Skip to main content
ARS Home » Midwest Area » Ames, Iowa » National Laboratory for Agriculture and The Environment » Soil, Water & Air Resources Research » Research » Publications at this Location » Publication #431484
Research Project: Optimizing Carbon Management for Enhancing Soil and Crop Performances

Location: Soil, Water & Air Resources Research

Title: Moving beyond book values: Using machine learning to improve the accuracy of manure nutrient concentration predictions

Author
item Chatterjee, Amitava
item BOHL BORMANN, NANCY - University Of Minnesota

Submitted to: Journal of Environmental Quality
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/13/2026
Publication Date: 6/11/2026
Citation: Chatterjee, A., Bohl Bormann, N. 2026. Moving beyond book values: Using machine learning to improve the accuracy of manure nutrient concentration predictions. Journal of Environmental Quality. https://doi.org/10.1002/jeq2.70203.
DOI: https://doi.org/10.1002/jeq2.70203

Interpretive Summary: Manure is commonly used to meet crop nutrient demand in the United States. However, since manure nutrient levels can vary across farms, it is critical to know the total nitrogen, phosphorus, and potassium content of the manure so that it can be used effectively to meet crop production needs. Separate machine learning models were developed for each nutrient in both solid and liquid manure. These models allow growers to accurately determine the amount of manure that should be applied to a field, ensuring there are sufficient nutrients to meet crop demands without overapplication.

Technical Abstract: Animal manure is an excellent source of nutrients, but its concentrations vary within and across farms. Using published nutrient concentration averages could lead to an excess or insufficient supply relative to crops’ nutrient demand. For this study, manure nutrient concentrations were obtained from the ManureDB database. Regional variabilities of manure total nitrogen (N), phosphorus (P as P2O5), and potassium (K as K2O) concentrations were determined based on animal source and physical form (solid vs. liquid) of the manure across the United States. Three machine learning (ML) models, Random Forest (RF), Extreme Gradient Boosting (XGB), and a hybrid of the two, were used to predict total NPK concentrations using ammonium-N (NH4-N), pH, moisture percentage, region, and animal source as predictor variables. Models had higher predictability for the total N concentration (R2 = 0.8 to 0.9) compared to total P and K (R2= 0.4 to 0.7). Models provided better predictions for the total N concentration of liquid manure than for solid manure. For solid manure, animal source emerged as the dominant factor, while moisture percentage and NH4-N had a secondary influence. For liquid manure, NH4-N was the best predictor of total N and P concentrations, followed by moisture percentage. These findings demonstrate that the ML models using the ManureDB database could be used to predict primary nutrient concentrations in manure.