Hasan (2025) Global Synthetic Crop Yield, Meteorological, and Climate Teleconnection Dataset for Machine Learning Benchmarking

Identification

• Journal: Mendeley Data
• Year: 2025
• Date: 2025-12-05
• Authors: Hasan, Raza
• DOI: 10.17632/y7hkz2zfcc

Research Groups

• Raza Hasan

Short Summary

This paper introduces a synthetic dataset (1990-2023) designed to benchmark machine learning models, particularly Spatial-Temporal Graph Neural Networks (ST-GNNs), in understanding the impact of global climate teleconnections on regional weather and crop yields.

Objective

• To generate a high-fidelity synthetic dataset that explicitly models global climate teleconnections (e.g., ENSO, NAO) and their influence on regional meteorological conditions and crop yields, specifically for benchmarking Spatial-Temporal Graph Neural Networks (ST-GNNs) in capturing these complex relationships.

Study Configuration

• Spatial Scale: Global, with data generated for 15 distinct regions.
• Temporal Scale: 1990–2023

Methodology and Data

• Models used: Physics-informed Python simulation script (GlobalSyntheticDataGenerator) incorporating: sinusoidal functions with Gaussian noise for global signals (ENSO, NAO), region-specific coupling coefficients based on real-world atmospheric physics, and a non-linear biological stress function for yield calculation (Yield = BaseYield * TempStress * WaterStress * TechTrend + Noise).
• Data sources: Synthetic data generated based on real-world atmospheric physics principles, latitude/longitude centroids for regional baselines, and simulated global climate indices (ENSO, NAO).

Main Results

• A high-fidelity synthetic dataset was successfully generated, covering global agricultural production, local meteorological conditions, and climate teleconnection indices from 1990 to 2023.
• This dataset explicitly incorporates physics-informed correlations between global climate drivers (ENSO, NAO) and regional weather patterns, providing a robust benchmark for Spatial-Temporal Graph Neural Networks (ST-GNNs).
• The generation process involved simulating global climate signals, coupling them to 15 regions with physics-informed coefficients, simulating local temperature and rainfall, and calculating yield using a non-linear biological stress function.

Contributions

• Creation of a novel synthetic dataset that uniquely integrates physics-informed global climate teleconnections and their regional impacts, offering a distinct advantage over standard agricultural datasets.
• Provides a specialized benchmark for evaluating Spatial-Temporal Graph Neural Networks (ST-GNNs) in their ability to model complex, long-distance climate-agriculture interactions.

Funding

• Not specified in the provided text.

Citation

@article{Hasan2025Global,
  author = {Hasan, Raza},
  title = {Global Synthetic Crop Yield, Meteorological, and Climate Teleconnection Dataset for Machine Learning Benchmarking},
  journal = {Mendeley Data},
  year = {2025},
  doi = {10.17632/y7hkz2zfcc},
  url = {https://doi.org/10.17632/y7hkz2zfcc}
}