Source Water Treatability: Fixing It Before It Reaches the Plant

Stratification sets up in June. By August the hypolimnion has run out of oxygen and the sediment surface starts releasing material. Ferric iron oxides dissolve, and the phosphate they were binding goes into solution alongside Mn²⁺ and Fe²⁺, at porewater concentrations one to two orders of magnitude above the overlying water. Your intake sits somewhere in that column. Where it sits decides how much you're buying.

That's one seasonal process, and it reaches the plant as separate purchase orders. Alum against the algal carbon that phosphorus grew. Oxidant to precipitate manganese. Carbon for geosmin, toxin testing, overtime, a heavier sludge haul. Each gets approved in a different month on its own justification. Nobody adds them up. In the programs we've audited, the person managing the reservoir and the person signing chemical invoices rarely sit in the same meeting.

Treatment has a ceiling, and the reservoir sets it. A plant corrects what arrives; it can't unmake the load. Push enhanced coagulation hard against low-SUVA algal carbon and the dose climbs while TOC barely moves. Oxidize a live bloom and you rupture cells you'd have been better off settling out intact. Both are the plant meeting a limit that raw water wrote.

So here's the budget question worth asking. How much of this year's chemical spend is actually a reservoir cost? Answering it takes a number most utilities don't have: what the sediment releases, and during which weeks. Without it, the next capital cycle commits to a plant-side answer on faith.

Water's safe and your customers are angry anyway. Complaints cluster over three or four days. Everyone describes the same earthy, musty, dirty-lake taste, and every finished-water number sits comfortably inside its limit. Nothing's wrong with the water in any regulatory sense. Geosmin and 2-methylisoborneol do this. The human nose picks them up around 5 to 10 ng/L, roughly a thimble of compound in a whole reservoir, and your most sensitive customers smell it well before an average nose does. Phones start ringing while a taste panel still rates the water fine.

Two populations make it. Planktonic cyanobacteria in the photic zone track the nutrient-driven bloom dynamics you'd expect. Benthic mats behave differently. They grow on sediment and shallow littoral hard substrate in low light, and a surface grab won't see them; neither will a sonde profiling open water. We've worked clear-looking reservoirs that delivered severe events from benthic production while the open water looked fine, with the producers confirmed by geosmin-synthase qPCR on periphyton scrapes. Timing misleads too. A population can dump a pulse when it collapses, so the event follows the end of a bloom, not its peak.

Chlorine won't help. Both molecules are saturated, cage-like alcohols with no double bond to attack. Destroying them takes hydroxyl radicals, meaning ozone/H₂O₂ or UV/H₂O₂, and those radicals react just as happily with the reservoir's organic matter. The ozone route can form bromate where source bromide is present. PAC adsorbs both, and background organics compete for its pores, so real-water doses run well above what a clean-water isotherm predicts. None of it changes next week's load.

There's no MCL here, which is exactly what makes it dangerous. The bill lands in the political ledger instead: complaint volume, staff hours, customers who switch to bottled water, confidence eroded in the year you file a rate case.

Finished water picks up a brown-black tint, laundry and fixtures stain, and the calls start. Utilities manage to the Secondary Maximum Contaminant Level, 0.05 mg/L. It's aesthetic, and it isn't federally enforceable, which tells you nothing about what it costs you. Manganese drives oxidant demand, shortens filter runs, and builds deposits in the distribution system that let go later on a hydrant flush. There's also a health-based lifetime advisory at 0.3 mg/L and a spot on EPA's contaminant candidate list, so that floor may not stay purely aesthetic.

It was released from the sediment weeks earlier. Once the hypolimnion goes anoxic, microbes work down the redox ladder, and manganese oxides reduce a rung above iron. Soluble Mn²⁺ shows up in porewater and bottom water ahead of Fe²⁺ and the phosphate riding with it. That's a cheap early-warning gauge. It buys lead time on the more damaging iron-and-phosphorus release behind it.

Where the withdrawal port sits decides whether you buy the problem. Draw from or near the anoxic layer during those weeks and manganese enters the plant. Draw above it and most of it doesn't, at essentially operating cost, where the structure allows and where you know when the layer forms and how deep it sits. A monthly surface grab won't tell you that. It reads clean in mid-August while the bottom water is already loaded.

Holding the redox line with oxygenation carries its own trap. Where sulfate reduction has already stripped reactive iron to monosulfide and pyrite, that iron won't re-bind phosphate when oxygen returns, on management timescales. A system sized without checking underperforms no matter how well it's engineered. Total manganese and total iron on a lab report say nothing about that balance. Measure the sediment's sulfide-to-reactive-iron balance before the design, not after a board asks why the plant budget spiked.

