Machine Learning (AutoML)-Driven Wheat Yield Prediction for European Varieties: Enhanced Accuracy Using Multispectral UAV Data

We are pleased to announce the publication of a scientific article titled "Machine Learning (AutoML)-Driven Wheat Yield Prediction for European Varieties: Enhanced Accuracy Using Multispectral UAV Data" in the journal Agriculture (MDPI), Volume 15, Issue 14, July 2025.

The publication highlights research conducted within the TALLHEDA project, focusing on applying machine learning and UAV-based remote sensing to achieve highly accurate wheat yield predictions for European varieties.

The study analyzed five European wheat cultivars across 400 experimental plots, combining 20 different nitrogen, phosphorus, and potassium (NPK) fertilizer treatments, with yield variations ranging from 1.41 to 6.42 t/ha.

This diversity of conditions was key to building robust and generalizable predictive models.

The article emphasizes key aspects of the research, including:

• The use of a DJI P4 Multispectral UAV to capture high-resolution imagery at three critical growth stages: Heading (9 May), Flowering (20 May), and Grain Filling (6 June).

• Derivation of 65 vegetation indices from multispectral and RGB imagery to serve as input features for machine learning models.

• Application of the AutoML framework PyCaret to automatically evaluate and tune 25 different regression algorithms.

• Top-performing models achieved exceptional accuracy — the Support Vector Regression (SVR) model reached R² = 0.95 on 9 May and R² = 0.91 on 6 June, while the Multi-Layer Perceptron (MLP) Regressor attained R² = 0.94 on 20 May.

• Confirmation that multispectral indices consistently outperformed RGB indices in correlation with yield across all measurement dates, due to their sensitivity to chlorophyll content, canopy structure, and water content.

  
  

This peer-reviewed publication marks a significant milestone for the TALLHEDA team, setting a new benchmark for predictive accuracy in European wheat yield estimation through the integration of UAV technology, comprehensive spectral data, and AutoML methodologies.

📄 Read the full article here:https://www.mdpi.com/2077-0472/15/14/1534