THE IMPACTS OF WATER CONSERVATION STRATEGIES ON WATER...

THE IMPACTS OF WATER CONSERVATION STRATEGIESON WATER USE: FOUR CASE STUDIES1

Yushiou Tsai, Sara Cohen, and Richard M. Vogel2

ABSTRACT: We assessed impacts on water use achieved by implementation of controlled experiments relatingto four water conservation strategies in four towns within the Ipswich watershed in Massachusetts. The strate-gies included (1) installation of weather-sensitive irrigation controller switches (WSICS) in residences and muni-cipal athletic fields; (2) installation of rainwater harvesting systems in residences; (3) two outreach programs:(a) free home indoor water use audits and water fixture retrofit kits and (b) rebates for low-water-demand toiletsand washing machines; and (4) soil amendments to improve soil moisture retention at a municipal athletic field.The goals of this study are to summarize the effectiveness of the four water conservation strategies and to intro-duce nonparametric statistical methods for evaluating the effectiveness of these conservation strategies in reduc-ing water use. It was found that (1) the municipal WSICS significantly reduced water use; (2) residences withhigh irrigation demand were more likely than low water users to experience a substantial demand decreasewhen equipped with the WSICS; (3) rainwater harvesting provided substantial rainwater use, but these volumeswere small relative to total domestic water use and relative to the natural fluctuations in domestic water use;(4) both the audits ⁄ retrofit and rebate programs resulted in significant water savings; and (5) a modelingapproach showed potential water savings from soil amendments in ball fields.

(KEY TERMS: water conservation; water demand management; water resource planning; nonparametricstatistics; controlled experiments.)

Tsai, Yushiou, Sara Cohen, and Richard M. Vogel, 2011. The Impacts of Water Conservation Strategies onWater Use: Four Case Studies. Journal of the American Water Resources Association (JAWRA) 1-15. DOI:10.1111/j.1752-1688.2011.00534.x

INTRODUCTION

The Ipswich watershed, situated north of metropol-itan Boston, MA, has experienced unnaturally low orno flows during some summer months in recent yearsowing in part, to increases in public water supplies(Canfield et al., 1999; Zarriello and Ries, 2000). The

ongoing streamflow depletion has raised awareness ofthe importance of water demand management amongthe water authorities, and as a result, the Massachu-setts Department of Conservation and Recreation(MDCR) launched a project, funded by the U.S. Envi-ronmental Protection Agency (USEPA), in an attemptto identify and pilot strategies that could help restoreinstream flows to the Ipswich River. In coordination

1Paper No. JAWRA-09-0196-P of the Journal of the American Water Resources Association (JAWRA). Received December 19, 2009; acceptedFebruary 8, 2011. ª 2011 American Water Resources Association. Discussions are open until six months from print publication.

2Respectively, Research Associate, Department of Civil and Environmental Engineering, Tufts University, Medford, Massachusetts 02155;Water Resources Specialist, Massachusetts Department of Conservation and Recreation, Boston, Massachusetts 02114; and Professor, Depart-ment of Civil and Environmental Engineering, Tufts University, Medford, Massachusetts 02155 (E-Mail ⁄ Tsai: [email protected]).

Re-use of this article is permitted in accordance with the Terms and Conditions set out at http://wileyonlinelibrary.com/onlineopen#OnlineOpen_Terms

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 1 JAWRA

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

AMERICAN WATER RESOURCES ASSOCIATION

with four communities in the Ipswich watershed,four water conservation projects were designed tosimultaneously meet immediate municipal needs anddemonstrate innovative water conservation strategiesthat could be evaluated with real-world data. Thefour projects are (1) installation of weather-sensitiveirrigation controller switches (WSICS) at residencesand at municipal athletic fields, (2) installation ofrainwater harvesting systems at residences, (3) town-administered programs to provide (a) home indoorwater use audits and fixture retrofit kits and(b) rebates for low-water-demand toilets and washingmachines, and (4) soil amendments to improve mois-ture retention and reduce water demand at a munici-pal athletic field.

The primary goal of this study is to evaluate theeffectiveness of four water conservation pilot strate-gies on water use. As is inherent to many small-scalepilots, the datasets for these demonstration projectstend to be small, variable, and exhibit nonnormal dis-tributions. A secondary goal of this study is to demon-strate the application of mostly nonparametricstatistical methods for their ability to enable sensibleinferences to be drawn, in some cases, even from thevery small samples.

Vickers (2001) has reviewed approaches relating towater conservation strategies for municipal, indus-trial, and residential uses. Hilaire et al. (2008) havesummarized factors impacting the efficiency of wateruse in the urban landscape: water conservation strate-gies, landscape design, economic and noneconomicincentives, irrigation ⁄ water application and reusetechnologies, and people-plants relationship. Most pre-vious research on water conservation strategiesinvolves price incentives. Literature on the price elas-ticity of water use – impact of water price on waterdemand – is so well developed that meta-analysis isnow possible (e.g., see the meta-analysis of 64 previousstudies by Dalhuisen et al., 2003). A review of researchrelating to nonprice water conservation strategies, inwhich price incentives are not used, reveals fewerstudies. We note three general approaches to non-price water conservation research: (1) behavioralapproaches, (2) retrospective analyses, and (3) con-trolled experiments. Examples of the first approachare provided by Corral-Verdugo and Frias-Armenta(2006) and others who have evaluated the impact ofsocial norms (an understanding of the attitudes andbehavior of others) on water conservation behavior.Similarly, Atwood et al. (2007) and others have identi-fied the key behavioral, community, and othersocioeconomic factors that impact water conservation,such as gender, environmental attitudes, and neigh-borhood features. Gilg and Barr (2006) have provideda review of research that summarizes behavioral atti-tudes toward water conservation. Most previous

behavioral research on water conservation consists ofcontrolled experimental designs based on a combina-tion of surveys and multivariate statistical analyses.

A second approach to nonprice water conservationresearch involves a retrospective analysis of previouswater use behavior using available data. For exam-ple, Kenney et al. (2004) showed the importance ofwater-use restrictions in reducing water demandsduring a drought experienced by eight Colorado cit-ies. Most retrospective research on nonprice waterconservation strategies has developed multivariaterelationships for predicting residential water demandas a function of conservation efforts in addition tonumerous other factors or explanatory variables. Forexample, some of the combinations of explanatoryvariables considered for predicting water demand, inaddition to conservation efforts, include price, house-hold appliances, landscape features, metering, andclimate (Bamezai, 1995); price, weather, and demo-graphic characteristics (Kenney et al., 2008); priceand public information (Wang et al., 1999; Smith andWang, 2008); price, weather, household income,municipalities, public information, and education(Michelsen et al., 1999); price, public information,weather, household characteristics, water use restric-tion, and ration (Renwick and Green, 2000); or price,public information, weather, household characteris-tics, use restriction, ration, and month (Renwick andArchibald, 1998). For those cited studies, the demandelasticity in response to conservation efforts rangedfrom 0.03 to )4.51 for indoor strategies and 0(unresponsive) to )4.81 for outdoor strategies.

On the other hand, the price elasticity of waterdemand reported in previous research on priceapproaches to water conservation varies. For example,Espey et al. (1997) found that price elasticity rangedfrom )0.02 to )0.75 for 75% of price elasticity esti-mates, whereas Brookshire et al. (2002) found esti-mates ranging from )0.11 to )1.59, and althoughDalhuisen et al. (2003) concluded that price elasticityof water demand is relatively elastic, the authorscautioned that price elasticity varied depending onfunctional form selection, aggregation level, data char-acteristics, and estimation issues. In conclusion, thesestudies indicate that the effectiveness of both nonpriceand price approaches varied drastically, thus we areunable to judge from previous research whether non-price or price approaches are more effective. Moreover,Dalhuisen et al. (2003) has concluded that price elas-ticity in East United States is insignificant; therefore,in the context of our analysis, it is probably safe toview economic incentives to be relatively ineffective incomparison with other incentives considered here.

