Impact statement
In 2020, the Dutch drinking water sector faced urgent challenges due to upcoming European Union (EU) regulations lowering the permissible lead levels in drinking water. Key questions emerged: Do brass components like water meters and shutoff valves pose health risks? Should they be replaced? What materials are safe for future use? And how can utilities distinguish their contribution to lead exposure from that of private plumbing systems? The prospect of a nationwide replacement program was daunting – both logistically and financially. Complicating matters further, premise plumbing lie outside utility control. To tackle these dilemmas, the sector launched a coordinated effort combining modeling, lab tests and field studies. This proactive approach delivered clear guidance on managing existing and future assets, especially brass components, and established protocols for responding to lead exceedances. These rapid, sector-wide results were made possible by the strong collaboration and knowledge-sharing across Dutch utilities. The insights gained now offer a valuable blueprint for other European countries preparing for the full enforcement of the revised Drinking Water Directive by 2036.
Introduction
The revised European Union-drinking water directive (EU-DWD) has led to a stricter norm for lead (Pb) in drinking water. In view of long-term public health, the norm has been reduced from 10 μg/L to 5 μg/L, see Directive 2020/2184, Annex I, Part B. As a supporting measure, newly installed components that come in contact with drinking water (such as pipes and fittings) must belong to the European positive list (EUPL) of materials (Commission Implementing Decision, 2024/3672024). The EU member-states have until 2036 to enforce the new law, offering room for transition. In the Netherlands (NL), the norm has already been enforced since 2023 (see Drinkwaterbesluit, 2024, Bijlage A, Tabel II).
The 10 Dutch drinking water utilities are responsible for ensuring the compliance with the lead norm. The utilities are responsible for all materials in the drinking water supply chain until (and including) the point of delivery (water meter). Thus, the drinking water utilities are owners of articles such as shutoff valves, water meters and lead-soldered copper service lines – all of which may leach lead. Articles downstream of the water meter in the premise plumbing system are the responsibility of the premises owner. The division of responsibilities between the drinking water utility and premises owner is complicated by the fact that the utility is responsible for the satisfaction of water quality norms at the point of delivery (water meter) as well as at tap (see Drinkwaterbesluit, 2024, Artikel 13). This is typically tested through samples collected at the kitchen tap.
In the NL, lead is typically introduced during distribution, as illustrated in Figure 1(a), based on measurements between 2012 and 2022 (REWAB, 2025). It shows how lead concentrations in the water leaving the treatment facility rarely show norm exceedance (0.2% above 5 μg/L) in contrast to the water sampled at premises (12% above 5 μg/L). There could be many possible sources – service lines, appurtenances and plumbing articles. As of 2021, there were 1,700 lead service lines in the NL and a further 23,000 service lines where further assessment is necessary to exclude lead as the pipe material (van Steen et al., Reference van Steen, Vertommen and Slaats2021). These are a considerable fraction of the about 7 million service lines present in the NL. There are hundreds of thousands of premises where the supplied water comes in contact with either lead-soldered copper service lines (lead in solder was only phased out around 1995 in the NL) and/or lead-containing brass articles. Extrapolating the results of a past study, it may be estimated that hundreds of thousands of premises might still contain lead plumbing downstream of the water meter (see Baron et al., Reference Baron, Leroy, Wagner, van den Hoven, Brink, Miller, Crosbie and Jackson1995, Table 11.5). Moreover, it cannot be ruled out that premises owners might unknowingly introduce lead leaching articles in the plumbing. For mainland Europe, it is acknowledged that legacy lead plumbing is present in many countries (see Jarvis and Fawell, Reference Jarvis and Fawell2021, Figure 1).
(a) Lead in regulatory samples in the NL above the detection limit at treatment facilities (left) and consumers’ taps (right). (b) Simulations show how lead concentrations at the tap fluctuate wildly across time.
Figure 1. Long description
Panel a is a schematic of a water distribution system. On the left, a water source leads through a treatment facility. A pie chart for the treatment facility shows 97 percent (1818 samples) are below the detection limit, 3 percent (64 samples) are above the limit but under 5 micrograms per liter, and 0 percent (3 samples) are at or above 5 micrograms per liter. On the right, the system reaches premise plumbing. A second pie chart shows 15 percent (298 samples) below the detection limit, 73 percent (1402 samples) above the limit but under 5 micrograms per liter, and 12 percent (225 samples) at or above 5 micrograms per liter.
Panel b contains two vertically stacked graphs sharing an x-axis representing time in days from 0 to 7.
1. The top graph shows water demand in liters per second (l p s). It features vertical blue spikes indicating when the tap is open, reaching up to 0.15 l p s, and flat sections where the tap is shut.
2. The bottom graph shows lead in water in micrograms per liter on a logarithmic y-axis from 10 super minus 2 to 10 super 2. A dashed horizontal line marks the norm at 0.5 micrograms per liter. Red data points fluctuate wildly between 0.02 and 50 micrograms per liter. Arrows indicate peaks labeled Higher than norm, points on the line labeled Norm, and troughs labeled Lower than norm. The data shows that lead concentration varies significantly each time the tap is opened.
