Understanding nitrogen export from agricultural regions is essential for tackling environmental impacts in the Mississippi River Basin and the Gulf of Mexico. Iowa—a major corn, soybean, and livestock producer—is one of the largest contributors of nitrate pollution entering the river system. This new study applies simulation decomposition (SimDec) to Iowa’s food–energy–water (IFEW) system to uncover how different nitrogen sources drive county-level nitrogen surplus.

The results provide a multi-level perspective unavailable in earlier work, enabling policymakers and researchers to see how low-level variables (such as specific livestock groups) feed into aggregate nitrogen categories and ultimately into the nitrogen exported from Iowa counties.

Why this study matters

Iowa’s agriculture is interconnected:

• Corn and soybean production feeds the ethanol and livestock industries.
• Livestock produces manure nitrogen (MN), a key environmental input.
• Fertilizer use (commercial nitrogen, CN) supports crop yields.
• Nitrogen fixation (FN) and grain nitrogen removal (GN) depend on crop patterns.
• Weather shapes crop yield and water movement.

The model structure shown in Fig. 1 (page 3) highlights these interactions visually, emphasizing how nitrogen moves through crops, livestock, energy production, and water systems.

Earlier GSA studies (such as Raul et al. 2020) operated at state level and used traditional Sobol’ indices, which could not analyze dependent variables. This study updates the dataset, performs analysis at the county level, and introduces SimDec, which can evaluate dependent, aggregate variables without redesigning the model.

Methods Overview

County-level computational model

The team assembled annual county-level data (1968–2019) for all 99 Iowa counties using the USDA NASS API. The model calculates:

• Grain nitrogen (GN) from yields and planted areas (p. 3).
• Fixation nitrogen (FN) based on soybean area (p. 3).
• Manure nitrogen (MN) using a detailed livestock breakdown (beef, dairy, pigs), with nitrogen content and lifecycle factors for each group (Table 1, p. 3).
• Commercial nitrogen (CN) from a GIS-integrated dataset (p. 4).

Nitrogen surplus (NS) is defined as:
NS = CN + MN + FN − GN (p. 4)

This serves as the key output for sensitivity analysis.

Sensitivity analysis using the simple binning method

The study applies the simple binning method developed by Kozlova et al. (2025), which computes first- and second-order sensitivity indices using only one simulation matrix. It captures dependent variables and is more computationally efficient than Sobol’ indices.

Low-level inputs

Figure 2 (p. 5) shows that the most influential low-level variable is:

• x12: pigs in the final growth stage, contributing 46% of total variance.

This aligns with known livestock nitrogen contributions. Large negative second-order effects between x10 and x11 (p. 5) reveal strong correlations among pig groups.

Aggregate variables

When inputs are aggregated into MN, CN, GN, and FN, the sensitivity structure becomes clearer (Fig. 3, p. 5):

• MN (manure nitrogen): first-order effect 50%, combined effect 52%
• CN (commercial nitrogen): 23%
• GN (grain nitrogen): 8%
• FN (fixation nitrogen): 6%

This demonstrates that MN overwhelmingly drives nitrogen surplus at the county level.

Simulation decomposition results

The core SimDec visualization decomposes the output distribution by the states of MN and CN. Both are divided into low, medium, and high states, producing nine scenarios (Table 2, p. 6).

Key findings from SimDec (Fig. 4, p. 6)

• MN is the primary driver: low and medium MN states produce similar nitrogen surplus ranges; high MN dramatically shifts the distribution upward.
• CN interacts with MN: CN has modest influence at low MN, but becomes more influential when MN is high.
• The highest nitrogen surplus occurs in the “high MN + low CN” scenario.

These heterogeneous relationships—visible only through SimDec—cannot be detected using traditional GSA approaches.

Multi-level perspective through a Sankey diagram

Figure 5 (p. 7) provides a multi-level sensitivity profile showing how low-level inputs contribute to aggregate variables and then to nitrogen surplus:

• x12 (final-stage pigs) flows heavily into MN and then into NS.
• Crop area variables (Acorn, Asoy) influence multiple aggregates (GN, FN, MN), confirming their cross-cutting impact.
• CN receives no flow from low-level variables because it is an external dataset, not model-generated.

This multi-layered view enables policymakers to trace high nitrogen surplus back to specific livestock categories and agricultural patterns.

Implications for Iowa’s nitrogen management

The findings highlight that:

• Regional variation in nitrogen surplus is large and driven primarily by livestock manure production.
• Southeastern counties show higher FN due to soybean prevalence, while northwestern counties have higher CN due to intensive corn farming (p. 6).
• Localized nitrogen management strategies are necessary, especially for regions with high manure production.

The study shows how SimDec enables a unified approach to sensitivity and uncertainty analysis—particularly valuable for policy, where model inputs are often dependent and multi-level.

Reference

Jeong, T., Kozlova, M., Leifsson, L. T., & Yeomans, J. S. (2025). Simulation decomposition analysis of the Iowa food–water–energy system. Environmental Modelling & Software, 188, 106415. https://doi.org/10.1016/j.envsoft.2025.106415

‍