Predicting Drinking-Water Biofilm from Hydraulic and Operational Data: A Proof-of-Concept Model
A 2026 laboratory study predicted biofilm thickness from seven operating variables with R² 0.91. Why that is a research foundation rather than a network diagnostic, and why it says nothing about disinfectant efficacy.
Water utilities measure flow, pressure and temperature continuously but almost never measure biofilm. A 2026 study asked whether the first could predict the second, and under controlled laboratory conditions the answer was a qualified yes. This guide explains what was modelled, how well it worked, and why it is a research foundation rather than a field diagnostic. It is not evidence about any disinfectant.
The question the study asked
Biofilm on distribution-pipe walls affects water quality, disinfectant decay and hydraulic roughness, but its thickness and distribution across a network are almost unknown, because direct measurement in a live main is impractical. A 2026 paper in ACS ES&T Water (Glynis, Blokker, Kapelan and Savić; published online 6 August 2026; doi 10.1021/acsestwater.6c00651) set out to build the first model that predicts mean biofilm thickness in a drinking-water pipe from routinely measurable operational variables, so that utilities might eventually infer where biomass is accumulating from data they already collect.
The work was done at KWR Water Research Institute and Delft University of Technology with the University of Exeter. It used a purpose-built laboratory facility: a 50-metre coil of plasticised PVC pipe of 13.2 mm internal diameter, chosen because the plasticiser encourages biofilm growth, fed with local tap water that carried no disinfectant residual. Biofilm was grown from December 2024 to October 2025 under conditioning flows of 50 to 400 litres per hour and inlet temperatures between about 12 and 24 °C, and thickness was measured 219 times over the eleven months using a hydraulic residence-time method, which infers the volume the biofilm occupies from how it changes the pipe's effective cross-section.
What the model found
A random-forest model using seven variables describing hydraulic, thermal and limited chemical conditions predicted biofilm thickness with a coefficient of determination of 0.91 on data it had not seen. Flow rate, water temperature and the stability of conditions over time were the dominant predictors. Explainability analysis (feature importance and SHAP values) showed that the shear stress the biofilm had been conditioned under, and the time it had been allowed to recover since the last disturbance, shaped the result. A second, meta-analysis step characterised the biofilm's mechanical structure: a stable, shear-resistant base layer plus an outer layer that thins readily when flow increases. Operating conditions governed how much of each formed.
Two of those findings are directly useful as concepts even before any model is deployed. First, hydraulic history matters: a biofilm grown under steady low flow is structurally different from one grown under variable flow, and a sudden increase in velocity strips the outer layer while leaving the base. Second, recovery time matters: the interval since the last flushing or velocity event predicts how much removable biomass has re-accumulated. Both are consistent with the field observation that flushing produces transient improvements and with the monitoring principle that a bulk-water sample reflects recent hydraulics as much as underlying biofilm.
Why this is proof-of-concept, not a diagnostic
The authors are explicit that the model is a foundation 'subject to further validation', and the reasons are structural, not incidental. The training data came from one pipe material, one diameter, one water and one laboratory over eleven months. Real networks contain cast iron, cement-lined, PE and PVC mains of many ages and diameters, with tuberculation, sediment and varying nutrient loads, and the model has not seen any of that. The water carried no disinfectant residual; almost all UK distribution water carries chlorine or chloramine, and building water systems may carry chlorine dioxide, each of which changes both biofilm growth and the relationship between operating conditions and biomass. The thickness measurement method itself, hydraulic residence time, is a research technique that is not available in a live main. And a random-forest model interpolates well within the range of conditions it was trained on and unreliably outside it; a network with flows or temperatures beyond the laboratory range is outside it.
The practical status is therefore: a demonstration that operational data contain enough information to predict biofilm accumulation under controlled conditions, and a set of hypotheses about which variables matter most. A utility cannot download this model and rank its mains by biofilm risk. What a utility or a large building operator can do is recognise that flow history, temperature and time-since-disturbance are candidate risk indicators worth recording alongside microbiological results, so that future field validation, and their own verification data, have something to correlate against.
