SURVEY OF DATA RELATED TO MUNICIPAL WATER SYSTEMS IN UTAH

 

 

Paul Harms

pharms@cc.usu.edu

CEE 6440:  GIS in Water Resources

Department of Civil and Environmental Engineering

Utah State University

Fall 2002

 

 

 

 

Introduction

 

Relating secondary or ancillary data to the water use data of municipal water systems allows one to examine why different water systems use different amounts of water.  If relationships are found between the ancillary data and the water use data, then estimates of existing water demands or recent water use may be made. 

 

Examination of ancillary data may also allow one to divide water systems into categories of similar water use.  This would allow better use of random sampling when estimating statewide water use.  A National Research Council report recently suggested random sampling as an alternative to the exhaustive collecting of water use numbers for the National Water-Use Information Program of the U.S. Geological Survey1.

 

Another potential benefit of these data relationships is to improve predictions of future water demands, using projections of the related data.

 

Numerous models have correlated data with water use.  IWR-MAIN (Institute for Water Resources – Municipal and Industrial Needs), originally developed through the U.S. Army Corps of Engineers, has a default model that correlates residential water use with household income, household size (persons), housing density, average maximum daily temperature, rainfall, the marginal price of water, and the fixed charge or rate premium of the water2.  Studies in the early 1990s for the Wasatch Front Water Demand/Supply Model found that residential water use in the Wasatch Front metropolitan areas correlated with persons per household, lot size, assessed value, soil type, season, and whether a dual system was used3.  A recent study by the Utah Division of Water Resources found that the number of persons in the home is the most influential factor in determining the per capita use rate4.    

 

This study examined data available on a sample of municipal water systems in Utah – data not requiring a survey of individual households – to see if this data correlated with water use.  It generally compared averages of the ancillary data with averages of the water use data at individual systems, rather than comparing a time series of the ancillary data with a time series of a system’s water use. 

 

 

Methods

 

According to the Utah Division of Drinking Water, there are 454 community drinking water systems in the state, some inactive.  Most of these are small, many are tiny.  For example, on the list of water systems, the Salt Lake City Corporation for Culinary Water, population 323,141, is immediately preceded by the S and W Trailer Park, population 200.  Water systems are required to report to the Utah Division of Water Rights the amount of water they diverted each month from each of their water sources, including whether these amounts were metered or estimated.  The completeness and legibility of these reports varies considerably.  The Division of Water Rights compiles what information they get, and makes it available on the internet:  http://waterrights.utah.gov/cgi-bin/wuseview.exe?Startup. 

 

From the Division of Water Rights web pages, municipal water systems were chosen partly by region – in order to have the sample well-scattered over the state.  Water systems representing towns were used; those owned by individuals or homeowners’ associations or trailer courts were excluded.  Systems were chosen largely by the completeness of the data available.  Systems with grossly inconsistent data (from year to year) were generally excluded, though it was not known whether the inconsistencies were due to errors or to actual year-to-year changes.  Approximately half of the systems chosen initially were not used due to incomplete data.  The selection process generally came down to looking at the information for the municipal water systems of several of the towns in each county, and trying to find a town that had reasonably complete Division of Water Rights data.

 

The resulting scatter of the 60 selected municipal water systems is shown below:

 

 

 

 

 

 

 

In an effort to decrease the effects of anomalous years or inaccurate data, total yearly water use for each system was generally taken as the average of the last five years for which there was consistent data – preferably 1997 to 2001.  For the selected systems, the yearly use figures varied from 16 acre-feet by Kingston in Piute County, to 88,000 acre-feet by the main system in Salt Lake City. 

 

The Division of Water Rights data also includes population estimates made by each water system.  For this study, population was also taken as the average of the last five years for which there was consistent data – preferably 1997 to 2001.  For the selected systems, these figures varied from 140 at Woodruff in Rich County, to 314,000 at the Salt Lake City system. 

 

In the Water Rights data, total water use is divided into domestic, commercial, industrial, institutional, stock, wholesale, other, and unmetered use.   For this study, the fraction of the total use that was domestic use was obtained by dividing the five-year average domestic use by the five-year average total use.  Numerous systems were excluded from the study because they did not provide this usage information.

