Measuring Escherichia coli in a single-grab sample of stored drinking water is often used to characterize drinking water quality. However, if water quality exhibits variability temporally, then one-time measurement schemes may be insufficient to adequately characterize the quality of water that people consume. This study uses longitudinal data collected from 193 households in peri-urban Tanzania to assess variability in stored water quality and to characterize uncertainty with different data collection schemes. Households were visited 5 times over the course of a year. At each visit, information was collected on water management practices, and a sample of stored drinking water was collected for E. coli enumeration. Water quality was poor for households, with 80% having highly contaminated (>100 CFU per 100 mL) water during at least one visit. There was substantial variability of water quality for households, with only 3% of households having the same category (low, medium, or high) of water quality for all five visits. These data suggest a single sample would inaccurately characterize a household's drinking water quality over the course of a year and lead to misestimates of population level access to safe drinking water.
Keywords: Monte Carlo simulation; drinking water quality; low- and middle-income country; sustainable development goals; temporal variability; water monitoring.