EPA/600/R-17/209 | September 2017 | www.epa.gov/research
United States
Environmental Protection
Agency
oEPA
Proximity of Private Domestic
Wells to Underground Storage
Tanks: Oklahoma Pilot Study
Office of Research and Development
National Risk Management Research Laboratory | Groundwater, Watershed, and Ecosystem Restoration Division
-------
Proximity of Private Domestic Wells to
Underground Storage Tanks: Oklahoma
Pilot Study
James W. Weaver
United States Environmental Protection Agency,
Office of Research and Development,
Ada, OK 74820,
Andrew Murray
Oak Ridge Institute for Science and Education,
Cincinnati, OH
Fran Kremer
United States Environmental Protection Agency
Office of Research and Development
Cincinnati, OH
l
-------
Abstract
For protecting drinking water supplies, the locations of areas with reliance on private domestic
wells (hereafter referred to as "wells") and their relationship to contaminant sources need to
be determined. A key resource in the U.S. was the 1990 Census where the source of domestic
drinking water was a survey question. Two methods are developed to update estimates of the
areal density of well use using readily accessible data. The first uses well logs reported to the
states and the addition of housing units reported to the Census Bureau at the county, census
tract and census block group scales. The second uses housing units reported to the Census and
an estimated well use fraction. To limit the scope and because of abundant data, Oklahoma was
used for a pilot project. The resulting well density estimates were consistent among spatial
scales, and were statistically similar. High rates of well use were identified to the north and east
of Oklahoma City, primarily in expanding cities located over a productive aquifer. In contrast,
low rates of well use were identified in rural areas without public water systems and
Oklahoma's second largest city, Tulsa, each attributable to lack of suitable ground water. High
densities of well use may be expected in rural areas without public water systems, expanding
cities and suburbs, and legacy areas of well usage. The completeness of reported well logs was
tested by counts from neighborhoods with known reliance on wells which showed reporting
rates of 20% to 98%. Well densities in these neighborhoods were higher than the larger-scale
estimates indicating that locally high densities typically exist within analysis units. A Monte
Carlo procedure was used to determine that 27% of underground storage tanks that had at
least one well within a typical distance of concern of 300 m (1,000 ft).
2
-------
QA Statement
This project was performed under quality assurance project plan ORD Project QA ID #G-
GWERD-0019367. In the section on "Data and Methods", the methods for assessment of the
quality of data are described in the "Positional Accuracy" subsection. The "Data Error and
Method Evaluation" subsection presents results on well position error, public land survey
system location accuracy, 1990 census sampling error, and accuracy of historical application of
the net housing unit method.
3
-------
Contents
Abstract 2
QA Statement 3
List of Figures 5
List of Tables 7
Introduction 8
Data and Methods 12
Well Log Data 16
Water supplies in Oklahoma 16
Estimation Methods for Counties, Census Tracts and Census Block Groups 16
Cities and Neighborhoods 19
Positional Accuracy 19
Underground Storage Tank Data 21
Results and Discussion 23
Data Error and Method Evaluation 23
State-Wide Inferences 38
County-level Estimates 39
Census Tract Scale Estimates 48
Census Block Group Scale Estimates 49
City and Neighborhood Scale Analysis 53
Example Cities without Public Water Supplies 53
Individual Neighborhoods 57
Coexistence of Public and Private Water Supplies 60
Distances between Underground Storage Tanks and Private Domestic Wells 66
Conclusions 69
Acknowledgements 72
References 73
Appendix: Fluctuation of Private Domestic Well Use in Oklahoma Counties 79
4
-------
List of Figures
Figure 1. Mapofthe United States showing the location of Oklahoma 13
Figure 2. Major (green) and alluvial (brown) aquifers of Oklahoma (Oklahoma Water Resources Board,
1998) 13
Figure 3. Density of housing units using private domestic we 11 use inferred from 1990 Census on a
census block group spatial basis, with locations of cities discussed in the text 14
Figure 4. Locations of Oklahoma cities discussed as examples in the results section 15
Figure 5. Spatial relationships for public land survey system (PLSS) placement in census block group
error estimates. PLSS unit entirely contained within a census blockgroup (left). PLSS unit partly
outside census blockgroup (center). Illustration of error estimate which equals (B + C + D)/(A + B + C +
D) (right) 21
Figure 6. Frequency distribution ofcharacteristicdimension of census blockgroupsand relationship to
public land survey system units. "Max Well Errors" follow from the size distribution of reported well
locations within PLSS units of which 99.5% are smallerthanthe smallest characteristic dimension of the
census blockgroups(blue line). The red line representsthe median probability of a well plottinginan
adjacent census block group as determined by Monte Carlo analysis 25
Figure 7. Standard sampling error estimates for Oklahoma counties determined from equations
presented below Table 1 29
Figure 8. Comparison ofcounty-levelestimated privatedomesticwell use and that inferred fromthe
U.S. Census for 1970. Becausethe watersupply question was not asked in placesofgreaterthan 50,000
in population in 1960, the 1970 estimatesexcludethreeof Oklahoma's 77counties, namely Comanche,
Oklahoma, and Tulsa 31
Figure 9. Error estimates (Census Value -Estimate)/Census Value for county-level PDW estimates given
by county for 1970 32
Figure 10. Comparison of county-level estimated private domestic we 11 use and that inferred from the
U.S. Census for 1980 33
