Lake buoys will warn us about E. coli outbreaks – with math

  • USGS scientist Joseph Levitt secured a buoy equipped to monitor water quality at Weirs Beach on Lake Winnipesaukee in June. Sanborn Ward / U.S. Geological Survey

  • The USGS weather station near Weirs Beach gathers data to help determine what factors produce bacteria outbreaks in the water. Richard Kiah / U.S. Geological Survey

Published: 8/9/2016 1:43:36 AM

Plenty of people have been chasing cartoon monsters in a virtual world during this Pokémon Go-filled summer. But at the local U.S. Geological Survey office, they’ve been chasing bacterial monsters in a virtual beach.

Actually, that should be Virtual Beach, since it’s a brand name of a mathematical modeling package developed by the EPA to estimate water quality.

The USGS, in partnership with health and environmental folks in state government, hopes that software package will let officials take real-time data gathered by experimental buoys on Pawtuckaway Lake and at Weirs Beach on Lake Winnipesaukee and use it to predict when dangerous amounts of E. coli bacteria are going to appear.

“We’re hoping to find something unique that correlates with an outbreak,” said Joseph Ayotte, supervisory hydrologist for the USGS New England Water Science Center, on Commerce Way in Pembroke. “Keep your fingers crossed.”

With any luck, one day we’ll be able to check a smart phone app that tells us whether any state beaches should be avoided that day – a sort of E. coli-in-the-water forecast. That’s not a crazy idea, since it’s being done in the Midwest in some lakes, but it’s not clear how realistic it is in New England, where local conditions vary more than in America’s flat heartland.

Hence the buoys, built by Richard Kiah, chief of Hydrologic Network Operations for the USGS’s office, and his team.

The buoys continuously measure things like water pH, temperature, depth, dissolved oxygen, flow and “specific conductance” (how well it conducts electricity, which is surprisingly important). Once a minute, they bounce the data off geosynchronous satellites to base stations.

That data is combined with data from on-shore weather stations (wind speed, air temperature, etc.) and then compared against daily bacteria readings, using multivariate analysis to spot patterns and correlations.

The idea is to find something like this: When the wind is blowing in a certain direction at a certain speed and the water level is X amount different from usual and the oxygen level is so high and the air temperature has changed by that amount – or some other combination of factors – then there’s a high chance that bacteria will be reproducing like mad and the beach should be closed.

That sounds well and good, but if you’re going to put complicated buoys into the lake (they cost some $20,000 a year to monitor, Kiah estimated), why not just use them to count the E.coli?

Sorry, that’s too slow.

“The best case is that it takes 24 hours to culture the bacteria” so it can be counted, Kiah said.

In other words, when state officials closed Weirs Beach last week because of high E. coli levels, they made the decision based on water samples taken a day earlier. People were swimming in that water while lab techs were waiting for the bacteria to multiply in Petri dishes so they could be analyzed. Hence the hope for mathematical models based on real-time data.

Before we go, a quick comment about the word “model,” which has nothing to do with the 1/20th-scale plastic Batmobile I built as a kid.

Models are mathematical creations, designed when experiment and experience find that certain real-world happenings (certain water and weather conditions) correlate with other real-world happenings (bacterial blooms).

A model expresses that relationship in mathematical form.

Weather forecasting models, for example, take temperature and air pressure and humidity and other factors from a zillion places and plug the data into a big formula that spits out a numerical result that can be translated into a physical prediction about, say, the likelihood of rain in my town tomorrow.

People often heap scorn on models, usually when they give an unpopular answer, but that’s lazy excuse-making. Weather forecasts have improved by many orders of magnitude in my lifetime, and you can thank mathematical modeling for much of that.

It’s true that models are always inexact, because they can never incorporate every single piece of data in the world, but they are still incredibly valuable. As statisticians like to put it: “All models are wrong, but some are useful.”

Let’s hope the Virtual Beach model proves to be useful. I love swimming in lakes.

Concord Monitor Office

1 Monitor Drive
Concord,NH 03301


© 2021 Concord Monitor
Terms & Conditions - Privacy Policy