Geosmin and MIB: what triggers taste-and-odor complaints at the plant
Geosmin and MIB are produced by cyanobacteria and some actinomycete bacteria, they taste like dirt and mildew, and the human nose can pick them up at absurdly low concentrations, often in the single-digit nanograms-per-liter range, well below anything that shows up as a visible problem in raw water turbidity. Every operator picks that up in year one. What gets missed is timing, and timing is what actually determines whether your phone rings on a Tuesday afternoon.
The compound isn't released when you think it is
Geosmin and MIB are intracellular. The cyanobacteria that make them, mostly species in genera like Anabaena, Oscillatoria, and Phormidium for geosmin, and various cyanobacteria and actinomycetes for MIB, hold the compound inside the cell while it's alive and dividing. A healthy, actively growing bloom can sit in your source lake for days without pushing much into the water column at all. The release happens on cell lysis: senescence, grazing, or a sudden stressor that ruptures cells and dumps the intracellular pool into the water.
That stressor is often something the plant did, or something the lake did on its own right around the time everyone assumed the problem was already past. A copper sulfate or algaecide application that kills a bloom fast will often produce a worse taste-and-odor event than if the bloom had been left to senesce on its own, because you've lysed a large standing population of cells all at once. The same thing happens naturally during fall turnover, when a stratified lake mixes and a bloom that's been sitting comfortably in the epilimnion gets stressed by temperature and light changes it wasn't built for.
So the pattern utilities see again and again: raw water clarity looks fine, chlorophyll readings from two weeks back showed a bloom building, and then a cold front or a treatment decision triggers die-off, and geosmin shows up in finished water a few days later when nobody in the plant was watching for it because the bloom itself seemed like old news.
What correlates with a complaint spike
Three conditions show up together often enough in the literature and in utility monitoring logs that they're worth tracking as a set rather than one at a time: a cyanobacteria-dominant bloom (not just any algae, chlorophyll-a alone won't distinguish bloom-forming cyanobacteria from diatoms or green algae), a period of thermal destabilization (turnover, a strong storm mixing event, or a sharp cold snap), and nutrient loading heavy enough that the bloom was dense rather than sparse. Sparse blooms lyse too, but a sparse population releasing its intracellular geosmin pool doesn't push finished-water concentrations near the complaint threshold the way a dense late-summer bloom does.
The practical problem for most source-water programs is that none of this is visible from a single sample point. A grab sample at the raw water intake tells you what's happening at the intake, not what's building across the lake, let alone across the other lakes in the service area feeding different plants. By the time a bloom is visible from the shoreline near the intake, it's frequently already past the growth phase and heading toward the senescence that releases the compound. Catching the buildup, and the drop-off that follows, means watching chlorophyll-a and turbidity trends across the whole water body on a cadence tight enough to see the turn, not just the peak.
That's the gap a weekly clarity read across every lake in a region is built to close: instead of one intake sample, you get a numbered trend memo per water body, so a chlorophyll spike followed by a sharp drop (the signature of a die-off event) shows up on the record before the taste-and-odor calls start rather than after. You can check what that looks like for your region on the Water Quality Retrieval site.
If your source-water program is still reacting to taste-and-odor complaints instead of seeing the die-off coming, it might be worth a look at what a standing weekly read on every lake you draw from would catch.