 

The water systems also provided an estimate of the percentage of their connections that had dual systems – secondary or gray water systems providing untreated water for outdoor use.  This water was generally not included in the total use figures.  If the dual system percentages varied from year to year, a five-year average was again used for this study.  

 

An example of “good” data (mostly complete data form, all water sources metered rather than estimated) is provided by the North Logan Culinary System: 

 

 

 

In separate available data, the amount of water a system obtained from surface water, from springs, and from wells was listed:  http://waterrights.utah.gov/cgi-bin/libview.exe?Modinfo=Viewpub&LIBNUM=50-1-212.  The fraction of water that each system obtained from wells was figured from this.  Also, the systems that obtained over 90% of their water from surface water were found from this data.  This data represented one year – usually 1999 (the most recent year of available information).

 

An example of this data source is shown, along with a map illustrating proportions of drinking water sources throughout the state.

 

 

 

 

Water price data was obtained from Tim Pine of the Utah Division of Drinking Water as part of a 2001 Water System Survey report that he is writing.  No data was available on 18 of the 60 systems.  Some of the pricing has changed in recent years – the rate structure used by a system may have been different in 1998, for example.

 

Water price was listed in the data as an unchanging base price charged up to a base limit (of gallons), then as an overage rate per certain number of gallons (usually 1000) that was charged, in addition to the base price, for water used beyond the base limit.  Up to four overage blocks were listed.  As an example, each connection at Escalante Culinary Water is charged an unchanging $17 for the first 15,000 gallons used each month, then an additional $1.50 per 1000 gallons for the next 15,000 gallons, then $2 per 1000 gallons beyond that. 

 

Some water systems had a base limit of zero – they had no flat charge but used rates throughout.  For this study, the rate structure used by a system was determined to be “flat charge” (charge not dependent on amount used – no overage blocks), “uniform rate” (one overage block), “increasing block rate” (two to four overage blocks, rates increasing with each), or “decreasing block rate” (rates decreasing with each overage block).  In the Escalante example above, the rate structure would be an increasing block rate. 

 

Besides rate structure, this study also listed each system’s base limit in gallons, and the rate per 1000 gallons in the first overage block (for systems that did not use a flat charge).  No data on seasonal rate changes was found. 

 

The latitude for each water system was obtained from a GIS map of municipal water systems in Utah.  For the selected systems, latitude varied from 37.06° at Kanab in Kane County (the Arizona border is 37°) to 41.98° at Lewiston in Cache County (the Idaho border is 42°).

 

The average annual maximum temperature, average annual precipitation, and average summer (June, July, August, September) precipitation for each water system was figured from climate summaries available from over 200 weather stations in Utah:  http://www.wrcc.dri.edu/summary/climsmut.html, and see image below.  Most systems had complete data for their town.  If a system’s town lacked some data, data from a nearby station (e.g., another town ten miles distant) was occasionally used.     

 

 

For the systems selected, average annual maximum temperature varied from 54.1°F at Randolph in Rich County, to 77.7°F at St. George. 

 

Utah is a fairly arid state throughout, with much of the precipitation falling on the mountains (see map).  However, average annual precipitation in the selected towns still varied from 5.42 inches at Hanksville in Wayne County to 23.57 inches in Bountiful.

 

 

For one part of this study, municipal water systems were grouped according to “town size”.  Systems that are part of the large metropolitan areas in Weber, Davis, Salt Lake, and Utah counties were classified as “metro”.  Other systems were classified as “small” or “mid”, with a population of 5000 being the dividing point between those categories.  Census values for the towns were used rather than the systems’ populations. 

 

 

Results

 

Once the average values were determined for the total yearly water use and for the population of each municipal water system, the gallons per capita per day (GPCPD) value was determined.  This was called “Total GPCPD”.  The average fraction of the total use amount that was listed as being for domestic use was then used to determine the “Domestic GPCPD”.  The mean, standard deviation, and coefficient of variation values for these are given in the following table – for all the systems together and for the systems as grouped by town size.

 

 

 

Total GPCPD

Domestic GPCPD

 

Total Use

 

number

mean

stan.dev.

coef. var.

mean

stan.dev.

coef. var.

