Global River Discharge Product: Revolutionizing Streamflow Estimation

This groundbreaking project leverages cutting-edge satellite technology and advanced machine-learning techniques to create a comprehensive Global Discharge Product.
This product is live at https://qtwsa-app.onrender.com/.
Project Highlights
• Innovative Proof of Concept
• Developed a satellite-based streamflow estimation model for the Continental United States (CONUS)
• Utilized GRACE-FO satellite's Terrestrial Water Storage Anomalies (TWSA) data
• Optimized hydrologic models at each gauge, considering crucial factors such as model parameters, training data exclusion, and data lag

Advanced Machine Learning Application
• Regionalized optimized models using state-of-the-art machine learning regression techniques
• Developed multiclass, multilabel classification models to predict when and where the global product would fail, enhancing its reliability
• Employed a diverse array of ML algorithms, including XGBoost, Random Forests, Artificial Neural Networks, Support Vector Regression, and Gaussian Processes

Multi-Source Data Integration
• Incorporated data from multiple satellite missions and sources, including:
• GRACE and GRACE-FO's TWSA mission
• IMERG mission precipitation data
• MODIS snow cover and land cover data
• Temperature data

Cutting-Edge Feature Engineering
• Performed extensive feature engineering on these diverse datasets to extract relevant information
• Applied sophisticated feature selection, engineering, and extraction methods to enhance model accuracy and efficiency
• Developed novel features from multi-source satellite data to capture complex hydrological processes

Explainable AI Integration
• Implemented ExplainableAI techniques to improve model interpretability
• Provided valuable insights and cautions for product users, ensuring responsible and informed application
Impact and Potential
This project represents a significant leap forward in global streamflow estimation, offering an invaluable tool for hydrologists, climate scientists, and water resource managers worldwide. By combining multi-source satellite data with advanced machine learning and explainable AI techniques, we've created a robust and interpretable solution for understanding and predicting global water dynamics.

The Global Discharge Product has far-reaching applications, including:
• Improved water resource management
• Enhanced flood prediction capabilities
• More accurate assessment of climate change impacts on global water systems
Our innovative approach, which integrates data from multiple satellite missions and employs sophisticated machine learning techniques, contributes to the advancement of hydrological science and supports informed decision-making in water-related issues on a global scale.