Wastewater epidemiology · COVID-19

How many people have COVID right now?

An interactive tool for estimating current COVID-19 prevalence from wastewater surveillance data, calibrated to published expert estimates across two years of data.

What is this? Public health researchers measure SARS-CoV-2 (the COVID-19 virus) in wastewater as an early warning system. Because infected people shed the virus in their feces, the concentration in sewage reflects how many people in a community are currently infected — even those who never get tested.

This tool converts those wastewater readings into an estimated number of people currently infectious. The underlying model was fitted to 11 paired data points spanning two years of wastewater signal levels and independent expert prevalence estimates, and explains 95% of the variance in that dataset.

Where to find your local wastewater reading: Visit WastewaterSCAN and find your nearest monitoring site. Look for the "N Gene:PMMoV" value — that's the number to enter below. You can use the slider or type directly into the number field.

Enter your local readings
Find this at data.wastewaterscan.org — select your nearest monitoring site and look for the N Gene:PMMoV reading. Use ↑↓ arrow keys in the number field for fine adjustments.
Used to detect whether COVID is currently rising, falling, or stable — which adjusts the estimate to account for people still infectious from earlier in the trend.
Estimated prevalence
Very lowModerateHigh
Currently infectious
of the US population
Uncertainty range
plausible bounds (±25%)
Est. new infections/day
national estimate
Prevalence curve — log scale
Model fit (power law) Uncertainty band (±25%) Your current reading Calibration data points
Power law curve fitted to 11 calibration data points from 2023–2026, shown on a log-log scale.

Data sources & credits

Wastewater data: WastewaterSCAN (Stanford & Emory), a national wastewater surveillance program monitoring SARS-CoV-2 and other pathogens across 140+ US sites.

Prevalence estimates: @JPWeiland on X, who publishes independent estimates of COVID infectious prevalence derived from wastewater and other surveillance data. All 11 calibration data points used to fit this model are drawn from his published estimates (2023–2026).

Tool methodology: Developed with assistance from Claude (Anthropic). The model uses a power law fit (log-log regression) across 11 paired signal/prevalence observations, achieving R²=0.953.

Disclaimer: This is an independent estimation tool, not an official public health product. Estimates carry significant uncertainty and should be interpreted as order-of-magnitude guidance, not precise counts.

Calibration data points & methodology
Date Signal Trend JPWeiland estimate Model prediction
Jun 22, 202617.33Stable<40,000 new/day (~1 in 1,196)1 in 1,093
Apr 10, 202637.09Declining1 in 550 infectious1 in 535
Mar 29, 202651.39Declining135,000 new/day; ~1 in 5001 in 394
Mar 10, 202673.90Declining325,000 new/day; ~1 in 2001 in 280
Nov 15, 202497.41Declining185,000 new/day; 1 in 1801 in 216
Apr 19, 2024109.80Declining200,000 new/day; 1 in 1641 in 193
Jul 18, 2025162.70Rising209,000 new/day; ~1 in 1601 in 133
Sep 5, 2025323.20Declining480,000 new/day; 1 in 701 in 70
Nov 27, 2023398.50Rising620,000 new/day (~1 in 90)1 in 58
Sep 6, 2024444.80Declining870,000 new/day; 1 in 381 in 52
Aug 2, 2024642.00Stable900,000 new/day; 1 in 371 in 37

How the model works: The N Gene:PMMoV ratio normalizes SARS-CoV-2 genetic material in wastewater against Pepper Mild Mottle Virus (PMMoV), a stable human fecal indicator virus. This normalization corrects for dilution effects and population size differences between monitoring sites.

Model fitting: A power law relationship (log-log regression) was fitted across 11 paired observations of wastewater signal and prevalence estimates published by @JPWeiland between 2023 and 2026. The resulting equation is: 1-in-N = 15,914 × Signal^(−0.939), with R²=0.953. The near-linear exponent of −0.939 means that halving the signal roughly doubles the estimated 1-in-N, with a slight flattening at very low signal levels. On the log-log chart, this appears as a straight line — which is why the log scale makes the fit so much easier to read.

Phase detection: When today's signal differs from the signal 10 days ago by more than 10%, the tool detects a trend automatically. During a decline, people infected when the signal was higher are still contagious today — so the effective signal used in the model is a weighted average of today's reading and the reading from 10 days ago (40%/60% weighting). During a rise, the weighting reverses (60%/40%). During stable periods, today's signal is used directly.

Uncertainty: The ±25% uncertainty band reflects typical residual error in the model — most predictions fall within this range of Weiland's actual estimates. Estimates should be treated as order-of-magnitude guidance. "Infectious" per Weiland's framing means currently capable of transmitting; total infected (including those recovering) would be somewhat higher.