Population

(ac-ft/yr)

small

38

289.13

133.58

0.46

235.86

126.58

0.54

64660

21180

mid

9

275.49

114.67

0.42

191.44

76.34

0.40

114863

38533

metro

13

188.21

81.74

0.43

131.68

47.69

0.36

748129

181726

 

 

 

 

 

 

 

 

 

 

total

60

265.22

126.58

0.48

206.62

114.56

0.55

927652

241438

 

The water systems from the small towns are seen to have much higher GPCPD values on average.  Perhaps of more significance, in terms of estimating and forecasting, is that the small town water systems are more variable in their water use.  The other data gathered also indicated that the water use at small systems is much less predictable than the water use at large systems. 

 

The above table also shows that, while most of the systems are small, most of the population and water use are in the few large systems.

 

The range for “Total GPCPD” was from 72 at North Ogden to 685 at Levan in Juab County.  The range for “Domestic GPCPD” was from 48 at Green River in Emery County to 638 at Levan. 

 

Population correlated with water use, as is commonly assumed.  Even with population, however, the small systems showed less correlation.  The correlation coefficients, sample sizes, and whether the correlation coefficients are significant for the given sample sizes, are given in the table below.  Significance levels were set at 1.96/√(n-3). 

 

 

 

 

 

 

r

n

sig.?

Total Use/Population Corr. Coef. =

 

0.9898

60

yes

Small Total Use/Population Corr. Coef. =

 

0.8640

38

yes

Mid Total Use/Population Corr. Coef. =

 

0.9589

9

yes

Metro Total Use/Population Corr. Coef. =

 

0.9592

13

yes

 

The population/water use relationship is also shown in the charts below.  The second chart only includes the small systems, and shows their variability.

 

 

One would think that a lower “domestic fraction” (the fraction of the total use listed as being for domestic use) would lead to a higher GPCPD, since it would indicate that more water is being used for things besides residences.  This held true for the metropolitan systems, but not for the majority of the systems.

 

Domestic Fraction:

 

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Domestic Fraction Corr. Coef. =

0.0207

60

no

Metro Total GPCPD/Domestic Fraction Corr. Coef. =

-0.7912

13

yes

 

The “dual fraction” (the fraction of a water system’s connections that have a secondary system of untreated water for outdoor use) correlated more closely with GPCPD than did any of the other data.  This is understandable, since the untreated water is not included in the total use figures, and since a recent Division of Water Resources study found that, in Utah, 63% of residential water use is for outdoor use4.  Having more connections with a dual system generally led to a lower average GPCPD.  This again held true primarily with the metropolitan systems, with the small town systems being more variable.

 

Dual Fraction:

 

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Dual Fraction Corr. Coef. =

 

-0.4444

60

yes

Metro Total GPCPD/Dual Fraction Corr. Coef. =

-0.9004

13

yes

 

 

Several data types were investigated to see if variations in climate in Utah were primary factors in the variations in GPCPD.  These data types included latitude, average maximum temperature, average annual precipitation, and average summer (June, July, August, September) precipitation.  None of these was found to have a strong effect on GPCPD.  Annual precipitation was found to have some negative correlation with GPCPD among the metropolitan systems, which was somewhat surprising since one would not expect precipitation to vary enough in the metropolitan region to have an effect.  The temperature and summer precipitation results also indicated a possibility of weak effects on GPCPD in the metropolitan region.  Some of the results figured from these data types are shown below:

 

Latitude:

 

 

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Latitude Corr. Coef. =

 

-0.0817

60

no

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Average Maximum Temperature:

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Temperature Corr. Coef. =

 

-0.0145

60

no

Metro Total GPCPD/Temperature Corr. Coef. =

0.5516

13

no

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Average Annual Precipitation:

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Ann. Precip. Corr. Coef. =

 

-0.0320

60

no

Metro Total GPCPD/Ann. Precip. Corr. Coef. =

-0.6298

13

yes

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Average JJAS Precipitation:

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/JJAS Precip. Corr. Coef. =

 

0.0367

60

no

Metro Total GPCPD/JJAS Precip. Corr. Coef. =

-0.4400

13

no

   

 

 

 

The water source – well, spring, or surface – of a system was also considered as a possible factor in the amount of water used.  A number of relationships may be postulated.  For example, a well may cost more to operate, leading to less water used.  On the other hand, surface sources may be more immediately vulnerable to dry conditions, leading to less water used. 

