Sage Journals: Discover world-class research

Abstract

Accurate estimation of water demand in buildings is essential for designing safe, efficient, and sustainable water supply systems. Conventional design approaches often lead to significant overestimations of water demand, resulting in water supply systems that are frequently oversized. This study introduces the Water Demand Estimation Model (WDEM), a novel stochastic model developed specifically for application to non-residential buildings. The model integrates statistical data on sanitary appliances and user behaviour using extensive Monte Carlo simulations to generate realistic scenarios of simultaneous appliance usage. In addition to appliance properties, WDEM accounts for building occupancy, an essential factor in design flow rate estimation. It provides a set of user-friendly design equations, as an essential step towards future application in practice. Application of WDEM to three diverse case study buildings revealed substantial reductions in estimated design flow rates, ranging from 63% to 73%, compared to current UK design guides. These findings demonstrate WDEM’s effectiveness in estimating water demand and thereby avoiding system oversizing, which is crucial for designing water supply systems that are cost-effective and have improved water quality.

Practical application

The WDEM provides a new approach to estimate water supply design flow rates by combining occupancy-based usage with appliance-efficiency ratings to derive explicit design equations. The outcomes include preliminary design equations which, once validated, should support improved estimation of design flows without running the simulation. In practice, WDEM may support faster early-stage decisions, potentially smaller pipework and storage volumes, reduced capital costs and pumping energy, and lower stagnation and water quality risks.

Keywords

Water demand estimation water supply systems design flow rate non-residential buildings WDEM

Introduction

A reliable water supply system is essential for all types of buildings, ensuring adequate water delivery to all appliance draw-off points. The primary challenge when designing these systems is accurately estimating the maximum simultaneous water demand. This becomes even more challenging in large-scale buildings, where the number of possible combinations of simultaneous water demand increases exponentially with the number of appliances. As a result of population growth and urbanisation, large-scale buildings, especially high-rise buildings, have become increasingly common in recent years.^1–3 Thus, it is important to provide an accurate approach to estimate water supply design flow rates, which form the basis for sizing pipework and all other components in both water supply and wastewater drainage systems. In designing water supply systems, when the water demand estimate is inaccurate, this can lead to either (i) underestimation, resulting in a system that is too small and an inadequate service provision; or (ii) overestimation, resulting in oversizing the entire system beyond actual needs. Such oversizing not only increases costs,^4,5 but also increases the risk of biofilm formation and microbial contamination, such as Legionella,^6,7 ultimately leading to unsustainable systems.⁸

Conventional design methods mainly originated from Hunter’s approach, introduced in the 1940s, which employed probability principles and fixture units (FUs) to estimate water demand.⁹ Although influential, Hunter’s method is increasingly recognised as leading to systems that are oversized due to outdated appliance and user behaviour parameters. The significant evolution in water appliance efficiency and user behaviour (notably enhanced by environmental awareness and commitments to water conservation) underscores the limitations of traditional estimation methods.^10–12 Therefore, many international design codes have been recognised as leading to oversizing of building water systems, including those used in the USA,¹³ the Netherlands,¹⁴ the UK,¹⁵ and Brazil.¹⁶

Most studies on water demand, including those involving artificial intelligence (AI), have focused on estimating water demand at large temporal scales, for example, daily consumption.^17–19 By contrast, only a few address the precise estimation of water demand in litres per second (l/s), which is essential for accurately sizing water supply systems such as the recent work by Fernandez et al.²⁰ In addition, research on oversizing of water supply systems has mainly focused on residential buildings, while far fewer studies address non-residential buildings, which represent a substantial share of the built environment and are often large-scale.

To address oversizing and support better design of water supply systems, this paper introduces a novel stochastic model specifically developed to estimate water demand in non-residential buildings. The WDEM model integrates up-to-date statistical data on sanitary appliances and user behaviour using Monte Carlo simulation to generate realistic scenarios of simultaneous appliance usage. By applying the WDEM to representative case study buildings, this study quantifies the potential reduction in design flow rates and demonstrates possible improvements in water supply system design and operation.

Statistical models for estimating design flow rate

Various statistical models have been used in international design codes to estimate water demand in buildings; differing in inputs, methodologies, complexity, and accuracy. These models are broadly categorised into three main types: (i) probabilistic; (ii) empirical; and (iii) stochastic models. Most international design codes for water supply systems use one of these approaches. Examples of international design codes and their associated modelling methods are provided in Table 1.

Table 1.

