50 Years of Water Body Monitoring: The Case of Qaraaoun Reservoir, Lebanon

TL;DR

This paper tackles unreliable reservoir monitoring in the QR in Lebanon due to sensor failures by introducing a sensor-free remote-sensing workflow that estimates reservoir volume from satellite-derived water surface. It combines a new water segmentation index WCWI, derived from $AWEInsh$ and $NDWI$, with an SVR model trained on a bathymetric-ground-truth dataset to map water surface area to relative volume, enabling near real-time estimates from Sentinel-2 and Landsat imagery. The approach achieves shoreline segmentation accuracy >95% and volume-prediction performance with $R^2$ > 0.98 and $MAPE$ < 6% across multiple years, with MAE around a few million cubic meters. It produces a long-term, multi-decadal time series (>50 years) and a public dashboard to visualize trends, supporting climate-change insights and more resilient water management; the methodology is generalizable to other water bodies and sensor-limited regions.

Abstract

The sustainable management of the Qaraaoun Reservoir, the largest surface water body in Lebanon located in the Bekaa Plain, depends on reliable monitoring of its storage volume despite frequent sensor malfunctions and limited maintenance capacity. This study introduces a sensor-free approach that integrates open-source satellite imagery, advanced water-extent segmentation, and machine learning to estimate the reservoir's surface area and, subsequently, its volume in near real time. Sentinel-2 and Landsat 1-9 images are processed, where surface water is delineated using a newly proposed water segmentation index. A machine learning model based on Support Vector Regression (SVR) is trained on a curated dataset that includes water surface area, water level, and water volume derived from a reservoir bathymetric survey. The model is then able to estimate the water body's volume solely from the extracted water surface, without the need for any ground-based measurements. Water segmentation using the proposed index aligns with ground truth for over 95% of the shoreline. Hyperparameter tuning with GridSearchCV yields an optimized SVR performance, with an error below 1.5% of the full reservoir capacity and coefficients of determination exceeding 0.98. These results demonstrate the method's robustness and cost-effectiveness, offering a practical solution for continuous, sensor-independent monitoring of reservoir storage. The proposed methodology is applicable to other water bodies and generates over five decades of time-series data, offering valuable insights into climate change and environmental dynamics, with an emphasis on capturing temporal trends rather than exact water volume measurements.

Figures (6)

• Figure 1: Workflow of the proposed pipeline, from data acquisition and preprocessing to water surface detection and volume estimation.
• Figure 2: Comparison of water segmentation results using different water indices: (a) $NDWI$, (b) $AWEInsh$ and (c) $WCWI$ composite index on a Sentinel-2 imagery from 17 October 2023 where yellow outlines represent the nominal lake contour and red indicate detected water extent.
• Figure 3: Water segmentation results in red versus ground truth values in blue: (a) and (b) Sentinel-2 imagery for 1 March 2023; (c) and (d) Landsat 8 imagery for 21 January 2023; (e) and (f) Sentinel-2 imagery for 26 September 2024; and (g) and (h) Landsat 8 imagery for 8 June 2024. The nominal lake contour is shown in yellow.
• Figure 4: Time series (1973–2025) of water surface area and storage volume in the Qaraaoun Reservoir. July 2025 recorded an exceptionally low volume ($49.2\times10^6\,$m³), representing a 66% decline from July 2024 and 56% from July 2023, highlighting an emerging drought signal (see https://geoai.cnrs.edu.lb/qaraaoun).
• Figure 5: Historical Landsat-3 satellite image of the Qaraoun Reservoir captured on 17 January 1982, believed to represent the minimum recorded water extent in the past five decades.