Secondary Effluent to Reuse: Why Filtration, Carbon and RO Are a Train, Not a Menu

Suggested readership: design teams planning a reuse or recycling scheme from municipal secondary effluent, and plant owners deciding how far down the train they really need to go. Reading time: ~8 min. Companion tools: Media filter · UF / MF · GAC adsorber · RO for reuse

Each unit has exactly one job

A reuse train from secondary effluent is four units doing four different jobs. The recurring design error is treating them as interchangeable "polishing steps" and sizing each with a rule of thumb from a different decade:

Skim any stage and the failure appears downstream, not at that stage. That is the design logic of the whole train, so it is worth following one flow through all four.

Stage 1 — media filter: insurance for everything downstream of it

On 10,000 m³/d at 10 m/h, a rapid-gravity sand/anthracite filter needs 41.7 m² — four cells of 10.4 m² (4.2 × 2.5 m). Clean-bed headloss is a modest 0.38 m through 1.0 m of media; the filter is washed on terminal headloss near 1.9 m, and the backwash (45 m/h, 469 m³/h per cell, 62.5 m³ per event) is the real design driver because it sets the cell count: one cell out of service washing while the rest carry the flow.

A media filter ahead of membranes is not about polishing the effluent — it is about protecting the SDI. Membrane makers quote feed limits around SDI < 3 (and turbidity < 5 NTU for UF); secondary effluent fresh from a good clarifier might sit at SDI 4–6. One deep-bed filter closes that gap and multiplies membrane cleaning intervals by an order of magnitude. Skimp here and the UF below becomes a very expensive particle filter.

Stage 2 — UF/MF: the RO's bodyguard

UF at 92 % recovery turns 10,000 m³/d of feed into 9,200 m³/d of permeate and 800 m³/d of reject, needing 7,667 m² of membrane — 192 modules at 40 m² each, at a conservative 1,200 L/m²·d (~50 LMH). The flux number is deliberately conservative: the modules are not there to make water at maximum rate but to hand the RO a feed with SDI < 3, turbidity < 1, and no bacteria.

The question at this point in the train is whether UF earns its place over a second media stage. The honest answer: UF earns it when the RO is large and expensive enough that irreversible fouling is the dominant risk. Below roughly 2,000–5,000 m³/d of RO feed, a well-run media filter plus cartridge guards is defensible; above it, the UF's absolute particle barrier pays for itself in RO element life.

Stage 3 — GAC: where the organics leave

The carbon adsorber is sized on empty-bed contact time, and its geometry is surprisingly rigid. At 30 min EBCT and 10 m/h, two vessels in series (15 min each) hold 104.2 m³ of carbon — 46.9 t — in 5.2 m diameter vessels, with a clean-bed headloss of 1.8 m across the pair.

Carbon's economics are in the usage rate, which is a direct function of how much COD you ask it to remove:

• At 30 mg/L COD removed on our example (150 kg/d), capacity of 0.2 kg COD/kg carbon gives 750 kg/d of carbon use — 150 g/m³, mid-range of the 50–250 g/m³ advisory band.
• That means the lead vessel is exhausted in ~31 days and the plant sends 274 t/yr of spent carbon to regeneration or disposal.

Two design truths hide in those numbers. First, a GAC is a consumable reactor, not a tank: the operating cost line is 274 t/yr of carbon, and the correct question at design stage is whether you can afford the CAPEX of regeneration or must price virgin replacement. Second, GAC quietly becomes biological activated carbon once colonised — the bacteria on the carbon degrade what adsorption misses — which is free capacity but also means the bed needs the same dissolved-oxygen and temperature attention as the biology upstream.

Stage 4 — RO: the arithmetic of concentration

RO at 75 % recovery turns 5,000 m³/d of feed into 3,750 m³/d of permeate and 1,250 m³/d of concentrate. The mass balance that governs everything:

• Permeate TDS 35 mg/L (from 800 feed); concentrate TDS 3,200 mg/L.
• Concentration factor at the membrane wall is 4.0× (1/(1−R)), but the mean factor along the vessel is only 1.85× — and it is the mean that sets permeate quality. Using the wall factor for permeate TDS is the classic over-estimate.
• At 28 °C, salt passage is 2.38 %; the RO needs 8,681 m² — 240 elements in 40 vessels and lands on 17.6 LMH actual flux.
• Feed pressure is only 8.6 bar (net driving pressure 6.4 bar) — and specific energy 0.42 kWh/m³. Reuse RO on secondary effluent is a low-pressure, low-energy animal compared with seawater.

Then the two scaling checks that decide whether the last element lasts:

LSI. Calcium and alkalinity concentrate 4× in the brine (160 mg/L Ca, 480 mg/L alkalinity), and the concentrate's Langelier index lands at +0.9 — scaling territory. The design answer is antiscalant, lower recovery, or acid. Silica concentrates to 60 mg/L against a ~120 mg/L solubility limit at the brine pH — under the line, but only just; silica does not respond to antiscalant the way carbonate does, and it is the number that usually caps recovery in the long run.

Boron and ammonia. RO rejects divalents at 99 %+, but boron only 40–70 % at neutral pH (95 %+ at pH 10) and ammonia passes almost freely. If the reuse standard includes those, the RO cannot do it alone — the train needs pH adjustment for boron and nitrification (not RO) for ammonia. This is where "train, not menu" stops being a slogan: the target standard decides which stages exist and in what order.

Sizing from the standard backwards

The professional way to design a reuse train is from the reuse standard backwards: pick the final barrier the standard demands (RO for salts, UV/AOP for trace organics, chlorination for residual), then add only the stages needed to keep that barrier alive (media → UF to protect RO, GAC when organics would foul or form disinfection by-products). Every stage you skip saves money; every stage you skip that the barrier needed costs you the barrier.