Examples of international design codes for water supply system design and their related statistical models.²¹

Code	Country	Statistical model
BS EN 806	UK	Probability
BS 8558 (BS 6700)	UK	Probability
CIBSE guide G	UK	Probability
CIPHE/IoP	UK	Probability
IPC/UPC/NSPC	USA	Probability
ISSO - 55	Netherlands	Empirical
SIMDEUM	Netherlands	Stochastic
SANS 10252-1	South Africa	Empirical
UNE 149201	Spain	Empirical
DIN 1988-300	Germany	Empirical
DS 439	Denmark	Empirical
SNiP 2.04.01-85	Russia	Empirical

Probabilistic models rely on assumptions and probability distributions to derive mathematical expressions, tables, and graphs. Hunter’s method was among the first successful probabilistic approaches. Hunter used factors such as average appliance flow rate, duration of water use, and interval between operations to estimate the design flow rate. Despite ongoing debates regarding its accuracy, Hunter’s method remains the foundation for many international design codes.²² Empirical models are more straightforward in their setup and application, typically generating a single result for a given set of parameters. These models derive formulae, tables, and charts by exploring mathematical relationships between input and output data.²¹ Despite their simplicity and accuracy, their applicability is limited by a short validity period, as changes in parameters such as user behaviour or appliance efficiency directly affect model performance. Stochastic models, on the other hand, represent the uncertainty in a dynamic system using probability theory.²³ They are traditionally used to test scientific theories when processes, observations, or boundary conditions are subject to stochasticity.²⁴

In response to concerns about fixture unit methods, there has been a sustained effort to develop improved approaches for estimating design flow rates. For instance, in the United States, the Water Demand Calculator (WDC) offers a modern alternative to Hunter’s method for sizing water supply systems in residential buildings.^25–28 The WDC is now recognised in practice, having been codified as Appendix M of the Uniform Plumbing Code (UPC), published by the International Association of Plumbing and Mechanical Officials (IAPMO).²⁹ In the Netherlands, the Simulation of water Demand, and End-Use Model (SIMDEUM), an end-use stochastic model has been developed and applied by Dutch water utilities and researchers.^14,30–32 This approach underpins procedures for deriving design rules for peak cold and hot water demand in water supply systems, informing current practice in the Netherlands.

To improve water supply system design in the UK, this study presents the WDEM, a stochastic modelling framework for estimating building design flow rate. Conceptually, WDEM aligns with SIMDEUM in adopting a stochastic, end-use style approach that represents variability in occupant behaviour and appliance use. This differs from the WDC, which is primarily a probabilistic sizing tool and has been developed and calibrated using practice and datasets from the USA. In contrast to both, WDEM has been developed in two formats (i) in a building-specific form using the installed or proposed sanitary appliances and their flow-rate characteristics, or (ii) in a generalised form in which design flow is related to user numbers and overall appliance water-efficiency level, consistent with UK design guidance. Explicit integration of sanitary appliance efficiency, together with dynamic occupancy patterns, allows WDEM to reflect key characteristics of modern plumbing systems and to capture the variability and uncertainty of building water use.

Water Demand Estimation Model (WDEM)

After evaluating various modelling approaches and assessing resource availability and accessibility, a stochastic framework was adopted to develop WDEM, with a specific focus on non-residential buildings. This choice was supported by previous research demonstrating the effectiveness of stochastic models in simulating water demand and improving the accuracy of design flow rate estimation, for example, in Italy,^33,34 in Australia,³⁵ in Brazil,³⁶ and in Mexico.³⁷ The primary focus of the study is the design of main water supply pipes, particularly in high-rise buildings, where accurate demand estimation is critical for ensuring system efficiency, sustainability, and maintaining water quality.

The WDEM is implemented in MATLAB and is designed for application to any non-residential building, using sanitary appliance parameters to estimate the design flow rate across the whole water supply system. Recognising that oversizing most commonly occurs in large-scale buildings, particularly in the main supply pipes, the model has also been generalised to estimate the total design flow rate directly for buildings. In this general form, WDEM provides closed-form design equations as explicit functions of building occupancy, enabling straightforward sizing without running the full simulation.

The model integrates user behaviour with sanitary appliance data, employing probability distributions to reflect potential simultaneous usage scenarios. The three-stage framework underpinning the WDEM first defines building types and allocates sanitary appliances based on user numbers in line with UK standards. It then uses a binomial distribution to characterise random water usage events and applies Monte Carlo simulations, involving extensive random sampling to generate realistic water use scenarios. This simulation approach is well-established for capturing uncertainty in complex systems³⁸ and underpins the model’s ability to reflect real-world variability in water demand. Finally, regression analysis is performed on simulation outputs to derive new design equations for estimating flow rates. Figure 1 summarises the methodological stages of the WDEM, from initial assumptions through to regression analysis and equation derivation.

