Generative AI for Synthetic Weather Generation and Long-Range Yield Prediction

• Introduction to Generative AI in climate and agricultural research
• Understanding weather–crop interactions and yield variability
• Overview of climate, crop, soil, and yield data sources
• Data preprocessing, seasonal alignment, and missing-value handling
• Development of rainfall anomalies, heat-stress days, and Growing Degree Days

• Retrieve historical weather data using the NASA POWER API
• Integrate sample weather and crop-yield datasets
• Visualize temperature, rainfall, and crop-yield relationships

Hands-on Tools: Google Colab, Python, pandas, NumPy,
Matplotlib, NASA POWER, and FAOSTAT

📅 Day 2: Generative AI for Synthetic Weather Generation

• Importance of synthetic weather data in agricultural forecasting
• Introduction to VAE, GAN, TimeGAN, and diffusion-based weather models
• Preparing multivariate climate time-series data for generative modelling
• Generating normal, drought, heatwave, and excess-rainfall scenarios
• Validating synthetic weather using statistical and temporal indicators

• Prepare temperature and rainfall sequences for AI modelling
• Generate synthetic seasonal weather using a lightweight Conditional VAE
• Compare real and synthetic weather distributions and correlations

Hands-on Tools: Google Colab, Python, TensorFlow/Keras,
scikit-learn, SciPy, and pandas

📅 Day 3: Long-Range Yield Prediction & Explainable Climate-Risk Modelling

• Fundamentals of seasonal and long-range crop-yield prediction
• Application of Random Forest, XGBoost, LSTM, and Transformer models
• Integration of observed and synthetic weather with crop-yield data
• Explainable AI for identifying key weather and crop-growth factors
• Yield uncertainty, drought risk, heat stress, and climate-smart decision-making

• Train an XGBoost-based crop-yield prediction model
• Interpret climate-variable effects using SHAP analysis
• Generate expected, adverse, and optimistic yield scenarios

Hands-on Tools: Google Colab, Python, XGBoost, SHAP,
scikit-learn, pandas, and Matplotlib