T7 – Modelling the relationship between sand mining and river turbidity and bank erosion using satellite data

Short description

After water, sand is the most consumed natural resource in the world, as it is a key ingredient for making concrete. Most sand mining occurs in large tropical river systems where monitoring and enforcement capacity is limited, allowing illegal or unregulated sand mining to proliferate. These activities can substantially alter river morphology, increase turbidity, destabilize riverbanks, and disrupt sediment transport, ultimately affecting ecosystems and infrastructure [1]. Because field monitoring is difficult in many regions, satellite remote sensing has emerged as a promising tool to detect sand-mining activity and assess its environmental impacts.

Recent studies have demonstrated the potential of satellite observations to identify sand-mining activity and associated disturbances. For example, remote sensing has been used to detect dredging vessels and nighttime mining activity using combinations of optical imagery and nighttime lights [2], while other studies have used Sentinel-1 SAR and deep learning approaches to map mining hotspots and estimate extraction budgets in major river deltas [3]. Additional work has shown that satellite-derived turbidity or suspended sediment estimates can capture water-quality disturbances caused by dredging operations [4]. Despite these advances, most studies focus either on detecting mining activity or on documenting environmental impacts, rarely integrating both within a unified monitoring framework. Moreover, many existing case studies are observational in nature: currently, there are no models that directly link sand mining activities to river turbidity (which can impact ecosystems) and bank erosion (which can lead to river bank collapse).

This thesis aims to develop an integrated remote-sensing-based framework that quantitatively links sand-mining activity to its environmental impacts in large river systems. Specifically, the work will derive a reproducible proxy of mining intensity from multi-source satellite observations and assess its relationship with indicators of water-quality disturbance and morphological instability. By bridging activity detection and impact quantification within a unified modeling approach, the thesis seeks to provide a scalable methodology to support monitoring, regulation, and management of river sand mining in data-scarce regions.

PROPOSED ACTIVITIES

Literature review and conceptual framework.
Conduct a targeted review of remote-sensing studies on sand-mining detection and its environmental impacts, with emphasis on identifying commonly used indicators (e.g., vessel detection, turbidity anomalies, riverbank change) and methodological limitations.

Data acquisition and preprocessing
Collect and preprocess multi-source satellite datasets for the selected study region (e.g., Mekong or Volta deltas), including Sentinel-1 SAR for activity detection, Sentinel-2 optical imagery for water quality and turbidity proxies, and Landsat/Sentinel archives for longer-term morphological analysis.

Derivation of a mining-intensity proxy from remote sensing
Develop a reproducible workflow to identify probable sand-mining hotspots and convert their temporal recurrence, spatial persistence, or activity frequency into a mining-intensity index (e.g., number of boats x cargo capacity x days of mining). This index will serve as the explanatory variable in the empirical modelling phase, avoiding the need for direct estimates of extracted volumes (which cannot be easily measured).

Extraction of remotely sensed impact indicators
Quantify one or more impact variables associated with sand mining, focusing on robust and measurable indicators such as turbidity anomalies and, if feasible, bank retreat or channel-margin instability. These indicators will be derived at spatial and temporal scales consistent with the mining-intensity metric to support comparative analysis.

Empirical modelling and evaluation of impact-response relationships
Build and evaluate a statistical model relating mining intensity to the selected impact indicator(s), while accounting for major confounding factors such as seasonality, discharge, or reach-scale morphology where possible. The final analysis will assess whether a consistent quantitative relationship can be identified and discuss its usefulness for monitoring and management of river sand mining.

References

  1. Rentier, Eline S., and L. H. Cammeraat. “The environmental impacts of river sand mining.” Science of the Total environment 838 (2022): 155877.
  2. Duan, Hongtao, Zhigang Cao, Ming Shen, Dong Liu, and Qitao Xiao. “Detection of illicit sand mining and the associated environmental effects in China’s fourth largest freshwater lake using daytime and nighttime satellite images.” Science of the total environment 647 (2019): 606-618.
  3. Kumar, Sonu, Edward Park, Dung Duc Tran, Jingyu Wang, Huu Loc Ho, Lian Feng, Sameh A. Kantoush, Doan Van Binh, Dongfeng Li, and Adam D. Switzer. “A deep learning framework to map riverbed sand mining budgets in large tropical deltas.” GIScience & Remote Sensing 61, no. 1 (2024): 2285178.
  4. Chowdhury, Masuma, Ignacio De La Calle, Irene Laiz, and Ana B. Ruescas. “Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques.” Remote Sensing 17, no. 22 (2025): 3716.

Relevant courses and knowledge

Natural Resources Management 1 (optional: machine learning or other AI courses from computer science)

Number of students

1

Requisites

The student should be comfortable with statistical analysis (e.g., extreme values), and data handling. Proficient coding skills (Python preferred) and interest in remote sensing are needed.