Figure 1.

WDEM methodology and derivation of design equations.

Number of users and provision of sanitary appliances

Stage 1 of the model involves establishing relationships between user numbers ( $N$ ) and sanitary appliance provisions. The number of users significantly impacts total water demand as it increases the likelihood of simultaneous appliance usage. A recent study by Josey and Gong³⁹ used a stochastic model to show that building occupancy strongly influences residential water demand characteristics, including peak demand, average use, peak hourly consumption, and stagnation. To further validate the direct impact of user numbers on water demand, user count and flow rate were monitored in one of three case study buildings at Heriot-Watt University. A Q-Scan TwinComm people counter was installed at the building’s main entrance, recording each entry and exit, while ultrasonic flow meters measured water flow every 5 seconds. The data indicated, as anticipated, a strong correlation between water consumption and occupancy levels, as shown in Figure 2. During periods of high occupancy, such as mid-day, noticeable spikes in water demand were recorded, reflecting increased water usage. For overnight periods when the building was unoccupied, water consumption dropped to zero. The cumulative functions of water consumption and user numbers, Figure 3, further illustrate the direct correlation between occupancy and water demand. This strong correlation highlights the importance of occupancy data in estimating water demand and supports the incorporation of user count as a key input parameter in the WDEM.

Figure 2.

Relationship between occupancy and water consumption in the case study building.

Figure 3.

Cumulative functions of occupancy and water consumption patterns in the case study building.

Accordingly, the WDEM was developed to accurately estimate water supply design flow rates by accounting for essential usage variables. For this purpose, the number of appliances installed in non-residential buildings was derived based on user numbers in accordance with the UK standard BS 6465.⁴⁰ This standard provides recommended numbers of appliances depending on anticipated occupancy levels and building type. Further details regarding the number and type of appliances, male-female ratio, and building capacity can also be found in BS 6465. At this stage, for any specified number of users, the corresponding numbers and types of appliances are defined and prepared for simulation.

Generating appliance-use events

In stage 2, the model simulates water usage via probability distributions for appliance operations. Each appliance usage event is treated as a Bernoulli trial, defined by binary outcomes: active (on/in use) or inactive (off/idle). Each appliance has a defined usage probability (p-value), a value that is statistically independent of other appliances. Accordingly, water usage in this context can be defined as a Bernoulli trial, assuming that water usage follows a binomial process under the following conditions: (i) the experiment/event can be repeated for any specified number of trials; (ii) each event is independent; and (iii) each trial has two possible outcomes (on or off). In statistical terms, the probability of an appliance being active or inactive is defined as follows:

P (O n) = p

(1)

P (O f f) = 1 - p = f

(2)

where:

$P (O n)$ : probability the appliance is in use

$P (O f f)$ : probability the appliance is not in use

$p$ : per-appliance usage probability (p-value)

$1 - p$ : per-appliance inactivity probability = $f$ (failure)

The probability that exactly $k$ appliances of the same type are simultaneously active out of $n$ , follows the binomial distribution:

P (k) = (\begin{array}{l} n \\ k \end{array}) p^{k} f^{n - k} k = 0, 1, \dots . n

(3)

where:

P(k): probability that exactly k appliances are simultaneously active out of n

n: total number of appliances of the same type considered

k: number of active appliances out of n

p: per-appliance usage probability (p-value)

f: per-appliance inactivity probability, defined as $f = 1 - p$

$(\begin{array}{l} n \\ k \end{array})$ : binomial coefficient, representing the number of ways to define k active appliances from n, calculated from the following equation:

(\begin{array}{l} n \\ k \end{array}) = \frac{n!}{k! (n - k)!}

(4)

Equation (3) may also be expressed in compact form as:

P (k) = b (k; n, p)

(5)

where

b (k; n, p)

explicitly denotes the binomial distribution with parameters

n

(number of appliances) and

p

(usage probability), with

k

as the random variable of active appliance count.

Model parameters and assumptions

In the WDEM approach, the design flow rate is estimated using three primary inputs, selected for their impact on overall building water demand. This section defines each parameter, states the associated modelling assumptions and explains how each input affects the predicted design flow rate.

