Landscape structure, climate variability, and soil quality shape crop biomass patterns in agricultural ecosystems of Bavaria

Abstract

Understanding how environmental variability shapes crop biomass is essential for improving yield stability and guiding climate-resilient agriculture. To address this, we compared biomass estimates from a semi-empirical light use efficiency (LUE) model with predictions from a machine learning–remote sensing framework that integrates environmental variables. We applied a combined LUE and random forest (RF) model to estimate the mean biomass of winter wheat and oilseed rape across Bavaria, Germany, from 2001 to 2019. Using a 5 km2 hexagon-based grid, we incorporated landscape metrics (land cover diversity, small woody features), topographic variables (elevation, slope, aspect), soil potential, and seasonal climate predictors (mean and standard deviation of temperature, precipitation, and solar radiation) across the growing season. The RF-based approach improved predictive accuracy over the LUE model alone, particularly for winter wheat. Biomass patterns were shaped by both landscape configuration and climatic conditions. Winter wheat biomass was more influenced by topographic and landscape features, while oilseed rape was more sensitive to solar radiation and soil properties. Moderately diverse landscapes supported higher biomass, whereas an extreme landscape fragmentation or high variability showed lower values. Temperature thresholds, above 21 °C for winter wheat and 12 °C for oilseed rape, were associated with biomass declines, indicating crop-specific sensitivities under Bavarian conditions. This hybrid modeling approach provides a transferable framework to map and understand crop biomass dynamics at scale. The findings offer region-specific insights that can support sustainable agricultural planning in the context of climate change.

Plant phenotyping relevance

作物バイオマスという植物形質を、LUEモデルとランダムフォレスト・リモートセンシングの統合手法で広域推定し、予測精度を比較・評価しているため、単なる環境要因研究ではなく形質推定手法の適用が中心です。

Code and data availability

The paper's own biomass datasets, RF/LUE code, and models are not publicly deposited in the supplied text; the data availability statement defers to Supplementary Material without naming a repository or accession. The cited URLs (Landklif LULC map, Augsburg climate data, BGR soil SQR, Copernicus SWF, Bavarian yield/GEN

No evidence-backed public reproduction asset is currently recorded.