. The objective of this research was to develop a multiple linear regression model to predict SOC levels in select southeastern South Dakota soils and the region. Conventional Till (CT), No-Till (NT), and Native Grass (NTVG) management systems were studied within South Dakota Major Land Resource Area 102B, 102C, and McCook County, South Dakota. It was hypothesized that NTVG treatments would have the highest SOC levels, followed by NT treatments, and CT treatments would have the least. Samples were analyzed for pH, electrical conductivity (EC), total nitrogen (TN), total carbon (TC), SOM, soil inorganic carbon (SIC), particle size, color, and water stable aggregates. Management was found to have a significant effect on soil pH, EC, TN, TC, and SOC compared to native conditions (p<0.05). Multiple linear regression (MLR) was used to build the full SOC prediction model which was then reduced using stepwise selection in R 3.5.0. The final reduced model that was produced by stepwise selection is defined as SOC = 3.25 - 0.811(Conventional Tillage) - 0.939(No-Tillage) - 0.548(10-20 Depth) - 0.918(20-40cm Depth) + 0.0396(Moisture) - 0.288(Temperature). Although this model did not result in an acceptable Shapiro-Wilk p-value, the model did not have multicollinearity issues, and approximately 67% of the variation in SOC was explained by the model. To create a model that includes all management variables, filtering the data set to include only specific data points before running MLR analysis is an option. One proposed filtered model incorporates No-Till management and Corn-Soybean rotation data points. The resulting filtered model is defined as SOC = -0.0885 - 0.473(10-20cm Depth) - 0.082(20-40cm Depth) + 0.067(Moisture) - 0.267(Temperature) + 0.156(pH). This model produced an acceptable Shapiro-Wilk p-value (p=0.969), displayed approximately normal residuals, and did not exhibit multicollinearity. Approximately 64% of the variation in SOC was explained by the model. Based upon these results, filtering the data set is an appropriate method for data analysis and model construction. Developing an electronic application for use via website or mobile device as a means of sharing this information with producers is a viable option. Additional data is needed to improve the models, to meet the assumptions of multiple linear regression when using stepwise selection, and to increase the applicability of the model to producers in southeastern South Dakota and the region before the electronic application is constructed. Furthermore, more data points are needed to validate the proposed models.

Description

Includes bibliographical references

Format

application/pdf

Number of Pages

134

Publisher

South Dakota State University

Rights

In Copyright - Educational Use Permitted
http://rightsstatements.org/vocab/InC-EDU/1.0/

Share

COinS