(i) The number of appliances, which directly influences the overall design flow rate, since a higher number of appliances increases the total demand and therefore increases the simulated total flow rate. In the generalised version of the model, appliance numbers are linked to building scale and occupancy, because larger user populations typically require a greater provision of sanitary appliances. Overall, the model is most sensitive to occupancy level and the associated number of sanitary appliances. (ii) Appliance discharge values $(q)$ , which represent the flow rate of each individual appliance. Because the total instantaneous demand in each simulation iteration is obtained by summing the flow rates of all appliances operating concurrently, the model is highly sensitive to individual appliance flow rates. The individual appliance flow rate is one of the main input parameters when the model is applied to a specific case, whereas in the generalised model, appliance flow rates are defined by water-efficiency levels. The appliance discharge values were determined with a market and efficiency-informed approach that integrates water efficiency ratings and current product availability. In Europe, including the UK, appliances are classified into five distinct efficiency groups (A – E) based on flow rate.^41,42 Average flow rates within each efficiency category (groups A–E) were adopted in the model for simulating design flow rates. (iii) Per-appliance usage probability (p-value), which represents the likelihood that an appliance is in use. This parameter reflects how often and for how long an appliance is typically used and therefore depends on appliance type and user behaviour. In this study, p-values were adopted from recommendations in the Chartered Institute of Plumbing and Heating Engineering (CIPHE) guide.⁴³ This guide provides comprehensive information on sanitary appliances and reflects UK plumbing systems. It defines p-values based on the duration for which an appliance is in use (t) and the average time between successive uses (T). Importantly, the $t$ values are combined with different $T$ values to represent varying levels of usage intensity and to provide $p$ -values for different building types (low, medium and high use). Using UK-specific data and accounting for building type allows the model to adopt the p-values from the CIPHE guide directly, without introducing additional assumptions that could increase uncertainty.

In this stage, the flow rate calculation framework is established using the full set of input parameters. In the generalised model, these inputs are expressed in terms of user counts $(N)$ , from which the corresponding number and types of sanitary appliances are defined.

Resultant probability distribution and cumulative distribution function CDF

The Monte Carlo technique provides an optimal means of generating the broad range of simultaneous use scenarios necessary for robust water demand estimation. In this study, 400,000 iterations were run to capture a wide range of appliance use combinations and represent realistic usage. This high repetition rate significantly reduced variability between successive runs and produced stable outputs. In each iteration, appliances were randomly assigned on/off states, and the corresponding total flow rate was calculated as the sum of the flow rates of all active appliances. Thus, varying combinations of simultaneously active appliances produced a distribution of total flow rates. Initial simulation outputs generated a wide range of flow rate values, ranging from zero (no appliance active) to a maximum (all appliances active). To characterise these outcomes, the simulated flow rates were counted across all iterations to quantify how frequently each combination of appliance uses occurred in the simulation, forming an empirical probability distribution. These probabilities were then cumulatively added and plotted against their respective flow rate values to generate a Cumulative Distribution Function (CDF). The CDF provides a full probabilistic representation of simulated water flow rates and supports design at a chosen reliability, which is the probability that demand does not exceed the design flow rate. Thus, reliability ranges from 0% (always fails to meet demand) to 100% (always meets demand). By using this comprehensive CDF, flow rates at any desired reliability level can be accurately identified. For practical design applications, the 99^th percentile flow rate ( $Q_{0.99}$ ) was selected as the design criterion $(Q_{d})$ . Employing the 99^th percentile ensures consistency and comparability with current industry standards and guidelines, which commonly adopt a similar reliability measure for system sizing. Figure 4 illustrates an example CDF derived from these Monte Carlo simulations and demonstrates how the 99^th percentile design flow rate is extracted (in this case 0.82 l/s).

Figure 4.

Example of a CDF and extraction of 99^th percentile simulated flow rate.

However, the model is designed to accept mixed appliances and can be applied to a specific building by using the installed or proposed combination of sanitary appliances, regardless of flow rate levels. The generalised form of the model expresses the design flow rate as a function of the number of users and appliance water-saving levels. Where a building contains appliances with different water-saving efficiencies, the input can be appliance-specific flow rates entered directly in WDEM or professional judgement can be applied by selecting a representative efficiency level (i.e. between the design lines) that best represents the appliance mix.

In the generalised model, design flow rates were simulated for all efficiency groups (A–E). For each efficiency group, the same method was applied iteratively to calculate new design flow rates across various occupancy levels (up to 2000 users). Figure 5 presents the simulated design flow rates as a function of user numbers for each appliance efficiency category.

Figure 5.

Simulated $Q_{0.99}$ flow rates as a function of user numbers for each appliance efficiency group (A–E).

