# Operation and maintenance costs of drinking water plants

## Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Terence Nelligan, Environment Accounts and Statistics Division, Sharon Wirth, Business Survey Methods Division, Cindy De Cuypere, Environment Accounts and Statistics Division and Lenka Mach, Social Survey Methods Division

Drinking water plants are a crucial element of the Canadian economy as
they treat water that is supplied to commercial, institutional, industrial
and residential users. In 2007, 86% of households received their
water from a municipal water source. ^{14} That year drinking water plants produced 5,617 million
cubic metres (Mm^{3}) of potable water. Surface water supplies provided 88%
of the water and served 24 million Canadians. Groundwater supplies
provided 10% of the water and served 3.5 million Canadians.
The remaining 2% came from groundwater under the direct influence of
surface water (GUDI sources) which served 460,000 people in 2007. ^{15}

Operation and maintenance (O&M) costs data collected by the 2007 Survey
of Drinking Water Plants include expenditures on materials (chemicals and
replacement parts), labour and energy for the acquisition and treatment of
water, but exclude water distribution costs. In 2007, $807 million
was spent on O&M. The largest component of these expenses was labour costs
($302 million), while material and energy costs represented $198 million
and $199 million respectively. Other costs accounted for the remaining $108 million. ^{16}

O&M costs associated with the acquisition and treatment of water will vary by treatment technology, annual production volumes and the type of source water processed. This study analyzes the data collected by the survey to illustrate how O&M costs vary by these factors. It examines two drinking water treatment systems: conventional systems, which treated the most surface water and unfiltered systems, which treated the most groundwater.

### What you should know about this study

Data sources

The primary data source for this article was the 2007 Survey of Drinking Water Plants, a new survey conducted by Statistics Canada. The survey provides national and regional information related to the production of drinking water. With a target population of all drinking water plants serving communities of 300 or more people, it collects data on the volumes of water withdrawn and treated, treatment type, capital and operating costs and water quality. The treatment plants covered by the survey provided water for about 85% of the Canadian population. For further information on data quality, concepts and methodology please refer to: Survey of Drinking Water Plants (survey no. 5149).

### Conventional treatment of surface water

Surface waters such as lakes, rivers and streams are susceptible to pollution
and almost always require treatment by coagulation, flocculation, sedimentation,
granular media filtration and disinfection. This combination of processes
is referred to as 'conventional' treatment. ^{17} Table 2 shows that in 2007,
over half (53.5%) of the total volume of surface water treated at drinking
water plants (2,640 Mm^{3}) was treated by conventional systems.
This water served 13 million people. ^{18} Because of their dominance
in the treatment of surface water, the analysis in this article focuses on
conventional treatment plants. Survey data were used to develop a model explaining
how O&M costs vary with annual production volumes for conventional treatment
plants treating surface water.

### Unfiltered systems treating groundwater

With respect to groundwater supplies, in most cases the ground can protect
the water from microbial contamination. Treatment approaches therefore differ
from those applied to surface water supplies. ^{19} Table 2 shows that in 2007,
about one third (34%) of the total volume of water treated from groundwater
supplies (232 Mm^{3}) was produced by systems that applied disinfection
as well as other processes. These are referred to as unfiltered systems treating
groundwater. ^{20} This water
served one million people. ^{21} Because these systems were among the dominant ones
in the treatment of groundwater, the analysis in this article focuses on unfiltered
systems. Survey data were used to develop a model to explain how O&M costs
vary with annual production volumes for unfiltered systems treating groundwater.

### O&M cost models developed for selected systems

For conventional plants treating surface water, ordinary least squares regression was used to produce a model to estimate annual O&M costs based on the volume of water treated. The following regression model was estimated:

Annual O&M costs (CAN$) = exp(9.18727 + 0.55085*ln(annual production volume in megalitres))

This model was developed using data from 274 treatment plants,
of which 22 had imputed data for either total O&M costs or annual
production volume. The coefficient of determination (R^{2}) for the
model was 0.73, meaning that the volume of water treated explains 73%
of the variability in total O&M costs.

For unfiltered systems treating groundwater, the following regression model was estimated using a similar procedure:

Annual O&M costs (CAN$) = exp(7.77325 + 0.71328*ln(annual production volume in megalitres))

This model was developed using data from 126 plants, of which 18 had
imputed data for either total O&M costs or annual production volume. The
R^{2} for this model was 0.65, meaning that the volume of water
treated explains 65% of the variability in total O&M costs.

### Note to data users regarding the regression models and confidence bands

The variable of interest in this study, total annual O&M costs, is related to the independent variable, annual production volume in megalitres; however, this relationship does not satisfy the assumptions needed for fitting a linear regression model. Thus transformations were applied to the data and it was determined that ln(total annual O&M costs) is linearly related to ln(annual production volume) and this relationship satisfies the assumptions for using ordinary least squares estimates of the regression parameters. Regression models were then developed to show the estimated average annual total O&M costs for a given annual production volume.

