Setting a turbidity alert threshold for intake draw decisions

A stone water intake tower standing in open reservoir water, with hills in the background.

Most utilities inherit their turbidity alert level. Someone picked 10 NTU or 25 NTU years ago because it sounded reasonable, wrote it into the SOP, and nobody has revisited it since. That's a problem if your source lake has changed since then, or if the number was never tied to anything your plant cares about.

A threshold that works is one you can trace back to a real constraint: coagulant dosing capacity, filter run time, or a permit limit on raw water turbidity before you're required to switch intakes or blend sources. If you can't explain why the number is what it is, it's not a threshold, it's a guess with a decimal point.

Start with what your plant can actually handle

Before you pick an NTU number, go talk to the treatment side. Ask what raw water turbidity starts to strain your coagulant dosing, and at what point filter runs get short enough to matter. Some plants can ride out a spike to 40 or 50 NTU with an adjusted dose and barely notice. Others start losing filter run time past 15 NTU because of how their media is sized. There's no industry-standard number here that applies to your plant, only the one your operators already know from experience, even if it's never been written down.

Once you have that number, your alert threshold should sit below it, not at it. If 30 NTU is where your plant genuinely struggles, your alert needs to fire at whatever turbidity level gives you enough lead time to do something, whether that's switching to a secondary intake, pulling from a deeper draw, or calling in extra coagulant before the water reaches the plant.

Account for the lag between a reading and a decision

This is the part most SOPs skip. A turbidity reading from a grab sample or a fixed sonde tells you what the water looked like at that location, at that moment. By the time that number reaches someone who can act on it, and by the time the intake responds, the water that triggered the alert may already be past the structure. Storm-driven turbidity spikes on a lake can move fast, especially after a heavy inflow event stirs up bottom sediment near a tributary mouth.

Build the lag into the threshold itself rather than into the response plan. If your typical time from reading to switched intake is two hours, and turbidity on your lake has historically climbed 10 to 15 NTU in that window during runoff events, your alert level needs that much headroom under your plant's hard limit. Pull your own storm-event records if you have them. A threshold set from one bad year on a different lake isn't a threshold, it's a hand-me-down.

One number rarely covers the whole year

Spring runoff, summer algal turbidity, and fall turnover don't behave the same way, and a single year-round NTU trigger tends to be too loose in spring and too tight in a calm August. Some utilities run a seasonal threshold: a tighter number through spring melt and early summer bloom season, a looser one once the lake settles into late-summer clarity. That only works if you're watching the lake often enough to know which season you're in and whether this week looks normal for it.

That's the gap a once-a-month boat sample can't close. If you're only reading turbidity on the body of water with your own intake, and checking it every few weeks, you're setting a threshold against incomplete information about how your own lake behaves across a season, let alone the smaller lakes upstream feeding it.

A weekly read across every lake in the region, not just the one with your intake, is a different way to build that baseline. If you want to see what a numbered weekly clarity memo for your whole source area looks like, the main Water Quality Retrieval page walks through it.

Get in touch if you want to talk through what a workable alert threshold looks like for your plant's specific limits.

← Back to the blog