Regression analysis and establishing design equations

Regression analysis was conducted to identify the most appropriate mathematical functions to represent the flow rate data. Three types of regression function were considered: (i) linear; (ii) polynomial (curve-fitting); and (iii) segmented (piecewise) regression. The linear regression model did not adequately capture data variability, particularly in the lower user range (0–300 users), resulting in insufficient fit across all appliance efficiency groups. Although flow rate generally increases with user numbers, the data show a clear change in slope around

N = 300

. This reflects the BS 6465 step change in appliance provision recommendations so that, beyond this point appliances are added in fixed steps. To address this limitation, a second-order polynomial regression was explored. However, despite an improved fit, this did not fully capture the distinct change in trend observed at the defined breakpoint (

N = 300

). Therefore, segmented regression was employed, allowing the dataset to be divided at the breakpoint, with separate linear equations fitted to each segment. The statistical performance indicators show a strong agreement between the simulated and fitted design flow equations with coefficients of determination (R²) of 0.95–0.99, and low errors as indicated by the mean absolute error (MAE) of 0.013–0.072 l/s and root mean square error (RMSE) of 0.016–0.087 l/s. This approach provided the best fitting design equations across all the appliance groups and occupancy levels as summarised in Table 2 together with the corresponding R², MAE and RMSE values.

Table 2.

Final design equations with R², MAE and RMSE values for segmented linear regression.

Appliance group	User number N	Design equations	R²	MAE l/s	RMSE l/s
A	N < 300	Q = 0.167+1.30 × 10 ⁻³ ⋅ N	0.96	0.013	0.016
A	N ≥ 300	Q = 0.372+6.9 × 10 ⁻⁴ ⋅ N	0.99	0.023	0.027
B	N < 300	Q = 0.294+2.60 × 10 ⁻³ ⋅ N	0.98	0.045	0.053
B	N ≥ 300	Q = 0.706+1.2 × 10 ⁻³ ⋅ N	0.99	0.039	0.048
C	N < 300	Q = 0.427+2.99 × 10 ⁻³ ⋅ N	0.95	0.045	0.053
C	N ≥ 300	Q = 0.896+1.5 × 10 ⁻³ ⋅ N	0.99	0.057	0.069
D	N < 300	Q = 0.55+3.6 × 10 ⁻³ ⋅ N	0.95	0.055	0.065
D	N ≥ 300	Q = 1.134+1.8 × 10 ⁻³ ⋅ N	0.99	0.049	0.058
E	N < 300	Q = 0.70+5.1 × 10 ⁻³ ⋅ N	0.96	0.072	0.087
E	N ≥ 300	Q = 1.455+2.5 × 10 ⁻³ ⋅ N	0.99	0.058	0.073

Impact of WDEM on reducing overestimation of design flow rate

While the WDEM can, in theory, be applied to any building type given appropriate sanitary appliance input data, this stage of the research focuses on its application to non-residential buildings, assessed based on occupancy and appliance water efficiency levels. For this purpose, a comparative analysis was conducted between WDEM-derived flow rates and those obtained from widely used UK design guides. Three diverse case study buildings (A, B and C) at Heriot-Watt University were selected to represent varying building scales, occupancies, and functionalities: (i) Building A - a small, two-storey office building primarily serving university staff and visitors, accommodating approximately 50 occupants; (ii) Building B - a medium-sized, three-storey mixed-use building comprising offices, lecture rooms, social areas, self-catering facilities, and a coffee shop, with a maximum occupancy of approximately 450 people; and (iii) Building C - a large, three-storey mixed-use facility incorporating offices, retail spaces, a canteen, and coffee shops, with an occupancy of up to 1200 people.

Design flow rate based on the design guides

To calculate the code-based design flow rate, two widely used UK design guides were selected: (i) British Standard BS 8558⁴⁴; and (ii) CIPHE guide.⁴³ Detailed information representing the installed sanitary appliances within each building was gathered initially through building floor plans and manufacturer manuals and then verified through site visits. During these visits, appliance quantities were confirmed, and appliance discharge rates were measured manually to ensure accuracy. From these gathered data, total loading units (LUs) were computed for each building, enabling estimation of water supply design flow rates from guides. Table 3 summarises the calculated design flow rates from both BS 8558 and the CIPHE guide for each building. The results indicate that there is no significant difference between the two guides, with the flow rates closely aligned across all three buildings.

Table 3.

Code-based water supply design flow rates calculated from two UK design guides.