A third approach to nonprice water conservationresearch, and the approach used here, involvesthe use of controlled experiments combined with

TSAI, COHEN, AND VOGEL

JAWRA 2 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

statistical methods. Here controlled experiments areperformed with actual water conservation methods.For example, Karpiscak et al. (2001) estimated watersavings by monitoring a water conservation demon-stration house. The water savings reported byKarpiscak et al. (2001), however, may not be an accu-rate response to a single water conservation strategybecause the synergistic effects associated with multi-ple water conservation practices implemented insidethe demonstration house were not considered. Buch-berger and Wells (1996) monitored residential waterdemand at four households over a one-year periodand used that information to develop stochastic mod-els of residential water demands. Although theirwork did not deal directly with water conservationefforts, such research could provide important inputsto future water conservation strategies. Mayer et al.(2003, 2004) and Ayres Associates (1996) haveemployed t-tests to assess water savings due to vari-ous water conservation strategies in an experimentalgroup relative to a control group.

There are a few examples of the type of researchperformed here, in which designed experiments areused to evaluate the effectiveness of water conserva-tion technologies and programs using hypothesis tests(Ayres Associates, 1996; Mayer et al., 2003, 2004).Those studies employed traditional parametric statis-tical methods, and the applicability of the t-test usedin these studies was not assessed by an investigationof probability distributions of the datasets. Theresearchers assumed that the data arose from a nor-mal distribution without performing normalitychecks. Here, we are careful to confirm the suitabilityof statistical methods before their application to con-trolled experiments to assess the effectiveness of eachof four independent water conservation strategies.We begin by providing an overview of the four conser-vation strategies considered and reviewing the statis-tical methods employed.

METHODOLOGIES

Design of Water Conservation Strategies

Four water conservation strategies designed toreduce water use were implemented in the IpswichRiver watershed by MDCR, with funding from theUSEPA. Due to the critical contribution of outdoorirrigation to the summertime streamflow deficit (Ips-wich River Watershed Action Plan, 2003), thesewater conservation strategies piloted and evaluatedhere have a strong emphasis on reducing lawn andathletic field irrigation. The installation of WSICS at

residences and municipal athletic fields, the installa-tion of rainwater harvesting systems, and the intro-duction of moisture-retaining soil amendments at anathletic field are all strategies designed to mitigatewater withdrawals for irrigation purposes during thesummer months. In addition, the home audit ⁄ retrofitand appliance rebate programs aim to mitigate with-drawals for indoor water use, year round. Each casestudy was designed in cooperation with one or moremunicipality in the watershed, based on an opportu-nistic assessment of water conservation needs andprogrammatic resources.

This section, along with Table 1, summarizes thewater savings hypothesis and evaluation design foreach of the four demonstration projects. The WSICSare designed to only trigger an irrigation cycle whenthe soil moisture is low, as estimated from regionalweather conditions and local rainfall. By deliveringwater optimally, such technology should reduce over-all irrigation demand by eliminating extraneouscycles triggered by automatic timers that are insensi-tive to weather conditions. The rainwater harvestingsystems store rainwater, providing a direct alterna-tive to the use of public drinking water for nonpo-table outdoor uses. We thereby anticipated that thesystems would reduce demand on household publicwater consumption. The moisture-retaining soilamendments were designed to extend the time thatmoisture remains available to the turf roots withinthe soil. As a result, we anticipated that the fieldcould tolerate reductions in irrigation volume withoutcompromising turf health. The audit ⁄ retrofit programwas anticipated to reduce water use in participatinghouseholds by leading to the direct repair of leaksand the replacement of faucets and water fixtureswith more efficient alternatives. The rebate programwas anticipated to similarly reduce household wateruse by encouraging the conversion to water-efficienttoilets and washing machines.

A summary of evaluation design for all four waterconservation strategies is documented in Table 1.This table includes the sample sizes associated withthe control and experimental populations, the timeperiods associated with the installation and thepre- and postexperiment evaluations, the timeperiods excluded from the analysis, and a list of theconfounding factors.

Statistical Methods

A wide range of statistical methods are considereddue to the different experimental designs and natureof the four water conservation strategies, which weredesigned in accordance with towns’ specific needsand administrative abilities. Nonparametric statistical

THE IMPACTS OF WATER CONSERVATION STRATEGIES ON WATER USE: FOUR CASE STUDIES


TA

BL

E1.

Su

mm

ary

ofE

valu

ati

onD

esig

nfo

rth

eF

our

Wate

rC

onse

rvati

onS

trate

gie

s.

Co

nse

rv

ati

on

Str

ate

gie

sa

nd

Ev

alu

ati

on

Desi

gn

Sa

mp

leS

ize

Da

teo

fIn

sta

lla

tio

no

rS

tart

of

Imp

lem

en

tio

nP

re

Perio

dP

ost

Perio

do

rS

imu

lati

on

Perio

d

Tim

eP

erio

dE

xclu

ded

Fro

mA

na

lysi

sC

on

fou

nd

ing

Fa

cto

rs

Res

iden

tial

WS

ICS

–E

valu

ati

on1:

Com

pari

son

ofex

per

imen

tal

gro

up

’sou

tdoo

rw

ate

ru

sep

rior

toan

daft

erW

SIC

Sin

stall

ati

on;

wea

ther

con

trol

led

for

by

com

pari

ng

toco

ntr

olgro

up

’sou

tdoo

rw

ate

ru

sed

uri

ng

sam

eti

me

per

iod

s

Con

trol

:71

Exp

erim

ent:

9(2

hou

seh

old

sin

the

Pil

otG

rou

pfo

rE

valu

ati

on2

bel

oww

ere

excl

ud

edfr

omth

isan

aly

sis,

as

they

join

edth

est

ud

yin

2006-2

007)

Su

mm

erof

2005

2001-2

004

qu

art

erly

wate

ru

sed

ata

from

ded

icate

dou

tdoo

rw

ate

rm

eter

s

2006-2

007

qu

art

erly

wate

ru

sed

ata

from

ded

icate

dou

tdoo

rw

ate

rm

eter

s

2005

wate

ru

sed

ata

wer

eex

clu

ded

,bec

au

se2005

was

atr

an

siti

onal

yea

r

Part

icip

ati

onw

as

vol

un

tary

.S

elf-

sele

ctio

nm

ay

bia

ssa

mp

leto

ward

(1)

hig

hin

itia

lw

ate

ru

sers

(exp

ect

pos

itiv

esa

vin

gs)

or(2

)co

nse

rvati

ve

init

ial

wate

ru

sers

(exp

ect

neg

ati

ve

savin

gs)

Res

iden

tial

WS

ICS

–E

valu

ati

on2:

Com

pari

son

ofex

per

imen

tal

gro

up

’sact

ual

wate

ru

sep

rior

toW

SIC

Sin

stall

ati

onto

sim

ula

ted

wate

ru

sew

ith

WS

ICS

for

sam

ep

erio

d,

usi

ng

kn

own

his

tori

cw

eath

erre

cord

s

Exp

erim

ent:

11

Su

mm

erof

2005

(9p

art

icip

an

ts)

Su

mm

ers

of2006

or2007

(2p

art

icip

an

ts)