To the best of our knowledge, there are only a few recent European studies (Döhla et al., Reference Döhla, Jaensch, Döhla, Voigt, Exner and Färber2021; Gartner et al., Reference Gartner, Leban and Kosec2024) where the revisions in the EU-DWD are explicitly reflected upon. This suggests that European nations are yet to take concrete actions to safeguard the stricter norm. In contrast, there exist numerous scientific studies emanating from the United States focusing on lead, as a consequence of the Lead and Copper Rule and subsequent amendments (Environmental Protection Agency, 2025). A major difference between the United States and the NL is how lead service lines are viewed. In the United States, conditioning the water (with orthophosphates, for example) to reduce leaching of lead from lead service lines (to promote scale growth) is viewed as an acceptable mitigation strategy, especially in the presence of chlorine (Edwards and McNeill, Reference Edwards and McNeill2002; Schock et al., Reference Schock, Cantor, Triantafyllidou, Desantis and Scheckel2014; DeSantis et al., Reference DeSantis, Schock, Tully and Bennett-Stamper2020). Thus, a lot of attention has been paid in the United States on understanding the fundamental aspects of lead leaching from lead service lines. Unfortunately, this extensive body of lead research in the United States is not directly relevant for the non-chlorinated water supply in the NL. In the NL, the preferred mitigation strategy was and is replacement, underlined by large-scale replacement campaigns in the past.
Given the situation in the NL and the state of current affairs brought upon by the EU-DWD, the following practical questions exist for the Dutch drinking water utilities:
-
1. What policy should a drinking water utility adopt with regard to existing and future assets that leach lead (with a special focus on brass)?
-
2. What should the drinking water utility do when an incidental norm exceedance is encountered?
While field measurements reflect reality, they also pose a challenge while interpreting the findings. In the NL, the standard way to test for lead in drinking water is by collecting a 1-L sample at the kitchen tap (without any pre-flushing) at a random day and time – the random daytime (RDT) sampling. This methodology is appropriate for assessing lead exposure at a community level (see Triantafyllidou et al., Reference Triantafyllidou, Burkhardt, Tully, Cahalan, DeSantis, Lytle and Schock2021, Table 3). However, it is inappropriate for assessing the situation at a household level. This is illustrated in Figure 1(b). One sees that the lead concentration at a tap fluctuates by orders of magnitude owing to the stochastic nature of water usage. The graph at the top shows moments at which a tap has been opened and the graph at the bottom shows the lead concentration in the water out of the tap when open. The lead concentration at a random moment can be much higher or lower than the average concentration over a longer period. In fact, Jarvis et al. (Reference Jarvis, Quy, Macadam, Edwards and Smith2018) provide an empirical example where the consumption pattern seemingly affects lead exposure among inhabitants in the same household. The impact of water use patterns is considered to be one of the most understudied contributors to lead variability (Triantafyllidou et al., Reference Triantafyllidou, Burkhardt, Tully, Cahalan, DeSantis, Lytle and Schock2021). Thus, the moment at which a random sample is collected can have a huge impact on the conclusion drawn about exposure or presence of lead.
As a result, the following research questions were formulated to address the aforementioned practical questions:
-
1. Modeling: Can a computational model that integrates plumbing system geometry, stochastic water demand, and time-dependent lead leaching accurately capture their complex interplay?
-
2. Validation: Can the validity of a complex lead-leaching model be established through empirical calibration and one-to-one experiments?
-
3. Application: Can a validated model of lead leaching in plumbing systems be effectively deployed to generate insights applicable to real-world scenarios, such as exposure risk assessment and sampling strategies?
This paper demonstrates how a composite framework, including a numerical model, laboratory experiments and field measurements, could help water utilities make informed decisions on the risk of lead in drinking water. The research presented here has been documented extensively in a Dutch report (Dash et al., Reference Dash, Galama-Tirtamarina, Dignum, ten Kate and van der Leer2024).
Methods
Mechanistic model LEadGO
This study relies largely on a mechanistic model schematically illustrated in Figure 2(a). The modeling framework is called LEadGO which was developed during the course of this study (Dash et al., Reference Dash, Hillebrand, Mitrovic and Blokker2025). It comprises of three key ingredients:
-
1. The plumbing system (lengths/diameters of pipes and location of various consumption points);
-
2. The water consumption time series (dependent on the number of inhabitants and their behavior); and
-
3. The location and characteristics of lead leaching (lead dissolution curve, as a first-order saturation growth model).
(a) The key ingredients of the modeling framework. (b) A brass sample in the experimental rig for brass. (c) Framework to validate the modeling approach with experiments.
Figure 2. Long description
The graphic is divided into three sections labeled a, b, and c.
Panel a: A conceptual diagram of the S I M D E U M modeling framework. At the top left, inputs include Number of people and Attitude of people. These feed into the S I M D E U M model to create demand patterns at each tap for one week. To the right, a plumbing schematic labeled 1 shows a network of blue and red pipes connected to icons for a washing machine, toilets, showers, and sinks. A green box notes Lengths/diameters of pipes and Locations of taps. Below the schematic, a bar graph labeled 2 shows Demand in litres/s over 7 days. Next to it, a line graph labeled 3 shows Dissolved lead concentration as an asymptotic curve against Stagnation period, reaching an Equilibrium lead concentration. An arrow points from this model to a specific tap in the plumbing schematic.
