Recently, IoT technologies have been used to monitor critical infrastructure in remote settings. Prior research has supported active O&M campaigns, quantified groundwater extraction rates, and evaluated service delivery approaches. In this study, continuous data collection was used to examine the operating characteristics of rural water infrastructure.
397 randomly-selected groundwater pump sites were observed within Plateau State, Nigeria over 12 months in 2021. 200 of these sites were instrumented with in-situ sensor systems, including 100 handpump sensors, 50 AC electrical sensors, and 50 water level cistern sensors. Bi-monthly phone calls and site visits were used to observe pump functionality statuses and served as ground-truth data over the study period.
An automated expert classifier system generated statuses for instrumented pumps on a weekly basis. Classifier statuses were compared to ground-truth statuses, showing overall high accuracy (82.4%), with good sensitivity (88.9%) but poor specificity (14.3%). The classifier was able to accurately detect running pumps, but did not perform well in detecting failures.
Varied responses were seen in pump usage as a function of rainfall, with handpump use decreasing significantly, AC pump usage decreasing to a lesser degree, and DC pump usage increasing in response to local rainfall.
A statistical comparison of the 200 instrumented to 197 non-instrumented sites showed significant overall functionality level differences due to a baseline functionality criteria for sensor installation, but similar repair and failure rates on a bi-monthly basis. This suggests that in terms of functional change, the sensor-enabled group statistically represented the larger group of water points in Plateau State.