decreasing climate-induced water supply risk through improved

63
Sectoral Applications Research Program Climate Program Office Oceanic and Atmospheric Research National Oceanic and Atmospheric Administration American Water Works Association Kearns & West George Washington University University of Colorado-Boulder Hazen and Sawyer July 2013 Decreasing Climate-Induced Water Supply Risk Through Improved Municipal Water Demand Forecasting

Upload: lydat

Post on 14-Dec-2016

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Decreasing Climate-Induced Water Supply Risk Through Improved

Sectoral Applications Research Program

Climate Program Office

Oceanic and Atmospheric Research

National Oceanic and Atmospheric

Administration

American Water Works Association

Kearns & West

George Washington University

University of Colorado-Boulder

Hazen and Sawyer

July 2013

Decreasing Climate-Induced Water Supply Risk Through Improved Municipal Water Demand Forecasting

Page 2: Decreasing Climate-Induced Water Supply Risk Through Improved

i

Table of Contents

Acknowledgements ...................................................................................................................................... iii

I. Executive Summary ................................................................................................................................... 1

II. Introduction and Overview ....................................................................................................................... 4

Forecasting and Uncertainty ..................................................................................................................... 5

Climate Change as a Source of Forecast Uncertainty ............................................................................... 6

III. Reasons to Examine Water Demand ....................................................................................................... 9

Background ............................................................................................................................................... 9

The Importance of Water Demand for Operations and Planning ............................................................. 9

Role of Demand Forecasting in Management and Planning ................................................................... 10

Implications of Climate-Induced Changes in Demand to Strategic Planning ......................................... 14

Summary ................................................................................................................................................. 15

IV. Project Approach and Methods............................................................................................................. 16

Project Team ........................................................................................................................................... 16

Pre-workshop interviews..................................................................................................................... 16

Surveys and Workshops in Two Regions ............................................................................................... 18

East Coast Focus in Washington, DC ................................................................................................. 18

Midwest/Western Focus in Denver, Colorado .................................................................................... 20

V. Current State of Water Demand Forecasting ......................................................................................... 22

Basics of Water Demand Forecasting ..................................................................................................... 22

Demand Forecasting Methodologies ...................................................................................................... 22

How Factors that Affect Demand Are Addressed in Models ................................................................. 24

How Uncertainty Is Addressed in Models .............................................................................................. 25

Current State of Water Demand Forecasting Models ............................................................................. 26

Summary ................................................................................................................................................. 28

VI. Risks Associated with Models and Methods ........................................................................................ 29

Page 3: Decreasing Climate-Induced Water Supply Risk Through Improved

ii

Limitations of Existing Models .............................................................................................................. 29

One Potential Approach to Identify Risks—Extreme Value Analysis ................................................... 30

Summary ................................................................................................................................................. 31

VII. What Utilities Should Be Doing Now ................................................................................................. 32

Collect Additional Weather and Demand Data ....................................................................................... 35

Analyze the Data and Translate It into Actionable Information ............................................................. 38

Evaluate Potential Changes in Demand .................................................................................................. 39

Evaluate Potential Changes in Demographics in the Service Area ......................................................... 43

Understand and Incorporate Uncertainty into Forecasting ..................................................................... 44

Plan for Drought So the System Can Cope ............................................................................................. 47

Summary ................................................................................................................................................. 47

VIII. Recommendations for Future Research ............................................................................................. 49

Understanding Baseline Conditions and Potential Changes ................................................................... 49

Potential Impacts of Demand on Appropriate System Design ................................................................ 50

System Data ............................................................................................................................................ 51

System Revenues .................................................................................................................................... 52

Data and Research Integration ................................................................................................................ 52

Historical Drought/Water Shortage Analyses ......................................................................................... 52

Value of Information Studies .................................................................................................................. 53

Social Science Research ......................................................................................................................... 53

Tools for Investment Decisions .............................................................................................................. 53

Summary ................................................................................................................................................. 54

IX. Summary and Conclusions ................................................................................................................... 55

X. References .............................................................................................................................................. 57

Page 4: Decreasing Climate-Induced Water Supply Risk Through Improved

iii

Acknowledgements

We would like to acknowledge NOAA for providing the funding for this project and we would

like to thank Nancy Beller-Simms from NOAA for her support. We would also like to thank the

participants in the two project workshops and the five webinars for their time and expertise:

Washington, DC, Workshop Denver, CO, Workshop

Alison Adams, Tampa Bay Water Sarah Deslauriers, Carollo

Veronica Blette, EPA Ben Dziegelewski, University of Illinois

Erica Brown, AMWA Rick Holmes, SNWA

Jim Chelius, American Water Pam Kenel, Black & Veatch

Roger Cooke, Resources for the Future Alfredo Rodriguez, Aurora Water

Pat Davis, OWASA Sean Senascall, Tacoma Water

Bill Davis, CDMSmith Lorna Stickel, Portland Water Bureau

Ron Harris, Newport News Waterworks David Yates, NOAA-NCAR

Rick Palmer, University of Massachusetts

Paul Peterson, Arcadis

Tom Rockaway, University of Louisville

Thurlough Smyth, New York City DEP

Roland Steiner, WSSC

Jennifer Warner, Water Research Foundation

Doug Yoder, Miami Dade Water and Sewer

Project Team

Alan Roberson and Craig Aubuchon, American Water Works Association

Abby Arnold, Elana Kimbrell, and Dani Ravich, Kearns & West

Emmanuel Donkor, Refik Soyer, and Tom Mazzuchi, GWU

Erik Haagenson, Balaji Rajagopalan, and Scott Summers, CU

Jack Kiefer, Hazen and Sawyer

Page 5: Decreasing Climate-Induced Water Supply Risk Through Improved

1

I. Executive Summary

Water demand forecasts are critical tools for water system managers

and planners. Water system managers (and their planning and

engineering staff) have to contend with many uncertainties in

planning, designing, and operating a water system to meet

customers’ demands. Water demand and the resultant water sales

generate the revenues that are the economic engine for any water

system, whether large or small, urban or rural. Since accurate water

demand forecasting is inextricably linked to a water system’s

finances and the system’s long-term sustainability, the financial

implications of water demand forecasting to a system are significant.

Accurate water demand forecasting depends on a variety of factors. Weather conditions,

economic/business cycles, and new connections or the loss of a large industrial/commercial

customer can affect short-term demands that impact day-to-day system operations. Over longer

terms, other factors can influence demand, including trends in population, housing, density of

land use, employment, mix of industries, water efficiency and conservation programs, and

climate change and variability.

The focus of this project was to develop a better understanding of the potential risk posed by

climate change and variability to demand forecasting for a water system; and then, to develop

recommendations to help reduce climate-induced risk arising from inaccurate forecasts. What

has been realized, however, is that long-term weather trends, caused by climate change and/or

weather cycles, are only part of the picture, and water system planners, engineers, management,

and water boards must invest adequate time and resources to understand all the factors that

influence water demand forecasts, and the interplay between them. Moreover, the

interrelationship between the different factors is system-specific; therefore, stakeholders of each

water system need to develop a better understanding of the appropriate factors for their system.

Although there is no simple answer and no clear path to reducing climate-induced risk in

developing demand forecasts for water systems, this project yielded several recommendations

Page 6: Decreasing Climate-Induced Water Supply Risk Through Improved

2

that water systems can implement now in order to develop a better understanding of climate-

induced risk in water demand forecasting. Many of these recommendations involve improved

collection and analysis of typical water system demand data. Some of the recommendations are

time and resource intensive while other are less so. The recommendations fall into six general

categories that need further investigation by water system managers to determine how to

appropriately implement them in their systems, noting that many system-specific factors would

impact implementation:

1. Collect additional weather and demand data.

2. Analyze the data and translate it into actionable information.

3. Evaluate potential changes in demand.

4. Evaluate potential changes in demographics in the service area.

5. Understand and incorporate uncertainty into forecasting.

6. Plan for drought so that the system can cope with it.

A water system manager faces many competing priorities in operating and managing a water

system, including managing the system finances and optimizing future capital investments. The

above recommendations offer a starting point for the water system manager when considering

the investment of time and resources necessary to improve and optimize a system’s long-term

water demand forecast.

Many utilities today still develop demand forecasts using the simple product of estimated per-

capita demand and a projection of population—known as the ―per-capita‖ or ―gpcd‖ method.

However, increasingly complex constraints in source water availability, variability in water sales

and revenues, and concerns about climate change and other emerging uncertainties, have led to

more emphasis on evaluating, understanding, and modeling the factors that influence water use

over both short-term and long-term intervals. In particular, traditional per-capita approaches to

forecasting water demand neglect—and are incapable of—measuring the effects of principal

factors that can produce variability in water use, such as weather and climate, the price of water,

land use, and several socioeconomic variables other than population. Additionally, the observed

reductions in per-capita use, for example, due to increases in water efficiency, the effects of

pricing, and recessionary economic pressures have been largely unanticipated by some systems.

Page 7: Decreasing Climate-Induced Water Supply Risk Through Improved

3

While few could predict the impacts of the latest recession, reliance on simple forecasting

methods is partly to blame for some systems not being adequately prepared for decreases in

water sales and revenues.

Long-term water demand forecasts have traditionally

assumed long-term normal weather (or stationary

mean) patterns for future forecast scenarios. However,

if ―stationarity is dead,‖ (i.e., past weather patterns

may not be the same in the future), then climate change

may be particularly problematic for long-term water

demand forecasts (Milly 2008). New and ―drifting‖

climate regimes may lead to changes in outdoor

watering patterns that would ultimately impact water demand, as well as other structural shifts in

water use. Along the way, greater variability in precipitation and temperature in the future may

also increase demand uncertainty over short-term planning horizons.

Improving forecasts is critical to optimizing future investments, as most water systems are faced

with a myriad of investment decisions, ranging from finding new water sources to supply a

growing population and economic growth; to additional treatment to comply with new

regulations; to rehabilitating and/or replacing distribution system pipes that have reached the end

of their useful life. Water system managers need to make the investment in developing a better

understanding of the potential risk posed by climate change and variability in order to improve

demand forecasting for their systems so that future capital investments will be optimized.

“If ‘stationarity is dead,’

then climate change may

be particularly

problematic for long-term

water demand forecasts”

Page 8: Decreasing Climate-Induced Water Supply Risk Through Improved

4

II. Introduction and Overview

The basic mission of a water utility is to meet the water demands of a community. Water demand

and water sales are the primary source of a utility’s revenue. As such, water demand and the

resultant revenues serve as the economic engine for a water system. Expectations, or forecasts, of

water demand therefore play a crucial role in a system’s financing, as well as its short- and long-

term operations. But a water utility’s role has a much greater impact on the community it serves

in that the benefits of water service are fundamental for the community’s well-being and its long-

term viability. Those benefits include protection of public health and safety, quality of life, and

economic viability. In short, water demand forecasting also affects a community’s long-term

viability, so it must be done accurately.

Every year, a water system planner predicts how much water its customers will buy (water sales)

and subsequently predicts the resultant revenues as part of the annual budgeting process. The

financial implications of this prediction are felt every year and a system planner finds out

throughout the year the accuracy of the predictions. Lower-than-expected water sales result in

lower-than-expected revenues and vice versa. Many water systems are seeing declining water

sales and revenues in the face of rising operation and maintenance costs, especially for

infrastructure repair and replacement (AWE 2012). Persistent errors in short-term forecasts and

revenue instabilities can result in the need to make frequent changes in rates and can also affect

bond ratings and a utility’s ability/cost to borrow.

Forecasting water demands over medium and longer terms is critical for developing and funding

Capital Improvement Plans (CIPs) and for planning future investments in system infrastructure.

Errors and inaccuracies in long-term forecasts can result in large costs to utilities and rate payers,

in the form of investments in stranded capital assets, insufficient supply reliability, or a reduced

level of service due to supply and/or system capacity limitations. In some cases, state regulatory

agencies could end up making decisions on future supply potentials that could significantly

impact a system’s supply portfolio. In some private water rights cases, exactly who ends up

holding the ―empty bucket‖ is only likely to be resolved in court. These types of problems can

take considerable time and resources to resolve.

Page 9: Decreasing Climate-Induced Water Supply Risk Through Improved

5

Because of these implications, system managers should be thoughtful about evaluating demand

and invest more time and resources for the development of additional technical capacity (both

internal and external) for developing more accurate demand forecasts for both the short-term and

the long-term.