Case study buildings	Code-based design flow rate (l/s)
Case study buildings	BS 8558	CIPHE
Building A	1.21	1.22
Building B	2.29	2.40
Building C	4.41	4.28

Comparison of flow rates obtained from WDEM and the design guides

Following the calculation of code-based design flow rates, WDEM-based design flow rates were determined for all the case study buildings. For this purpose, the installed appliance efficiencies in each building were assessed to assign the appropriate WDEM design equations. Most appliances in case study buildings B and C correspond primarily to efficiency group A. Building A featured a mix of appliances mainly within groups A and B, with a few appliances classified in group C. Accordingly, flow rates for buildings B and C were compared with the group A design curve, and building A with the group B design curve at the relevant occupancy levels.

For visual comparison, Figure 6 plots design flow rate (l/s) against the number of users. The triangle, square and diamond markers show average code-based flow rates for buildings A, B and C, respectively. The dotted trendline is a best-fit through these code-based points, illustrating how the conventional method scales with increasing occupancy. The two lower curves are the WDEM-derived design curves; the solid line is group A, and the dashed line is group B. The circular markers on the WDEM curves denote the exact design flow rates for the case study buildings at their corresponding user counts.

Figure 6.

Comparison of flow rates obtained from the design guides and WDEM results.

The results show that the WDEM curves generally predict lower design flow rates compared with the current design codes across all observed occupancies. The WDEM-based design flow rates are approximately 0.5 l/s for building A, 0.7 l/s for building B, and 1.2 l/s for building C, indicating design flow reductions of 63%, 72%, and 73%, respectively, for the case study buildings. These reductions demonstrate how the application of WDEM reduces total design flow rates across different building scales. In addition, the widening gap between the code-based trendline and the WDEM curves, especially at higher user counts, highlights the growing risk of oversizing when current design codes are applied to large-scale buildings. This systematic overestimation underscores the potential of WDEM as a robust and evidence-based approach for sizing water supply systems more accurately, once fully validated.

Conclusion

This study introduced WDEM, an innovative stochastic approach developed to address persistent oversizing in water supply systems for non-residential buildings. The WDEM incorporates contemporary data on sanitary appliance efficiency and user behaviour, using extensive Monte Carlo simulations to produce robust and realistic water usage scenarios. Through a comprehensive comparison of WDEM-derived flow rates with those obtained from the UK design guides in three case study buildings, this research clearly demonstrates the significant reductions in design flow rates, ranging from 63% to 73% depending on the building size. These reductions indicate the model’s effectiveness in reducing predicted design flow rates and mitigating the risk of oversizing in water supply systems in buildings. The WDEM is proposed as a framework for estimating building-specific design flow rates in non-residential buildings. Alternatively, the generalised form suggests initial design equations to estimate the design flow rate based on occupancy and the relevant appliance-efficiency rating. The fitted design equations showed strong agreement with the simulation outputs (R² = 0.95–0.99) and low errors (MAE = 0.013–0.072 l/s; RMSE = 0.016–0.087 l/s). These design equations enable more reliable estimation of building design flow rates and help reduce the risk of oversizing pipework during system sizing. Thus, WDEM supports more robust and sustainable design decisions and represents a step forward in water demand modelling, aligning system sizing with modern efficiency standards and user behaviours.

While the model performs well compared with current design guides, validation against measured flow rates from real buildings was beyond the scope of this study. Future work should therefore focus on collecting measured flow rate data and comparing the results with WDEM predictions, to allow validation before adoption. This will allow uncertainty bounds to be quantified and any differences between simulated and measured demand to be examined, supporting model refinement and improvement based on real-world operating conditions. In addition, although the model demonstrates strong potential for broader application in non-residential buildings, further work is recommended to assess its suitability and extend its applicability to other building types and usage patterns.

Footnotes

ORCID iD

Sarwar Mohammed

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Nethercote

. Theorising vertical urbanisation. City 2018; 4813: 657–684. https://doi.org/10.1080/13604813.2018.1549832

Sharif

Gormley

. Integrating the design of tall building, wastewater drainage systems into the public sewer network: a review of the current state of the art. MDPI-Water 2021; 13: 3442. https://doi.org/10.3390/w13223242, Epub ahead of print 2021.