What it does not say about disinfectants or ChloroKlean
The study did not use a disinfectant, did not compare disinfectants and did not measure disinfection efficacy against anything. It is not evidence that chlorine, chloramine or chlorine dioxide controls biofilm, nor evidence that any of them does not, and it is not evidence about any ChloroKlean product. It is cited in this centre because it bears on verification: it shows, from a different direction than microbiological sampling, that biofilm accumulation is governed by hydraulics and temperature that a treatment programme does not control, and that a monitoring scheme should record those variables if it wants to interpret its results.
Evidence and uncertainty
Published biofilm studies are valuable for understanding mechanisms, but their conditions may not match a particular installation. Species, surfaces, deposits and operating conditions should be recorded when interpreting evidence.
For safety-critical systems, decisions should be documented through the relevant risk assessment and management plan.
Choosing the next question
A useful next step is to identify what is known, what is inferred and what needs verification. This avoids treating a general reference as a site diagnosis.
Where a product is considered, confirm the intended use, lawful GB BPR route and label directions independently of this educational guide.
A proportionate biofilm-management approach
Use this sequence to frame investigation and control; it is not a dosing protocol.
Define the system and risk
Map wetted surfaces, operating conditions, users and relevant legal or sector guidance.
Gather evidence
Review inspection, operational, residual and microbiological records rather than relying on one indicator.
Address contributing conditions
Consider cleaning, hydraulics, nutrients, stagnation and equipment condition alongside any lawfully supplied biocide programme.
Verify and review
Document the intervention and review results through the site’s written scheme or hygiene plan.
Expert Insights
"Biofilm control is a system-management question: chemistry, surfaces, flow, cleaning and verification all matter."
ChloroKlean Technical Team
Technical review team
About the Reviewer
Gavin Owen
Managing Director, ChloroKlean
Gavin Owen leads ChloroKlean's technical and commercial operations, bringing over 20 years of experience in industrial chemical distribution and water treatment. He oversees product development, regulatory compliance strategy, and the company's BPR compliance programme across PT2, PT4, PT5, and PT11 product types. Gavin works directly with water treatment professionals, facilities managers, and public health engineers across healthcare, leisure, food processing, and industrial sectors.
Frequently Asked Questions
Common questions about this topic, answered by our technical team.
Scope and safe-use note
- This is general educational information, not a dosing instruction or a product label.
- Use only a biocidal product that is authorised for its intended use or lawfully supplied under applicable GB BPR transitional arrangements, and follow its label, Safety Data Sheet and site risk assessment.
- Investigate system design, cleaning, monitoring and microbiological findings with a competent person where there is a health risk.
Published evidence about a disinfectant or another product does not establish efficacy, authorisation or an appropriate use pattern for any ChloroKlean product.
Related Resources
Continue exploring our knowledge base and product information.
Biofilm learning hub
Browse the connected biofilm guides.
Biofilm monitoring and verification
The cornerstone guide to what each method can show.
Planktonic bacteria vs biofilm
Why a water sample under-represents attached growth.
Disinfectant residuals and biofilm control
What a residual can and cannot tell you.
Legionella prevention in water systems
Where flow, temperature and stagnation sit in the UK framework.
Sources & References
This article references guidance from the following authoritative sources:
- Biofilms: survival mechanisms of clinically relevant microorganisms
Industry Standard - Frontiers in Microbiology (PMC)
- Legionella and the prevention of legionellosis
WHO - World Health Organization
- ACOP L8: Legionnaires' disease
HSE - Health and Safety Executive
- Biocidal Products Regulation
ECHA - European Chemicals Agency
- Modeling of Drinking Water Biofilm Thickness (published online 6 August 2026)
Industry Standard - ACS ES&T Water 6(9):6019-6032
External links open in a new window. ChloroKlean is not responsible for the content of external websites.