Forecasting and Uncertainty

Forecasts of water demand depend on a number of

factors that are assumed to influence water use. If a

forecast is based on a mathematical model, this means

that a predicted value of demand is some function of

predicted future values of important variables, such as

population or job creation.

Forecasts can be inaccurate for a number of reasons.

The two most fundamental reasons involve errors in

assumptions of the forecasting model and errors in predicting the future values of variables

contained in the forecasting model. For example, if future water demand is assumed to be a

function solely of population, but other factors influence water use, then even a perfect

prediction of future population will not result in an accurate forecast. On the other hand, if it

turns out that demand really does depend only on population, then an imperfect prediction of

future population will also lead to inaccuracies in forecasting water use.

In reality, both of these sources of uncertainty are likely to occur, which typically makes perfect

forecast accuracy unattainable (except by chance). Making predictions about the future always

involves uncertainties—the key is identifying and understanding the sources of those

uncertainties. This amounts to learning more about why water demands vary, making

improvements in how forecasting models are designed, and identifying the best ways to portray

forecasts that are inherently uncertain.

“Making predictions about

the future always involves

uncertainties—the key is

identifying and

understanding the sources

of those uncertainties.”

Page 10: Decreasing Climate-Induced Water Supply Risk Through Improved

6

Climate Change as a Source of Forecast Uncertainty

Historically, several factors and trends have made forecasting water demands difficult: increased

indoor plumbing efficiency; economic boom and bust cycles; fluctuations in prices for land and

housing; new and emerging industries; evolving water-use attitudes; and normal day-to-day and

month-to-month variability in weather. Looking forward, another complicating factor will likely

be climate change and/or weather extremes.

One only needs to review the weather statistics for 2012 to see examples of climate change and

extreme weather patterns. The year 2012 was the warmest of any year in the 1895-2012 period of

record for the United States (NCDC 2012). Every state had above-average annual temperatures

and 19 states had record warm annual averages. Much of the United States was drier than

average for 2012. The area of drought in the United States during 2012 roughly equaled the

drought of the 1950s. The drought peaked in July, when according to the Palmer Drought

Severity Index (PDSI), 61.8 percent of the United States was in moderate drought.

The 2013 Draft Climate Assessment Report from the National Climate

Assessment and Development Advisory Committee (NCADAC) provides

some insight into future weather conditions (NCADAC 2013a). In this

report, the NCADAC concludes that U.S. temperatures will continue to

rise, with an increase of 2o – 4

oF predicted for most areas. The report also

concludes that the chances of record-breaking high temperature extremes

will continue to increase as the climate continues to change.

For water resources and precipitation, the NCADAC concluded that precipitation and runoff

increases have been observed in the Midwest and the Northeast and are predicted to continue or

develop in the northern states (NCADAC 2013b). Parallel decreases have been observed and are

projected to continue in the southern states. Droughts are predicted to intensify in most of the

United States, with long-term reductions in water resources in the Southwest, Southeast, and

Hawaii in response to both rising temperatures and changes in precipitation.

Page 11: Decreasing Climate-Induced Water Supply Risk Through Improved

7

Changes in both average and extreme temperatures will impact demand forecasting. Long-term

water demand forecasts have traditionally assumed long-term normal weather (or stationary

mean) patterns for future forecast scenarios. If, in fact, ―stationarity is dead‖ as previously

mentioned, then climate change may be particularly problematic for long-term water demand

forecasts. New and ―drifting‖ climate regimes may lead to long-term changes in outdoor

watering patterns that would ultimately impact water demand, as well as other structural shifts in

water use. Along the way, greater variability in precipitation and temperature in the future may

also increase demand uncertainty over short-term planning horizons.

However, for water utilities, it should be noted that continued data collection and long-term

statistics are needed before a change can be demonstrated to be attributable to climate. For some

data elements such as stream flow and temperature, long-term datasets are available, but are not

available for other data elements. It should also be noted that the U.S. Geologic Survey (USGS)

is under significant budget pressures (like all federal agencies) and maintaining its existing

network of stream gages in the future will be challenging. More data collection, research, and

statistics are needed to better understand the relationship between climate and weather at a

watershed level to factor into water resource planning and water demand forecasting.

The range of possible future climate conditions and weather extremes is uncertain at this time.

However, it is prudent to evaluate potential changes to future water demand and demand

forecasts that may be attributable to differences in future climate.

Goals and Objectives

The goal of this report is to improve water demand forecasting by increasing the awareness of

water system managers and demand forecasters to the potential implications of climate change

for water demand forecasting. This report is not intended to resolve the debates surrounding

climate change and all of the potential implications of climate change for water systems. Rather,

the objectives of this research were to:

Conduct a literature search and review of the existing research on water demand

Conduct case studies using extreme value analysis on the potential impacts of climate

change to water demand at two water systems (Aurora, Colorado, and Tampa, Florida)

Identify knowledge gaps and research needs related to demand forecasting

Page 12: Decreasing Climate-Induced Water Supply Risk Through Improved

8

Develop a list of recommendations for what water system managers should be doing now

to improve their own demand forecasts

Conduct outreach to water systems on the need to improve their own demand forecasts

In order to address the above objectives, this report is organized as follows:

Section III examines the primary reasons to look at water demand (or why improving

demand forecasts is important) and addresses the implications of climate change on

water demand.

Section IV details the project approach and research methods for conducting outreach

on demand issues to water system managers and experts.

Section V presents a review of the current state of water demand forecasting, existing

research, and example approaches.

Section VI provides results of the Aurora and case study, and examines risks

associated with model methods and climate change.

Section VII summarizes what systems are doing now and what they can be doing

better to improve their own demand models.

Section VIII provides recommendations for future research.

Section IX provides a summary and conclusions for this research.

Section X lists the references used in this report.

Page 13: Decreasing Climate-Induced Water Supply Risk Through Improved

9

III. Reasons to Examine Water Demand

Background

Water system managers and planners face a myriad of competing issues (AWWA 2013), such as:

Rehabilitating or replacing infrastructure

Lack of public understanding of the value of water

Capital costs and availability

Water supply and scarcity

Aging workforce/talent attraction and retention

Regulation and government oversight

Water security and emergency preparedness

Climate risk and resiliency

Managing finances and optimizing investments are two critical priorities for all water systems.

All water system managers need to balance the short-term delivery needs of their customers with

the long-term planning necessary to build and maintain required infrastructure, including

adequate source waters and treatment and storage capacity to meet increasing future demands.

This section provides some reasoning and justification as to why system managers should be

thoughtful and deliberate in evaluating demand and why more time and resources should be

invested in order to develop more technical capacity for deriving more accurate demand

forecasts. The section also provides an overview of the importance of demand for system

operations and planning and how demand forecasting is used (or should be used) in the strategic

planning of water systems.

The Importance of Water Demand for Operations and Planning

Water demand and resultant water sales represent a water system’s economic engine. Demand

projections form the basis of several complex financial and strategic decisions. Those that link to

capital investments (noting that daily and seasonal projections are needed for water system

operations) are usually prepared for the short term (one to five years) and the long term (15+

Page 14: Decreasing Climate-Induced Water Supply Risk Through Improved

10

years), but each projection has a different use and each may be performed by a separate group or

division of water system staff.

For example, water system staff typically develop an

operating budget for the upcoming year or financial planning

period (usually five to ten years), and then this budget is

refined by management and ultimately approved by the water

system’s governing body. Inherent to the development of

these budgets is a projection of water demand that translates

into a projection of gross revenues, and, in conjunction with utility costs, projections of net

revenue requirements. Additionally, if a system has sold bonds to pay for capital improvements,

short-term demand forecasts are important for predicting the bond coverage ratio for future years

for the system. For many water systems, meeting a specific bond coverage ratio is an important

financial and strategic goal.

Meanwhile, water demand forecasts also drive long-term investment and planning strategies,

which may be derived from asset management plans, and are commonly expressed in capital

investment plans, system master plans, and/or urban water management plans. These plans

typically require a longer view on the adequacy and reliability of the water system. These plans

also detail when the next new source of water supply and/or the next increment of treatment

capacity might be needed in the future.

Role of Demand Forecasting in Management and Planning

Demand projections are one of the primary methods by which most water system managers

attempt to align short-term and long-term priorities and objectives. Several demand forecasting

methods are available, which vary in complexity and data requirements. Adopted techniques will

typically differ from utility to utility depending on a host of factors, including the adequacy of

existing supplies, the diversity of the customer base, internal technical capabilities, and

availability of data upon which to build forecast assumptions (Kiefer 2006).

Page 15: Decreasing Climate-Induced Water Supply Risk Through Improved

11

Traditionally, water demand forecasts have been prepared using relatively simple methods, such

as taking the simple product of an estimate of the per capita demand and a projection of

population—known as the ―per-capita‖ or ―gpcd‖ method. In some areas, water systems are

required to use this method for demand forecasting

(SFWMD 2012). In South Florida, the average per-

capita daily use is calculated for the last five years or

period of record. This method of calculation is

adequate for gradual decreases in per capita demand but

may not adequately account for more rapid decreases.

Additionally, increasingly complex constraints in

source water availability, financial capacity, and

concerns about climate change and other emerging uncertainties, have led to more emphasis on

evaluating, understanding, and modeling the factors that influence water use over short-term and

long-term horizons. In particular, traditional per-capita approaches to forecasting water demand

neglect—and are incapable of—measuring the effects of factors that can produce variability in

water use, such as weather and climate, the price of water, land use, and several other

socioeconomic variables other than population. In fact, observed reductions in per-capita use—

for example, due to increases in water efficiency, the effects of pricing, and recessionary

pressures—have been seen by many systems over the past 30 years (Rockaway et al. 2011). In

some cases, systems have modified their rate structures to account for declining water use;

however, in some cases, the decline was not anticipated by some systems. Reliance on simple

forecasting methods is partly to blame. Addressing factors that influence demand in the demand

forecasting process permits a means in which to evaluate and consider future demand in the

context of long-term investment and planning strategies.

Water demand is inherently difficult to forecast because water is a complex, multidimensional

commodity that operates in legal, economic, and hydrologic dimensions (Olmstead 2010).

Water is used for a variety of purposes. Some purposes are essential for public health, like

drinking, cooking, and bathing. Other uses include water for cooling, irrigation, production of

goods and services, and aesthetic purposes. Demand patterns vary significantly on a daily,

“Water demand is

inherently difficult to

forecast because water is

a complex,

multidimensional

commodity.”

Page 16: Decreasing Climate-Induced Water Supply Risk Through Improved

12

weekly, and seasonal basis. Water treatment and distribution systems must be designed to meet

peak demands during the height of summer irrigation season (for most of the United States),

while having adequate sources to meet average annual demands. Factors that can affect short-

term demands and demand forecasts include:

Weather conditions and extremes

Restrictions on outdoor use due to water shortage

Economic/business cycles (such as recessions)

New connections or the loss of a particular customer (particularly a commercial or

industrial customer that is a large water user, for example, a factory shuts down)

Over longer time horizons, several factors can influence demand, including trends in:

Population

Housing and housing mix (e.g., single-family detached homes versus multifamily

developments)

Density of land use and lot sizes

Employment and mix of industries

Disposable incomes and economic output

Price of water and sewer service

Water efficiency and conservation

programs

Re-use of treated wastewater

Climate change and variability

Climate change represents a relatively new source of uncertainty in the planning process for

many water systems. To date, the understanding of the potential impacts of climate change on

source waters and watersheds, public health, and infrastructure investments is evolving (Means

et al. 2010, AWWA Climate Change Committee 2011). These analyses have generally

considered the impacts of climate change on water supply and show that climate change is

emerging as an additional consideration in the planning and design of water infrastructure.

However, from a planning perspective, water supply and water demand represent the two sides

“Climate change

represents a relatively

new source of uncertainty

in the planning process for

many water systems.”

Page 17: Decreasing Climate-Induced Water Supply Risk Through Improved

13

of the water budget and changes in either demand or supply influence the strategic planning for

the other.

One of the difficulties in incorporating climate change into demand forecasts is a clear

understanding of time frames in forecasting and the distinction between weather variability and

climate change. Put succinctly, ―Climate is what you expect, weather is what you get‖ (Miller

and Yates 2006). Water managers have historically dealt with weather variability in short-term

operations through the use of a safety factor or supply buffer to deal with potential drought or

increased demand.

