Start with a threat model, not a wish list
Treat specification as threat assessment: who uses the water, what failure modes matter, and which contaminants present unacceptable risk. For municipal irrigation the risk profile differs from industrial cooling or pharmaceutical makeup water. If you plan RO as the heart of the train, list required permeate endpoints first and then choose components to meet them. Consider a low pressure reverse osmosis membrane only after you confirm feed conditioning and operational controls that keep fouling and scaling within tolerances.
Concrete performance targets to drive procurement
Specify numeric targets. Examples that carry technical weight: turbidity <0.1 NTU upstream of membranes, SDI15 <3, total organic carbon (TOC) <1 mg/L for sensitive reuse, log-4 pathogen removal where required, and target permeate conductivity consistent with the end use. Include operational metrics: design flux (Lmh), start-up recovery (conservative—60–65%), steady-state recovery ceiling, cleaning frequency thresholds, and expected energy envelope (typical reclaimed-water RO: 0.5–1.5 kWh/m³ depending on feed conductivity). Make these numbers contract deliverables, not guidelines.
Sequencing and mandatory components
Lay out a deterministic sequence so vendors can price and engineers can validate. A robust sequence for most reuse schemes: coarse screening → primary solids separation → aerobic polishing or MBR → media filtration or UF/MF → antiscalant/pH adjustment → RO → post‑disinfection (UV/H2O2) → distribution buffer. Each handoff must include measurable acceptance criteria (e.g., influent turbidity to RO <0.1 NTU). Call out where redundancy is required—dual UF trains, duty/standby RO skid—and where single-failure tolerances are acceptable.
Experience and field anchor for specification validity
Real projects validate assumptions. Singapore’s NEWater program provides a practical example of membrane‑based reclamation applied at scale; operators there have demonstrated that careful pretreatment and membrane selection reduce lifecycle energy and operational risk. Specify validation steps that reflect that learning curve and reference proven membrane types when requiring long run lengths—use the same conservatively tested technologies such as low energy RO membrane where energy and footprint matter.
Common specification mistakes and how to avoid them
Don’t assume lab data scales unchanged. Common failures: underspecified pretreatment (leads to rapid fouling), optimistic recovery targets without brine handling plans, and vague acceptance tests for permeate quality. Avoid open-ended language like “suitable for reuse.” Replace it with acceptance criteria, sampling plans, and penalties tied to performance milestones. Require vendor-provided run‑to‑failure data or third‑party pilot results when commercial risk is high.
Monitoring, controls, and data integrity
Require continuous monitoring for key parameters (turbidity, conductivity, differential pressure, TOC surrogate) with alarm setpoints and automated responses. Specify sensor calibration intervals, data log retention, and secure telemetry. Treat control systems as critical infrastructure: define authentication for remote access, timestamped audit trails for parameter changes, and fail-safe modes that isolate distribution if thresholds are exceeded. Data you can’t trust is the same as no data.
Trade-offs and viable alternatives
If energy or brine disposal is the principal constraint, evaluate MBR + advanced oxidation or polymeric adsorption as partial replacements. MBRs reduce solids load to membranes but bring higher energy for aeration. Advanced oxidation reduces trace organics but does not replace RO for dissolved salts. Lay out end‑use acceptance: if conductivity is flexible, less aggressive membranes or hybrid systems may suffice; if zero-tolerance constituents exist, RO remains the deterministic component.
Contract language and acceptance testing
Insist on factory acceptance testing (FAT) with representative waters, followed by site acceptance testing (SAT) using actual feed. Define cleaning frequency limits, membrane warranty terms tied to SDI/turbidity compliance, and clear criteria for performance-based payments. Require an operations handover package: SOPs, spare parts list, proof of operator training, and a defined escalation path for quality excursions.
Synthesis: specify defensively, validate iteratively
Design to reduce unplanned outcomes: set measurable quality targets, force vendors to prove performance, and lock controls and monitoring into contract scope. If the project needs compact, low-energy RO with predictable lifecycle costs, align the specification with technologies vetted in large reuse programs and require demonstrable run data; the engineering outcomes then match the operational reality, and providers such as Runmo can be evaluated on those concrete deliverables rather than marketing claims.