The confidence bands shown in the charts were obtained using the output of SAS PROC REG and are the Working-Hotelling 95% confidence bands for the true regression line relating the expected O&M costs of all plants to the volume of water they treat annually. In order to preserve confidentiality, the estimated average annual O&M costs and the confidence bands extend beyond the minimum and maximum observed values for treated volume used to develop the model. The data presented in the charts result from application of the models to arbitrary values of annual water volumes in order to illustrate the relationship between annual production volumes and annual O&M costs.

When the models are applied to estimate ln(cost) for a given ln(volume), the estimation is more accurate the closer the ln(volume) is to the mean ln(volume) treated by the plants used in the development of the model. The models are more accurate in estimating O&M costs for small volumes and less accurate for estimating costs for large volumes.

The Survey of Drinking Water Plants was a census of plants serving at least 300 people. Only the data from plants that responded were used in this study. The plants that responded to the survey were assumed to have been selected by stratified simple random sampling. The model was estimated using both a design-based approach (SUDAAN procedure REGRESS) and a model-based approach (SAS procedure REG with a WEIGHT statement), with nearly identical results.

### Model results for conventional plants treating surface water

Chart 5 illustrates the model estimates for O&M costs associated with a range of annual production volumes for conventional plants treating surface water. The chart shows that the estimated average annual O&M cost for plants producing 100 megalitres (ML) is about $123,000. The lower and upper confidence bands show that the true average annual O&M cost for all plants producing 100 ML is between $106,000 and $143,000 (with 95% confidence). For plants producing higher volumes, the confidence bands for the annual O&M cost widen. For example, the estimated average annual O&M cost for plants producing 50,000 ML is $3.79 million, with the true figure falling between $3.13 and $4.59 million (with 95% confidence).

Using the model, Chart 6 shows the estimated relationship between average O&M costs per ML and total annual production volume for conventional plants treating surface water. The chart shows that the estimated average annual O&M cost per ML for plants producing 100 ML is $1,235/ML, with the true value falling between $1,063/ML and $1,436/ML (with 95% confidence). The cost per ML declines when more water is produced. The estimated average annual O&M cost per ML for plants producing 50,000 ML is just $76/ML.

In 2007, the average conventional plant in Canada produced 5,706 ML
(15.63 ML per day) and served about 26,000 people. ^{22} The model estimates
O&M costs for this average plant to be $204/ML. This figure increases
to $268/ML for plants producing annual volumes half of the average and
falls to $147/ML for plants producing annual volumes twice the average.

### Model results for unfiltered systems treating groundwater

Chart 7 illustrates results of the model applied to a range of annual production volumes for unfiltered systems treating groundwater. The chart shows that the estimated average annual O&M cost for plants producing 100 ML is $63,000, with the true value falling between $52,000 and $76,000 (with 95% confidence).

Chart 8 shows the estimated relationship between O&M costs per ML and total annual production volume for unfiltered systems treating groundwater. The estimated average annual O&M cost per ML for plants producing 100 ML is $635/ML, with the true value falling between $527/ML and $764/ML (with 95% confidence).

In 2007, the average unfiltered system treating groundwater in Canada
produced 817 ML (2.24 ML per day) and served about 5,000 people. ^{23} The model estimates
O&M costs for this average plant to be $350/ML, a figure which increases
to $426/ML for plants producing annual volumes half of the average and
falls to $287/ML for plants producing annual volumes two times the average.

### Annual O&M costs to treat surface water versus groundwater for the selected systems

Chart 9 shows the upper and lower confidence bands for the two selected systems studied.

The models suggest that O&M costs per ML for treating surface water are higher than those treating groundwater for annual production volumes lower than 900 ML, the point where the upper limit for groundwater meets the lower limit for surface water (with 95% confidence).

Application of the regression models shows how O&M costs vary by treatment
process applied and source water type for various levels of annual production
volumes. The model results provide context for the survey data, which indicate
that the average O&M cost for all conventional plants was $161/ML. ^{24} The model described in this article shows that O&M
costs for conventional plants treating surface water can vary significantly
from this average, moving from $1,400/ML to $35/ML for the range
of annual production volumes of 100 ML to 170,000 ML as
shown in Chart 9.

Similarly, for unfiltered systems treating groundwater, the model shows
that O&M costs can vary from $764/ML to $85/ML for the range
of annual production volumes of 100 ML to 19,000 ML. The
average O&M cost from the aggregated survey data for all unfiltered systems
was $128/ML. ^{25}