Corn Nitrogen Monitoring with Hyperspectral Data and Ensemble Learning

A study in Heilongjiang Province, China, used hyperspectral data and a stacked ensemble learning model to monitor canopy nitrogen content in the maize cultivar Jinboshi. The optimized model achieved a prediction R² of 0.826 and RMSE of 0.450. The dataset, shared by Haoquan Kong under a CC-BY-4.0 license, was last updated on 2026-05-21.

Use Cases

• Developing nitrogen prediction models based on optimized hyperspectral bands and vegetation indices.
• Comparing feature selection methods like Genetic Algorithm and Successive Projections Algorithm for spectral data.
• Building stacked ensemble models using base learners like XGBoost and GBDT for agricultural regression tasks.
• Applying interpretability techniques to analyze the contribution of red-edge and near-infrared spectral features to model predictions.

Strengths

• Model performance is quantified with specific metrics (R²p = 0.826, RMSEP = 0.450).
• The description details a specific methodology including feature optimization and a stacked ensemble model.
• The study was conducted in a defined geographic region (Heilongjiang Province, China) on a specific cultivar (Jinboshi).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The primary data file is a 13.1 MB DOCX document, which may not be a standard data format for direct analysis.

Provenance

Source

Haoquan Kong via figshare

Collection Method

Hyperspectral data collection and analysis from field experiments on maize.

Freshness

Last updated 2026-05-21 05:42:36; freshness should be verified.

Geography

Heilongjiang Province, China