Search

Broken water pumps continue to impede efforts to deliver clean and economically-viable water to the global poor. The literature has demonstrated that customers’ health benefits and willingness to pay for clean water are best realized when clean water infrastructure performs extremely well (>99% uptime). In this paper, we used sensor data from 42 Afridev-brand handpumps observed for 14 months in western Kenya to demonstrate how sensors and supervised ensemble machine learning could be used to increase total fleet uptime from a best-practices baseline of about 70% to >99%. We accomplish this increase in uptime by forecasting pump failures and identifying existing failures very quickly. Comparing the costs of operating the pump per functional year over a lifetime of 10 years, we estimate that implementing this algorithm would save 7% on the levelized cost of water relative to a sensor-less scheduled maintenance program. Combined with a rigorous system for dispatching maintenance personnel, implementing this algorithm in a real-world program could significantly improve health outcomes and customers’ willingness to pay for water services. [ABSTRACT FROM AUTHOR]

Summary: The Standardized Precipitation Index (SPI) has become a popular measure of drought across the globe. Several techniques have been employed in attempts to estimate the time at which any changes in drought patterns occur using categories defined by SPI values. Log-linear models, odds ratios and multiple confidence intervals have all been used while trying to estimate such changepoints and have met with varying degrees of success, in part due to small cell counts and the large number of inferences that must be carried out. The purpose of this paper is to provide an alternative method for the detection of a changepoint in SPI data, specifically in the matrix of transitions from one drought class to another. A resampling algorithm is provided that both detects a potential changepoint and determines its significance under the assumption that at most one changepoint is present. Simulation results are provided to validate the method as well as the suggested metrics for detection. Data from Kenya is used to provide four examples where the technique is implemented on observed rainfall data. [Copyright &y& Elsevier]