2003-2

004

act

ual

an

nu

al

aggre

gate

dou

tdoo

rw

ate

ru

se

2003-2

004

sim

ula

ted

an

nu

al

aggre

gate

dou

tdoo

rw

ate

ru

se

Mu

nic

ipal

WS

ICS

–C

omp

ari

son

ofact

ual

wate

ru

sep

rior

toW

SIC

Sin

stall

ati

onto

sim

ula

ted

wate

ru

sew

ith

WS

ICS

for

sam

ep

erio

d,

usi

ng

kn

own

his

tori

cw

eath

erre

cord

s

5ball

fiel

ds

Su

mm

erof

2005

2003-2

004

act

ual

an

nu

al

aggre

gate

dw

ate

ru

se

2003-2

004

sim

ula

ted

an

nu

al

aggre

gate

dw

ate

ru

se

Rain

wate

rh

arv

esti

ng

–C

omp

ari

son

ofd

omes

tic

(pu

bli

c)su

mm

erw

ate

ru

sep

rior

toan

daft

erin

stall

ati

onof

rain

wate

rh

arv

esti

ng

syst

em;

com

pari

son

ofra

inw

ate

ru

sean

dra

infa

llca

ptu

reef

fici

ency

bet

wee

n0.7

6an

d3.0

3m

3sy

stem

s

0.7

6m

3(2

00-g

all

on)

tan

k:

26

3.0

3m

3(8

00-g

all

on)

tan

k:

11

1.3

8m

3(3

65-g

all

on)

tan

k:

12.2

7m

3(6

00-g

all

on)

tan

k:

1

Mid

-Ap

ril

2006

2004-2

005

dom

esti

cw

ate

ru

se2006-2

007

dom

esti

cw

ate

ru

se2006-2

007

rain

wate

ru

se

Su

bst

an

tial

dif

fere

nce

insc

ale

bet

wee

nd

omes

tic

an

dra

inw

ate

ru

se



TA

BL

E1.

Con

tin

ued

.

Co

nse

rv

ati

on

Str

ate

gie

sa

nd

Ev

alu

ati

on

Desi

gn

Sa

mp

leS

ize

Da

teo

fIn

sta

lla

tio

no

rS

tart

of

Imp

lem

en

tio

nP

re

Perio

dP

ost

Perio

do

rS

imu

lati

on

Perio

d

Tim

eP

erio

dE

xclu

ded

Fro

mA

na

lysi

sC

on

fou

nd

ing

Fa

cto

rs

Au

dit

⁄ret

rofi

tan

dre

bate

pro

gra

ms

–C

omp

ari

son

ofw

inte

rw

ate

ru

sep

rior

toan

daft

erau

dit

⁄ret

rofi

tev

ent

orin

stall

ati

onof

one

orm

ore

rebate

-qu

ali

fyin

gap

pli

an

ce;

part

icip

an

tsgro

up

edin

tom

utu

all

yex

clu

sive

cate

gor

ies

by

typ

eof

part

icip

ati

on.

Con

trol

gro

up

use

dto

con

trol

for

exte

rnal

cau

ses

ofp

rev.

pos

td

iffe

ren

ces

Au

dit

⁄Ret

rofi

t:99

Reb

ate

-toi

let:

87

Reb

ate

-wash

er:

527

Reb

ate

-toi

let

an

dw

ash

er:

30

Au

dit

⁄Ret

rofi

tan

dR

ebate

(toi

let

an

d⁄o

rw

ash

er):

32

Con

trol

gro

up

:5,0

50

Vari

able

,beg

inn

ing

Sep

tem

ber

2003

for

Au

dit

⁄Ret

rofi

tsan

dJu

ly2003

for

Reb

ate

s

Qu

art

erly

win

ter

wate

ru

sefr

omF

ebru

ary

2001

toau

dit

date

orap

pli

an

cein

stall

ati

ond

ate

Con

trol

gro

up

esta

bli

shed

for

each

pre

v.

pos

tp

erio

d

Qu

art

erly

win

ter

wate

ru

sefr

omau

dit

date

orap

pli

an

cein

stall

ati

ond

ate

toM

ay

2007

Con

trol

gro

up

esta

bli

shed

for

each

pre

v.

pos

tp

erio

d

On

lyw

inte

rqu

art

ers

incl

ud

ed:

(beg

an

onor

aft

erO

ctob

er19

an

den

ded

onor

bef

ore

Ap

ril

14)

Soi

lam

end

men

ts–

Com

pari

son

ofsi

mu

late

dw

ate

ru

sebet

wee

nco

ntr

olan

dex

per

imen

tal

fiel

ds

du

rin

gsa

me

per

iod

,u

sin

gk

now

nh

isto

rica

lw

eath

erre

cord

san

dk

now

nw

eath

ertr

igger

sfo

rir

rigati

onon

bot

hfi

eld

s

Con

trol

:1

Exp

erim

ent:

12003-2

007

sim

ula

ted

wate

ru

sed

ata

for

con

trol

fiel

d

2003-2

007

sim

ula

ted

wate

ru

sed

ata

for

exp

erim

enta

lfi

eld



methods are often recommended over parametricmethods (Helsel and Hirsch, 2002) when sample sizesare limited and ⁄ or in cases when a probability distri-bution cannot be determined for the random variableof concern. Here, we used mostly nonparametrichypothesis tests, because most of the datasets wereeither too small and ⁄ or they violated various assump-tions required for parametric hypothesis tests to bemeaningful. We assumed, throughout our analyses,that the type I error probability a was 5%.

We used nonparametric confidence intervals forthe true population median because the probabilitydistributions of the original random variables couldnot be confirmed for small samples. Such confidenceintervals for the true population median, shown inmany subsequent figures, are used to assess whetherthe median estimated from one sample differs fromthe median estimated from another sample. Helseland Hirsch (2002) suggested that the nonparametricinterval for the median can be estimated using thebinomial probability distribution. The probability ofan observation being above or below the median isequal so that p = 0.5. For a sample size n, the cumu-lative probability p(x) of x observations exceeding themedian is then

pðxÞ ¼Xx

y¼0

n!

y!ðn� yÞ! 0:5yð1� 0:5Þn�y; 8x¼ 1;2; . . . ;n:

ð1Þ

The lower bound of the interval can be estimatedusing the (x + 1)th smallest observation, where x cor-responds to p(x) = 0.025, which reflects a 2.5% proba-bility in each tail of the distribution of x. The upperbound of the interval can be estimated using the(n ) x)th smallest observation. The resulting confi-dence intervals for the median reflect the distribu-tions of the estimates of medians drawn from anydataset of length n. For cases where the sample sizesare large (n > 20), one may use a normal approxima-tion to the binomial distribution in Equation (1) lead-ing to the rank corresponding to the lower bound ofthe interval estimate of:

Rl ¼n� Z0:025 �

ffiffiffinp

2for n>20; ð2Þ

and the upper bound of the interval estimate is theRuth smallest observation, where

Ru ¼nþ Z0:025 �

ffiffiffinp

2þ 1 for n>20; ð3Þ

and Z0.025 = 1.96.In some instances, we were able to employ hypoth-

esis tests based on the assumption of a normal distri-bution. To check whether observations of a sampleare normally distributed, the normal probability plotcorrelation coefficient (PPCC) was computed andchecked against its critical value given in table 18.3.3of Stedinger et al. (1993). The normal quantiles wereestimated using Blom’s unbiased, plotting position fornormal variates (Stedinger et al., 1993):

pi ¼i� 3=8

nþ 1=4; ð4Þ

where i is the ith observation when ranked in ascend-ing order.

