Shallow aquifer monitoring using handpump vibration data

Journal article

We present a novel technology for monitoring changes in aquifer depth using handpump vibration data. This builds on our previous works using data to track handpump usage and facilitate handpump maintenance systems in rural parts of Kenya. Our motivation is to develop a cost-effective and scalable infrastructure to monitor shallow aquifers in regions where handpumps are already part of water infrastructure, but where traditional sources of groundwater monitoring data may be limited or non-existent. The data is generated using accelerometer sensors attached to the handles of nine handpumps in the study site in Kenya, instrumented for a year. These time-series data from handpumps are individually modelled using machine learning methods to track the changes in the water level with respect to the bottom of the rising main. Results show promise in modelling handpump vibration data with machine learning approaches to provide useful aquifer monitoring information from the “accidental infrastructure” of community handpumps. This technology is intended to complement existing hydrogeological modelling, and one of our key future goals is to integrate these machine learning outputs with hydrogeological information to develop more refined and robust models for shallow aquifer monitoring.

Authors: Manandhar, A; Greeff, H; Thomson, P; Hope, R; Clifton, DA

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Elsevier
• Journal: Journal of Hydrology X
• Article number: 100057
• Publication date: 2020-06-07
• Acceptance date: 2020-05-18
• DOI: 10.1016/j.hydroa.2020.100057
• EISSN: 2589-9155
• Language: English
• Keywords: novel techniques; LSTM; water column; FFR; machine learning; aquifer monitoring; handpumps