Sharma, P. 2025. Modeling nitrous oxide emissions at different spatial and temporal scales. , Michigan State University, East Lansing, MI.

Agricultural soils are the largest anthropogenic source of nitrous oxide (N₂O), a potent, long-lived greenhouse gas and ozone-depleting substance. Yet despite decades of research, practical mitigation in row-crop systems is constrained by three gaps: (1) national and global inventories report totals rather than crop-resolved emissions and rarely attribute changes to drivers such as harvested area, fertilizer inputs, or nitrogen-use efficiency, leaving emission-intensity patterns and their causes poorly understood; (2) existing prediction tools are difficult to scale or perform inconsistently, especially for short, high-magnitude flux events; and (3) management guidance evaluates practices in isolation instead of prescribing region-specific bundles tuned to climate and soils. This dissertation advances methods to quantify, explain, and mitigate agricultural N₂O emissions across multiple spatial and temporal scales by integrating global data products, process-based models, and machine learning into a single, decision-relevant framework.
In Chapter 1, I provide an overview of the research motivation and a comprehensive literature review. In this chapter, I synthesize the biogeochemical controls on N₂O (nitrification, denitrification, and their sensitivity to mineral N, labile C, temperature, moisture, pH, and texture). I review the strengths and limitations of measurement techniques and IPCC inventory tiers, and compare process-based models approached with machine learning.
In Chapter 2, I present the first global, spatially explicit, crop-resolved assessment of N₂O emissions and intensities (emissions per unit of production). By coupling crop-specific activity data with different emission factors, I map where intensity is high, where it is falling, and why. Decomposition analysis attributes change to crop type, yield, area, and nitrogen use efficiency, revealing hotspots for targeted mitigation and cases. This chapter provides a decision-ready baseline for prioritizing mitigation by crop and country, rather than relying solely on national totals.
In Chapter 3, I develop a Ensemble Modeling Systems (EMS) that blends the strengths of process-based models with a stacked ensemble of machine-learning algorithms to predict daily N₂O fluxes. The framework is trained and evaluated on a large, diverse dataset of manual chamber measurements from long-term U.S. field sites encompassing multiple crops and management regimes. It reproduces both the timing and the size of emission peaks with high agreement to observations, including at locations not used for training. By design, the framework preserves interpretability: feature attribution identifies when and where moisture, temperature, substrate supply, and management jointly control emissions.
In Chapter 4, I apply the hybrid framework across a large Midwestern cropland domain over a roughly decade-long period to evaluate a group of realistic management pathways, including conventional tillage versus no-till, the presence (with corn-soybean rotation) or absence of winter rye (continuous corn), and partial removal of maize stover and/or rye biomass. The analysis shows that mitigation outcomes are context-dependent: no-till tends to reduce N₂O with the strongest benefits in drier or transitional climates and in the non-growing season; cover crops can either mitigate or elevate emissions depending on soil and climate; and partial removal of cover-crop biomass can lower emissions where excess labile carbon would otherwise stimulate denitrification. A combination of management that jointly manages mineral N supply, soil structure, and carbon inputs achieves more durable reductions than any single practice alone.
In Chapter 5, I synthesize the dissertation’s findings into actionable guidance. Together, these contributions deliver (1) a global, crop-resolved view of agricultural N₂O that links field-scale mechanisms to national inventories; (2) a hybrid modeling system that improves daily flux prediction and clarifies drivers; and (3) a regional scenario analysis that translates modeling advances into actionable, place-based mitigation strategies.

Associated Treatment Areas:

• GLBRC Biofuel Cropping System Experiment
• MCSE Main Cropping Systems Experiment

Download citation to endnote bibtex