 

In this study, the fraction of water obtained from wells did not correlate with GPCPD.  Only four of the 60 systems were found to rely almost exclusively on surface water.  These four did have low GPCPD values, but the association with surface water use is uncertain due to the small sample size.  Some of these results are shown below:

 

Well Fraction:

 

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Well Fraction Corr. Coef. =

 

-0.0816

60

no

Dom. GPCPD/Well Fraction Corr. Coef. =

 

-0.1789

60

no

 

 

 

 

 

 

 

 

Metro Total GPCPD/Well Fraction Corr. Coef. =

0.4122

13

no

Metro Dom. GPCPD/Well Fraction Corr. Coef. =

0.4904

13

no

 

>90% Surface Water Source:

Tot.GPCPD

Dom.GPCPD

 

 

 

mean

 

mean

 

# yes:

4

 

169.70

 

127.34

 

# no:

56

 

272.04

 

212.28

 

 

Water pricing was also examined for effects on water use.  No strong correlations were found.  Among the 42 systems that had data, none used a flat charge, 25 used a uniform rate, 15 used an increasing block rate, and 2 used a decreasing block rate.  One might expect an increasing block rate to encourage more water conservation than the uniform rate.  This was not apparent among the systems in general, but may be indicated among the metropolitan systems (though again the sample size is small). 

 

A lower base rate limit may be expected to lower water use, since a customer’s water bill would then be affected more by the amount used.  For example, a customer that started paying a per-gallon fee after using their base limit of 1000 gallons may use less water than a customer of another system that did not start paying a per-gallon fee until after using a base limit of 20,000 gallons.  The data in this study indicated, however, that if there was any relationship between base rate limit and GPCPD, it tended to be in the opposite direction.

 

The amount charged per gallon may also be expected to affect water use.  Like with the base rate limit, no significant correlation was found.  Unlike with the base rate limit, the relationship did tend in the expected direction (higher charge – lower GPCPD).

 

Some of these results are shown below.  With the “base rate limit” and the “rate per 1000 gallons”, the metropolitan data was similar to the overall data.

 

Rate Structure:

 

 

 

 

 

 

 

 

 

 

 

 

Total

Dom.

# no data:

 

18

 

 

 

GPCPD

GPCPD

# flat charge:

0

 

 

 

mean

mean

# uniform rate:

25

 

 

uniform

275.31

216.99

# increasing block:

15

 

 

increasing

287.86

228.21

# decreasing block:

2

 

 

decreasing

297.56

232.25

 

 

 

 

 

 

 

 

# metro uniform:

6

 

 

metro uniform

230.09

157.63

# metro increasing:

3

 

 

metro increasing

90.01

75.31

 

 

Base Rate Limit:

 

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/Base Limit Corr. Coef. =

 

-0.2661

42

no

 

 

Rate Per 1000 Gallons:

 

 

 

 

 

 

 

 

 

 

r

n

sig.?

Total GPCPD/RatePer1000 Corr. Coef. =

 

-0.2771

42

no

 

 

For the 42 systems in this sample that had pricing data, the average “base rate limit” was 20,429 gallons, with the range varying from 0 gallons at South Ogden to 100,000 gallons at four water systems.  The median was 12,000 gallons.  With the “rate per 1000 gallons”, the average was $0.98, with the range varying from $0.25 at Enoch in Iron County to $4.00 at Hanksville in Wayne County.  The median here was $0.80.

 

Since the metropolitan water systems appeared to have the strongest relationships between water use and the ancillary data, some multiple linear regression equations were tried with the data from these systems.  The best equation found (in terms of “standard error of the estimate” or “mean square error”, se , and the “adjusted coefficient of multiple determination”, R2adj ) for “Total GPCPD” was one that used only the dual system fraction:

 

Total GPCPD = 248.966 – 161.851*(dual fraction).

 

se  = 36.2095.               R2adj = 0.80374.

 

The best equation found for total water use was one that included population, dual fraction, average maximum temperature, average annual precipitation, and average summer precipitation:

 

Total Use = -17359.8 + 0.278205*(pop) – 1516.78*(dual) + 417.821*(temp) + 214.045*(annPrecip) – 3574.23*(sumPrecip).

 

se = 1916.91.               R2adj = 0.993422.

 

With all of the equations tried, there was a 0.99 confidence level that a linear regression existed. 

