GLoRS (Global Long-term Reservoir Sedimentation) - Source Code

Computational pipeline and research notebooks for generating the GLoRS global reservoir sedimentation dataset.

---

Version: 1.0 | Primary Authors: Sanchit Minocha & Faisal Hossain
Affiliation: University of Washington, Seattle, WA 98105, US

---

📌 Table of Contents

• Overview
• Repository Structure
• Prerequisites
• Usage
• Dataset
• Citation
• Contact

---

📖 Overview

This repository contains the source code and computational pipeline used to generate the GLoRS dataset. The pipeline leverages the RECLAIM (Reservoir Estimation of Capacity Loss using AI-based Methods) framework to predict annual reservoir sedimentation and capacity loss at a global scale by integrating observed records with multi-decadal satellite-based Earth observations.

---

🌐 Interactive Web Application

Explore the dataset visually through our interactive portal:

👉 GLoRS - Interactive Web Application

---

📂 Repository Structure

File / Directory	Description
apply_reclaim.ipynb	Core Pipeline: Generates the GLoRS predictions using the RECLAIM framework. This is the primary notebook used for global dataset production.
paper_global_reclaim_application_v1.ipynb	Exploratory Analysis: Used for data exploration, preparing manuscript plots/figures, and calculating aggregated lifetime statistics.

Pipeline Highlights:

• 🛠️ Data Ingestion: Processes global sources including GRILSS, RAT 3.0, GLC-SHARE, and HWSD.
• 🧪 Feature Engineering: Automated log-scaling and area-based transformations.
• 🤖 ML Ensemble: Application of CatBoost, XGBoost, and LightGBM models via the pyreclaim library.

---

🛠️ Prerequisites

To run the notebooks, ensure the following frameworks and libraries are installed:

1. Core Frameworks

• pyreclaim: Core machine learning logic for capacity loss prediction.
• RAT 3.0: Reservoir Assessment Tool for satellite-derived hydrological time-series.

2. Python Environment

The environment should include pyreclaim and standard data science libraries:

pip install geopandas pandas numpy catboost xgboost lightgbm scikit-learn matplotlib seaborn pyreclaim

---

🚀 Usage

Follow these steps to generate GLoRS predictions:

1. Environment Setup: Ensure pyreclaim is installed and properly configured in your Python environment.
2. Data Acquisition:
   • Obtain Reservoir Dynamics (RAT outputs) and Static Data (catchment boundaries, GRILSS metadata, etc.).
3. Path Configuration:
   • Open the notebooks and update the input/output file and folder paths to match your local directory structure.
4. Execution:
   • Run apply_reclaim.ipynb to generate the annual sedimentation rate predictions (GLoRS).
   • Use paper_global_reclaim_application_v1.ipynb for further analysis and visualization.

---

🗺️ Dataset

The output products (CSV and GeoJSON) generated by this code are available at:

👉 GLoRS Dataset (OSF)
👉 DOI: https://doi.org/10.17605/OSF.IO/2KF3R

---

📝 Citation

If you use this code or the resulting dataset in your research, please cite:

Minocha, S., & Hossain, F. (2026). Hidden sediments, lost capacity: How fast is the world losing reservoir storage in a changing climate? Proceedings of the National Academy of Sciences. [under review]

---

✉️ Contact

For any inquiries or support, please contact:

Sanchit Minocha (msanchit@uw.edu)
Faisal Hossain (fhossain@uw.edu)
Email: saswe@uw.edu

Affiliation:
Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98105, US