Microcystin sits mostly inside the cell until lysis. Cylindrospermopsin carries a large dissolved fraction even in healthy populations, so it slips past cell-removal steps and adsorbs less readily. Anatoxin-a often comes off benthic mats on the sediment, where your surface monitoring isn't looking. Saxitoxins are potent at low concentration and analytically awkward across their congeners. Intracellular versus dissolved is the split that decides which barrier matters, and it shifts while your sample bottles are still in transit to the lab.

That's why you can't oxidize your way through a live bloom without paying for it. You convert a manageable intracellular load into a dissolved one you now have to chase. Chlorine handles dissolved microcystin at adequate CT and pH below about 8, does little for cylindrospermopsin, and lyses intact cells on the way through. Ozone is stronger on both, ruptures cells too, and its demand climbs with the dissolved organic carbon a bloom pours into the water. PAC needs dose and contact time your hydraulics may not give it.

EPA's ten-day drinking water Health Advisories put microcystins at 0.3 µg/L for children under school age and 1.6 µg/L for everyone older. Cylindrospermopsin sits at 0.7 and 3.0. They're advisory at the federal level, and some states have made their own numbers binding. Recreational guidance is a separate set of values for a separate exposure route, and mixing the two up in a public notice creates its own problem. Anatoxin-a and the saxitoxins have no final federal advisory, so state guidance fills the gap and varies by jurisdiction.

Lysing a bloom upstream of an intake is genuinely risky. Lyse a standing crop, drop that biomass to the sediment, and you drive a pulse of oxygen demand as it decays. That deepens the anoxia and accelerates the phosphorus release feeding next summer. Warning time doesn't come from one instrument. Satellite chlorophyll screens and needs local validation, ELISA is fast enough to run daily during an event, and LC-MS/MS confirms congeners and defends the number in a compliance file. The failure mode here is a do-not-drink notice, and one of those reshapes a utility's governance for years.

Jar tests call for more alum than they did five years ago. Sludge handling gets heavier, filter runs get shorter, and the plant absorbs it as the cost of doing business. Turbidity isn't driving that dose. Dissolved organic carbon is. NOM carries a negative charge, and the metal coagulant gets consumed roughly in proportion to how much shows up. That's why the Stage 1 D/DBPR frames enhanced coagulation as a TOC-removal requirement and not a turbidity one.

How removable that carbon is depends on where it came from. Watershed humics are aromatic, high molecular weight, high SUVA₂₅₄, and they respond to enhanced coagulation. Algal carbon produced inside the reservoir is hydrophilic and low-SUVA, and it won't come out no matter how hard you push the dose. So a climbing dose against climbing UV₂₅₄ tells a humics story. A climbing dose against flat or falling UV₂₅₄ points at algal carbon your coagulant is chasing and mostly missing. Conflating those two is how a plant overdoses against a fraction that was never going to floc.

Whatever survives coagulation meets chlorine downstream and forms THMs and HAAs. You're reading one variable twice. Bromide keeps that honest. Where the raw water carries it, chlorine shifts formation toward brominated species, so a bromide-loaded source can sit in DBP trouble at a TOC where a low-bromide source is comfortable.

Every extra milligram per liter of coagulant becomes metal-hydroxide and organic solids somebody has to dewater and haul. That's a truck, a contract and a tipping fee, scaled to a load you didn't choose. The algal share of that load traces back to phosphorus, much of it released from the sediment weeks before your August dose climbed. Your coagulant trend is partly a sediment measurement nobody's taken.

Start with who holds the pen. A utility can retrofit its own filters on its own schedule. Dosing a reservoir it doesn't own is a different problem entirely. That means a state permit, usually NPDES or an aquatic pesticide application permit, plus a lake association, a recreation constituency, and a flood-control operator whose priorities aren't yours. We've watched technically correct recommendations sit three years on permitting and politics alone. A control you can't permit in time won't help you this summer.

Speed is the second axis. Moving the intake port changes water quality the day you move it, where the structure allows. Oxygenation works within a season and stops working when it stops running. Phosphorus inactivation is a capital event, and its durability depends on the dose against the mobile sediment pool, on watershed control, and on lake type; deep stratified systems hold substantially longer than shallow polymictic ones. Watershed load reduction takes years and shifts the equilibrium. Fast reactive treatments buy weeks to months. That's a real answer during an event and a poor foundation for a decade.

Then there's the operator and the ten-year bill. An oxygenation system is an asset with power draw, maintenance and somebody's time. GAC is capital plus continuous media regeneration. Enhanced coagulation is a chemical line that scales with the incoming load, every year, forever. Read the cost basis, not the sticker.

Frame the whole call as reliability. Some controls lower the average load; some cap the worst week. A plant barrier you own outright is worth paying for precisely because you control it during the event, and a reservoir control that cuts the load decides how often that barrier gets tested. What you're choosing is a portfolio. Get the mix wrong and you'll fund the same fix twice, once at the reservoir and once at the plant, and still be exposed in the August that matters.