Angel

. Urban expansion: theory, evidence and practice. Build Cities 2023; 4: 124–138. https://doi.org/10.5334/bc.348

Hobbs

Anda

Bahri

. Estimating peak water demand: literature review of current standing and research challenges. Results Eng 2019; 4: 100055. https://doi.org/10.1016/j.rineng.2019.100055

Kragh

Rose

Buhl

. The risks of oversizing domestic water piping – measurement of peak flows in multistory residential buildings. Build Serv Eng Res Technol 2025; 46: 361–374. https://doi.org/10.1177/01436244251318169

Spina

Sbaraglia

Magini

, et al. Studying a hospital distribution network with a stochastic end-uses demand model. Procedia Eng 2014; 89: 909–915. https://doi.org/10.1016/j.proeng.2014.11.524

Carlson

Boczek

Chae

, et al. Legionellosis and recent advances in technologies for legionella control in premise plumbing systems: a review. Water (Switzerland) 2020; 12: 1–22. https://doi.org/10.3390/w12030676

Mohammed

Gormley

Kelly

. Environmental impact of building drainage systems: analysis of embodied carbon emissions in terms of code-based design. Sustain Times 2025; 17: 8207. https://doi.org/10.3390/su17188207, Epub ahead of print 1 September 2025.

Hunter

. Plumbing manual. National Bureau of Standards, 1940.

10.

Wong

Mui

. Stochastic modelling of water demand by domestic washrooms in residential tower blocks. Water Environ J 2008; 22: 125–130. https://doi.org/10.1111/j.1747-6593.2007.00087.x

11.

Wong

Liu

. Demand analysis for residential water supply systems in Hong Kong. Hong Kong Inst Eng Trans 2013; 15: 24–29. https://doi.org/10.1080/1023697x.2008.10668114

12.

Omaghomi

Buchberger

Cole

, et al. Probability of water fixture use during peak hour in residential buildings. J Water Resour Plan Manag 2020; 146: 1–10. https://doi.org/10.1061/(asce)wr.1943-5452.0001207

13.

Buchberger

Omaghomi

Wolfe

, et al. Peak Water Demand Study: Probability Estimates for Efficient Fixtures in Single and Multi-family Residential Buildings, 2015, vol 68.

14.

Blokker

EJM

Vreeburg

JHG

Van Dijk

. Simulating residential water demand with a stochastic end-use model. J Water Resour Plan Manag 2010; 136: 19–26. https://doi.org/10.1061/(asce)wr.1943-5452.0000002

15.

Tindall

Pendle

. A comparison of UK domestic water services sizing methods with each other and with empirical data. Build Serv Eng Res Technol 2017; 38: 635–649. https://doi.org/10.1177/0143624417719009

16.

Oliveira

Cheng

Goncalves

, et al. Modelling of water demand in building supply systems using fuzzy logic. Build Serv Eng Res Technol 2011; 34: 145–163. https://doi.org/10.1177/0143624411429381

17.

Kavya

Mathew

Shekar

, et al. Short term water demand forecast modelling using artificial intelligence for smart water management. Sustain Cities Soc 2023; 95: 104610. https://doi.org/10.1016/j.scs.2023.104610

18.

Shu

Xia

Han

, et al. Long-term water demand forecasting using artificial intelligence models in the Tuojiang River basin, China. PLoS One 2024; 19: 1–23. https://doi.org/10.1371/journal.pone.0302558

19.

Maußner

Oberascher

Autengruber

, et al. Explainable artificial intelligence for reliable water demand forecasting to increase trust in predictions. Water Res 2025; 268: 122779. https://doi.org/10.1016/j.watres.2024.122779, Epub ahead of print 2025.

20.

Fernandez

Jack

Dass

. Data-driven method for estimating fixture use probability in simultaneous peak water flow calculations. Build Serv Eng Res Technol. 2025; 0(0): 1–18. https://doi.org/10.1177/01436244251383647

21.

Jack

. An assessment of the validity of the loading units method for the sizing of domestic hot and cold water services, 2017. https://www.cibse.org/knowledge-research/knowledge-portal/assessing-the-validity-of-the-loading-units-method-for-sizing-domestic-water-services-2017/

22.

Buchberger

Omaghomi

Wolfe

, et al. Peak water demand study, 2017. https://www.iapmo.org/WEStand/Documents/PeakWaterDemandStudy-ExecutiveSummary.pdf

23.

Nelson

. Stochastic modeling. McGraw-Hill, 1995.

24.

Hartig

Calabrese

Reineking

, et al. Statistical inference for stochastic simulation models - theory and application. Ecol Lett 2011; 14: 816–827. https://doi.org/10.1111/j.1461-0248.2011.01640.x

25.

Buchberger

Blokker

Cole

. Estimating peak water demands in hydraulic systems I - current practice. In: WDSA 2012: 14th Water Distribution Systems Analysis Conference, 2012 in , p. 1165.

26.

Omaghomi

Buchberger

. Estimating water demands in buildings. Procedia Eng 2014; 89: 1013–1022. https://doi.org/10.1016/j.proeng.2014.11.219

27.