The hypothesis tests used in this study, corre-sponding to the various types of comparisons, are doc-umented in Table 2. The sign test was chosen overthe sign rank test and the paired rank-sum testbecause the latter two assume a symmetrical distri-bution of the observations and most of our datasetsare asymmetrical.

WATER SAVINGS ASSOCIATED WITH WATERCONSERVATION STRATEGIES

The following sections summarize the effectivenessat reducing water demand of the four water conserva-tion strategies.

TABLE 2. The Hypothesis Tests Used in This Study Are Presented by Shaded Cells.

ComparisonBetweenor Among

One Sample and TwoDependent Samples Two Independent Samples

More Than TwoIndependent

SamplesMore Than Two

Dependent Samples

Nonparametric tests Sign test1,2 Rank-sum test (orWilcoxon-Mann-Whitney test)2

Kruskal-Wallis test1,2 Appropriate test isnot available1

Parametric tests t-test1,2

Paired t-test1,2Two-sample t-test1,2 One-way ANOVA1,2 Two-way ANOVA or

multi-way ANOVA1

1Zar (1999); 2Helsel and Hirsch (2002).



Weather-Sensitive Irrigation Controller Switches

A total of 11 WSICS were evaluated on residentialproperties and 5 in municipal athletic fields. Thesedevices (Weather Reach WR-7� by Irrisoft�, Logan,UT, USA) contain an on-site rain gage and receivecontinuous solar radiation, temperature, relativehumidity, and wind data from a regional weatherstation (town of Ipswich) via wireless transmission.Based on this information, the WSICS device isdesigned to deliver water only when needed by thelandscape.

Residential WSICS

Approximately 150 residences in the town of Read-ing, MA, have exclusive outdoor water meters.Among this group, nine households that met ourexperimental group criteria had WSICS installed dur-ing the summer of 2005, and two during the followingtwo summers. Criteria included continuous owner-ship and use of an automatic irrigation system since2001. An additional 71 households with dedicatedoutdoor meters meeting these criteria were selectedas the control group. For this analysis, quarterly out-door water use records were obtained from the Read-ing Water Department for all households in the studyfrom January 2001 through November 2007.

For the nine residences whose WSICS wasinstalled in 2005, a single value representing historic(‘‘pre’’) water use (pre-experimental condition) wasobtained by averaging the annual outdoor water usefrom 2001 to 2004, and a single value representingwater use during the experimental period (‘‘post’’)was obtained by averaging the annual outdoor usefrom 2006 to 2007. Data from 2005 were excludedfrom the analysis due to this being a transitionalyear. Because a PPCC normality test determined thatthe control group was not well approximated by anormal distribution, the nonparametric rank-sumhypothesis test was used to compare the water use ofboth the control and experimental groups as shownin Figure 1. There is no statistically significant differ-ence between the water use of the control and experi-mental groups in either the ‘‘pre’’ or ‘‘post’’ periods,which can be seen visually in Figure 1 with the over-lapping confidence intervals. However, a visualassessment of Figure 1 also suggests that the WSICSmay have reduced the variability of water use amongthe experimental group, especially among high waterusers.

Rank-sum tests applied to the rainfall records froma nearby water treatment plant suggest that typicaltotal rainfall and number of days of rain betweenMay 15 and October 15 (the approximate irrigation

season) were statistically indistinguishable duringthe ‘‘pre’’ and ‘‘post’’ periods. Thus, we were comfort-able calculating ‘‘savings’’ for each household by sub-tracting the ‘‘post’’ from the ‘‘pre’’ period water use.The results of a rank-sum test do not show that thewater savings for households with the WSICS weredifferent than for the control group. The large rangeassociated with the confidence interval (Figure 2) forthe median of the experimental group is due to thesmall experimental sample size and large variation inresponse to the WSICS installation within the group.Nevertheless, Figure 2 illustrates that although theaverage household in the control group saw a drop inwater demand of 3.27 m3 ⁄ year between the twotime periods, the average WSICS household saw a

"Post" period"Pre" period

350

300

250

200

150

100

50

0

Control (71 Households)Experiment (9 Households)

Group

⊕Narrower boxes: 95% confidence intervals for the medians

: mean

Wat

er U

se(m

3/y

ear/

hou

seh

old)

FIGURE 1. BoxPlots Comparing Annual Outdoor WaterUse in the Control and Experimental Groups in Both the‘‘Pre’’ (2001 to 2004) and ‘‘Post’’ (2006 to 2007) Periods.

ExperimentControl

200

150

100

50

0

-50

-100

-150

0


Group

24.0715.57

40.69

3.27

⊕ : meanNarrower boxes: 95% confidence intervals for the medians

Wat

erS

avin

gs (

m3/y

ear/

hou

seh

old)

FIGURE 2. BoxPlots Showing Water Savings During the ‘‘Post’’Period Relative to the ‘‘Pre’’ Period, in Both Groups. For eachhousehold, this value represents ‘‘post’’ period water use subtractedfrom ‘‘pre’’ period water use. A value <0 implies more water wasused during the ‘‘post’’ than ‘‘pre’’ period.



reduction of 40.69 m3 ⁄ year. Although this differenceis not statistically significant, it reflects the fact thathouseholds with high ‘‘pre’’ period water demand sawa large reduction in water use postinstallation. Asshown in Figure 3, when only the highest ‘‘pre’’ per-iod water users (90th percentile; annual use>261.6 m3) are included in the analysis, the watersavings for the experimental group is significantlygreater than the control group. These results suggestthat households with high irrigation water demandsare more likely to reduce their water use due to theWSICS installations. Our analysis also highlights theimportance of increasing the sample size of the exper-imental group of households in any future studies.

Retrospective Analysis

A retrospective analysis of the WSICS comparedactual outdoor water used by each experimentalhousehold in 2003 and 2004 to the estimated volumeof water that would have been applied by the WSICSduring that same period. This analysis required cal-culating the number of irrigation cycles that wouldhave been triggered for each system, based on:(1) weather data from that period, (2) the algorithmused by the WSICS units to trigger irrigation cyclesbased on weather data, and (3) each system’s individ-ual ‘‘evapotranspiration (ET) threshold.’’ ET thresh-olds are used to set the tolerance for how muchestimated ET should be allowed before an irrigationcycle is triggered to replenish the loss. The number oftriggered irrigation cycles was then converted to avolume for each household by multiplying it bythe appropriate per-cycle volume. The latter wasdetermined at each residence by reading the water

meter before and after a test irrigation cycle. Thisapproach was only applied to 2003 and 2004 to coin-cide with the years for which the extensive weatherdata needed in the algorithm was available. A PPCCnormality hypothesis test suggests that the nonpara-metric sign test is preferred over a parametric testfor assessing the difference between the actual andsimulated water uses. Although positive overall meanand median water savings (22.60 and 29.28 m3 ⁄household ⁄ year, respectively) are reported when com-paring simulated with actual use, we conclude fromthe nonparametric sign test that the savings is notsignificantly different from zero, owing to the largevariation in the small sample. When this analysis isapplied only to water users with high actual wateruse (use >261.6 m3), during the years their useexceeded this threshold, the average savings is statis-tically significant at 135.8 m3 ⁄ household ⁄ year. How-ever, this sample consisted only of one year of datafor each of three households.