However, over the long-term, excess capacity may not be available to provide sufficient buffers,

or stranded capacity may lead to financial challenges in meeting debt payments and the ability to

maintain high bond ratings. It is no surprise then, that a recent study on the impacts of climate

change on infrastructure planning and design found that ―most [water systems] anticipate

needing to make changes to their demand forecasting modeling‖ (Means et al. 2010). Figure 1

conceptualizes a few of the climate-induced changes in demand along the two dimensions of

time and water system orientation.

Figure 1. Climate-Induced Changes in Demand. Source: AWWA

Short Term Long Term

Inter

(within

watershed)

Intra

(within water

system)

Water Rights and

Demand:

Minimum In-Stream

Flows

Agriculture Industry

Power Generation

Seasonal Peak

(summer irrigation)

Drought Restrictions

Land Use Patterns:

Residential

Industrial

Environmental

Source Water Supply:

Quality

Quantity

End Use Technologies

Demographics/

Population

Page 18: Decreasing Climate-Induced Water Supply Risk Through Improved

14

Figure 2, reprinted from the Water Services Association of Australia (WSAA), illustrates several

of the direct and indirect factors that influence water demand and demand forecasting.

Figure 2. Factors Influencing Water Demand. Source: Original Figure from Water Services

Association of Australia Occasional Paper No. 9 – Urban Water Demand Forecasting and

Demand Management.

Implications of Climate-Induced Changes in Demand to Strategic Planning

Climate change will likely exacerbate existing pressures and situations of supply and/or demand

stress (Kiefer et al. 2013 forthcoming). For example, regions that are expected to receive less

precipitation and experience warmer temperatures could see a lengthening of the irrigation

season and higher summer peaking factors. Coupled with population growth, some urban water

systems may experience more frequent regional conflicts involving competing demands from

agriculture, power production, and in-stream uses of water. Anticipating these possible changes

will be important for designing long-term adaptive strategies. Thus, demand forecasts and the

informational characteristics of demand models will become even more important from a

strategic perspective.

The Water Utility Climate Alliance (WUCA), a committee of ten large water systems, published

a white paper in 2010 on Decision Support Planning Methods (DSPMs) for water systems and

how to incorporate climate change uncertainties into long-term planning (WUCA 2010). This

Page 19: Decreasing Climate-Induced Water Supply Risk Through Improved

15

white paper presents five distinct DSPMs, with a special

emphasis on the availability and familiarity of traditional

scenario planning. Scenario planning allows a water

system manager to better understand the risk and

exposure of potential investment decisions to different

scenarios based on changes in climate, demand, or

business conditions. By incorporating climate change

into demand projections, and allowing for both micro

and macro dynamic feedback from climate change on

land use, population growth, and supply availability, water managers can more accurately

develop potential scenarios that model the potential variability in demand. For example, one of

the earliest demand forecasts to include both conservation and climate scenarios found that in the

Washington, DC, metro area, future climate-induced demand could likely be offset by

conservation programs including regulatory policies for appliance efficiency and appropriate

pricing signals (Boland 1997). By more explicitly dealing with future uncertainties, this type of

scenario planning will allow water systems to more appropriately justify infrastructure

investments, implement adaptation practices, and generate political support when discussing rate

and cost of service studies.

Summary

Water system managers and staff need to understand both their current demand and their future

demand predictions. All water systems need to balance the short-term delivery needs of their

customers with the long-term planning necessary to build, finance, and maintain the required

infrastructure, ranging from source waters, treatment plants, transmission mains, distribution

system pipes, storage tanks, booster pumping stations, and any other needs. Accurate water

demand forecasts are critical for short-term and long-term service and financial sustainability of

any water system.

“Accurate water demand

forecasts are critical for

short-term and long-term

service and financial

sustainability of any water

system.”

Page 20: Decreasing Climate-Induced Water Supply Risk Through Improved

16

IV. Project Approach and Methods

The American Water Works Association (AWWA) was the lead organization for this project.

Established in 1881, AWWA is the oldest and largest nonprofit, scientific, and educational

association dedicated to safe and sustainable water in the world. With more than 50,000

members worldwide and 43 sections in North America, AWWA advances public health, safety,

and welfare by uniting the efforts of the entire water community.

As a member-driven association, AWWA drew on the expertise of its members to provide a

diversity of perspectives on water demand forecasting and the potential implications of climate

change. The sections below describe the project team and the process used to solicit and obtain

input related to water demand and water demand forecasting.

Project Team

A multi-disciplinary team was assembled for this project. Kearns & West (K&W), a firm

specializing in stakeholder engagement, assisted in planning the two project workshops in 2011

by conducting surveys to workshop attendees to develop the priority demand/climate issues.

K&W also facilitated the two project workshops and additional follow-up webinars to continue

solicitation of expert judgment on the project report outline and drafts of the project report.

George Washington University (GWU) conducted a literature search and a review of the existing

research and recent studies on water demand. University of Colorado-Boulder (CU) conducted

two case studies using extreme value analysis on the potential impacts of climate change to water

demand at two water systems (Aurora, Colorado, and Tampa, Florida). These case studies are

intended as examples to guide other utilities in conducting similar extreme value analysis to feed

into their improved water demand forecasts.

Hazen and Sawyer, a multi-disciplinary engineering consultant, provided technical information

on water demand forecasting, and provided additional information during the editing of this

report.

Page 21: Decreasing Climate-Induced Water Supply Risk Through Improved

17

Pre-workshop interviews

AWWA engaged a broad range of stakeholders throughout the study. The stakeholder input

helped inform the study of issues to be addressed in the topic of water demand and to develop

recommendations on how to address these issues. AWWA, working with K&W, conducted

interviews with representatives from key stakeholder groups prior to the two workshops in 2011

and designed the workshops based on stakeholder input. Additionally, in order to inform the

discussion at the two workshops, AWWA and K&W conducted an electronic survey of

workshop participants. Below is a summary of the interviews, the survey results, and the

workshops.

K&W interviewed eight experts, representing consultants

(3), academics (4), and utilities (1). The experts were

interviewed on a number of topics including their area(s) of

expertise, key themes related to the current state of

modeling, suggested topics to consider during the

workshops, and recommendations for additional participants.

From the interviews, K&W found that overall, the quality of water demand forecasting models is

highly dependent on funding and data availability. Lack of data in a consistent format from

different sources and different data systems is a significant obstacle to developing a useful model

for a utility, as well as different data formats hindering the development of national-scale

models. Many existing models are not able to adequately define important variables and/or

disaggregate water use spatially or by sector (e.g., by region, family versus multifamily,

commercial versus industrial). Also, some utilities do not use mathematical or statistical models

because they do not see the need, do not have staff trained to use them, and/or data is not readily

available in a format that is easy to model.

In general, the interviews suggest there is a need for standardized data and possibly a central

portal for data access, as well as models that account for climate change (using regional weather

patterns). More modeling expertise and funding is needed, as well as the inclusion of behavioral

variables in demand forecasting, and a better understanding of industrial water use.

Page 22: Decreasing Climate-Induced Water Supply Risk Through Improved

18

The interviews indicated that increasingly, larger utilities are seeking to model socioeconomic

and demographic variables as drivers in their demand forecasts. However, the majority of

utilities, especially the smaller utilities, still focus on cost as the main driver underlying the

characteristics of their demand models and demand forecasting approaches.

Surveys and Workshops in Two Regions

K&W assisted AWWA in hosting two workshops, one with an East Coast focus and the other

with a Midwest/Western focus, to help understand respective concerns and recommendations,

and how they differ from each other.

East Coast Focus in Washington, DC

The first workshop was held on March 30, 2011, in Washington, DC. The goal of the workshop

was to initiate an interactive dialogue, and to give presentations on methodologies used in

forecasting water demand and climate change. Prior to the workshop, K&W sent a ten-question

online survey to participants, to generate thinking and help inform discussion at the workshops.

There were 14 respondents to the survey sent in advance of the East Coast meeting. Participants

in the survey and workshop were primarily from the Washington, DC, area, Florida,

Massachusetts, and New York. The following summarizes the results of the East Coast survey,

and then identifies the objectives of the East Coast workshop and some of the main discussion

points and recommendations.

Pre-Workshop Survey

Participants of the East Coast survey indicated that the main reasons people use water demand

forecasts are for water conservation, revenue forecasts, long-term supply planning, regulatory

planning, and infrastructure planning. The climate-related concerns raised were based on

potential changes in water flow, duration, and intensity. The survey results indicated concern

about extreme events, timing of floods and drought, changes in stream flow, sea-level rise, and

changing consumption patterns. Forecasts in the East generally use population demographics

combined with per-capita data, based on billing information. The results from the East Coast

survey indicated that sophisticated models are not widely used, and historical climate data is

Page 23: Decreasing Climate-Induced Water Supply Risk Through Improved

19

used minimally by the majority of respondents. The water demand data that people are most

interested in are peak daily demand and total annual demand. The survey found that most people

wanted to attend the workshops to learn new methodologies and best practices for improving

their water demand forecasting capabilities, including how to factor in climate change.

Workshop

The following topics and recommendations were discussed at the March 30, 2011, workshop in

Washington, DC:

1. Review existing research on water demand related to climate change

Presentations were given on related topics

2. Discuss current models’ strengths and weaknesses

It was suggested that models break down data by customer type, and account for

population densities

3. Identify knowledge gaps and list future research topics

Share best practices for data collection and model methodologies; standardize the data

collection process; and develop data templates

Study the effects on water demand of new water efficient fixtures; utility water use

(such as continuously running water through pipes to prevent freezing); reuse of water;

and, green building practices

Predict demographic and behavioral responses to climate change (rising sea levels)

Consider utility zoning

Include population density in models

Look to the financial and insurance industries/communities for lessons learned about

diversification and risk

Partner with energy utilities

4. Develop recommendations for how water utilities can reduce the uncertainties in water

demand forecasting

Better acknowledge and communicate the possibility of error and the confidence

intervals to decision makers

Page 24: Decreasing Climate-Induced Water Supply Risk Through Improved

20

Focus on how to incorporate climate change into a forecast; potentially compare

climate change models to both seasonal forecasts and longer term supply forecasts

Midwest/Western Focus in Denver, Colorado

The second workshop, held on July 12, 2011, in Denver, Colorado, was focused around the same

main topics as the workshop held on the East Coast and many of the same issues and

recommendations were discussed.

Pre-Workshop Survey

The Midwest/Western survey sent in advance of the July workshop had nine respondents, who

primarily reside in the intermountain west. Participants stated that their main reasons for using

water demand forecasts were for long-term and short-term planning, which encapsulate many of

the same purposes noted in the East Coast survey. In the West, complex models seem to be used

more frequently, and historical climate data more commonly incorporated, in order to

characterize and model drought cycles and decreasing stream flows. Climate concerns included

extreme events, stormwater runoff, and effects of precipitation on infrastructure. The water

demand data that people are most interested in are total annual demand, followed by peak daily

demand. The other results of the survey were largely the same, including the respondents’ goals

for the workshop.

Workshop

Additional notes from the July 12, 2011, workshop included the following:

1. Climate change models are often too complex to be useful for decision making; there is a

need to develop ―actionable science‖

2. Develop recommendations for utilities based on a ―profile type,‖ depending on their

local/regional issues, their size, funding, etc.

3. Consider alternative water sources

4. Improve tools to analyze data

5. Assess rate structures and their relationships with demand

Page 25: Decreasing Climate-Induced Water Supply Risk Through Improved

21

After the two workshops, a series of meetings were held via webinars to review the draft report

sections on the priority research plan and the recommendations for water systems. These

webinars were held in October and December of 2011, and in January, April, and May of 2012.

Through an interactive format, these webinars provided a mechanism for the workshop attendees

to provide additional input on the recommendations for what utilities should be doing now to

improve water demand forecasting and on recommendations for future research. Additional

input was also given through reviews of the draft report.

Page 26: Decreasing Climate-Induced Water Supply Risk Through Improved

22

V. Current State of Water Demand Forecasting

Basics of Water Demand Forecasting

There are a number of ways to forecast water demand, which vary in analytical rigor and

requirements for data. In the most general sense, the intent of a demand forecast is to make a

prediction of future water use. However, the actual dimensions of the problem can be numerous

and more complex depending on considerations related to agent or purpose specificity, temporal

scale, and spatial extent (Kiefer et al. 2013). In other words, how water use is defined (e.g., total

use in a service area, water use of particular user types or sectors or geographic areas, annual,

monthly, or seasonal demand) will influence the choice among alternative demand modeling and

forecasting methods.