Panel b: A horizontal photo of a brass pipe sample integrated into an experimental rig. The pipe section is labeled C W 6 0 2 and is connected to various valves, pressure gauges, and silver corrugated hoses on either side.
Panel c: A validation flowchart. On the left, photos of pipe segments lead to a box for Lead dissolution parameters estimation. This feeds into a graph of dissolved lead concentration and a schematic of a pipe network labeled Modelling with estimated parameters. On the right, a photo of a complex metal cage rig is labeled HomeWaterLab. Arrows from both the model and the lab measurements point to a final box: Compare results and determine appropriateness of modelling framework.
The lead concentration at the tap depends on the interplay between the three ingredients. The package WNTR (Klise et al., Reference Klise, Hart, Moriarty, Bynum, Murray, Burkhardt and Haxton2017) forms the backbone for all the hydraulic and water quality calculations. Mathematically, the lead dissolution curve is described using the equation
$ C=E\left(1-{e}^{-t/T}\right) $
, where
$ C $
is the lead concentration [μg/L],
$ t $
is time [s],
$ E $
is the equilibrium lead concentration [μg/L] and
$ T $
is the characteristic dissolution timescale [s]. The timescale is dependent on other parameters such as the dissolution rate
$ M $
[μg/(m
$ {}^2\cdot $
s)] and pipe diameter,
$ D $
[m] via the mathematical relationship:
$ T=\frac{DE}{4M} $
.
Similar models have been used in the past to study the dissolution and propagation of lead in drinking water (Abokifa and Biswas, Reference Abokifa and Biswas2017; Burkhardt et al., Reference Burkhardt, Woo, Mason, Shang, Triantafyllidou, Schock, Lytle and Murray2020; Hatam et al., Reference Hatam, Blokker, Doré and Prévost2023; Hatam et al., Reference Hatam, Blokker and Prevost2025). Such a model tackles one of the most understudied contributors to lead variability – the impact of water use patterns (Triantafyllidou et al., Reference Triantafyllidou, Burkhardt, Tully, Cahalan, DeSantis, Lytle and Schock2021).
In the present study, a typical Dutch single-family residence is considered as the basis for the plumbing geometry. The water demand patterns are based on average water consumption behavior for a Dutch household with two adults – generated using SIMDEUM (EJM Blokker et al., Reference Blokker, Vreeburg and van Dijk2010). The lead leaching characteristics are based on actual laboratory experiments (performed in the present study) on components found in Dutch plumbing. It is assumed that the incoming water from the distribution network is free of lead, which is reasonable for the Dutch drinking water network (Figure 1[a]).
Experiments in experimental rigs
The mechanistic model is supported by the following experiments:
-
1. Tabletop stagnation experiments: Lead-releasing components are filled with water and sealed for a fixed duration. By analyzing the water following a series of predetermined stagnation periods, the lead released as a function of stagnation duration is derived (Kuch and Wagner, Reference Kuch and Wagner1983). The lead dissolution curve is used to extract parameters for the first-order saturation growth model. These experiments provide input values for the mechanistic model.
-
2. Experimental rigs for brass: Rigs at two locations (‘Weesperkarspel’ and ‘Leiduin’ – water treatment facilities of the utility Waternet), contain various brass components (two each of CW614, CW617, CW602, CW724R and one piece of PVC as a reference for correction), which are subject to water flow of 125 L per day (programmed flushing of rig multiple times a day for 1–2 min). An example of a brass sample is shown in Figure 2(b). Two types of experiments were performed. ‘Run-time’ tests involved collecting water samples following 16 h of stagnation, with sampling being performed every 4 weeks. These ‘run-time’ tests allow determining whether the lead leaching tendencies change following extended operation. A second test involved collecting water samples following 0.5, 1, 2, 4, 8, 16 h of stagnation. These were performed after conditioning in the rig for 3 months. The outcomes aid extracting the lead dissolution curves. These experiments provide input values for the application of the mechanistic model.
-
3. Premise plumbing rig: This rig called the HomeWaterLab (Dash et al., Reference Dash, Summeren and Blokker2023) mimics a Dutch premise plumbing system with 11 points where water can be used. This facility is similar to the Home Plumbing Simulator in the United States (see Lytle et al., Reference Lytle, Formal, Cahalan, Muhlen and Triantafyllidou2021, Figure 1). Water consumption in HomeWaterLab can be automatically controlled through inputs in an Excel file containing SIMDEUM-generated water demand patterns. These experiments are used for validating the mechanistic model.
The construction and operation of the experimental rigs for brass were inspired by the European norm for assessing metal release (Nederlands Normalisatie Instituut, 2008, 2010). Deviations from the norm (in design and operation) exist since the rigs were used to estimate the urgency to replace existing assets and not to assess the suitability of a material for market introduction (material certification). This research activity was a proactive step taken by the utility (in 2023) before the EUPL was published (in 2024). It helps assess existing assets (lead dissolution curves translated to exposure at the tap) and future assets (anticipating which types of brass may be purchased/installed).
