SimultaneousPeak Water Demands in Residential Areas

ReportNo WSAA 67

September 1993




Waterdistribution networks must have sufficient capacity to ensure that minimum headrequirements at consumer supply offtakes are met during the peak demand period.To achieve this, the design of networks, using computer modelling, requires aclear understanding of the water usage patterns of consumers during thisperiod.


Existingmodels used by the Water Authority of Western Australia have recognised thatthe peak water usages by individual consumer services occur at varying times,giving a diversity effect, but have assumed that the distribution of demand ateach service can be described statistically as a normal distribution.


Theaim of this study was to assess diversity effects during times of peak wateruse in the Perth summer and involved the recording of instantaneous water useby a sample of 84 Perth households continuously over a period of three months.


Themajor findings of the study are:


1.       During the timeof peak water usage (typically at around 6.30 pm on a day with hightemperatures and easterly winds) a large proportion of households are stillusing little or no water.


2.       Consequently thetraditional models for water demand which assume a normal distribution areparticularly inappropriate. There is a need for a skew distribution with anextended upper tail.


3.       The Gammadistribution with a shape parameter around 0.2 to 0.3 provides a good model forthe observed demand during peaks. The model is best in the upper tail, theregion most critical in a description of the diversity effect.


4.       No evidence wasfound for correlation of demand by nearby households. Pressure and long andshort services also did not have a measurable effect on peak consumption.However the relatively small sample size (84 households) suggests that thisshould not be taken as conclusive evidence that correlation can always beignored.


5.       The effect ofmoving from normal distribution models to Gamma distribution models has asimilar effect on the general shape of diversity curves to that of introducingcorrelation to the normal distribution model. Consequently it is likely thatearlier studies which used normal distribution models may have interpreteddistribution effects as correlation.


In addition the study developed methodologies for theoptimal collection and analysis of instantaneous demand data.


Copiesof the Report are available from WSAA, price $A30. Orders may be placed throughthe Bookshop at or by email to