4.4 Billion Lack Safe Drinking Water

Greenwood, E. E. Lauber, Thomas, van den Hoogen, Johan,  Donmez, Ayca, Bain, Robert E. S.,  Johnston, Richard, Crowther, Thomas W.,  Julian, Timothy R.  Mapping safe drinking water use in low- and middle-income countries;  Science 385, 784–790 (2024).

 A recent study (cited above) found that more than 4.4 billion people in low- and middle-income countries lack access to safely managed drinking water.

According to the authors more than half the world’s population lacks safe drinking water and fecal contamination affects almost half the population of low and middle income countries. (The analysis excludes the richer nations which probably also have some portion of their populations without clean and safe drinking water.)  The findings either show that previous global estimates for lack of safe drinking water availability at 2.2 billion  have greatly underestimated the problem, or their model for prediction overstates it.

The availability of safe drinking water is far from universal, but exactly how it varies geographically and why this occurs is not well understood. Greenwood et al. combined Earth Observation data, geospatial modeling, and household survey data and linear regression modeling to estimate that only one in three people in low- and middle-income countries have access to safely managed drinking water. 

Using four criteria for safe drinking water: they must be improved, consistently available, accessible where a person lives and free from contamination. The authors used environmental data combined with survey responses about the 4 criteria from 64,723 households across 27 low- and middle-income countries between 2016 and 2020. If a household failed to meet any of the four criteria, it was categorized as not having safe drinking water. 

The authors trained a machine-learning algorithm using survey data and global geospatial data. “The resulting model was best at predicting drinking water accessibility on premises (R² = 0.48, MAE = 0.18) and fecal contamination (R² = 0.44, MAE = 0.16) followed by an improved drinking water source (R² = 0.25, MAE = 0.11). Model performance was worse when predicting populations with drinking water availability (R² = 0.01, MAE = 0.09) based on leave-one-country-out cross-validation R².” (Greenwood et al 2024) The low  R squared (R²) could reflect limited associations between predictors and a household member’s subjective assessment of experiencing water insufficiency which is influenced by their practices of water storage or use of multiple water sources when in need  to estimate that 4.4 billion people lack access to safe drinking water, of which half are accessing sources tainted with the pathogenic bacteria Escherichia coli.

 R-squared ( R²) is a statistical measure that indicates how well a model's independent variables explain the variation in a dependent variable. It's also known as the coefficient of determination. An R squared of 1.0 means that the model perfectly predicts the data. The closer the R squared is to 1.0 the better the fit. Not so much here. There are anomalies in the calculation that could generate a low R squared and still have the model be predictive. This data sample is incredibly small to build this type of model. Variations in water availability and quality change over time and location and population density. The predictive validity of the model is brought into question. 

The United Nations, General Assembly in its 64th session: 2009-2010 adopted a resolution that Safe Drinking Water and Sanitation was a Human Right. The WHO/UNICEF Joint Monitoring Program for Water Supply ,Sanitation and Hygiene (JMP) has reported country, regional and global estimates of progress on drinking water, sanitation and hygiene (WASH) since1990. The JMP uses a mathematical extrapolation of available data to create their estimates.

The 2017 report was the first this was followed by progress in 2019, 2021 and 2023. The 2023 update estimated (using country provided data) that in 2022, 27% of the global population (2.2 billion people) lacked “safely managed drinking water”– meaning water at home, available, and safe. 43% of the global population (3.5 billion people) lacked “safely managed sanitation” – meaning access to a toilet or latrine that leads to treatment or safe disposal of excreta. 25% of the global population (2.0 billion people) did not have access at home to a handwashing facility with soap and water.

Greenwood et al estimated that 88% of all people living in LMICs use an improved drinking water source. Their results emphasize that access to an improved drinking water source does not always provide safe drinking water as almost half of the LMIC populations (48%) were estimated to be exposed to fecal contamination in their primary drinking water source. Their predictions show that more than half of the populations of Oceania, sub–Saharan Africa, southeastern Asia, and Latin America and the Caribbean may be exposed to  drinking water contaminated with E. coli.

Lack of accessibility of drinking water on premises was the second most common subcomponent limiting safe drinking water coverage, with an estimated 36% of the overall LMIC population not having water on premises. This is especially true in sub–Saharan Africa where Greenwood et al estimated that over 650 million people lack drinking water services on premises. However, the bottom line is that Greenwood et al used  limited data and AI to create an model  with limited correlation to estimate the number of people on earth without access to safe drinking water. The WHO/UNICEF Joint Monitoring Program using limited data and estimating techniques found that number of people without access to safe drinking water to be half of the Greenwood estimate. It’s probably billions and it is humanity’s big challenge, just using an AI trained model, Earth Observation data, and geospatial modeling does not necessarily make the estimate better.

Read Tim Smedley’s book “The Last Drop” for a feel of what this means in people’s lives.