In summary, two approaches were used: (1) com-paring outdoor water use in households whereWSICS were installed to outdoor water use in controlhouseholds, both prior to and after installation; and(2) the retrospective analysis, comparing actual wateruse to theoretical water use had the WSICS beeninstalled in 2003 and 2004. Both approaches confirmthat even though overall water savings for the experi-mental group is greater than that for the controlgroup, the difference in the savings between the twogroups was not statistically significant owing to thehighly variable savings in the experimental group.WSICS were, however, likely to result in water sav-ings when installed at residences with high outdoorwater demands. Although we did not assess the effi-ciency of individual watering regimes prior to WSICSinstallation, the significant response to the systemsamong the highest water users suggests over-water-ing by these households prior to the WSICS installa-tion, as WSICS systems are designed specifically toreduce unnecessary irrigation.

Municipal WSICS

In addition to residential WSICS, five municipalathletic fields across two municipalities (Reading andMiddleton, MA) were equipped with WSICS in thesummer of 2005. A retrospective analysis was con-ducted using the same methodology as describedabove for the residential participants. Hypotheticalwater use was derived by simulating irrigation trig-gers that would have been signaled by the WSICS,had they been installed during 2003 and 2004, usingweather records from that period and each field’sWSICS ET thresholds and irrigation cycle volumes.

ExperimentControl

200

150

100

50

0

-50

-100

-150

0


Group

155.9

-23.32

⊕ : meanNarrower boxes: 95% confidence intervals for the medians

Wat

er S

avin

gs (

m3/y

ear/

hou

seh

old)

FIGURE 3. Comparison of the Annual Outdoor WaterSavings Between the Control and Experimental Groups

Among the 90th Percentile of ‘‘Pre’’ Period (2001 to 2004)Water Users (annual use >261.6 m3).



This simulated use was compared with actual wateruse for each of the five fields aggregated for 2003 and2004 (Figure 4). Theoretical water savings wereobtained by subtracting simulated use from actualuse for each field for each year. Nonparametric testswere used again due to a sample size of 10 (two yearseach, for five fields). The sign test indicates that asignificant positive water savings would haveresulted from the WSICS installations. A box plot ofthe theoretical water savings (Figure 5) indicates thatthis statistically significant average savings wasapproximately 0.11 m3 ⁄ m2 ⁄ year (equal to 121,000gallons ⁄ acre ⁄ year).

Rainwater Harvesting

Rainwater harvesting systems are designed to cap-ture runoff from rooftops and store the water for non-potable uses, such as lawn and garden watering. Oneintent of such systems is to reduce demand on publicwater supplies by replacing potable water that wouldotherwise be used for these outdoor purposes. A totalof 39 rainwater harvesting systems were installed onresidential properties mid-April 2006 in the town ofWilmington, MA, based on a lottery among 150 inter-ested households. The systems consist of a storagetank, a pressure pump to aid in water distribution, aspigot for a hose, and a water meter to measure flowpumped from the tanks. Two different sizes of storagetanks were installed: twenty-eight 0.76 m3 (200-gal-lon) and eleven 3.03 m3 (800-gallon) tanks. Two ofthe participants with 200-gallon tanks upgraded theirstorage capacity to 1.38 m3 (365 gallons) and 2.27 m3

(600 gallons), respectively, using their own funds.Except where otherwise noted, the households withupgraded systems were excluded from the analyses.The rainwater systems were in use during the sum-mers of 2006 and 2007. Total rainwater use from thetime each system was turned on in the spring towhen it was decommissioned in the fall was recordedfor each household for 2006 and 2007. The distribu-tion of the rainwater use observations for both groupsis well-approximated by a normal distribution. Allhouseholds used the rainwater systems, and a two-sample Student’s t hypothesis test on sample meansindicates that those with 3.03 m3 tanks used signifi-cantly more rainwater than those with 0.76 m3 tanks(Figure 6).

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

FullerMeadow

ElementarySchool

CoolidgeMiddleSchool

SymondsWay Ball

Field

Pearl St.Soccer Field

Parker Jr.High

Wat

er A

pplic

atio

n Pe

r Uni

t Are

a (m

3 /m2 /y

ear)

Actual Water UseSimulated Water Use

FIGURE 4. Actual Water Use (without WSICS)and Simulated Water Use (with WSICS) Aggregated

for 2003 and 2004 for Each Ball Field.

0.30

0.25

0.20

0.15

0.10

0.05

0.00 0

0.11

0.07

⊕ : meanNarrower box: 95% confidence interval for the medians

Wat

er A

pplic

atio

n p

er U

nit

Are

a (m

3/

m2/y

ear)

FIGURE 5. Box Plot of Theoretical Water Savings (actual–simulated water use) for Each Ball Field, Each Year (2003and 2004). Mean per-unit-area savings is 0.11 m3 ⁄ m2 per

year (equal to 121,000 gallons ⁄ acre ⁄ year).

20072006800200800200

20

15

10

5

0

0.76 m (n=26)3.03 m (n=11)

Tank Capacity

⊕- : median

: mean

Rai

nw

ater

Use

(m

3/s

umm

er/h

ouse

hol

d) 3

3

0.76 m3 3.03 m3 0.76 m3 3.03 m3

(200 gal.) (800 gal.) (200 gal.) (800 gal.)

2006 2007

FIGURE 6. The Data and 95% Confidence Intervals for the Meanof the Total Rainwater Used From Both Sizes of Harvesting

Systems During the Summer Watering Seasons of 2006 and 2007.



To assess whether the use of rainwater resulted ina decrease in domestic water use, domestic water usebefore and after the installation of the rainwater har-vesting system was compared for each residentialparticipant. The visual comparison of the domesticwater use and the rainwater use in Figure 7 showsthat the volumes of rainwater used were generallyless than the fluctuation in domestic water use fromyear to year, making reductions in domestic wateruse due to rainwater difficult to discern. A rank-sumtest confirmed that, regardless of the size of thetanks, rainwater systems could not be shown toimpact summer domestic water use.

However, a written survey completed by all partici-pants who attended a meeting at the conclusion ofthe study suggests qualitatively that rainwater was afrequent substitute for domestic water among therainwater harvesting participants. The survey askedparticipants to allocate the proportion of the rainwa-ter they used across seven usage activities (one cate-gory was defined flexibly as ‘‘other’’ to captureuncommon uses) and to state for each whether theywould have used an equivalent or greater amount ofdomestic water for that purpose if they did not haveaccess to stored rainwater. All respondents (19 of 37households that were in the program; i.e., 50% of par-ticipants) estimated that at least some of their rain-water uses were direct substitutes for domestic waterthat they otherwise would have used for the samepurpose.

Twenty-five households were able to provide esti-mates of the roof area contributing to their rainwatercollection system. For each of these households, thetotal volume of rain falling on the contributing areawas estimated by multiplying contributing area bydaily rainfall depth recorded at a nearby facility forthe days the system was in use. Rainfall capture

efficiency was defined as the ratio of total volume ofrainwater used relative to the total volume of rainthat fell on the contributing roof area. Each house-hold has a unique rainfall capture efficiency, basedon the combined influences of system storage capac-ity, frequency of system use, and the pattern (distri-bution, intensity, etc.) of rainfall events. A rank-sumtest of ‘‘rainfall capture efficiency’’ by system size(Figure 8) suggests that, in 2007, households with800-gallon systems had statistically higher efficien-cies than those with 200-gallon systems, whereas in2006 the two groups had statistically equivalent effi-ciencies. The efficiencies of both groups improved in2007 relative to 2006, which might be explained by adifference in rainfall patterns between the two yearsor might indicate a learning curve as participants getused to system operation. As a final observation, thetwo households with modified systems (365- and 600-gallon systems) demonstrated a relatively high rain-fall capture efficiency among all the study partici-pants. A possible explanation is that the participantswho took extra care to tailor their systems to theirspecific needs were able to increase their systems’efficiency.

