A machine learning approach to estimate domestic use of public and private water sources in the United States.
Journal:
Water research
PMID:
39923400
Abstract
In the United States, people obtain water for household use from one of two sources. Public water systems, which are subject to rules and regulations under the Safe Drinking Water Act, or private sources such as domestic wells, which are not subject to federal regulation and are generally the responsibility of the homeowner or occupant. Public water systems are required to treat their drinking water and conduct regular testing to ensure the delivery of safe water to consumers. From a public health perspective, it is essential to know who is drinking what water to determine risk and impacts from water-borne disease and contamination. We present a new machine-learning approach to estimating water supply source (public or private) at the census block level for the year 2020. While previous studies have largely focused on spatially delineating either public or private water supply, our method incorporates data from both universes, resulting in more accurate modeling results. The utilization of machine learning and additional explanatory data that have not been considered in prior studies results in the most accurate and up-to-date estimate of the count and location of users supplying household water from either a private source or a public water supply. We estimate that 14.1 % of US housing units are supplied by private wells and 84.9 % of housing units are served by a public water system as of 2020.