There is one major difference between the tabletop experiments and the experimental rigs for brass with respect to the leaching mechanisms. In the experimental rigs, the brass articles are installed ‘in-line’ with a stainless steel pipe. Lead leached into stagnant water from brass can diffuse longitudinally, allowing for more lead to be leached into the system. Such a mechanism is not possible when an article is filled with water and sealed off (like in the tabletop experiments). This behavior has been emphasized previously by Triantafyllidou et al. (Reference Triantafyllidou, Raetz, Parks and Edwards2012), implying that the ‘in-line’ measurements significantly increase lead concentrations and are more representative of reality. However, from the perspective of modeling where the diffusion is not accounted for, the tabletop experiments offer a more representative input.
In the Supplementary Material, the water compositions at Weesperkarspel and Leiduin are available (Supplementary Tables S1–S2); so are the outcomes of the ‘run-time’ tests (Supplementary Figures S1–S4) and lead dissolution curves from the rig experiments (Supplementary Figures S5–S8) and tabletop experiments (Supplementary Figures S9–S10).
Validation of the mechanistic model LEadGO
These experiments collectively were used to validate the mechanistic model. The validation framework is schematically illustrated in Figure 2(c) involving the following steps.
-
1. Lead dissolution curves for various specimens were gathered. Tabletop stagnation experiments were performed on a lead pipe, a copper pipe with lead soldering and a water meter containing brass. The experimental rigs for brass offered information on leaching of various types of brass.
-
2. The lead pipe from the tabletop experiments was integrated into the HomeWaterLab facility. HomeWaterLab was operated for 5 weeks with unique weekly water demand patterns. A total of 25 composite water samples were collected across 5 weeks and five end use points.
-
3. A computer model of HomeWaterLab was made in LEadGO. The location of the lead pipe in the model corresponded with the location in HomeWaterLab. The lead leaching characteristics in the model were based on tabletop stagnation experiments. Five weeks of simulations were performed with the same water demand patterns used for the experiments, to facilitate a one-to-one comparison (25 composite water samples in simulations).
Applying LEadGO: Lead concentrations at the tap
LEadGO was applied to various informed scenarios. Only six examples are considered for brevity. The six examples have three combinations of leaching locations (shutoff valve, water meter and kitchen tap) and two brass types (CW602 and CW617). CW602 contains 1.7–2.8% lead by weight whereas CW617 contains 1.6–2.2% lead by weight. The equilibrium lead concentrations, E, for CW602 and CW617 were measured and found to be 223.7 μg/L and 19.3 μg/L respectively. The leaching rates, M, were measured and found to be 0.014 μg/(m2
$ \cdot $
s) for both. The CW602 parameters were based on the experimental rigs for brass whereas the CW617 parameters were based on the tabletop experiments. Simulations were run for 20 weeks, each week with a different water demand pattern.
Using profile sampling to localize lead leaching components
One powerful method to localize lead leaching components is through profile sampling (also known as sequential sampling). Triantafyllidou et al. (Reference Triantafyllidou, Burkhardt, Tully, Cahalan, DeSantis, Lytle and Schock2021) provide a good overview of this methodology. This sampling technique involves three key steps: (i) flushing the system long enough to rid it of existing leached lead, (ii) facilitating stagnation to allow a disturbance-free leaching and (iii) collecting multiple samples in a sequence.
In our study, we reflect on the robustness of this sampling method through several routes:
-
1. Numerical modeling: It has been previously discussed in Dash et al. (Reference Dash, Steen and Blokker2022). Lead is assumed to leach from a lead service line. To simulate profile sampling, a stagnation period of 6 h as well as sample collection following the stagnation (20 samples of 300 mL) were built into the existing water demand patterns. The stagnation was not preceded with flushing to rid the plumbing of existing leached lead. The resulting lead concentrations in the 20 virtual samples were studied on 140 simulated days.
-
2. Experiments in HomeWaterLab: The lead pipe tested in the tabletop experiments was installed in HomeWaterLab. The system is flushed, and a stagnation period of 6 h follows. Thereafter, 20 samples of 180 mL were collected successively at one tap.
-
3. Field measurements: Profile sampling was tried out at one residence in two rounds. In the first round of sampling, five samples of 1,000 mL were collected. In the second round, 2 samples of 1,000 mL were followed by 10 samples of 100 ml (decomposition of the third liter).
Results
Long term behavior of lead leaching from brass
The long-term leaching behavior of four types of brass were investigated. Only the results of CW617 and CW724R are considered for brevity – see Figure 3. Lead concentrations leached from CW617, when it is brand new, are extremely high (>100 μg/L). After a couple of months of conditioning, the concentrations dropped to about 10 μg/L. Even after about 8–10 months of conditioning, the leached concentrations from CW617 did not fall below this plateau of about 10 μg/L. A qualitatively similar observation had been made in the past (see Latva et al., Reference Latva, Kaunisto and Pelto-Huikko2017, Figure 4). For CW724R, the lead leached from a brand new specimen is about 5 μg/L, significantly lower than CW617. With more conditioning, the leached concentrations reduced further to under 2.5 μg/L. Lead leaching from brass is known to be sensitive to the chloride-to-sulfate mass ratio (CSMR) in water (see Triantafyllidou, Reference Triantafyllidou2006, Chapter 4). The CSMR at Weesperkarspel and Leiduin were 8.8 and 1.5, respectively. Nonetheless, there do not seem to be major differences in the lead concentrations measured at the two sites.