At a fundamental level, a demand forecast represents a set of calculations, which defines a

formula and embodies a set of assumptions. In short hand, one can generalize the set of

calculations symbolically as a function, Q=f (X), where Q is the measure of water demand to be

forecasted, X represents a set of factors that are part of the calculation and thus influence the

forecast, and the term f(*) defines mathematically how X relates to Q. Therefore, future or

forecasted values of Q are a function and conditioned on future or forecasted values of X.

Unfortunately, and in the case of most situations that involve human preferences and choices,

neither the ―true‖ nature of the dependencies on X are seldom known with certainty nor is the

proper definition of X. Even if one is skilled or lucky enough to have f(*) defined properly, then

one must have confidence in forecasts of X to have confidence in the forecast of Q. Limitations

on available data, less than perfect knowledge of underlying relationships, and inherent

uncertainty about the future make water demand forecasting both an art and a science.

Demand Forecasting Methodologies

Based on reviews of contemporary demand forecasting approaches found in Kiefer (2006) and

Billings and Jones (2008), one may classify demand forecasting into several different categories.

The Aggregate Per Capita Approach is a traditional approach to water demand forecasting that

relies exclusively on population projections. The aggregate per capita approach assumes a fixed

Page 27: Decreasing Climate-Induced Water Supply Risk Through Improved

23

Q = f(X)

value of water use per person (per-capita consumption) and multiplies this value by population to

calculate a forecast. (Using the formula previously discussed, the definition of X includes

population, an estimate of water use per person, and f(*) represents a simple multiplication of

these terms.)

Other Fixed Unit Use Coefficient methods define other fixed water use factors and drivers of

demand (other than population) to prepare a forecast. Examples of these methods include the use

of water use per acre coefficients and

projections of future developed acres, water

use per residential housing unit coefficient and

projections of future housing units, and water

use per employee coefficients and

projections of future employment. Many of these types of forecasts rely on disaggregation of

water use into user sectors and seasons, which may improve the informational qualities of a

forecast relative to the per-capita method. However, similar to the per capita method condition,

the demand forecast relies solely on counts of users or related proxies.

Time-Series and Trend-Based Models predict future water demand based on assignment of trend

parameters or statistical (autoregressive) relationships that link past values and systematic

repeating cycles of demand to future values of demand. Using the conceptual example above,

past values of Q are used to predict future values of Q, and the function f(*) defines how past

values relate to current and future values. Time-series models tend to be used to predict demands

over relatively short timeframes when longer-term influences may not be as significant.

Regression and Econometric Models are statistical models that explicitly estimate the parameters

of a function that relates changes in defined explanatory variables (X) to changes in water use.

This class of models uses cause-effect relationships among water use and specified factors that

affect water use to forecast demand. Econometric models can be considered a class of regression

models that specify variables, such as price and income, which according to economic theory,

would be expected to influence consumption.

End-Use Models account for and forecast water used by specific water-using fixtures,

appliances, or for specific purposes. In most cases, end-use models represent an accounting

Page 28: Decreasing Climate-Induced Water Supply Risk Through Improved

24

structure that is dependent on assumptions for water-using technologies, market saturation of

various water-using technologies, and behavioral factors, such as frequency of use. Some end-

use models are developed using regression analysis that specifies factors correlated with end-use

consumption (e.g., see Mayer et al. 1999). These types of models are well-suited for examining

the effects of increasing water efficiency through time, which result from plumbing standards

and codes and water utility conservation programs.

There can be considerable overlap among some of the forecasting approaches classified above,

particularly with regard to disaggregation of water use sectors, as well as hybrid models that

blend the features of different techniques. For example, unit use coefficients may be scaled

according to information and parameters obtained from the literature or by means of separate

regression models (sometimes called variable forecast factor approaches). Time series models

may also be blended with regression models to create forecasts based on past values of

consumption and exogenous factors. Furthermore, outputs from end-use models are sometimes

used to adjust the results of forecasts derived from other methods in order to account for

predictions of future water efficiency. Finally, traditionally less conventional methods, such as

artificial neural networks, are being used more often to model and characterize (or learn) patterns

of water consumption, which may hold some promise for demand forecasting.

How Factors that Affect Demand Are Addressed in Models

Some of the key factors that affect water demand were previously described in Section III. By

construction, different models will have different capabilities for addressing these factors or will

address them differently. For example, by relying only on population, the per-capita approach

cannot directly address factors other than population that influence water demand, and cannot

recognize differences in water use patterns across water use sectors. Sector-based fixed

coefficient methods may provide additional information on the structure of underlying demands,

but will generally also lack the ability to test and specify the effects of other factors, especially

economic and climatic factors.

Regression model and econometric approaches are able to incorporate multiple variables to

explain and predict water demand. However, the degree to which the multiple factors that

influence water use are specified depends on a host of modeling considerations, including but not

Page 29: Decreasing Climate-Induced Water Supply Risk Through Improved

25

limited to the availability of historical data to estimate numerical relationships and the existence

of projection data to effectively use these factors for the purposes of forecasting. Nevertheless,

predictive model-based approaches to water demand forecasting would seem to be the most ideal

for evaluating the potential effects of climate change. Notwithstanding data constraints, they are

capable of directly incorporating principal indicators of weather and climate, which can be used

to assess alternative scenarios. Disaggregation of data into water use sectors and time periods

further augments the capability to analyze climate change by providing an opportunity to isolate

impacts on the underlying components of water use.

How Uncertainty Is Addressed in Models

Water demand forecasting involves inherent uncertainties. As suggested in earlier sections, there

is always incomplete knowledge and understanding of the determinants of water use and how

they are best related in a mathematical sense to water use. In addition, future values of important

factors are not known with certainty and/or can be highly variable, which makes it virtually

impossible to achieve 100 percent forecast accuracy. In fact, even if one were to know the future

values of key factors with certainty, a demand forecast is likely to be wrong because of practical

and technical shortcomings related to the model being used. For example, if one were to predict

future population with 100 percent accuracy, the per-capita forecasting method may still produce

an inaccurate forecast because water consumption depends on more than just population. This

example rightly implies that the options available to address uncertainty are also affected by

choices about the design of the forecasting model.

In practice, and depending on the characteristics of the forecasting model, forecast uncertainty is

addressed through the use of scenarios, the application of statistical routines for estimating

forecast error, or both. Scenario analysis is very common, such as using high, medium, and low

population growth to create an envelope of future demands. Other types of scenarios such as hot-

dry, cool-wet weather scenarios are also often used, but require a model or other mechanism to

translate weather into predictions of water use. Oftentimes, extreme scenarios are combined in an

attempt to account for most future demand possibilities.

Page 30: Decreasing Climate-Induced Water Supply Risk Through Improved

26

Standard formulae exist to

calculate random and sampling

error associated with ordinary

least squares regression

procedures. However,

computational difficulties have

tended to limit applied uncertainty

analysis to evaluation of

conditional forecast error, which

assumes the model is accurate and

accounts only for uncertainty

about the future values of model

variables. Monte Carlo simulation

methods are sometimes used to

simulate potential values of

independent variables, given some

underlying knowledge or

assumptions regarding the type and shape of their respective distributions, which results in a

range of predicted demands.

Current State of Water Demand Forecasting Models

This section summarizes the results of a literature review on water demand forecasting,

conducted by researchers at The George Washington University (GWU) as part of this project.

The objective of the GWU research was to provide a guide to the literature on improving the

practice of demand forecasting for effective decision making.

GWU conducted a search of the water demand forecasting literature published from 2000 to

2010 and developed a bibliography of 79 papers from a cross-section of peer-reviewed journals.

These papers were then categorized into the appropriate type of model (qualitative extrapolative

methods versus nonparametric). The analysis of these models focused on three questions:

How practical are the models?

Figure 3. Forecast uncertainty can be analyzed using a statistical

demand model and assumptions about the distributions of variables that

affect water demand. Source: construct developed by Jack C. Kiefer.

Page 31: Decreasing Climate-Induced Water Supply Risk Through Improved

27

Are the forecasts reliable?

What is the best approach?

These papers were then synthesized in order to identify what the main focus of research has been

and to make proposals on how the practice of water demand forecasting can be improved. The

synthesis found that while a wide variety of methods and models have been used and have

attracted attention, applications of these models differ, depending on the forecast variable, its

periodicity, and the forecast horizon (Donkor et al. 2012).

The analysis found that a shift is ongoing from pure conventional methods to a focus on three

approaches:

1. Scenario-based and Decision Support System (DSS) models: approaches that

accommodate some amount of uncertainty in demand forecasting

2. Comparative assessment of performance between neural nets and conventional methods

3. Recognition of the need to improve forecast accuracy by using hybrid models

The results of the literature search and analysis indicated that it is difficult to answer the question

―Which model is best for water demand forecasting?‖ without specifying the periodicity of the

demand variable. The research found that neural networks and hybrid models are more

appropriate for short-term forecasts; but, for extended ones, where incorporating future scenarios

of a variable might be important, scenario-based and DSS models are more suitable. However,

the use of regression in modeling monthly demand follows the generally held view that short-to-

medium-term demand is typically influenced by weather variables while long-term forecasts are

more determined by socioeconomic factors.

Overall, improving forecast accuracy, accounting for uncertainty in long-term forecasts, and

maintaining system reliability now and in the future seem to have provided the impetus for the

current research in urban water demand forecasting.

Page 32: Decreasing Climate-Induced Water Supply Risk Through Improved

28

Summary

There are differences in how water system planners model and forecast demand for water. The

forecasting techniques vary in their sophistication and methodology to account for determinants

of water use. The choice of a particular forecasting methodology is affected by the data used to

model relationships among demand determinants and sector water use. The availability and

quality of data to support the development of models typically serves as a practical constraint on

the options that are applicable for forecasting. Furthermore, the specific goals and objectives of

any particular water demand forecasting effort may not immediately require one to enhance the

prediction and informational capabilities associated with more complex methods. However, the

array of uncertainties facing the water utility industry seems to require an emphasis on better

forecasting and more robust modeling capabilities.

Page 33: Decreasing Climate-Induced Water Supply Risk Through Improved

29

VI. Risks Associated with Models and Methods

Any forecast has some chance of being incorrect due to the fact that a forecast is an attempt to

predict the future. Water system planners and managers need to understand and mitigate the risks

of being wrong in either direction when predicting future demands. A demand forecast that turns

out to be high can result in stranded capacity and the

water system paying for debt for facilities that are not

producing the predicted revenues. A demand forecast

that turns out to be low can result in lower than

desirable levels of service, i.e., restrictions that might be

placed on water use might be unpopular with the

system’s customers.

The objective in water demand forecasting is to minimize the risk of being incorrect and to

provide for adaptive management so that the water system can accommodate the range of

potential outcomes and their probabilities of occurrence. Water system planners and managers

now need to incorporate the potential changes in demand from climate change, and incorporate

the uncertainties in future weather predictions with all of the other uncertainties previously

discussed, including population and employment predictions and changes in per-capita demand.

Limitations of Existing Models

As previously discussed, models are highly dependent on the quality of the data used to build

and validate the model. Improving modeling at a water system is an investment decision as it

takes additional resources to go beyond what is already being done. In other words, water system

managers need to evaluate whether the limitations in existing demand models are significant

enough to warrant the additional investment in the collection and analysis of existing data, and in

the development of improved models.

In most cases, the additional investment is justified. As previously discussed, for many water

systems, demand forecasting is simply multiplying the gallons per capita per day (gpcd) by the

projected population growth and job growth. However, the traditional per-capita approaches to

forecasting water demand neglect and are incapable of measuring the effects of principal factors

that can produce variability in water use, such as weather and climate, the price of water, land

“Each water system has a

unique set of data; there is

no single model that can

fit all systems.”

Page 34: Decreasing Climate-Induced Water Supply Risk Through Improved

30

use, and several other socioeconomic variables other than population. Past observed reductions

in-per capita use—for example, due to increases in water efficiency, the effects of pricing, and

recessionary pressures—have been largely unanticipated by many systems. Therefore, in many

cases, the additional investment in the collection and analysis of existing data, and in the

development of improved models is warranted to overcome the limitations of existing models.

One Potential Approach to Identify Risks—Extreme Value Analysis

Water system managers and planners are particularly interested in the high-impact, low-

probability water demand events that drive infrastructure investment decisions and the need to

fund such investments. Accurate predictions of peak-day and peak-hour demands are necessary

for planning capital improvements such as alternate sources of supply, treatment plant capacity,

transmission mains, storage tanks, and booster pumping stations.