 

It was interesting to see several instances where two systems a few miles from each other had widely different GPCPD values.  With this study, some of this appears to be due to dual systems.  In Tooele County, Grantsville with 185 GPCPD has a 0.38 dual system usage, while Stockton with 425 GPCPD has a 0.02 dual system usage.  In Rich County, Randolph with 507 GPCPD has no dual system usage, while Woodruff with 303 GPCPD has a 0.40 dual system usage.  In Weber County, Riverdale with 262 GPCPD has a 0.12 dual system usage, while nearby Roy with 96 GPCPD has a 0.88 dual system usage.

 

With the small systems in general, however, the wide variance in water use remains unexplained.  Some of this may be due to factors peculiar to individual systems; factors that have a larger relative effect on small systems than they might have on large systems.  Tiny Randolph and Woodruff, as examples, are reported to spill water from their systems in winter to keep pipes from freezing.  This raises their GPCPD values.  Other small systems may have other unusual factors affecting their water use; factors that perhaps could only be learned by studying individual systems more closely.

 

 

Problems

 

The main problem in this survey was with obtaining trustworthy data.  The primary data was the data reported by the municipal water systems to the Division of Water Rights.  In general, the water use values may be fairly accurate when systems have all of their sources metered.  The sample in this study included some systems that lacked meters and that provided estimated water use values. 

 

The water use values were probably more accurate overall than were the population values provided by the systems.  The population figures often appeared to be rough estimates since they were commonly rounded to the nearest hundred and were often left unchanged for several years.  The percentages of connections with dual systems also appeared to be rough estimates. 

 

The difficulty in finding satisfactory data led to a sample of water systems that was not random.  This means that the statistics given above are invalid, and should only be viewed as indicating possible trends.  Gathering a random sample could possibly be achieved by examining all of the water systems and eliminating those that did not meet certain criteria – such as having satisfactory data and representing towns.  A random sample could then be taken from this reduced population of water systems.  However, this would still just be a sample of the systems with satisfactory data.

 

Another problem in this study is that different data types sometimes cover different years.  The total use, population, domestic fraction, and dual fraction figures were averaged over five years, yet some of the figures were from one year (such as the water source percentages and the pricing data), and some were from more than five years (such as average precipitation and temperature).

 

A more subtle problem is that some correlations may exist but be obscured by stronger correlations.  One may have to control for some factors to find correlations among other factors.  For example, if the believed effects of dual systems were accounted for, then perhaps the effects of pricing policies, if any, would become evident.

 

 

Conclusions

 

Given the limitations listed above, the following relationships were suggested as possibilities by examination of the data collected in this survey:

 

Correlations significantly different from zero:

  • more population – more total use
  • more dual – less GPCPD
  • more metro annual precipitation – less GPCPD
  • lower metro domestic fraction – more GPCPD

 

Other trends:

  • smaller water system – more variable GPCPD
  • higher latitude – less GPCPD
  • higher metro and mid temperature – more GPCPD
  • more metro summer precipitation – less GPCPD
  • metro increasing block rates – less GPCPD
  • higher base rate limit – less GPCPD (counter-intuitive)
  • higher rate per 1000 gallons – less GPCPD
  • more surface water dependent – less GPCPD

 

There appeared to be real differences, besides size, between the large water systems and the small.  Grouping the systems this way appeared appropriate, and may be useful when random sampling is used.  However, while the large systems generally showed somewhat predictable relationships between water use and the ancillary data, the small systems generally did not.  For the majority of municipal water systems in Utah (i.e., the small ones), using water use averages over many systems may be valid, but using the ancillary data presented here for estimating or predicting water use at an individual system would not be valid. 

 

 

References

 

  1. National Research Council, Committee on USGS Water Resources Research.  2002.  Estimating Water Use in the United States:  A New Paradigm for the National Water-Use Information Program, Washington, D.C.:  National Academy Press.
  2. Baumann, Duane D., John J. Boland, et al. 1998.  Urban Water Demand Management and Planning, New York:  McGraw-Hill, Inc.
  3. Utah Division of Water Resources.  1993.  Wasatch Front Water Demand/Supply Model.  Salt Lake City, Utah:  Department of Natural Resources.
  4. Utah Division of Water Resources.  2001.  Identifying Residential Water Use.  Salt Lake City, Utah:  Department of Natural Resources.