Degree

Doctor of Philosophy (PhD)

Department

School of Plants, Environmental and Soil Sciences

Document Type

Dissertation

Abstract

The adoption of conservation management practices is critical for improving soil health, enhancing nutrient use efficiency, and sustaining crop productivity in row crop systems in Louisiana. This study evaluated the role of conservation agronomic practices, soil biochemical indicators, and machine learning predictive models to improve soil nutrient dynamics, soil health indicators, microbial communities (MC), and crop productivity on a corn (Zea mays L.) research plot scale and in a commercial forty-hectare cotton (Gassypium hirsutum L.)-corn-soybean (Glycine max L.) rotation system in northeast Louisiana. The objectives of the study were to evaluate soil nutrient dynamics and MCs under long-term, no-nitrogen (N), and no-till systems with different cover crop (CC) functional groups, assess the effect of best management practices (BMPs) on soil health indicators, MCs, and crop productivity, and compare machine learning predictive (ML) models for crop yield and soil health parameters using soil and management variables. The results indicated that for the long-term effect of conservation practices with no-N, no-till, and different CC functional groups, the nutrient dynamics changed over time for a system dominated by inorganic N forms to microbial-regulated transformations and stabilized soil carbon (C) and N pools. For the effect of BMPs compared to farmers practices (FP) and stale, crop yield remains similar across the different treatments, despite the difference in fertilization rates. However, the study shows seasonal and yearly variability for plant-available nutrients, soil health indicators, and MCC. The dbRDA multivariate analysis indicated that MC and soil indicators shifted across years and were associated with nutrient pools and enzymatic activity (NAGase), while treatment-level MCs overlapped, showing no differences. For ML models, the linear models showed limited predictive performance due to the complex interaction of the variables, in contrast to nonlinear ML models such as neural networks (NN), random forest (RF), and XGBoost that effectively capture the relationship among soil, management, and yield responses, demonstrating their potential as a decision-making tool. Collectively, these findings demonstrate that conservation practices can maintain crop productivity while promoting long-term improvements in soil biochemical indicators and the benefits of prediction models as a decision-making tool to enhance sustainable agriculture.

Date

5-27-2026

Committee Chair

Setiyono, Tri

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Share

COinS