These events, by definition, are ―extreme events,‖ and one approach that this research found to

model these events is the use of extreme value analysis (EVA). Climate change and more

extreme weather events will likely change the above peak demands, and will need to be

appropriately considered by water systems in future planning, design, and operations of their

systems. EVA has been used in a wide variety of disciplines including the financial industry, the

global reinsurance industry, civil engineering, ecology, water quality, and especially climatology

and hydrology. EVA has been used in hydrology to estimate and forecast flood frequency, model

financial loss related to flooding events, and to model extreme hydrological events in various

watershed sizes, but has seen limited use in the water sector, particularly for demand forecasting.

Research conducted by the University of Colorado-Boulder (CU) as part of this project using

EVA show one potential approach to identify and evaluate risks (Haagenson et al. 2013). The

objective of the CU research was to apply EVA techniques to water demand data at two case

study utilities (Aurora, Colorado, and Tampa, Florida) and show the potential impacts from

climate change on the water demand forecasts for these two case studies.

Focusing on Aurora, Colorado, the CU research used an EVA approach to predict the changes in

water demand due to potential climate change scenarios. Daily production data from 1990-2010

showed the critical season of high demand in June-August. Daily weather data were used to

develop weather attributes (hot/dry, wet/cold spells along with average weather) for June-

Page 35: Decreasing Climate-Induced Water Supply Risk Through Improved

31

August. A simple bootstrapping method was used to forecast weather trends, and then EVA was

used to generate projections of water demand extremes. This research found that under climate

change scenarios, exceedances increase over time for the warm/wet and the warm/dry cases,

relative to natural variability.

Summary

A forecast for anything has some chance of being incorrect due to the fact that any forecast is an

attempt to predict the future. Realistic forecasts of water demand extremes should be valuable to

water system managers (and their planning staff) during costly infrastructure decisions, as the

cost of being wrong could be significant. Different methods and models can be used for

forecasting, ranging from a simple scenario of 10 percent additional peak-hour and peak-day

demands, to a slightly more complicated scenario of 10 percent additional demand coupled with

a 10 percent decrease in water supply, to more computational-intensive approaches such as EVA.

Page 36: Decreasing Climate-Induced Water Supply Risk Through Improved

32

VII. What Utilities Should Be Doing Now

Water system managers are increasingly confronted by a variety of challenges. These challenges

include an increase in drinking water regulations from the Environmental Protection Agency

(EPA), an aging/transitioning workforce, and increased needs for investment in the aging

distribution system in the face of opposition to raising rates, especially in light of current

political and economic conditions (AWWA 2013). Business factors are a significant concern for

water system managers, but climate change introduces a new set of challenges for water system

managers and their planning staff for both long-term planning and for future operations and

maintenance of the water system.

A universal cookie-cutter approach for incorporating

potential impacts from climate change into water

demand forecasting cannot be easily developed.

Situation-specific approaches need to be developed

for demand forecasting that take into account local

considerations in terms of:

Availability of water

Characteristics and patterns of water use and related data

Characteristics of demand (sensitive to climate change) and influential explanatory

factors such as the temporal and spatial characteristics of temperature, precipitation, and

socioeconomic factors

Availability of internal and external modeling expertise

Understanding issues faced by management, the public, and the political leadership

Real or perceived importance of climate change relative to other planning challenges

(e.g., lack of new water sources) and objectives (e.g., reliability of existing supplies).

The last bullet warrants some additional discussion. At any given time, a water system is

presented with a number of risks that vary in terms of immediacy, severity of potential

consequences, and likelihood of occurrence. These risks can include the possibility of supply

loss (due to contamination, regulation, supply seasonality/drought, turbidity, reservoir operations

constraints, etc.), system or component failure, supply contract performance, stranded resources,

“A universal cookie-cutter

approach cannot be easily

developed.”

Page 37: Decreasing Climate-Induced Water Supply Risk Through Improved

33

political environments, personnel/labor relations, demand management requirements, and rate

affordability. Managers must be cognizant of the costs associated with eliminating or mitigating

any particular risk, and must justify risk-management efforts accordingly. Realistically,

achieving zero risk is not possible and eliminating certain risks may be cost prohibitive.

Decisions about how to manage the potential demand-side risks related to climate change will

need to be made within the context of the larger

portfolio of risks for each water system, and these risks

typically will be different for each system.

Water systems may face several potential impacts from

climate change on operations and maintenance for both

existing and future facilities. For example, sea-level rise

may result in a system having to relocate facilities

and/or modify operations, and water demand may also

change due to customer relocation and changes in

regional growth patterns. Increased weather variability

resulting in increased incidence of flooding, increasing numbers of other significant weather

events such as hurricanes, ice storms, etc., may also have a more pronounced direct impact on

the water system (i.e., flooding and/or other physical damage to the system). The relative

significance of these impacts will vary from utility to utility.

Water systems need to understand that if ―stationarity is dead,‖ the past may not be the best

predictor of the future, but it is important to understand past dynamics and use that information

to help inform future decision-making (Milly et al. 2008). While all models are built upon past

data, improving the measurement of the impacts and causal relationships between the factors that

are known and have been experienced and a system’s demand is critical. A system’s average

demand could change in the future due to increased market penetration of low-flow plumbing

fixtures. A system’s peak demand could also change due to changes in the weather.

In the future, climate and weather conditions are likely going to change for water systems.

Furthermore, future climate and weather variations may vary geographically. Some examples of

potential changes that may affect operations and planning include:

“…relying solely on the

past to predict future

water demands could be

problematic, especially

without a more in-depth

understanding of what

factors have influenced or

determined past patterns.”

Page 38: Decreasing Climate-Induced Water Supply Risk Through Improved

34

Springtime beginning earlier and ending later, which extends the watering season

Changes in total precipitation, or the annual/seasonal distribution of precipitation (e.g.,

fewer but more intense storms)

Warmer temperatures and longer periods of hot and dry spells

The availability of adequate water resources (and/or the lack of new water sources) generally

places water systems into three categories when considering actions to address the potential for

altered conditions and associated risks. The level of effort and expense (time and resources)

allocated toward increasing the understanding of water use patterns and improved demand

forecasting may be characterized as:

1. Wait and See—refers to systems having ample long-term water supplies and adequate

treatment and transmission capacity.

2. Start Thinking About It—refers to systems that are looking for a ―no-regret‖ strategy that

can be adopted now with minimal cost, while learning more about potential impacts of

climate change to their systems. Most systems in this category are also looking for

flexible adaptive management strategies that can be adjusted as more data is available

and translated into actionable information.

3. Should Be Thinking About It—refers to systems already resource constrained (i.e.,

resource shortages already exist) or nearing safe yields and where the possibility of

constraints or demand pressures is likely to worsen.

More research is needed to understand specific potential impacts to water systems so that water

systems can establish their risk tolerance and adaptive management positions. The research

needs are described in greater detail in the next section of this report. In the meantime, system

managers can take measures now to help reduce uncertainty in forecasting water demand. These

recommendations were developed by a broad range of stakeholders, including water system

managers, consulting engineers, and academics during the two workshops discussed in Section

IV and the subsequent webinars.

Page 39: Decreasing Climate-Induced Water Supply Risk Through Improved

35

These recommendations fall into six general categories that need further investigation by water

systems to determine the appropriate means of implementation by system managers and

planners. The six categories are discussed in more detail below.

Collect Additional Weather and Demand Data

The past and the present have to be better understood before one can make accurate predictions

about future conditions. Having complete and accurate data lies at the heart of modeling.

Unfortunately, a significant number of utilities collect, maintain, and store only limited data to

support water demand analysis and forecasting. To start the process of reducing uncertainty in

forecasting water demand, systems should improve their data collection processes for historical

water use data, weather information, and related factors that influence water demand.

Water system managers should also be aware of the time and resources necessary for identifying

the appropriate data to collect, how to collect it, and how to analyze it. The objective of any data

collection effort needs to be clear at the outset. Without a clear objective, a water system will

likely struggle not only with turning the data into actionable information for decision-making,

but also with justifying the cost of enhancing data collection efforts.

For weather data, temperature and precipitation are the most commonly collected parameters.

Several potentially important temperature variables will typically be correlated with water use,

for example (but not limited to):

Daily or average high and low temperatures

Frequency of hot days (e.g., number of days in a month with high temperatures exceeding

90 degrees)

Number of consecutive days the temperature is above a certain threshold

Understanding precipitation patterns is also important for analyzing water demand patterns.

Several potentially important precipitation variables will typically be correlated with water use,

for example (but not limited to):

Amount of precipitation (daily, weekly, monthly, annually)

Frequency of precipitation events

Page 40: Decreasing Climate-Induced Water Supply Risk Through Improved

36

Intensity of precipitation events (for example, two inches of rain in an hour is quite a

different event than two inches in 24 hours)

Number of consecutive days with or without precipitation

Data on a daily time-scale are typically available for

both of these parameters from the National Climatic

Data Center (NCDC) for several thousand weather

observation stations across the United States (NCDC

2013). Data for multiple timescales are available online

for download and analysis.

It should be noted that missing weather observations

continue to be a problem for data analysis (even from

official data sources) and may worsen due to lack of funding for continuing operations of

weather stations. Despite this problem, several statistical techniques such as imputation, partial

imputation, bootstrapping, partial deletion, and interpolation are used for handling missing

values in time-series data (Honaker and King 2006).

Many water systems may have only one weather station in their service area, especially those

serving a small geographic area. Larger water systems may have multiple available weather

stations, allowing them to decide upon the best data to represent historical conditions or to

weight weather observations from multiple locations to estimate the contours of weather across

their service area. Many water systems use various statistical techniques such as nearest neighbor

or other distance-weighting techniques to incorporate data from multiple weather stations. These

procedures are particularly well-suited for situations where climate and weather conditions can

vary significantly across a water system’s service territory.

Understanding seasonal weather patterns is crucial for evaluating the impact of climate on water

demand. During the summer, most systems experience more outdoor water use (i.e., watering

lawns, gardens, etc.) that can significantly contribute to system peaking patterns. Most (but not

all) systems experience peak-hour and peak-day demands during the summer, with increased

water use for bathing and washing clothes, further driving peak demand.

“Understanding seasonal

weather patterns is crucial

for evaluating the impact

of climate on water

demand.”

Page 41: Decreasing Climate-Induced Water Supply Risk Through Improved

37

Both average demand and peak demand (i.e., peak-day and peak-hour) data are important

benchmarks of water demand. Water treatment plants and distribution systems have to be

designed to meet peak demands, and in many cases, those peak factors are based on ―best

engineering judgment‖ and/or design requirements established by the state. Such peak use

factors could be refined and/or modified if sufficient historical data is available and these peak

demands can be correlated with weather conditions. Because of the importance of peak demand

for system design, investment in historical data collection and concentration on peak-use periods

might be easily justified.

Many systems use production data as a surrogate for water demand. This approach is generally

acceptable when it is the only data available, but system managers need to be aware that

production data will include non-revenue water, such as physical losses in the distribution system

and/or fire flows.

Water sales data can provide more disaggregated detail on water usage patterns, though customer

classification schemes can vary substantially from system to system. For example, some systems

have two categories—residential and commercial. Others may classify solely on the basis of

water meter size. More water systems are now classifying customers into multiple categories

such as single-family residential, multifamily residential, commercial, industrial, institutional,

and/or additional categories.

Several other more refined classifications are possible and have been implemented in many

water billing systems. In general, the more categories in which demand data are collected, the

greater the opportunity to better understand water usage patterns.

The emergence of Automated Meter Infrastructure (AMI) technology provides a new means for

collecting very detailed water consumption data. AMI technology can provide instantaneous

demand data, but this requires careful attention to data management considerations. For example,

if a system has 10,000 residential meters and each meter transmitted a flow (demand) reading

every hour, over 87 million data points would be generated every year if the data from every

meter were collected and analyzed. How would all of this data be collected, stored, and

analyzed? What would be the bands of ―normal‖ variability and what would be considered

―outliers‖? Developing the appropriate Quality Assurance/Quality Control (QA/QC) protocols

Page 42: Decreasing Climate-Induced Water Supply Risk Through Improved

38

quickly becomes very important, as well as the development of data analysis protocols. High

resolution data from a statistically representative sample of accounts could be used instead of the

data from all accounts, but the sample design would need to be appropriate for the power

(quality) of the data needed. System managers should not underestimate the time and resources

needed to translate this data into actionable information for informed decision-making. However,

as more AMI data becomes available, the benefits and costs for planning and evaluation will

become more transparent.