Lead concentration leached into water following 16 h of stagnation as a function of the number of months conditioned. (a) Results for CW617. (b) Results for CW724R.
Figure 3. Long description
Panel a, labeled C W 6 1 7, and panel b, labeled C W 7 2 4 R, both feature a logarithmic Y-axis for Lead leached in micrograms per liter, ranging from 10 super minus 1 to 10 super 3, and a linear X-axis for Months passed from 0 to 10.
In panel a, four data series (Weesperkarspel 1 and 2, Leiduin 1 and 2) start at a high concentration near 2 times 10 super 2 micrograms per liter at month 0. There is a sharp exponential decay over the first 2 months, after which all series stabilize and fluctuate around the 10 micrograms per liter horizontal dashed line for the remainder of the 10 months.
In panel b, the same four data series start much lower, between 5 and 10 micrograms per liter at month 0. They show a gradual decline over the first 4 months before stabilizing between the 2.5 and 5 micrograms per liter horizontal dashed lines.
Both panels include three horizontal dashed reference lines on the right side indicating 10 micrograms per liter, 5 micrograms per liter, and 2.5 micrograms per liter. Legends in the bottom-left of each panel identify the four water sources using colored triangles: blue upward for Weesperkarspel 1, purple downward for Weesperkarspel 2, red rightward for Leiduin 1, and green leftward for Leiduin 2.
(a) Validation of LEadGO through a comparison between measured and modeled average lead concentrations. (b) Situations considered for exploring the effect of brass appurtenance location on lead concentration at the cold water kitchen tap. (c) Statistics of lead concentration at the tap.
Figure 4. Long description
Panel a is a scatter plot comparing Modelled lead concentrations on the x-axis to Measured lead concentrations on the y-axis, both in micrograms per liter from 0 to 3. Blue x-shaped data points are clustered around a dashed 1 to 1 diagonal line. The background is divided into a red upper-left zone and a green lower-right zone by dotted lines.
Panel b shows three isometric plumbing diagrams representing different brass appurtenance locations. Each diagram features blue and red piping. A red circle and arrow indicate the specific brass location being tested. From left to right, the locations are:
1. Shutoff valve: Length 20 cm, Diameter 32 mm, Volume 160 ml.
2. Water meter: Length 20 cm, Diameter 32 mm, Volume 160 ml.
3. Kitchen tap: Length 20 cm, Diameter 10 mm, Volume 16 ml.
A red arrow at the bottom indicates the Cold Water kitchen tap.
Panel c contains two tables.
The first table shows Average lead concentration at the kitchen tap for brass types C W 6 0 2 and C W 6 1 7. For C W 6 0 2, concentrations are 0.23 at the shutoff valve, 0.28 at the water meter, and 6.76 at the kitchen tap. For C W 6 1 7, concentrations are 0.17 at the shutoff valve, 0.20 at the water meter, and 2.02 at the kitchen tap.
The second table shows the Percentage of time lead exceeds 5 micrograms per liter. For C W 6 0 2, values are 0.6 percent at the shutoff valve, 0.8 percent at the water meter, and 19 percent at the kitchen tap. For C W 6 1 7, values are 0.4 percent at the shutoff valve, 0.6 percent at the water meter, and 18 percent at the kitchen tap.
Validation of the mechanistic model LEadGO
The validation of LEadGO was achieved through a one-to-one comparison between the measured and modeled average concentrations at five consumption points, with 5 weeks of unique water demand patterns. It is observed that the order-of-magnitude of the measured and modeled lead concentrations are the same (<5 μg/L), as displayed in Figure 4(a). In this sense, the model is found to be sufficiently reliable for the present objectives (determining exposure with respect to the norm). The experimental work herein should be seen as a first rigorous attempt at validating the mechanistic model and there is certainly room for further improvement. For example, the model does not include aspects such as particulate lead and longitudinal dispersion, which can influence outcomes (Abokifa and Biswas, Reference Abokifa and Biswas2017; Burkhardt et al., Reference Burkhardt, Woo, Mason, Shang, Triantafyllidou, Schock, Lytle and Murray2020). The validation also relies on one type of lead source at one location. Additional experiments with variation in the lead component and placement will certainly test the robustness of the validation framework.
Applying LEadGO: Lead concentrations at the tap
Once sufficient confidence was gained in the mechanistic modeling framework, it was applied to various informed scenarios. The case studies and outcomes are shown in Figure 4(b) and (c) respectively.
We focus in our case study on the results of lead concentrations in water coming out the cold water kitchen tap, as this is the most commonly used tap for oral consumption. It is evident in Figure 4(c) that brass in the shutoff valve and water meter does not lead to an exceedance of the (new) norm of 5 μg/L, irrespective of the brass. When the brass is at the kitchen tap, the average lead concentration at the tap shoots up, despite the much lower contaminated water volume (16 mL vs. 160 mL for the shutoff valve and water meter). In this case, it does matter which brass is utilized. The use of CW602 (which leaches more) at the kitchen tap leads to an average exposure above the norm, in contrast to CW617. These examples collectively show that when it comes to exposure to lead through components in the plumbing system, location matters. In fact, the presence of a lead-releasing component in the final branch of the plumbing system toward the kitchen tap is the least desirable situation (due to longer stagnation periods and no escape routes to other branches of plumbing).
