Filter Runs Collapse After a Storm Event
A 4.0 MGD conventional plant treats a river supply. Raw water turbidity has climbed from 4 to 9 NTU with the first autumn storms; raw pH is 7.9 and total alkalinity is 140 mg/L as CaCO3. The plant feeds 22 mg/L of alum at a rapid-mix G of 700/s and flocculates in three tapered basins. Settled turbidity, normally 1.5 NTU, now reads 3.4 NTU. Filtered water is still 0.19 NTU, but runs on all six filters have fallen from 60 hours to 26 hours and backwash water use has doubled. A jar test at the current 22 mg/L alum gives settled turbidity of 3.4 NTU at pH 7.9, 1.1 NTU at pH 7.0, and 0.8 NTU at pH 6.5. The technician notes the jars were pH-adjusted with sulfuric acid.
1. Calculate the current alum feed in lb/day and the feed at 30 mg/L. Using the rule that 1 mg/L of alum consumes about 0.5 mg/L of alkalinity as CaCO3, determine whether alum addition alone can reach the pH 6.5 optimum, and explain how that changes your reading of the jar test.
Read the worked answer
Feed calculations (lb/day = mg/L x MGD x 8.34):
- Current: 22 x 4.0 x 8.34 = 734 lb/day of alum
- At 30 mg/L: 30 x 4.0 x 8.34 = 1,001 lb/day
Can alum alone reach pH 6.5?
- Alkalinity consumed at 30 mg/L: 30 x 0.5 = 15 mg/L as CaCO3
- Residual alkalinity: 140 - 15 = 125 mg/L as CaCO3
That is a heavily buffered water. The CO2 produced by alum hydrolysis will barely move the pH; from 7.9 you would expect a drop of a few tenths at most, not 1.4 units. So the pH values in the jar test were produced by the sulfuric acid, not by the coagulant, and the test tells you the optimum is pH-dependent but not that your existing chemical program can get there.
What that means:
- The 0.8 NTU result is real, but achieving it requires a deliberate pH-adjustment strategy (acid or CO2 feed with alkalinity and LSI monitoring downstream), not simply more alum.
- Ask the lab to run a confirmatory series without acid to see where the plant actually operates, and a series with acid to size the acid dose.
- Check that acid addition will not drive the distribution system water corrosive; 125 mg/L residual alkalinity gives room, but the LSI and lead/copper considerations still apply.
- An alternative worth pricing is a different coagulant or a coagulant aid that performs better at the existing pH.
2. Explain mechanistically why filter runs collapsed even though settled turbidity only rose from 1.5 to 3.4 NTU and filtered turbidity is still below 0.3 NTU. What two data sets would you pull first, and what would each tell you?
Read the worked answer
Mechanism: the problem is what kind of solids are reaching the filters, not their mass. Underdosed or pH-mismatched coagulation produces small, weak, slowly settling floc (pin floc) that carries through sedimentation. Those particles are captured almost entirely in the top inch of the bed, so headloss climbs steeply while effluent turbidity stays acceptable. Classic surface blinding: the filter is doing its job, but it is doing it in the first inch of media instead of in the bed.
Rising settled turbidity from 1.5 to 3.4 NTU plus a doubled backwash volume is consistent with the same cause.
Pull first:
- Particle counts and size distribution on settled water. Turbidity is a poor proxy for particle number when particles are small. A high count in the 2-10 micron range confirms pin floc carryover rather than a genuine increase in settleable solids.
- Headloss versus time curves for the last several runs on one filter. If the curve is steep and concave from the start, the solids are being captured at the surface. If the curve is normal but the run ends on effluent turbidity, the failure is penetration, not blinding, and the diagnosis is completely different (media, ripening, backwash).
Also worth a look: whether backwash is actually removing what is being captured, since doubling backwash water against mudballs is a sign the bed is agglomerating.
3. Recommend a 48-hour course of action. For each option, name the operational risk you would have to manage.
Read the worked answer
Immediate (0-8 hours):
- Confirm the settled-water particle count and headloss pattern above; verify the jar test was representative (same raw water, same temperature, same mixing energy).
- Shorten filter runs voluntarily rather than running to 26 hours of loading, and if any single filter exceeds 1.0 NTU, take it out of service. Combined filter effluent must stay under the 0.3 NTU 95th-percentile limit and your state's single-filter limit; 0.19 NTU leaves little margin if one filter breaks through.
Options and risks:
| Option | Risk to manage |
|---|---|
| Raise alum to 30-35 mg/L to improve sweep and enmeshment | Extra sludge load and possible pH depression; check that alkalinity and finished pH stay in the corrosion-control target |
| Add acid or CO2 to move pH toward 6.5-7.0 | Overfeed drives low pH, high finished-water corrosivity, and can violate the Lead and Copper Rule; requires tight pH monitoring and a feed control interlock |
| Feed a low-dose polymer as filter aid | Overshearing at rapid mix destroys the polymer; carryover can blind filters faster and may add to finished-water residuals |
| Reduce filtration rate by bringing all filters online / equalizing flow | Spikes loading on the remaining filters if done poorly; coordinate with distribution to avoid a pressure upset |
Do not simply increase backwash frequency and call it solved; doubled backwash water at 8 percent of production is a symptom, and more washing of a blinded bed wastes water without fixing the capture mechanism.