Residential Audit ⁄ Retrofit and Water ConservationAppliances Rebates

As part of a town-wide water conservation plan, inSeptember of 2003, the town of Reading, MA, beganoffering water customers free indoor water use auditsand water saving retrofit devices tailor-made to theresults of the audits. The town also began offeringcustomers rebates for eligible water-efficient appli-ances (washing machines and toilets) purchased on or

FIGURE 7. Comparison of Scale Between HouseholdDomestic Water Use and Rainwater Use.

Rat

io:

Rai

nwat

erus

e/R

ainf

allo

nCo

ntri

buti

ngro

of

20072006800600365200800600365200

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.76 m (n=16)1.38 m (n=1)2.27 m (n=1)3.03 m (n=7)

Tank Capacity

Narrower boxes: 95% confidence intervals for the medians⊕ : mean

3

3

3

3

0.76 m3 3.03 m3 0.76 m3 3.03 m3

(200 gal.) (800 gal.) (200 gal.) (800 gal.)

2006 2007

1.38 m3

(365 gal.)

2.27 m3

(600 gal.)

2.27 m3

(600 gal.)

1.38 m3

(365 gal.)

FIGURE 8. Rainfall Capture Efficiency in, 2006 and 2007.When the sample size n = 1, the interquartile box and

confidence interval for the median cannot be determined.



after July 1, 2003. The purpose of this study was toevaluate the effectiveness of these two programs atreducing town-wide water demand and the waterdemand of those households who chose to participatein either or both programs. Only winter water usedata were evaluated to isolate indoor water use andeliminate the confounding effect of year-to-yearweather variability on water use during the irrigationseason.

Participating households were grouped into fivemutually exclusive categories of participation: (1)audit ⁄ retrofit (AR), (2) audit ⁄ retrofit and any type ofrebate(s) (AR&R), (3) rebate-toilet(s) (RT), (4) rebate-washing machines(s) (RW), and (5) rebate-toilet(s)and washing machine(s) (RT&W). Participants in thesame category should not be interpreted to haveexactly the same level of participation. For example,the numbers of low-flush toilets for any two house-holds in the group RT may be different, and the num-ber of retrofit devices installed among households inthe group AR is variable. This variability did not hin-der analysis, as the intent of the study was not toevaluate savings associated with individual technolo-gies, but rather savings resulting from the programsas a whole, which naturally include varying levels ofparticipation.

Quarterly water use records for the entire townwere obtained from February 2001 through May2007. To isolate indoor water use, only quarters thatbegan on or after October 19 and ended on or beforeApril 14 of any year were included in the analysis.For each household, records dated before the installa-tion of a qualifying rebate device or date of audit areregarded as ‘‘pre’’ winter use, whereas those recordedafter are ‘‘post’’ winter use. Savings was determinedby subtracting the average of the ‘‘post’’ use recordsfrom the average of the ‘‘pre’’ use records. To controlfor factors other than participation in the water con-servation program that might trigger a change inwater use patterns, households that did not partici-pate in any program were included in a controlgroup. However, as participating households initiatedtheir participation across different years during thestudy window, a single date could not be selected toseparate ‘‘pre’’ and ‘‘post’’ time periods for the controlgroup. Therefore, we analyzed the control group fourtimes to coincide with the variable points of initiationfor the participating households. Specifically, ‘‘pre’’minus ‘‘post’’ water use was calculated for the controlgroup using each of the following four pre v. postgroupings of years: (1) 2001-2002 v. 2003-2007,(2) 2001-2003 v. 2004-2007, (3) 2001-2004 v. 2005-2007, and (4) 2001-2005 v. 2006-2007.

The normal PPCC hypothesis test results suggestedthat nonparametric hypothesis tests are preferred.Sign tests showed statistically significant winter

water savings in each conservation program categoryexcept AR&R (Figure 9 and Table 3a). However, theAR&R households (those participating in both theaudit ⁄ retrofit and rebate programs) did demonstratethe highest median and second-highest average sav-ings among the categories. The small sample size ofthis group likely explains our inability to detect a sta-tistically significant savings for this category. In con-trast to the households participating in theconservation programs, the control group householdsshowed no statistically significant changes in wateruse for any of the time frames defined.

To evaluate the effect of the two outreach pro-grams on town-wide water use, the overall per-house-hold median savings for participating at any level ineither program was multiplied by the number of par-ticipating households (Table 3b). The town saved3,950 m3 ⁄ quarter as a result of implementing bothprograms. Town-wide participation rates are shownfor each program and for those participating in bothprograms (number of participating households ⁄ num-ber of households in town). Participation rates are animportant factor in estimating the overall savingsthat another town might be able to achieve by imple-menting similar programs. However, it should benoted that Reading saw waves of new participationeach time the town conducted concerted outreachefforts during the course of the programs. We canassume, then, that the participation rates observed inReading are closely related to the particular level ofoutreach effort exerted by the town, and it followsthat other towns might be able to increase participa-tion rates with more intensive outreach efforts.

RT&W (n=30)

RW (n=527

)

RT (n=87)

AR&R (n=32)

AR (n=99)

Control

4 (n=505

0)

Control

3 (n=505

0)

Contro 2 (n=

5050)

Control

1 (n=505

0)

50

25

0

-25

-50

0

: meanNarrower boxes: 95% confidence intervals for the medians

⊕

Wat

er S

avin

gs (

m3/q

uar

ter/

hou

seh

old)

FIGURE 9. Winter Water Savings Among the Five Different WaterConservation Treatment Categories and the Control Group, Ana-lyzed Four Ways. Values <0 imply an increase in water use afterinstalling a water conservation device or receipt of an audit andretrofit kit. The five treatment categories are: audit ⁄ retrofit (AR);audit ⁄ retrofit and any type of rebate(s) (AR&R); rebate-toilet(s)(RT); rebate-washing machines(s) (RW); and rebate-toilet(s) andwashing machine(s) (RT&W).



Soil Amendments in Ball Fields

A portion of an 8-acre municipal athletic field com-plex in the town of North Reading, MA, was redevel-oped to maximize infiltration and minimize irrigationrequirements and application of fertilizer and pesti-cides by employing the following techniques: (1) soilenhancement with zeolite, an additive that retainsmoisture and nutrients; (2) use of drought-resistantturf; and (3) installation of a WSICS (see sectionon Weather-Sensitive Irrigation Controller Switches).The adjacent field, which has identical solar orienta-tion, drainage patterns, and original soil profile,received only the latter two treatments and was usedas a control to evaluate the effectiveness of the zeoliteadditive.

The field manager progressively adjusted theWSICS ET thresholds for each field in order to iden-tify the most conservative watering scheme that couldstill maintain healthy turf. These thresholds set thetolerance for how much estimated ET is allowedbefore an irrigation cycle is triggered. The optimalthresholds of the zeolite and control fields were foundto be 0.89 cm (0.35 inches) and 0.64 cm (0.25 inches),respectively. These settings were used to simulate thenumber of irrigation cycles that the WSICS would

have applied to each field over the five-year periodfrom 2003 to 2007, using historic weather data(see Retrospective Analysis under Weather-SensitiveIrrigation Controller Switches for methodology). Thenumber of cycles was then converted to a total annualvolume, based on the respective per-cycle volumesmeasured for each field. Savings was defined by sub-tracting the total per-acre irrigation volume applied tothe zeolite field from that applied to the control field,for each year. The optimum settings resulted in anestimated average annual per-unit-area savings ofapproximately 3.59 cm3 ⁄ cm2 (38,000 gallons ⁄ acre), or37% (Table 4). Such substantial savings suggest thatzeolite soil amendments may prove to be a very effec-tive means to reduce irrigation demands of athleticfields. However, these results are highly dependent onthe optimal ET thresholds observed for each field,based on trial and error and field observation over thecourse of a few months. To further refine the expectedsavings achievable through zeolite soil amendments,optimal watering thresholds could be verified by theuse of soil moisture sensors. Additionally, obser-vations over a longer time period that encompassgreater variability of weather patterns would helpverify optimal ET thresholds and refine long-termsavings estimates.