The results also offer insights relevant for RDT, especially in the context of exposure determination. The presence of brass in the water meter or shutoff valve can lead to an occasional (0.5–1%) incidental measurement of lead concentration higher than 5 μg/L at the tap, even if the average concentration is lower than 5 μg/L. When there is brass at the kitchen tap, the percentage of time that the lead concentration in the water coming out the tap exceeds 5 μg/L is significantly higher than when the brass is in the service line, implying higher detection chances with RDT sampling. For both brass types at the kitchen tap, the amount of time the lead concentration at the tap exceeds 5 μg/L is nearly identical, in contrast to the differences in weekly average concentrations with respect to the 5 μg/L. These results suggest exercising caution while interpreting RDT measurements.
Using profile sampling to localize lead leaching components
The location of a lead-releasing component clearly affects exposure. To be able to replace such components, one needs to know the location as well as who is responsible for replacement. Thus, it is important to localize lead-leaching components. In this section, we reflect on the robustness of profile sampling by means of modeling, experiments in HomeWaterLab and field measurements. A compilation of results from all three approaches is shown in Figure 5.
Robustness of profile sampling. (a) Numerical modeling. (b) Laboratory experiments. (c) field measurements.
Figure 5. Long description
Panel a features a plumbing schematic on the left with labels for cold water plumbing, hot water plumbing, sampling pathway, lead service line, distribution mains, and sampling location. To the right is a grid of six plots showing lead concentration in micrograms per liter on a logarithmic Y-axis from 0.1 to 100, against sampling volume. The plots are labeled Day 20, 30, 62, 72, 91, and 127. Blue squares represent concentrations less than 5 micrograms per liter, and red squares represent concentrations greater than or equal to 5 micrograms per liter. Rectangular boxes highlight lead leached during regular usage, while ovals highlight lead leached during strategic prolonged standstill.
Panel b is a line graph of total lead concentration in micrograms per liter versus volume of water sampled in liters. The data shows an initial peak at 0.2 liters, a trough between 0.5 and 1.2 liters, and a major peak reaching nearly 20 micrograms per liter at approximately 1.8 liters. A vertical tan line at 1.5 liters marks the expected location of the profile sampling peak due to the lead pipe.
Panel c is a scatter plot of lead concentration in micrograms per liter versus volume sampled in liters. Red diamonds represent profile sampling with the third liter composed of 100 m l sample volumes, showing a peak of 1.4 micrograms per liter near the water meter. Pink diamonds represent profile sampling with one liter sample volumes. Vertical dashed lines indicate the water meter and shutoff valve locations. A horizontal line at 0.5 micrograms per liter marks the measurement limit.
The following can be observed from profile sampling in the numerical simulations:
-
• The first section (light blue background) comprises the first 11 samples, each of which almost always has a lead concentration <5 μg/L. These samples belong to the piping between the kitchen tap and the service line. The few outliers in these virtual samples can be avoided by flushing the pipes prior to stagnation.
-
• The second section (yellow background) comprises three samples that always have a lead concentration of >100 μg/L. These elevated concentrations correspond to the lead that has been released from the service line during prolonged stagnation. It can be seen that these three samples always give the same value.
-
• The third section (gray background) comprises the last six samples with a lead concentration of <5 μg/L. This corresponds to water from the distribution network that has been modeled as lead-free water.
The robustness was further evaluated through laboratory tests in HomeWaterLab. An example is shown in Figure 5(b), where the peak of the profile corresponds reasonably well with the location of the lead-releasing component. The lead-releasing component, a lead pipe, has a volume of 60 mL. However, one sees that the profile of lead concentrations has a much broader peak than 60 mL. This observation is similar to that of Burkhardt et al. (Reference Burkhardt, Woo, Mason, Shang, Triantafyllidou, Schock, Lytle and Murray2020) and Lytle et al. (Reference Lytle, Formal, Cahalan, Muhlen and Triantafyllidou2021), which is attributed to the phenomenon of longitudinal dispersion. The height of the peak (20 μg/L) is significantly lower than concentrations encountered during the tabletop stagnation experiments (160 μg/L), possibly due to the dispersion. The area under the profile gives a total lead mass in the same order-of-magnitude as expected from the lead dissolution curves.
An example where the robustness of profile sampling in field measurements was demonstrated is shown in Figure 5(c). In the first round of sampling, the third sample showed increased lead concentrations. This increased lead concentration stemmed either from the water meter or shutoff valve. From the second round, it becomes evident that the released lead corresponds to the location of the water meter. The two iterations have nearly identical areas under the profiles.