Analyze the Data and Translate It into Actionable Information

As previously discussed, a data analysis plan is a critical component of any data collection effort.

As part of this effort, a system should have a clear understanding of the data life cycle, which

includes data collection; quality control and assurance; data management; data analysis; long-

term data archival; and, data retirement. Each component of the life cycle is dependent on how

the data will be applied to support utility functions. A water system planner should not

underestimate the time and resources needed to translate data into actionable information for

informed decision-making. The previous example of 87 million data points annually from a

medium-size system with 10,000 meters provides some insight into the criticality of developing a

data analysis plan as part of making the investment in the data collection effort.

The data analysis plan may dictate the collection, processing, and integration of ancillary data

such as the weather and demand data discussed previously. Knowing what questions need to be

answered by the data collection effort will drive the list of critical data elements to be collected,

as well as any metadata that will be required to support the analyses and conclusions reached. A

clear understanding of what is known about the causal relationships between the factors that are

known versus what hypotheses are being tested within a system will help determine what data

should be collected and how it should be stored and managed. Developing a data analysis plan

before data collection begins will improve the efficiency of the process, as well as its

effectiveness.

.

Page 43: Decreasing Climate-Induced Water Supply Risk Through Improved

39

Multiple data analyses could potentially be conducted based on the enhanced collection of

weather and demand data previously described. One potential analysis should be developing an

understanding of both the inherent buffer in peak demand (the ability to restrict outdoor demand)

from outdoor use and the potential for ―demand hardening‖ as a result of past outdoor use

restrictions. For example, if outdoor irrigation has been reduced through past conservation

measures, then the potential for additional reductions during a drought has likely been reduced. It

is important for a water system to determine what the ―soft‖ demand actually is, so that it can be

addressed accordingly.

Finally, a system should develop a plan for a regular update of water demand forecasts. These

forecasts should be updated, just as Capital Improvement Plans (CIPs) and cost-of-service

studies. The California Urban Water Conservation Council includes requirements for five-year

demand forecast updates as part of its Memorandum of Understanding (MOU) with its signatory

members (CUWCC 2013). A similar five-year update of demand forecasts (and resource

availability) is performed by the water systems supplying the Washington, DC, metropolitan area

by the Interstate Commission on the Potomac River Basin (ICPRB 2013). Systems should

consider adopting a five-year cycle for a top-to-bottom update of their water supply and demand

forecasts.

Evaluate Potential Changes in Demand

A water system should carefully evaluate potential changes in demand when developing demand

forecasts. Future demands can be impacted by a number of factors:

1. Increased water use efficiency of plumbing devices and appliances (such as dishwashers

and washing machines) and increased installation of such devices

2. Increased water conservation in industrial/commercial applications

3. Socioeconomic factors

a. Density of development

b. Mix and types of businesses

c. Population, employment, and housing

d. Future costs and pricing

e. Relative shift from single-family to multifamily

f. Other

Page 44: Decreasing Climate-Induced Water Supply Risk Through Improved

40

4. Attitudes/behaviors

a. Conservation ethic

b. Landscaping

5. Shift towards more water reuse

6. Climate change

The continued penetration into the marketplace of low-flow plumbing fixtures and water-

efficient appliances such as dishwashers and washing machines will likely continue to lower per-

capita demand for most systems until such time as it levels out. The impacts of the Energy Policy

Act of 1992 (EPACT92) on water use has been to reduce water use by 5 percent in the first

decade after EPACT92 (Billings 2008). Furthermore, it has been estimated that additional

reductions in water use over the next 10-20 years could be in the range of 15-25 percent. This is

a significant reduction from past water use patterns. Systems should develop an understanding of

the penetration of low-flow plumbing fixtures in their service area (i.e., the current installations

of such fixtures and appliances, approximately how many are being replaced annually, and the

approximate numbers of older higher-flow plumbing fixtures). Systems should then make

predictions on the future trend of that penetration and its implications on future per capita

demand.

Additionally, ultra low-flow plumbing fixtures that go beyond the regulatory requirements have

entered the marketplace, and their market share is continuing to grow. For example, waterless

urinals are being used in some new buildings and are being retrofitted into some existing

buildings. The use of more water-efficient plumbing fixtures and no water use for outdoor

irrigation in new construction are both part of the LEED scoring system developed by the US

Green Building Council (USGBC 2013).

Other demand-side management programs can also impact future demand forecasts. EPA’s

WaterSense program not only addresses irrigation, but also other residential and commercial

conservation measures (EPA 2013). Between all of these low-flow fixtures and water-efficient

appliances that are driven by the consumer and commercial marketplaces, a huge part of the

forecast (―passive savings‖) is out of the water system’s control. So the system needs to develop

an understanding of these potential impacts on water demand. A water system will likely have to

Page 45: Decreasing Climate-Induced Water Supply Risk Through Improved

41

invest time and resources to develop an understanding of all of the installations of low-flow

fixtures and water-efficient appliances in its service area.

Many commercial, industrial, and institutional customers of water systems have also

implemented demand-side management programs. In some areas, these customers have reduced

water use due to water supply constraints (i.e., a drought). Other customers have reduced water

use as part of corporate and institutional environmental stewardship. Within the beverage

industry, the Beverage Industry Environmental Roundtable (BIER) has designated water

stewardship as one of its focus areas (BIER 2013). The roundtable provides a mechanism for the

major beverage manufacturers to share best practices and to benchmark against each other

through the BIER-developed World Class Water Stewardship in the Beverage Industry. Other

commercial, industrial, and institutional customers have reduced their water demands for simple

financial reasons—they want to reduce their operating costs and in some cases, water bills can be

a substantial part of their costs.

Many water systems have already seen decreased water use from their commercial, industrial,

and institutional customers. In some areas, this trend may continue; but in other areas, the

majority of the demand-side reductions have already taken place, and the decreasing trend may

be ―flattening out.‖ Water systems need to have discussions with their major commercial,

industrial, and institutional customers to understand what demand-side management programs

have already been implemented and what programs might be implemented in the future at these

facilities as part of the system demand forecasts.

Water systems need to also consider several socioeconomic factors in their demand forecasts.

The US population is going to undergo several demographic shifts in the next 40-50 years

(Smithsonian 2013). The density of development and the type of development might be different

in the future. While suburban living will still appeal to many, others will gravitate to urban areas.

Similar shifts in the future could hold true for the mix and types of businesses in the service area

of a water system. Beyond demographics, another socioeconomic factor for water systems to

consider will be their customers’ ―willingness to pay‖ as water and sewer rates increase.

Depending on the economic status of the community, future water and sewer rate increases could

decrease demand as consumers become less able to afford to use as much.

Page 46: Decreasing Climate-Induced Water Supply Risk Through Improved

42

The attitudes and behaviors of the customers in the service area of a water system will also need

to be factored into water demand forecasts. More and more customers are embracing a

conservation ethic in many facets of their lives, including water use. Water systems will need to

understand the penetration of the water conservation ethic within their customer base when

developing water demand forecasts. For example, the Saving Water Partnership (SWP) is a

collaborative regional conservation program lead by Seattle Public Utilities and includes 18

water utilities purchasing wholesale water from Seattle. SWP has developed nine customer water

conservation use efficiency strategies that focus on education, outreach, and information transfer

(SWP 2013). The SWP goal is to hold total water use below a specified level despite population

growth being forecasted to increase by 3.9 percent between 2013 and 2018.

Changes in landscaping practices can also impact water demand forecasting. In some areas,

increased use of home sprinkler systems has led to increased peak demands in the early morning

hours. In Northern Virginia (typically not considered a water resource-constrained area), one

water system asks its customers to voluntarily implement a two-day watering schedule to reduce

peak summer demands, and it encourages the planting of water-wise/native landscaping, also

known as Xeriscaping (Loudoun Water 2013). In North Carolina, another water system is

providing incentives for customers to purchase and install smart irrigation controllers through a

Smart Irrigation Program (Charlotte-

Mecklenburg 2013). In more traditionally

water resource-constrained areas such as the

West and the Southwest, Xeriscaping is quite

popular (Denver Water 2013). Increased

implementation of Xeriscaping in a water

system’s service area would impact water

demand forecasts. Water systems need to

develop an understanding of changes in their

service areas in landscaping practices for system demand forecasts.

Increased water reuse is another factor to consider in water demand forecasts. Many water

agencies provide both drinking water and wastewater services in their communities. In many

Page 47: Decreasing Climate-Induced Water Supply Risk Through Improved

43

areas of the country, water reuse is increasing for primarily two reasons—either to provide an

alternate water resource for non-potable uses to reduce the drinking water demand or to provide

an alternative for wastewater discharge (or in some cases, for both). Reclaimed water can be

used for a variety of purposes, including irrigation, makeup water for cooling towers and/or other

industrial uses, and in some cases, for indirect potable reuse.

For water demand forecasting, the exact end use of the reclaimed water is not critical compared

to the fact that the reclaimed water has taken the place of a traditional drinking water use and the

demand has changed. A 2008 report by the WateReuse Association found 1,221 utilities with a

water reuse facility in their national database, primarily in Florida, Texas, and states in the West

and Southwest (WateReuse 2008). These water reuse facilities produced over 374 billion gallons

of reclaimed water in 2008. Water systems need to understand the use of reclaimed water in their

service areas and understand the projected reclaimed water system growth in their system

demand forecasts.

The potential impacts of climate change are additional factors to consider in demand forecasts.

Increased temperatures and changes in precipitation will change outdoor water use, which in

many areas, is the driver for determining how to meet system demand. Water systems need to

make some predictions on how outdoor watering patterns might change due to future changes in

the climate for their system demand forecasts.

Evaluate Potential Changes in Demographics in the Service Area

As previously discussed, for many systems, water demand forecasting is conducted by simply

multiplying gallons per capita per day (gpcd) times the projected population. Both gpcd and

population have their own underlying uncertainty and multiplying the two magnifies those

uncertainties.

Projections of population and job growth are fraught with uncertainty. The recent recession is

providing an opportunity for many water systems to take a hard look at past population and job

growth projections. The Water Research Foundation has identified water demand as one of its

ten focus areas in its research program and one of the resultant research projects is to develop an

understanding of the recent recession on water use and demand forecasting (WaterRF 2013).

Page 48: Decreasing Climate-Induced Water Supply Risk Through Improved

44

Water system managers and planners should actively engage demographers, urban planners and

land use planners to understand their population and jobs projection data and its inherent

assumptions and uncertainties. Water system planners and utility public relations staff should

talk to their customers to learn what influences demand and then apply that knowledge to future

forecasts.

Equivalent accounts are typically the starting point for forecasting water demand for many

systems. Equivalent accounts scale accounts according to the relative water use per account

across sectors. For assuming an average single-family account with 400 gallons per day and an

average industrial account of 2,000 gallons per day, the industrial account is ―equivalent‖ to five

single-family accounts. Systems should be aware of changing demographics and changing size

of families in their service area, and incorporate those changes into future demand forecasts.

Water system managers and planners should seriously consider breaking away from

conventional population and jobs forecasting and invest the time and resources into developing

detailed demand models. Systems need to better understand demand and model what makes

demand vary over both short- and long-term model run periods. Again, a system should not

underestimate the time and resources needed to build a demand model. A system should run the

model over both short- and long-term model run periods, and then translate the model results into

actionable information for informed decision-making.

Understand and Incorporate Uncertainty into Forecasting

Water systems managers and planners only need to look at past forecasts to see their inherent

uncertainty.

Page 49: Decreasing Climate-Induced Water Supply Risk Through Improved

45

Figure 4. Source: ICPRB, Demand and Water Resource Availability for 2040.

Figures 4 and 5 depict results from two major metropolitan areas that could be replicated at most

US water systems. Demands forecasted 20 or 40 years ago are not seen today for the many

factors such as conservation and demographics previously discussed.

Figure 5. New York City Water Demand. Source: New York City Department of Environmental

Protection.

What can one learn from these two examples? First, updating water demand forecasts on a

regular basis is prudent, as these examples are typical of many systems that have over-predicted

future water demands in the past. For example, demand in 2013 in the Washington, DC,

metropolitan area is approximately 500-510 million gallons per day (MGD) and past forecasts

had predicted demand in 2013 to be as high as 900 MGD.

