Thanks for posting this. I saw this graph first on Twitter and found the discussion that took place under the tweet by Brennan Spiegel very interesting. See here:
Invalid consumer key/secret in configuration

One of the people commented what I also would have said (well not in the same words):
Nathaniel Krefman
@NateKrefman
Replying to @BrennanSpiegel
and @Yale

"the authors leaned very heavily on an extremely smoothened viral RNA curve to get a correlation that looks near perfect. R-squared is 0.97? Give me a break! What’s R-squared when you plot the unsmoothened data in 2A?"
"Where’s the data for literally *any other* municipality that demonstrates that this correlation (in very noisy data) is reproducible?"

Another said:
oscar pazos
@opr_tw
Replying to @BrennanSpiegel
@PCMagalhaes
and @Yale

Yes it is amazing, but this graph is more accurate to real forecastibility..

Another said:

Timothy Hochberg
@TimHochberg
Replying to @BrennanSpiegel
and @Yale

This paper is very cool, but the graphs are LOWESS smoothed, which is a non-causal filter, and the daily data is quite noisy. So, it's unclear how well this could be used predict outbreaks in advance. It would be interesting to see this same data analyzed with causality in mind.

Anyway, that twitter thread is very interesting, I recommend reading it.

Dear all,
Sharing link to pre-print of a paper  that provides evidence that WBE provides an early warning by upto a week (see graph below)


Title: SARS-CoV-2 RNA concentrations in primary municipal sewage sludge as a leading indicator of COVID-19 outbreak dynamics
Authors: Jordan Peccia et al.

Abstract:
We report a time course of SARS-CoV-2 RNA concentrations in primary sewage sludge duringthe Spring COVID-19 outbreak in a north eastern U.S. metropolitan area. SARS-CoV-2 RNA was detected in all environmental samples and, when adjusted for the time lag, the virus RNA concentrations were highly correlated with the COVID-19 epidemiological curve (R² =0.99) and  local hospital admissions (R² =0.99).  SARS-CoV-2 RNA concentrations were a seven-day leading indicator ahead of compiled COVID-19 testing data and led local hospital admissions data by three days. Decisions to implement or relax public health measures and restrictions require timely information on outbreak dynamics in a community.

Regards
paresh