TABLE 3. (a) Sample Size, Mean and Median Water Savings for Each of the Five Participation Categories,(b) Participation Rates and Town-Wide Savings for Audit ⁄ Retrofit and Appliance Rebate Programs.

(a)

Savings (m3 ⁄ quarter ⁄ household)

AR AR&R RT RW RT&W

N 99 32 87 527 30Mean water savings 4.93 5.01 3.94 5.38 4.58Median winter water savings 3.96 9.20 1.89 5.66 7.08

(b) N

Participation RateBased on Number of

Households in Town (8,436)

Water Savings(m3 ⁄ quarter ⁄households) Town-Wide

Savings(m3 ⁄ quarter)Mean Median

All levels of participation 775 0.092 5.11 5.10 3,950AR 99 0.012 4.93 3.96AR&R 32 0.004 5.01 9.20Combined RT, RW, RT&W 644 0.076 5.15 5.19

TABLE 4. Simulated Irrigation Volumes Applied to Zeolite and Control Fields (2003 to 2007).

Threshold (cm)

Simulated Volumes in Year (cm3 ⁄ cm2 ⁄ year)

Mean2003 2004 2005 2006 2007

Zeolite field 0.89 3.68 5.11 8.37 5.35 9.30 6.36Control field 0.64 9.08 7.92 11.62 7.90 13.25 9.95Savings (control-zeolite) 5.40 2.80 3.25 2.56 3.95 3.59% Savings (savings ⁄ control) 59.52 35.40 28.00 32.35 29.82 37.02



CONCLUSIONS

The overall goal of this study was to evaluate theeffectiveness of four water conservation pilot strate-gies on water use. As is inherent to many small-scalepilots, the datasets for these demonstration projectstend to be small, variable, and exhibit nonnormal dis-tributions. A secondary goal of this study was to dem-onstrate the application of mostly nonparametricstatistical methods for their ability to enable sensibleinferences to be drawn, in some cases, even from thevery small samples.

Statistical hypothesis tests combined with con-trolled water conservation experiments were used toevaluate water savings associated with four waterconservation strategies implemented in communitiesin the Ipswich watershed in Massachusetts,designed for their combined ability to meet animmediate municipal need and pilot an innovativeconservation strategy. Our review of the literaturerevealed that controlled water conservation experi-ments combined with nonparametric statistical anal-yses of the type performed here are not commonlyreported. Instead, most previous research hasfocused on retrospective statistical analyses of wateruse as well as studies that sought to elucidatebehavior and attitudes concerning various waterconservation strategies. Our overall findings for eachof the four water conservation programs are asfollows:

1. Weather-sensitive irrigation controller switches:Residential water use patterns were variablyimpacted by the addition of the WSICS, withsome participants showing a decrease and othersshowing an increase in water use. The WSICSappeared to reduce the variability of water useamong residential participants, most notably bycausing a reduction in water use of the highesthistorical water users. Our findings underscorethat initial water use patterns are likely to be aprominent factor in determining whether wateruse will increase or decrease after WSICS instal-lation in a residential setting. Water users whorely on inefficient watering regimes, historically,are more likely to benefit from the WSICS, whichmay explain why the participants in our studywith the highest historical water use showedlarge and statistically significant water savingsafter installing the WSICS. In contrast to theresidential setting, WSICS installations at muni-cipal athletic fields resulted in consistent reduc-tions in water application, with an averagesavings of 0.11 m3 ⁄m2 ⁄year (121,000 gallons ⁄ acre ⁄year). This suggests that, prior to installation of

WSICS, ball fields in our study were more con-sistently overwatered than residential lawns.This is not surprising, given that towns gener-ally require a high level of turf performance ontheir athletic fields but lack the staff to fre-quently adjust irrigation settings in response toweather (such as reducing irrigation volumesafter or in anticipation of rain events). To ensuresufficient irrigation without frequent adjust-ments, systems are set to water frequently,regardless of need. Strict standards for turf per-formance and limited staff resources are commonin municipal settings, suggesting that the sav-ings observed at ball fields in this study arelikely transferable to other ball field sites.

2. Rainwater harvesting: Rainwater was used foroutdoor purposes by all participants, and thosewith 3.03 m3 systems (800 gallons) used signifi-cantly more than those with 0.76 m3 systems(200 gallons). Annual volumes of rainwater usedwere small compared with domestic water use,and reductions in domestic water use as a resultof substitution with rainwater could not be dis-cerned amidst the background fluctuations indomestic water use from year to year. However,a participant survey suggested that for everyhousehold, at least some of the rainwater usedwas a direct substitute for domestic water thatwould have been used for the same purpose.Rainfall capture efficiency was measured as theratio of rainwater used relative to the rain thatfell on the contributing roof area during themonths of system operation. Efficiency of bothsize systems improved in the program’s secondyear, which may indicate different rainfall pat-terns between the two years or that there is alearning curve as participants got used to systemoperation. In the second year, the larger systemswere more efficient than the smaller systems,whereas they were statistically equivalent thefirst year. A possible explanation is that as rain-fall capture efficiency improves, the impact ofsystem size becomes more pronounced. Twohouseholds that modified their systems’ size wereamong the most efficient, suggesting that effi-ciency may be improved by tailoring one’s systemto one’s needs.

3. Residential audit ⁄ retrofit and water conservationappliance rebates: Participation in two town-administered water conservation programs(a. free indoor water use audits and fixtureretrofit kits; b. low flow toilet and washingmachine rebates) was divided into five catego-ries. Four resulted in modest but significant posi-tive water savings averaging between 3.94 and5.38 m3 ⁄ quarter ⁄ household. Although the fifth



participation category (participation in bothprograms) showed no statistically significantwater savings, this group’s median and meansavings were ranked the highest and second-highest, respectively, among all five categories.The finding of nonstatistically significant savingsof this group appeared to result from thesmall sample size and large variation in watersavings among the participants. In the first fouryears of program implementation, 9.2% of thetown’s households participated in one or both ofthe programs, resulting in an overall averagesavings of approximately 3,950 m3 ⁄ quarter forthe town.

4. Soil amendments in ball field: The addition of amoisture and nutrient-retaining additive, zeolite,to the soil of a ball field resulted in healthy turfwith less water applied than to an adjacent con-trol field. Based on observed irrigation require-ments, the zeolite material was estimated tosave approximately 3.59 cm3 ⁄ cm2 ⁄ year (38,000gallons ⁄ acre ⁄ year). This represents a reductionof 37% in irrigation volume, suggesting promis-ing water savings from zeolite soil amendments.

Future research on all of the above strategiescould be used to verify or refine the results reportedhere. To address the specific constraints encounteredin this study, the following approaches are sug-gested. WSICS should be evaluated with larger resi-dential sample sizes and include an assessment ofhistoric irrigation efficiency. Additional size catego-ries of rainwater harvesting systems should be eval-uated for rainfall capture efficiency under a varietyof rainfall conditions and further investigationshould be made into the ability of such systems toreduce domestic water use. Town-administeredwater conservation programs such as Reading’sshould continue to be evaluated over longer timeframes to better understand the long-term potentialfor savings among participating households and atthe town level. Lastly, turf health on the soil-amended and control ball fields was determined byvisual inspection. Future research should employ amore sophisticated method for comparing the turfhealth.