Second, presenting future water demand forecasts as a straight line implies a mistaken level of

precision and does not appropriately present the inherent uncertainties in these forecasts.

Page 50: Decreasing Climate-Induced Water Supply Risk Through Improved

46

Figure 6. Long-Range Water Demand Forecast for Orange Water. Source: Orange Water and

Sewer Authority Long-Range Water Supply Plan Update.

Figure 6 shows the uncertainty bands for future forecasts for Orange County, North Carolina.

These bands can either be the ―high‖ or ―low‖ forecasts, or a specific confidence interval.

Forecasts of future water demand data are typically shown graphically as a single time-series line

or curve. This graphical single line representation gives a mistaken impression of confidence

surrounding these future predictions. Future demand forecasts should be presented to water

system managers with the appropriate uncertainty bands.

Water system managers and governing board members need to understand these uncertainties

and how these uncertainties might impact investments in future infrastructure. Avoiding stranded

investments (i.e., building infrastructure to meet future needs that turns out not to be needed until

later) is important from both the manager’s and governing board’s perspective. A balance needs

to be struck between water-supply risk and financial risk in all cases. Once the debt is incurred

for infrastructure construction, repayment of the debt is constant and continues whether or not

the demand increases as predicted or lags behind.

Page 51: Decreasing Climate-Induced Water Supply Risk Through Improved

47

Plan for Drought So the System Can Cope

A drought plan should address the consequences and operational challenges associated with both

acute and protracted water shortage events, and should also be integrated with a system’s

Emergency Response Plan (ERP) for other emergencies such as hurricanes, tornadoes, flooding,

and power outages. Some water systems are required to develop drought plans but others do not

face such a regulatory requirement.

The typical drought plan involves a series of increasingly stringent conservation and use

measures. One issue facing water systems is that conservation is sometimes regarded as being

equivalent to a source of future supply. This impairs the ability of the water system to deal with

future droughts because there is less excess capacity in the system. While this is a supply issue

(as opposed to demand issues, which are the focus on this project), water systems need to

understand the linkages between supply and demand that result from their conservation efforts.

If a system already has a drought plan, the plan should be revised on a regular basis to take into

account lessons learned from past droughts. A five-year review cycle is what is commonly used

for other Standard Operating Procedures (SOPs) such as Emergency Response Plans (ERPs).

Systems should consider adopting a five-year cycle for a top-to-bottom update of their drought

plans. In situations where the water system has undergone a major expansion or modification, the

EPA recommends revising the ERP after completion of the expansion/modification (EPA 2004).

A drought plan should be part of this revision.

Summary

Water system managers must consider a myriad of challenges when planning for future

operations and maintenance of their water systems. Business factors are a concern; climate

change introduces an additional set of challenges for water system managers and their planning

staff. A single ―one-size-fits-all‖ approach cannot be easily devised for incorporating potential

impacts from climate change into water demand forecasting. More research is needed to

understand specific potential impacts to water systems so that water systems can establish their

risk tolerance and adaptive management positions. However, system managers can take

measures now to help reduce uncertainty in forecasting water demand:

Page 52: Decreasing Climate-Induced Water Supply Risk Through Improved

48

1. Collect additional weather and demand data;

2. Analyze the data and translate it into actionable information;

3. Evaluate potential changes in demand;

4. Evaluate potential changes in demographics in the service area;

5. Understand and incorporate uncertainty into forecasting; and

6. Plan for drought so that the system can cope with it.

Before the future effects of climate change are upon us, water system managers and planners

should carefully consider the above recommendations. And, while all system managers can take

them into consideration, the implementation at any one system will be very system-specific

given the time and resources available to the system manager and planning staff as well as the

manager’s knowledge and understanding of the technical and policy issues affecting operations.

Page 53: Decreasing Climate-Induced Water Supply Risk Through Improved

49

VIII. Recommendations for Future Research

Water systems can (and should) take steps now to improve their data collection and analysis to

better address the uncertainties surrounding water demand forecasting, particularly in light of

climate change. The direct linkage between water demand and water sales (and system revenues)

requires developing a better understanding of future uncertainties surrounding infrastructure

needs, revenues, and the potential impacts from climate change for risk management. However,

more research is needed to help drinking water systems better understand water demand. This

chapter summarizes some recommendations for future research.

Understanding Baseline Conditions and Potential Changes

Situation-specific (e.g., service-area specific) research on baseline and changing conditions is

recommended for most medium to large water systems, as well as regional and national level

assessments that require aggregation of system data. However, regional and national studies of

baseline and changing conditions will be hampered by lack of uniform data elements and naming

conventions, which must also be addressed.

Water system managers and planners need to better understand both baseline conditions and

potential future changes to those conditions. Questions requiring research include:

What percentage of the service area has installed low-flow plumbing fixtures and what is

the net impact on water demand? How will future per-capita demands change as more

low-flow fixtures are installed and as more water-efficient dishwashers and clothes

washers penetrate the marketplace? How much is indoor residential water use expected

to decrease and when is that decrease expected to flatten out? How might upgrades to

cooling towers on commercial, industrial, and institutional buildings reduce water use?

What percentage of the service area has installed residential lawn sprinklers and what is

the net impact on water demand? How many existing homes are being retrofitted and

how many new homes are being built with lawn sprinkler systems? How many sprinkler

systems have ―smart‖ controllers versus manual controls?

How have changing demographics in the service area impacted water demand? How is

family size changing in the service area? How many households are ―downsizing‖ versus

staying in place? Is the average household smaller in number now than in the past? What

Page 54: Decreasing Climate-Induced Water Supply Risk Through Improved

50

do demographers project, and what are the uncertainties, surrounding their future

population, land use, and economic forecasts? How do community sustainability

initiatives impact future forecasts? How do changing economic conditions affect choices

about water use either directly or indirectly? How might climate change impact future

demographic patterns, e.g., people relocating to different parts of the country?

How have shifts in business and industrial customers (e.g., factories closing or relocating)

impacted water demand and what future shifts should be considered?

How might increasing water rates (driven by aging infrastructure, water quality

regulations, and other factors) further impact future water demand? How much will

consumers reduce their water use based on increasing water bills?

How will the increased use of Automated Meter Reading/Automated Metering

Infrastructure (AMR/AMI) help a system develop a better understanding of its non-

revenue water and customer water use characteristics?

How much is water re-use forecasted to replace potable water in the future?

For systems that provide both drinking water and wastewater services, what percentage

of the service area has installed rainwater harvesting and/or using gray water and what

are the impacts on wastewater flows? Are increasing costs and pricing for wastewater

service having an impact on water consumption?

Potential Impacts of Demand on Appropriate System Design

Water systems are typically designed using projections of water demand that are based on

historical trends, as well as a combination of best engineering practices and codes and standards

(such as the Ten-State Standards or state-level design standards). But more research is needed so

that system design can meet customer needs in a changing environment. For example, peak

factors used for peak-day and average-day demand may need revision as lot sizes decrease or the

population ages and ―downsizes,‖ leading to decreased outdoor irrigation demands. However,

increased temperature extremes combined with drought could lead to increased outdoor

irrigation demands. Situation-specific (e.g., service area specific) research on baseline and

changing conditions is needed. Design standards may need to be revised in the future, but data

collection and analysis is needed now to inform any potential revisions.

Page 55: Decreasing Climate-Induced Water Supply Risk Through Improved

51

System design questions that require further research include:

Have peaking factors (i.e., peak-hour and peak-day factors) changed over time? Are

these peaking factors projected to change in the future due to increased implementation

of low-flow plumbing fixtures, residential lawn sprinkler systems (with or without

―smart‖ controllers), and due to changing demographics?

Have the requirements for system storage changed over time and are those requirements

projected to change in the future?

For systems that provide both drinking water and wastewater services, will the projected

reduction in returned flows due to low-flow plumbing fixtures impact the design of

wastewater collection and treatment systems? What percentage of the service area has

installed rainwater harvesting and/or using gray water and what are the impacts on the

design of wastewater collection and treatment systems? What are the potential impacts

of reclaimed water used for outdoor irrigation?

System Data

As more water systems implement AMI/AMR, more

and more data is being collected; however, each

system has its own information management

protocols and specific naming conventions for data

elements. For example, researchers at Virginia Tech

found that regarding customer complaint data, one

system had 13 descriptors for water that appeared

―black‖ or contained something black (Whelton

2007). The lack of a common convention for

managing data elements inhibits comparisons

among systems, and makes regional and/or national aggregation of demand data difficult and

time intensive. Development of uniform conventions for managing water demand and

demographic data is necessary to allow for accurate comparison and aggregation of data across

systems.

A minimum set of requirements for collecting water demand and demographic data is needed to

―Development of uniform

conventions for managing

water demand and

demographic data is

necessary to allow for

accurate comparison and

aggregation of data across

systems.”

Page 56: Decreasing Climate-Induced Water Supply Risk Through Improved

52

inform demand forecasting and water resource planning, and would also be useful to the drinking

water community across a wide variety of data types and analyses. In general, more research is

needed to get systems ―on the same page,‖ in regard to research and planning.

Systems also need to ensure that their customer data is categorized correctly. Effective

management of customer billing information is critical. At a system level, research may be

needed to ensure that account information is up-to-date (e.g., that the ownership of a building has

not changed, or a building’s uses have not changed), and that single-family, multifamily,

commercial, industrial, and institutional accounts are correctly categorized. The need for uniform

data management conventions is increasingly acute with the volume of data generated by

AMR/AMI systems.

System Revenues

Future water demand forecasts are inextricably linked to forecasts of system revenues. More

detailed information is needed for both water demand and revenues projections. Improved data

will better inform the design of rate structures that strike an appropriate balance between fixed

account charges, tap fees, and commodity sales revenues.

Data and Research Integration

More research is needed on how systems may integrate data from external sources (such as

demographic data from the Bureau of the Census or climate modeling data from the National

Climatic Data Center) with utility demand, production, and planning data. More research is also

needed on the relationships and the linkages between water quantity (water resources) and water

quality and the potential water quality changes resulting from climate change.

Historical Drought/Water Shortage Analyses

Analyses of past droughts, especially droughts resulting in water use restrictions, could provide

useful insights on integrating climate change projections into water quantity and water quality

forecasts. This research would focus on the weather conditions that led to water-use restrictions

and the decision-making process that the water system used to implement those water-use

restrictions (as well as how and when those restrictions were ultimately lifted), yielding

important information on supply reliability. The research should also assess water demand

Page 57: Decreasing Climate-Induced Water Supply Risk Through Improved

53

during the period of restricted use versus the average daily demand. The results should show how

effective the water-use restrictions were and whether the demand returned to ―normal‖ pre-

drought levels or if a ―new‖ normal evolved after the restrictions were lifted. System-specific

research would be useful, but research on a regional and national level (requiring aggregation of

data) might inform a broader audience and be more useful from a policy perspective.

Value of Information Studies

Classical decision analysis can be used to estimate the value of new information for decision-

making. Since better information makes for more informed decisions, a value can be placed on

that information. Value of Information (VOI) studies are not typically applied to drinking water

decision-making; however, in a past drinking water context, one researcher used a VOI approach

to estimate the value of arsenic health effects as part of the regulatory development process for

the arsenic regulation (North 1994).

More research is needed to determine the applicability of VOI methods to water demand

forecasting. A VOI assessment could help decision-makers determine how much to invest in data

collection and analysis, and how much these efforts might be worth in evaluating future

investment decisions (i.e., how much should a system spend now to reduce future uncertainty by

a certain amount). VOI studies could also be used as a sensitivity analysis for investment

decisions on whether to spend additional money on new data collection and analysis projects.

However, VOI studies are not simple and require some informed judgment as part of the process.

Social Science Research

Most decision-makers at a water system come from either a technical or financial background;

therefore, limited social science research has been conducted in the drinking water community.

Research is needed on the effectiveness of ongoing conservation efforts, turf replacement

programs, customer education initiatives, and on the tolerance/acceptance of level of service

(e.g., frequency of water-use restrictions being imposed).

Tools for Investment Decisions

While water systems are a fundamental societal need, they are capital-intensive, requiring

significant investments to develop sources, to build and operate treatment plants, and to build

Page 58: Decreasing Climate-Induced Water Supply Risk Through Improved

54

and operate distribution systems. Significant expenditures are also needed to continually repair,

rehabilitate, or replace aging components of a water system (AWWA 2012). AWWA’s Buried

No Longer report estimated that the cost to restore existing infrastructure and to build new

infrastructure to serve a growing population will total at least $1 trillion between 2012 and 2037.

