/mixed-method-puso

Modeling Paddy Crop Failure Using Machine Learning

Primary LanguageJupyter Notebook

Modeling Paddy Crop Failure Using Machine Learning

This project utilizes Sentinel Data, Climate Data (ERA5), CHIRPS Rainfall Data, Radar Vegetation Index (RVI), and other geospatial datasets to model paddy crop failures using XGBoost. By integrating diverse geospatial and climatic datasets with a powerful gradient boosting algorithm, the project aims to predict crop failure risks, analyze contributing factors, and provide actionable insights for agricultural stakeholders.

📋 Objectives

  • Predict Crop Failures: Develop an XGBoost-based machine learning model to forecast paddy crop failures.
  • Analyze Influencing Factors: Evaluate the effects of climatic, vegetation, and geospatial data on crop health.
  • Support Decision-Making: Enable early intervention and risk management through data-driven insights.

🛠️ Key Features

  1. Integration of Multi-Source Data:

    • Fuse Sentinel-1 data, ERA5 climate reanalysis, CHIRPS rainfall data, RVI, and other geospatial data to create a comprehensive dataset.

  2. XGBoost Model:

    • Train and optimize a gradient boosting model using engineered features from geospatial datasets.
    • Evaluate the model’s performance using metrics such as accuracy, precision, recall, and F1-score.

  3. Spatial and Temporal Analysis:

    • Perform geospatial visualization of risk areas using GIS tools.
    • Conduct temporal analysis to study changes in vegetation and climate over time.

  4. Insights and Recommendations:

    • Identify critical factors contributing to crop failure and suggest mitigation strategies.

📊 Data Sources

1. Sentinel Data

2. Climate Data (ERA5)

3. CHIRPS (Climate Hazards Group InfraRed Precipitation with Station Data)

4. Radar Vegetation Index (RVI)

5. Other Geospatial Data

⚙️ Technologies Used

  • Programming Language: Python
  • Libraries and Tools:
    • Data Handling: pandas, numpy, xarray
    • Machine Learning: xgboost, scikit-learn
    • Geospatial Analysis: geopandas, rasterio, shapely, GDAL
    • Visualization: matplotlib, seaborn, folium, plotly
    • Remote Sensing: SentinelHub, earthpy, pyproj

🚀 Getting Started

Prerequisites

  1. Install Python (3.7 or above).
  2. Obtain necessary datasets:

Steps

  1. Data Collection: Download Sentinel, ERA5, CHIRPS, and other datasets.
  2. Preprocessing:
    • Align datasets spatially and temporally.
    • Extract features such as NDVI, RVI, rainfall indices, and soil properties.
  3. Feature Engineering:
    • Create a feature matrix by combining geospatial data with crop health metrics.
  4. Model Training:
    • Train the XGBoost model on the preprocessed data.
    • Perform hyperparameter tuning for optimal performance.
  5. Evaluation:
    • Validate the model using a holdout dataset and compute metrics like RMSE, precision, and recall.
  6. Visualization:
    • Generate geospatial heatmaps and dashboards to display risk areas.

🤝 Contributions

We welcome suggestions and contributions to enhance the project. Please submit issues or pull requests for improvements.

📜 License

This project is licensed under the MIT License.