Creaco

Blokker

Buchberger

. Models for generating household water demand pulses: literature review and comparison. J Water Resour Plan Manag 2017; 143: 04017013. https://doi.org/10.1061/(asce)wr.1943-5452.0000763

28.

Omaghomi

Buchberger

Wolfe

, et al. Peak water demand study, 2024. https://iapmostandards.org/

29.

IAPMO ANSI . Uniform plumbing code: Appendix M. International Association of Plumbing and Mechanical Officials (IAPMO), 2024. https://epubs.iapmo.org/2024/UPC/?utm.

30.

Agudelo-vera

Pieterse-Quirijns

Scheffer

, et al. New method to design domestic water systems. REHVA J. 2013; 06: 12–16.

31.

Blokker

Agudelo-Vera

Moerman

, et al. Review of applications for SIMDEUM, a stochastic drinking water demand model with a small temporal and spatial scale. Drink Water Eng Sci 2017; 10: 1–12. https://doi.org/10.5194/dwes-10-1-2017

32.

Blokker

EJM

Vreeburg

JHG

Pieterse-Quirijns

, et al. Simulating non-residential water demand with a stochastic end-use model. J Water Resour Plan Manag 2011; 137: 511–520. https://doi.org/10.1061/(asce)wr.1943-5452.0000146

33.

Gargano

Di Palma

de Marinis

, et al. A stochastic approach for the water demand of residential end users. Urban Water J 2016; 13: 569–582. https://doi.org/10.1080/1573062x.2015.1011666

34.

Piazza

Sambito

Freni

. Application of a stochastic model for water demand assessment under water scarcity and intermittent networks. Eng Proc 2024; 69: 36. https://doi.org/10.3390/engproc2024069036, Epub ahead of print 2024.

35.

Rathnayaka

Malano

Arora

, et al. Prediction of urban residential end-use water demands by integrating known and unknown water demand drivers at multiple scales I: model development. Resour Conserv Recycl 2017; 117: 85–92. https://doi.org/10.1016/j.resconrec.2016.11.014

36.

Ferreira

Tde V

Goncalves

. Stochastic simulation model of water demand in residential buildings. Build Serv Eng Res Technol 2020; 41: 544–560. https://doi.org/10.1177/0143624419896248

37.

Cortez-Lara

Hernández Gress

Ballinas-Gonzalez

, et al. A methodology to simulate peak water demand for water supply systems using semi-direct methods: a case study for a residential building in Mexico. Heliyon 2024; 10: e25104. https://doi.org/10.1016/j.heliyon.2024.e25104, Epub ahead of print 2024.

38.

Thomopoulos

. Essentials of Monte Carlo simulation: statistical methods for building simulation models. Springer, 2013.

39.

Josey

Gong

. The impact of building occupancy on water demand characteristics in residential buildings: a sensitivity analysis. Eng Proc 2024; 69: 119. https://doi.org/10.3390/engproc2024069119, Epub ahead of print 2024.

40.

BS 6465-1:2006+A1 . Sanitary installations-part 1: code of practice for the design of sanitary facilities and scales of provision of sanitary and associated appliances. BSI, 2009.

41.

Kelly

David

. Labelling and water conservation: A European perspective on a global challenge. Building Services Engineering Research & Technology. 2015; 36(6): 643–657. https://doi.org/10.1177/0143624415571758

42.

The Water Label. European water label. https://uwla.eu/ (2025, accessed 5 September 2025).

43.

The Institute of Plumbing . Plumbing engineering services design guide. The Institute of Plumbing (IoP) was renamed as the Chartered Institute of Plumbing and Heating Engineering (CIPHE) in 2008, 2002.

44.

BS 8558 . Guide to the design, installation, testing and maintenance of services supplying water for domestic use within buildings and their curtilages - complementary guidance to BS EN 806. BSI, 2015.

Optimising water demand estimation in non-residential buildings: A new design flow rate approach

Abstract

Practical application

Keywords

Introduction

Statistical models for estimating design flow rate

Water Demand Estimation Model (WDEM)

Number of users and provision of sanitary appliances

Generating appliance-use events

Model parameters and assumptions

Resultant probability distribution and cumulative distribution function CDF

Regression analysis and establishing design equations

Impact of WDEM on reducing overestimation of design flow rate

Design flow rate based on the design guides

Comparison of flow rates obtained from WDEM and the design guides

Conclusion

Footnotes

ORCID iD

Funding

Declaration of conflicting interests

References