Therefore, water systems need tools that effectively optimize their investment decisions and

generate accurate water demand forecasts. Issues to address include:

Relationships between past and future water demand, water sales, demographics,

weather, economic conditions, and investment decisions

Relationships between the past and future combinations of tap fees and commodity sales

for water system revenues

Uncertainties in variables important for each system (i.e., whether demographic issues are

more important than weather/climate or economic issues)

Translating the uncertainties in predicting future water demands, risks, and potential

impacts on investment decisions

Steps required by water system staff and governing boards to develop a better

understanding of the linkages between future water demands and investment decisions

Summary

More research is needed to help water system decision-makers better understand their current

water demand and the improvements needed for enhanced water demand forecasting. The above

recommendations are a starting point for identifying specific areas of future research, but by no

means an exhaustive list. More work is needed to develop a water demand research roadmap.

What is known now is that better tools are needed to help systems implement enhanced water

demand forecasting, especially as it relates to future investment decisions. Given the direct

linkage between demand and water sales and system revenues and their impacts on a

community’s viability, developing a better understanding of future uncertainties surrounding

infrastructure needs, revenues, and the potential impacts from climate change is prudent risk

management for water systems and the communities they serve. As part of their fiduciary

responsibilities, governing boards and elected officials also need to understand these linkages

and the potential impacts of climate change.

Page 59: Decreasing Climate-Induced Water Supply Risk Through Improved

55

IX. Summary and Conclusions

Operating and managing a water system involves juggling many competing priorities, such as:

Rehabilitating or replacing infrastructure

Lack of public understanding of the value of water

Capital costs and availability

Water supply and scarcity

Aging workforce/talent attraction and retention

Regulation and government oversight

Water security and emergency preparedness

Climate risk and resiliency

As shown above, matching supply and demand is just one set of competing priorities for water

system managers. Optimizing new and/or expanded sources of supply (new and/or expanded raw

water sources as well as new and/or expanded treatment plants) with new and/or expanded

transmission and distribution facilities to match increasing demands is challenging for water

utility managers and planners. Climate change now adds another level of complexity for water

demand forecasting. In order to incorporate climate change in demand forecasting, existing

forecasting needs to be improved before climate change is added to the mix.

Water system managers and planners do not have to wait until all of the uncertainties

surrounding climate change and weather extremes are resolved to improve demand forecasting.

Steps can be taken now to reduce the risk of ―being wrong‖ in their water demand forecasts.

The recommendations for water systems that resulted from this research fall into six general

categories. Water system managers and their planning staff will need to determine how to

appropriately implement these recommendations at their systems, noting that many system-

specific factors would impact potential implementation:

1. Collect additional weather and demand data.

2. Analyze the data and turn into actionable information.

3. Evaluate potential changes in demand.

4. Evaluate potential changes in demographics in the service area.

Page 60: Decreasing Climate-Induced Water Supply Risk Through Improved

56

5. Understand and incorporate uncertainty into future forecasts.

6. Plan for drought so the system is able to cope with it.

Every water system does not have the resources and/or the expertise to implement all six of the

above recommendations. However, appropriately managing the system finances and optimizing

future capital investments are two critical priorities for all water systems, and the above

recommendations will help improve the decision-making for future capital investments. These

recommendations offer a starting point for considering the investment of time and resources

necessary to improve and optimize a system’s long-term water demand forecast that drive many

capital investments.

This project also developed several future research recommendations. The water sector will need

to determine how to get this research funded and implemented. Ideally, this research would be

conducted by a blend of government agencies (such as the EPA and the Army Corps of

Engineers Institute for Water Resources), universities, and the research organizations affiliated

with the water sector (such as the Water Research Foundation). The Water Research Foundation

allocates 60 percent of its annual research budget into ten focus areas and water demand is one of

those focus areas (WaterRF 2013). The goal of this focus area is to increase the effectiveness of

water demand forecasting and the incorporating of their uncertainty into water systems

infrastructure and financial planning. But a sustained effort will be needed by a variety of

research organizations in order to address the research topics discussed in this report.

Water demand forecasting is a critical component of water system planning, and having accurate

demand forecasts will ensure that system demand does not exceed the system capacity during

peak demands, as well as minimize the possibility of ―stranded assets‖ due to building new

facilities too far ahead of future demand. It cannot be stressed enough that water system

managers need to make accurate water demand forecasts a priority when juggling all of the

issues inherent in operating and managing a water system.

Page 61: Decreasing Climate-Induced Water Supply Risk Through Improved

57

X. References

AWE. 2012. ―Declining Water Sales and Utility Revenues: A Framework for Understanding

and Adapting.‖ Accessed June 5, 2013. http://www.allianceforwaterefficiency.org/Declining-

Sales-and-Revenues.aspx.

AWWA. 2013. ―State of the Water Industry.‖ Accessed June 4, 2013.

http://www.awwa.org/resources-tools/water-utility-management/state-of-the-water-

industry.aspx.

AWWA. 2012. ―Buried No Longer: Confronting America’s Water Infrastructure Challenge.‖

Accessed June 6, 2013.

http://www.awwa.org/Portals/0/files/legreg/documents/BuriedNoLonger.pdf.

AWWA Climate Change Committee. 2011. ―Committee Report: Sustainability of Water

Resources Depends on Implementing Our Knowledge on Climate Variability.‖ Journal AWWA,

103:6:42.

BIER. 2013. Beverage Industry Environmental Roundtable – Water Stewardship. Accessed June

25, 2013. http://bieroundtable.com/water_stewardship.html

Billings, R.B., and C.V. Jones. 2008. ―Forecasting Urban Water Demand.‖ American Water

Works Association, Denver, CO.

Boland, J. 1997. ―Assessing Urban Water Use and the Role of Water Conservation Measures

Under Climate Uncertainty.‖ Climatic Change, vol. 37, pp. 157-176.

Charlotte-Mecklenburg Utilities. 2013. ―Smart Irrigation Program.‖ Accessed June 25, 2013.

http://charmeck.org/city/charlotte/Utilities/WaterSmart/Pages/SmartIrrigationProgram.aspx.

CUWCC (California Urban Water Conservation Council). 2013. ―Memorandum of

Understanding.‖ Accessed June 5, 2013. http://www.cuwcc.org/mou-main-

page.aspx?ekmensel=b86195de_24_0_7872_1

Denver Water. 2013. Xeriscape. Accessed June 25, 2013.

http://www.denverwater.org/conservation/xeriscape/.

Donkor, E.A., T.A. Mazzuchi, R. Soyer, and J. Alan Roberson. 2012. ―Urban Water Demand

Forecasting: A Review of Methods and Models.‖ Journal of Water Resources Planning and

Management. 10.1061/(ASCE)WR.1943-5452.0000314.

EPA. 2013. WaterSense. Accessed June 5, 2013. http://www.epa.gov/watersense/

EPA. 2004. Emergency Response Plan Guidance for Small and Medium Community Water

Systems to Comply with the Public Health Security and Bioterrorism Preparedness and Response

Act of 2002. EPA 816-R-04-002.

Page 62: Decreasing Climate-Induced Water Supply Risk Through Improved

58

Haagenson, E., B. Rajagopalan, S.R. Summers, and A.J. Roberson. 2013. ―Projecting Demand

Extremes Under Climate Change Using Extreme Value Analysis.‖ Journal AWWA, 105:2:37.

Honaker, J., and G. King. 2010. ―What to Do About Missing Values in Time-Series Cross-

Section Data.‖ American Journal of Political Science. 54:2:561.

ICPRB (Interstate Commission on the Potomac River Basin). 2013. Drinking Water. Accessed

June 5, 2013. http://www.potomacriver.org/2012/drinking-water

Kiefer, J. 2006. ―Prevailing Water Demand Forecasting Practices and Implications for

Evaluating the Effects of Climate Change.‖ Reproduced in Climate Change and Water:

International Perspectives on Mitigation and Adaptation, Copyright © 2010 American Water

Works Association, IWA Publishing.

Kiefer, J., Clayton, J., Dziegielewski, B., and J. Henderson. 2013 (forthcoming). ―Analysis of

Changes in Water Use under Regional Climate Change Scenarios.‖ Water Research Foundation,

Denver.

Loudoun Water. 2013. ―Conservation.‖ Accessed June 25, 2013.

http://www.lcsa.org/Residential-Customers/Conservation/.

Means, E.G., M.C. Langier, J.A. Daw, and D.M. Owen. 2010. ―Impacts of Climate Change on

Infrastructure Planning and Design: Past Practices and Future Needs.‖Journal AWWA, 102:6:56.

Miller, K., and D. Yates. 2006. ―Climate Change and Water Resources: A Primer for Municipal

Water Providers.‖ Water Research Foundation (formerly AWWA Research Foundation) and

University Corporation for Atmospheric Research, Denver and Boulder, CO.

Milly, P.C.D., J. Betancourt, M. Falkenmark, R.M. Hirsch, Z.W. Kundzewicz, D.P. Lettenmaier,

and R.J. Stouffer. 2008. ―Stationarity is Dead: Whither Water Management?‖ Science, 319

(5863) 573-574.

NCADAC. 2013a. ―NCADAC Draft Climate Assessment Report – Executive Summary.‖

Accessed June 4, 2013. http://ncadac.globalchange.gov/download/NCAJan11-2013-

publicreviewdraft-chap1-execsum.pdf.

NCADAC. 2013b. ―NCADAC Draft Climate Assessment Report – Chapter 3 Water Resources.‖

Accessed June 4, 2013. http://ncadac.globalchange.gov/download/NCAJan11-2013-

publicreviewdraft-chap3-water.pdf.

NCDC. 2013. Land-Based Station Data. Accessed June 5, 2013 http://www.ncdc.noaa.gov/data-

access/land-based-station-data.

NCDC. 2012. NOAA National Climatic Data Center, State of the Climate: National Overview

for Annual 2012. Accessed May 25, 2013. https://www.ncdc.noaa.gov/sotc/national/2012/13.

Page 63: Decreasing Climate-Induced Water Supply Risk Through Improved

59

North, D.W., F. Selker, and T. Guardino. 1994. ―The Value of Research on Health Effects of

Ingested Inorganic Arsenic.‖Arsenic Exposure and Health, Science and Technology Letters,

Northwood.

Olmstead, S.M. 2010. ―The Economics of Managing Scarce Water Resources.‖ Review of

Environmental Economics and Policy. 4:2:179.

Rockaway, T.D., P.A. Coomes, J. Rivard, and B. Kornstein. 2011. ―Residential Water Use

Trends in North America.‖ Journal AWWA, 103:2:76

Rosen, J.P. 2013. Personal communication.

Saving Water Partnership. 2013. ―Preserving the Customer Conservation Ethic.‖ Accessed June

25, 2013. http://www.savingwater.org/docs/2013WaterConsProgActionsStrategies.pdf.

Smithsonian. 2013. ―The Changing Demographics of America.‖ Smithsonian. Accessed June

25, 2013. http://www.smithsonianmag.com/specialsections/40th-anniversary/The-Changing-

Demographics-of-America.html.

SFWMD (South Florida Water Management District). 2012. ―Basis of Review for Water Use

Permit Applications.‖ Accessed June 4, 2013.

http://www.sfwmd.gov/portal/page/portal/xrepository/sfwmd_repository_pdf/bor_wu.pdf.

USGBC. 2013. ―LEED: Water Efficient Landscaping – No potable water use or no irrigation.‖

Accessed June 5, 2013. http://www.usgbc.org/credits/new-construction/v22/wec12.

WateReuse Foundation. 2008. ―National Database of Water Reuse Facilities Summary Report.‖

http://www.watereuse.org/files/s/docs/02-004-01.pdf. WateReuse Foundation, Alexandria, VA.

WaterRF. 2013. The Foundation: Research Programs: Focus Area Program. Accessed June 4,

2013. http://www.waterrf.org/the-foundation/research-programs/Pages/Focus-Area-

Program.aspx.

Water Utility Climate Alliance. 2010. ―Decision Support Planning Methods: Incorporating

Climate Change Uncertainties into Water Planning – January 2010.‖ Accessed June 3, 2013.

http://www.wucaonline.org/assets/pdf/actions_whitepaper_012110.pdf.

Whelton, A.J, et al. 2007. ―Using Customer Feedback for Improved Water Quality and

Infrastructure Monitoring.‖ Journal AWWA, 99:11:62.