Discussion
Long-term behavior of lead leaching from brass
The lead concentrations from the ‘run-time’ tests are in agreement with the choices for materials made in the EUPL. The EUPL shall succeed the 4MS common approach (a consortium initiated by four EU nations including the NL). A full historical overview of norm development falls outside the scope of our work. For the interested reader, the work of Estelle (Reference Estelle2016) offers a historical overview of the 4MS approach, including references to the norms EN 15664 and DIN 50930. The work in the European Commission and Directorate-General for Environment and Umweltbundesamt GmbH (2017) offers an overview of recommendations surrounding the update to the Drinking Water Directive in the context of materials in contact with drinking water. In the 4MS approach, both CW617 and CW724R were deemed acceptable for contact with drinking water (4MSI Common Approach, 2025). This has changed in the EUPL, where CW617 is not allowed. For metallic materials that come in contact with drinking water and leached lead, there is an additional allocation factor of 50% (see Commission Implementing Decision, 2024/3672024, Table 1, Annex V). This means that such metallic materials may not leach more than 50% of the normative lead concentration (50% of 5 μg/L = 2.5 μg/L), as illustrated in Figure 3.
A couple of consequential outcomes for drinking water utilities based on the observations are as follows:
-
• Increased lead leaching from brass components in the first months of usage is expected, a phenomenon that is known (Wuijts et al., Reference Wuijts, Slaats, Versteegh and Meerkerk2008; Versteegh and Boshuis-Hilverdink, Reference Versteegh and Boshuis-Hilverdink2010). This requires good communication toward consumers.
-
• While it is commercially referred to as ‘lead-free’, it is known that CW724R may contain up to about 0.05% Pb by weight (see Zoghipour et al., Reference Zoghipour, Tascioglu and Kaynak2024, Table 1), possibly due to recycling of brass. This suggests exercising caution while placing purchase orders.
Applying LEadGO: Lead concentrations at the tap
The simulations provide significant insights into lead exposure caused by brass components. There are two types of observations from our results: (a) absolute value of average lead concentration at the kitchen tap with respect to the norm of 5 μg/L and (b) role of location of the lead-leaching component.
The simulations presented herein are limited in range, which may raise questions on the robustness of our findings. Below, we discuss different parameters that affect observation (a).
-
• Drinking water consumption: It has been shown that reduced water consumption can lead to increased average lead concentrations, up to a factor of 3 (see Hatam et al., Reference Hatam, Blokker, Doré and Prévost2023, Figure 6). Given the lead concentration due to brass components in the service line is about 0.2 μg/L, we anticipate that variations in water consumption behavior will not lead to an exceedance of 5 μg/L.
-
• Drinking water composition: Water composition at the plumbing of the consumer may differ from that at treatment sites. For example, higher concentrations of copper in the water can affect the leaching mechanisms in brass (Maynard et al., Reference Maynard, Mast and Boyd2008). Uncertainty in drinking water composition was modeled by increasing the equilibrium lead concentration by a factor of 11.6 (while holding the same lead dissolution timescale). This leads to an increase in average lead concentration at the tap by the same factor. Even under such an exaggerated lead dissolution behavior, the average lead concentration at the kitchen tap due to brass components in the service line remains below 5 μg/L.
-
• Particulate lead: The tabletop experiments for brass water meter suggest that particulate lead is less prevalent in brass. See Figure S.9 in the Supplementary Material. In that sense, we expect our conclusions not to be affected severely by particulate lead.
-
• Temperature: While not relevant for the service line, it is known that lead leaching from brass is exacerbated at higher water temperatures (see Sarver and Edwards, Reference Sarver and Edwards2011, Figure 6), suggesting increased exposure via brass in hot water plumbing. While relevant from a human health perspective, the hot water plumbing falls outside the jurisdiction of a drinking water utility. This suggests that our conclusions for rehabilitation is robust for this parameter too.
The observation (b) regarding the role of lead leaching component location will stay robust to the parameters stated above. As a consequence of the simulated lead concentrations and the expected robustness of our findings, the urgency of immediate replacement of brass appurtenances in service lines is reduced. RDT measurements during regular monitoring campaigns can serve as an indicator for further investigation.
Using profile sampling to localize lead-leaching components
When RDT measurements show norm exceedance, profile sampling can serve as a useful tool. The height, breadth and location of the peak in the profile give information on the equilibrium lead concentration, volume of the lead-releasing component and location of the lead-releasing component respectively. This can additionally help identify who is responsible for replacing it while also offering an indication on the priority of urgency in which the replacement needs are to be approached. Of course, practitioners need to be aware that precise interpretation of the results in practice will be hindered (e.g., height and breadth of the peak affected due to dispersion). In 2023–2024, profile sampling was attempted by a few drinking water utilities in the NL, and several practical challenges in implementation were encountered (such as residents disturbing the stagnation or the sampler letting water flush away between two samples or not all samples being analyzed to save costs or collecting large sample volumes leading to dilution), which undermine the strength of profile sampling. To help Dutch practitioners make the most of profile sampling, a simple list of ‘good practices’ has been formulated (Dash et al., Reference Dash, Blokker, Dignum, ten Kate and Gardien2024).
Conclusions
Concise answers to the research questions are as follows:
-
1. Modeling: Using a mechanistic model which relies on integration of the underlying physicochemical processes, the interplay among these various processes are captured. While the model in the present study can certainly benefit from refinements, the current version captures enough detail to answer the questions at hand.
