Multi-Class Machine Learning to Quantify the Impact of Nitrogen Management Practices on Grassland Biomass

Aug 4, 2026

Grassland biomass yield reflects a complex interaction between management intensity and environmental conditions, while isolating the role of practices such as mowing and fertilization remains difficult. The study presents a multi-class machine learning framework predicting above-ground biomass on 150 permanent plots over eight years in Germany’s Biodiversity Exploratories. After data cleaning, nitrogen imputation, standardization, and categorical encoding, CatBoost classifiers are trained with Bayesian hyperparameter optimization and ADASYN oversampling to handle class imbalance. Model performance is evaluated using binary to five-class quantile schemes, showing mowing frequency and mineral nitrogen input as dominant predictors, with reduced accuracy at finer class resolutions requiring additional environmental information or sensing.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Abstract

Introduction

Grassland productivity drivers

Nitrogen availability and mowing regimes

Machine learning for agricultural prediction

Study approach

Article Multi-Class Machine Learning to Quantify the Impact of Nitrogen Management Practices on Grassland Biomass Sebastian Raubitzek 1, *, Margarita Hartlieb 2, Philip König 1, Judith Hinderling 3, Kevin Mallinger 4 Academic Editor: Pietro Iannetta Received: 30 April 2025 Revised: 24 June 2025 Accepted: 25 June 2025 Published: 30 June 2025 Citation: Raubitzek, S.; Hartlieb, M.; König, P.; Hinderling, J.; Mallinger, K. Multi-Class Machine Learning to Quantify the Impact of Nitrogen Management Practices on Grassland Biomass. Nitrogen 2025, 6, 52 . [https://doi.org/10.3390/](https://doi.org/10.3390/)[ ](https://doi.org/10.3390/)nitrogen6030052 Copyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://creativecommons.org/](https://creativecommons.org/)[ ](https://creativecommons.org/)[licenses/by/4.0/](licenses/by/4.0/)) . 1 Complexity and Resilience Research Group, SBA Research gGmbH, Floragasse 7/5.OG, 1040 Vienna, Austria 2 Ecological Networks, Department of Biology, Technical University of Darmstadt, 64289 Darmstadt, Germany; [margarita.hartlieb@gmx.at](margarita.hartlieb@gmx.at) 3 Institute of Plant Sciences, Plant Ecology, University of Bern, Altenbergrain 21, 3013 Bern, Switzerland 4 Christian Doppler Laboratory for Assurance and Transparency in Software Protection, Faculty of Computer Science, University of Vienna, Kolingasse 14–16, 1090 Vienna, Austria * Correspondence: [sraubitzek@sba-research.org](sraubitzek@sba-research.org) Abstract Grassland biomass yield reflects a complex interaction of management intensity and environmental factors, yet quantifying the relative role of practices such as mowing and fertilization remains challenging. In this study, we introduce a multi-class machine learning framework to predict above-ground biomass on 150 permanent grassland plots across eight years (2009–2016) in Germany’s Biodiversity Exploratories and to evaluate the influence of key management variables. Following rigorous data cleaning, imputation of missing nitrogen values, feature standardization, and encoding of categorical practices, we trained CatBoost classifiers optimized via Bayesian hyperparameter search and mitigated class imbalance with ADASYN oversampling. We assessed model performance under binary, three-class, four-class, and five-class quantile-based categorizations, achieving test accuracies of 0.76, 0.57, 0.42, and 0.38, respectively. Across all schemes, mowing frequency and mineral nitrogen input emerged as the dominant predictors, while secondary variables such as drainage and conditioner use contributed as well. These results demonstrate that broad biomass categories can be forecast reliably from standardized management records, whereas finer distinctions necessitate additional environmental information or automated sensing to capture nonlinear effects and reduce reporting bias. This work shows both the potential and the limits of machine learning for informing sustainable grassland management and explainability thereof. Frequent mowing and higher mineral nitrogen inputs explained most of the predictable variation, enabling a 76% accurate separation of low and high biomass categories. Predictive accuracy fell below 60% for finer class resolutions, indicating that management records alone are insufficient for detailed yield forecasts without complementary environmental data. Keywords: biomass prediction; grassland yields; fertilizer data; mowing practices; CatBoost; Bayesian optimization; ADASYN; feature importance analysis 1. Introduction Grasslands play a crucial role in global ecosystems, serving as a primary source of forage for livestock, supporting biodiversity, and contributing to carbon sequestration [1,2] . Understanding the factors that influence grassland productivity is essential for optimizing land manage

FAQs

What data and time span are used to train the multi-class models?

The framework is trained on 150 permanent grassland plots across eight years (2009–2016) in Germany’s Biodiversity Exploratories.

Which management variables are most influential for predicting biomass categories?

Across classification schemes, mowing frequency and mineral nitrogen input are the dominant predictors, while drainage and conditioner use also contribute.

Why does prediction accuracy drop for finer biomass classes?

Management records alone become insufficient for detailed forecasts at higher class granularity, indicating a need for complementary environmental information or automated sensing to capture nonlinear effects and reduce reporting bias.

Aug 4, 2026

Upload by Seraphina