About

Eleven million measurements, two blind spots

Fecal indicator bacteria decide whether the world's rivers, lakes, and beaches are called safe. This map is the first harmonized global picture of that record, and of where it goes dark.

Agencies almost never test water for pathogens directly. Instead they count indicator organisms: Escherichia coli in fresh water, intestinal enterococci in marine water. Those counts govern beach closures, drinking-water advisories, and compliance reporting worldwide. Until now the record behind those decisions had never been assembled and characterized at a harmonized global scale.

We compiled 11,110,309 surface-water observations from nine public sources, standardized to one schema, one indicator vocabulary, and one unit. The map plots the 189,872 monitoring sites those observations come from. Each dot is a place where somebody tested the water.

Two structural weaknesses emerge, and they are the reason this map exists.

Why it matters

For fresh waters, both gaps are widest in the low- and middle-income regions that carry the greatest waterborne-disease burden and most often drink untreated surface water. The monitoring gap is therefore a question of public-health equity, not just data coverage. The case for wider open-data practice, and for higher-frequency, event-based, predictive monitoring tied to accountable response, follows directly from what this map shows.

How to read the map

Open the Map tab and use the guided tour to walk the argument, or filter by source, indicator, water-body type, and year. The heat layer shows monitoring density; switch to points to see individual sites. Click any site to pull its full observation time series.

Methods

Harmonization, cadence, and the cost of sampling slowly

Nine national and subnational programs, each with its own schema, units, and indicator vocabulary, reduced to a single comparable record.

Harmonization and quality control

All concentrations were standardized to colony forming units per 100 mL. Most probable number values were treated as colony forming units, an approximation; qPCR-equivalent and nonconvertible units were dropped. Censored values were assigned a direction and filled at half the lower detection limit, or at the upper limit for right-censored values. Source-specific location descriptors were mapped to a controlled vocabulary of water-body types, and the marine and inland enterococci records were unified into a single enterococci class.

Repeated observations were averaged within each source, then deduplicated across sources using a fuzzy key with roughly 1 km coordinate tolerance, so records appearing in several portals are not double-counted. Filtering to a valid indicator, a plausible date and coordinates, and a non-negative bounded value reduced 11,151,268 harmonized records to 11,136,805. Groundwater is a distinct exposure pathway and was the only realm excluded; removing its 26,496 records yields the final analysis set of 11,110,309 surface-water observations.

Sampling cadence

For each site we computed the number of observations, the temporal span, and the median gap between consecutive distinct sampling days. Counting distinct days rather than raw records prevents same-day replicate sampling from inflating apparent frequency. Sites were binned by that median gap into daily, near-daily, weekly, monthly, quarterly, yearly, and longer classes.

The high-frequency analysis applies a stricter test for genuine near-daily monitoring: at least 30 distinct sampling days within a single recent calendar year (2018 or later), with a distinct-day median gap of at most three days, judged on each site's densest year.

Safety thresholds

Recreational safety uses realm-specific single-sample values from the 2012 US EPA recreational water quality criteria: 410 CFU/100 mL for E. coli in fresh water and 130 CFU/100 mL for enterococci in marine water.

Downsampling: what a slower schedule would have missed

To quantify the safety cost of coarse sampling, each near-daily site-year is treated as ground truth. We simulate monitoring programs at intervals from daily to quarterly: to every scheduled date we assign the nearest available observation, post a single-sample advisory that persists until the next scheduled sample, and score that standing advisory against the true daily state on every observed day, averaging false-safe and false-unsafe day rates over all phase offsets of the schedule.

The result is the paper's second blind spot. Slower schedules do not merely lose resolution; they systematically misadvise, and they err toward calling unsafe water safe.

Publication

Eleven Million Measurements, Two Blind Spots

The State of Global Surface-Water Fecal Indicator Monitoring

Mike Vlah and Matt Ross are at Colorado State University. Evan Thomas is at the University of Colorado Boulder.

Key points

• The first harmonized global surface-water fecal indicator dataset: 11.1 million measurements from nine public sources.
• Openly shared data concentrate in the US, Europe, and Canada, leaving the regions with the highest waterborne disease burden least visible.
• Genuine near-daily sampling is rare and almost entirely US-based, so most monitoring cannot catch the transient pulses that drive risk.

Abstract

Fecal indicator bacteria (Escherichia coli in fresh water and intestinal enterococci in marine water) are the primary tools used worldwide to judge whether surface waters are safe from waterborne pathogens, but that record has never been characterized at a harmonized global scale. We compiled, harmonized, and analyzed 11,110,309 surface-water fecal indicator observations (5.9 million freshwater, 5.2 million marine) from nine public sources, standardized to a common schema, indicator set, and unit.

Two structural weaknesses emerge. First, openly shared data are spatially unequal: the United States, Europe, and Canada hold 94.5% of records, while much of the Global South is sparse and large monitoring nations are absent from open repositories. Second, the record is temporally rigid: only about 1.5% of multi-sample sites achieve a near-daily cadence, itself almost entirely confined to the United States, even though contamination is a transient, weather-driven condition.

Using a temporal downsampling analysis, we simulated monitoring regimes across every near-daily record. On average, weekly sampling misadvises roughly one monitored day in nine, while monthly sampling misadvises one in seven, biased toward false reassurance. For fresh waters, these gaps are widest in the low- and middle-income regions that bear the greatest waterborne-disease burden and most often use surface water for drinking, making the monitoring gap a question of public-health equity. We argue for wider open-data practices and for higher frequency, event-based, predictive monitoring coupled to accountable response.

UNU-INWEH Policy Brief · July 2026

The World Cannot See Its Water

Data sharing and sampling frequency are policy failures, not technical ones.

Read the brief (PDF) 7 pages · UNU Institute for Water, Environment and Health

The map on this site shows where the world measures the safety of its water. The policy brief asks why the map looks the way it does, and concludes that the gaps are governed, not merely unfunded. The instruments, the laboratories, and the data portals already exist. What is missing is the mandate to publish, the obligation to sample when risk is highest, and an accountable actor positioned to respond.

What the brief argues

Official statistics inherit the blind spot. The WHO/UNICEF Joint Monitoring Programme reports one modelled national estimate per country per year, and for Africa, 24 of 54 countries have no safely-managed drinking water estimate at all. Turn on the SDG 6.1.1 layer on the map tab to see it. Where the estimate exists it is national and annual, so the granularity a regulator would need to act on a specific water body, in a specific season, is absent by construction.

Recommendations

• National regulators should mandate open publication of fecal indicator data as a condition of monitoring, in machine-readable form under FAIR principles, with location, method, detection limit, and sampling time attached.
• Health and environment ministries should replace fixed-calendar compliance sampling with risk-based, event-triggered sampling that captures rainfall and overflow events.
• Standard-setting bodies should state the sampling cadence their thresholds assume, making an implicit assumption explicit.
• SDG custodian agencies should publish data-sharing status and sampling frequency alongside every indicator estimate.
• Development banks and donors should fund monitoring as accountability infrastructure and condition disbursement on open publication.
• Governments should pair every monitoring obligation with a named, accountable responder. A measurement no one must act on is an expense, not a safeguard.

Citation

The views expressed in the brief do not necessarily reflect the official policy or position of the United Nations University.