-
2. Validation: Carefully formulated laboratory experiments helped with the validation of the mechanistic model, LEadGO. Lead dissolution experiments provide empirical inputs for the mechanistic model whereas composite sampling in a plumbing rig facilitate one-to-one comparison with modeling outcomes.
-
3. Application: A prudent selection of case studies allows the model to provide valuable insights on practically relevant scenarios. By being able to statistically study lead concentrations at the tap, a statement can be made on the risk posed by a certain component. Additionally, such a model serves as a test bed for sampling protocols, as shown for profile sampling.
The outcomes of the research activities have then been used to provide answers to questions relevant for the Dutch drinking water sector:
-
1. What policy should a drinking water utility adopt with regard to existing and future assets that may leach lead (with special focus on brass)?
-
• The present research shows that it is mainly lead leached closer to the end use points (e.g., a kitchen tap) that poses a higher risk for exposure. Lead leached from brass appurtenances in service lines leads to much lower lead concentrations at the tap.
-
• Thus, several utilities have decided not to pursue an accelerated rehabilitation strategy and to replace existing brass components at end of lifetime. Due to the presence of moving components, a water meter has a shorter lifetime than a shutoff valve, necessitating separate strategies.
-
• For future assets, a few utilities will select brass types that appear in the EUPL. However, a few utilities have also taken the decision to step away from brass entirely and are installing composite-based water meters (as a mitigating measure for future reductions in lead norms).
-
• For newly installed brass appurtenances, heightened lead leaching is expected in the first few months. Corresponding consumers are advised to flush water before consumption.
-
• A drinking water utility can also communicate to consumers to be mindful with choices of plumbing materials, since these have a higher influence on exposure. Specific advice could include recommendations to flush water at brass taps especially for stagnation periods exceeding a couple of hours.
-
2. What should the drinking water utility do when an incidental norm exceedance is encountered?
-
• As of now, the Dutch drinking water sector will continue relying on RDT measurements as the standard way of monitoring when it comes to detecting norm exceedance.
-
• If RDT measurements return a lead concentration above the legal threshold (leading to an incidental norm exceedance), the profile sampling technique will be deployed.
-
• Profile sampling (when deployed correctly) can help localize the lead-leaching components, as well as reveal whether the drinking water utility or the premises owner is responsible for replacing it.
Outlook
In closing, this study reflects the research-based efforts undertaken by the Dutch drinking water sector to support decision making in response to the EU-DWD. The sector has a clearer view on the way forward with respect to rehabilitation of existing assets as well as choices for future assets. Moreover, the sector is prepared for incidental norm exceedance with the help of profile sampling, a method which can help identify whether the utility or the premise owner is responsible for follow-up actions. The present study was possible thanks to the tightly knit nature of the sector. The findings and conclusions of this study can be used as a starting point by drinking water utilities of other nations, where the stricter norm for lead in drinking water is yet to be enshrined into the law.
Open peer review
To view the open peer review materials for this article, please visit http://doi.org/10.1017/wat.2026.10026.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/wat.2026.10026.
Data availability statement
The data that support the findings of this study are available from the corresponding author (E.J.M.B.) upon reasonable request.
Acknowledgements
We would like to thank the steering committee members from drinking water utilities of the projects for their guidance and inputs during the study. We are grateful to Prof. Dragan Savić for feedback on a draft manuscript. The current title and impact statement are reformulated versions of original text. The reformulation was done with the help of an AI tool (Microsoft Copilot version 2.20251007.34.0), to make the original text more concise and impactful.
Author contribution
Conceptualization: M.D., M.B.; formal analysis: A.D.; funding acquisition: M.B., investigation: A.D., M.D.; methodology: A.D., M.D.; project administration: A.D., M.D.; software: A.D., M.B.; supervision: M.B.; validation: A.D.; visualization: A.D.; writing – original draft: A.D.; writing – review & editing: A.D., M.D. and M.B. All the authors approved the final submitted draft.
Funding statement
This research was financially supported by the collective research program of 10 Dutch and 1 Flemish drinking water utilities (specifically projects 403660/001, 402045/316, 402045/357, 404300/077) and the European Partnership for the Assessment of Risks from Chemicals (PARC) under the Horizon Europe Research and Innovation Programme, Grant Agreement No. 101057014. Open access funding provided by Delft University of Technology.
Competing interests
The authors declare none.
Ethics and approvals
The research meets all ethical guidelines, including adherence to the legal requirements of the study country.
Open access
Comments
Dear Editor,
We submit our article titled “Dutch drinking water sector readiness for stricter EU lead norm” to be considered for publication in Cambridge Prisms: Water. Our study is a direct response to the stricter EU norm for lead in drinking water following the revised Drinking Water Directive. Our study brings together facets from scientific research (modelling and experiments), practical application, policymaking and we believe that our article will be suitable for a broad audience, as is the target of the journal. Our article reflects actions taken in the Dutch drinking water sector and we expect that our findings can be relevant for other European member states. We hope our article will be considered for publication and we look forward to hearing back the opinions of the reviewers.
For transparency purposes, AI (Microsoft Copilot) has been used for improving parts of the manuscript, specifically, the title and the impact statement.
Thanks,
Amitosh Dash, Marco Dignum, Mirjam Blokker