ACKNOWLEDGMENTS

This article was developed under Cooperative Agreement No.WS-97117501 awarded by the U.S. Environmental ProtectionAgency (EPA) to the Massachusetts Department of Conservationand Recreation (DCR). The views expressed in this document aresolely those of the authors; not those of EPA or DCR. Neither EPAnor DCR endorses any products or commercial services mentionedin this publication.

LITERATURE CITED

Atwood, C., R. Kreutzwiser, and R. de Loe, 2007. Residents’ Assess-ment of an Urban Outdoor Water Conservation Program inGuelph, Ontario. Journal of the American Water ResourcesAssociation 43(2):427-439.

Ayres Associates, 1996. An Evaluation of Davis Islands LandscapeIrrigation System Conservation Program. A Report Preparedfor the City of Tempa Water Department Water ConservationSection.

Bamezai, A., 1995. Application of Difference-in-Difference Tech-niques to the Evaluation of Drought-Tainted Water Conserva-tion Programs. Evaluation Review 19(5):559-582.

Brookshire, D.S., H.S. Burness, J.M. Chermak, and K. Krause,2002. Western Urban Water Demand. Natural ResourcesJournal 2(4):559-582.

Buchberger, S.G. and G.J. Wells, 1996. Intensity, Duration andFrequency of Residential Water Demands. Journal of WaterResources Planning and Management – ASCE 122(1):11-19.

Canfield, S., L. Claessens, C. Hopkinson, Jr, E. Rastetter, and J.Vallino, 1999. Long-Term Effect of Municipal Water Use on theWater Budget of the Ipswich River Basin. Biological Bulletin197:295-297.

Corral-Verdugo, V. and M. Frias-Armenta, 2006. Personal Norma-tive Beliefs, Antisocial Behavior and Residential Water Conser-vation. Environment and Behavior 38, 406, doi: 10.1177/0013916505282272.

Dalhuisen, J.M., R.J.G.M. Florzzx, H.L.F. de Groot, and P. Nijk-amp, 2003. Price and Income Elasticities of Residential WaterDemand: A Meta-Analysis. Land Economics 79(2):292-308.

Espey, M., J. Espey, and W.D. Shaw, 1997. Price Elasticity ofResidential Demand for Water: A Meta-Analysis. WaterResources Research 33(6):1369-1374.

Gilg, A. and S. Barr, 2006. Behavioural Attitudes Towards WaterSaving? Evidence From a Study of Environmental Actions.Ecological Economics 57(3):400-414.

Helsel, D.R. and R.M. Hirsch, 2002. Statistical Methods in WaterResources. In: Techniques of Water-Resources InvestigationsBook 4, Hydrologic Analysis and Interpretation, U.S. GeologicalSurvey, Chapter A3. U.S. Geological Survey, 522 pp, http://pubs.usgs.gov/twri/twri4a3/, accessed March 2011.

Hilaire, R.S., M.A. Arnold, D.C. Wilkerson, D.A. Devitt, B.H. Hurd,B.J. Lesikar, V.I. Lohr, C.A. Martin, G.V. McDonald,R.L. Morris, D.R. Pittenger, D.A. Shaw and D.F. Zoldoske,2008. Efficient Water Use in Residential Urban Landscapes.Hortscience 43(7):2081-2092.

Ipswich River Watershed Action Plan, 2003. A Publication of theMassachusetts Executive Office of Energy and EnvironmentalAffairs, produced by Horsley & Witten, Inc. http://www.mass.gov/Eoeea/docs/eea/water/wap_ipswich_2003.pdf, accessed March2011.

Karpiscak, M.M., G.W. France, K.J. DeCook, R.G. Brittain, K.E.Foster, and S.B. Hopf, 2001. Casa Del Agua: Water Conserva-tion Demonstration House 1986 Through 1998. Journal of theAmerican Water Resources Association 37(5):1237-1248.

Kenney, D.S., C. Goemans, R. Klein, J. Lowrey, and K. Reidy,2008. Residential Water Demand Management: Lessons FromAurora, Colorado. Journal of the American Water ResourcesAssociation 44(1):192-207.

Kenney, D.S., R.A. Klein, and M.P. Clark, 2004. Use and Effective-ness of Municipal Water Restrictions During Drought in Colo-rado. Journal of the American Water Resources Association40:77-87.

Mayer, P.W., W.B. DeOreo, E. Towler, and D.M. Lewis, 2003. Resi-dential Indoor Water Conservation Study: Evaluation of HighEfficiency Indoor Plumbing Fixture Retrofits in Single-FamilyHomes in the East Bay Municipal Utility District Service Area.



A Report Prepared for East Bay Municipal Utility District andthe United States Environmental Protection Agency.

Mayer, P.W., W.B. DeOreo, E. Towler, and L. Martien, 2004.Tampa Water Department Residential Water ConservationStudy: The Impacts of High Efficiency Plumbing Fixture Retro-fits in Single-Family Homes. A Report Prepared for TampaWater Department and the United States EnvironmentalProtection Agency.

Michelsen, A.M., J.T. McGuckin, and D. Stumpf, 1999. NonpriceWater Conservation Programs as a Demand Management Tool.Journal of the American Water Resources Association 35(3):593-602.

Renwick, M.E. and S.O. Archibald, 1998. Demand Side Manage-ment Policies for Residential Water Use: Who Bears the Conser-vation Burden? Land Economics 74(3):343-359.

Renwick, M.E. and R.D. Green, 2000. Do Residential WaterDemand Side Management Policies Measure Up? An Analysisof Eight California Water Agencies. Journal of EnvironmentalEconomics and Management 40:37-55.

Smith, Jr, W.J. and Y.D. Wang, 2008. Conservation Rates: TheBest ‘New’ Source of Urban Water During Drought. Water andEnvironment Journal 22:100-116.

Stedinger, J.R., R.M. Vogel, and E. Foufoula-Georgiou, 1993. Fre-quency Analysis of Extreme Events. In: Handbook of Hydrology,R. Maidment (Editor). McGraw-Hill, New York, Chapter 18,68 pp.

Vickers, A., 2001. Handbook of Water Use and Conservation:Homes, Landscapes, Business, Industries, Farm. WaterplowPress, Amherst, Massachusetts.

Wang, Y.D., J.S. Song, J. Byrne, and S.J. Yun, 1999. Evaluatingthe Persistence of Residential Water Conservation: A 1992-1997Panel Study of a Water Utility Program in Delaware. Journal ofthe American Water Resources Association 35(5):1269-1276.

Zar, J.H., 1999. Biostatistical Analysis (Fourth Edition). Prenticehall, New Jersey, 663 pp.

Zarriello, P.J. and K.G. Ries, III, 2000. A Precipitation-RunoffModel for the Analysis of the Effects of Water Withdrawals onStreamflow. Ipswich River Basin, Massachusetts, U.S. Geologi-cal Survey Water-Resources Investigation Report 00-4029, 107pp. http://pubs.usgs.gov/wri/wri004029/, accessed March 2011.



Date post:	03-Aug-2018
Category:	Documents
Upload:	hadat
View:	212 times
Download:	0 times

THE IMPACTS OF WATER CONSERVATION STRATEGIES ON WATER...

Documents