Figure 11. Error estimates (Census Value-Estimate)/Census Value for county-level PDW estimates
given by county for 1980 34
Figure 12. Comparison of county-level estimated private domestic we 11 use and that inferred from the
U.S. Census for 1990 35
Figure 13. Error estimates (Census Value -Estimate)/Census Value for county-level PDW estimates
given by county for 1990 36
Figure 14. Rural water districts added in Oklahoma 1889-1995 (data from OWRB, 1998) 37
Figure 15. Cumulative number of domestic wells reported to the Oklahoma Water Resources Board...39
Figure 16. Reported-welis estimate of private domestic weII density for 2010 on a county-wide spatial
scale. Oklahoma (top) and Cleveland (bottom) counties inthe centerof Oklahoma, and Delaware
County on the border with Missouri and Arkansas had the highest estimated well density in the state.. 42
Figure 17. Approximate serviceareas of publicwatersystems in Oklahoma (Oklahoma Water Resources
Board, 1998) 43
5
-------
Figure 18. County-level predicted private domestic weII use for 2010 from housing unit increase
compared against predicted well use from well logs reported to the OWRB 45
Figure 19. Changes in fraction of private domestic we 11 use from USGS water supply data for 1985 to
2010 for five Oklahoma Counties. Results for other Oklahoma counties appear in the Appendix 46
Figure 20. Net housing unit (NHU) method (top), and reported-wells(RW) method (bottom) 2010
results for census tracts(left) and census block groups (right) in Oklahoma 49
Figure 21. Percent change in estimated well density 1990to 2010 usingthe reported well log(RW)
method on the census block group level 51
Figure 22. Estimated change in fraction of welltototal watersupplyfrom 1990 to 2010 on the census
block group level 52
Figure 23. The City of Forest Park sits adjacent to Oklahoma City (Figure 4) and has no public water
supply system 56
Figure 24. The City of Nicoma Park lies to the east of Oklahoma City (Figure 4) and has no public water
supply system 56
Figure 25. The City of Enid's water supply does not cover its entire territory, and contains areas with
large numbers of private wells 61
Figure 26. Wells re ported to the Oklahoma Water Resources Board in 2011 and 2012 for Garfield
County (Enid). Drilling increased in August 2012 after imposition of an outdoor watering ban 62
Figure 27. City of Bethany, Oklahoma, which is contained within the city limits of Oklahoma City (Figure
4) 63
Figure 28. The City of Edmond public water supply has not extended throughout its entire territory....63
Figure 29. The City of Choctaw public water supply extends only through a portion of its territory 64
Figure 30. The location of water mains, publicwells, and watertanks in Edmond, Oklahoma (2009). ...65
Figure 31. Frequency distribution ofestimated numberof USTs with PDW within specified distances
from 10,000 Monte Carlo simulations of reported wells augmented with estimated locations of PDWsto
match wells-added estimate of PWDdensity. Symbols represent numbers of USTs with only reported
wells considered, plotted for comparison with median of the Monte Carloestimate 67
Figure 32. Changes in fraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Adair to Cherokee counties 79
Figure 33. Changes infraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Choctaw to Ellis counties 80
Figure 34. Changes in fraction of private domestic well use from USGS water supply data (1985-2010)
for Garfield to Jefferson counties 81
Figure 35. Changes infraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Johnston to McCurtain counties 82
Figure 36. Changes infraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Mcintosh to Okmulgee counties 83
Figure 37. Changes infraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Osage to Seminole counties 84
Figure 38. Changes infraction of private domestic we 11 use from USGS water supply data (1985-2010)
for Sequoyah to Woodward counties 85
6
-------
List of Tables
Table 1. Error estimates for the 1990 census inference of private domestic well use 27
Table 2. Oklahoma water sourcesfrom the 1990 census (U.S. DOC, 1993) 38
Table 3. Housing unit, population and private domestic well characteristics forthe state of Oklahoma.40
Table 4. Summary statistics for private domestic well density estimates (wells per km2) from the 1990
census, reported wells, and net housingunitestimatesforthe 77 counties, 1,046 censustracts, and
2965 census block groups of Oklahoma 44
Table 5. Statistical comparison of 1990 private domesticwell densityand estimates madethe RWand
NHU methods. Bold italic values differ statistically at 0.05 P value in the table. Quantity names are as
follows: Dens 90 = inference from 1990 census; RW 2000 = reported-welis result for 2000; RW 2010 =
reported wells result for 2010; NHU 2000 = nethousingunitresultfor2000; NHU 2010 = net housing
unit result for 2010; NHU 2010 (f2ooo) = net housing unit result for 2010 with the fraction of private
domesticwell use (fpdw) updatingfrom RW 2000 result; NHU USGS 2000 = net housingunit resultfor
2000 with/pdw updatingfrom USGS results; andNHU USGS 2010 = nethousingunitresultfor2010 with
fpdw updating from USGS results 47
Table 6. County, censustract, census blockgroup reported-wells method, and neighborhood-count
estimates of private domesticwell usage forthe cities of Lake Alu ma, Forest Park, and Nicoma Park, all
of which have no public water supply system 53
Table 7. Estimated private domesticwell densityforcentral Oklahoma neighborhoods, based on county,
censustract, census block group, and neighborhood counts 58
Table 8. Results of 10,000 Monte Carlosimulationsofthe distance between USTsand reported and
estimated private domesticwell locations. The distances were binned into categories and the counts
represent the number of USTs with at least one well within the specified distance 68
7
-------
Introduction
Throughout the world, public water supplies may be limited by municipal expansion that
outpaces the extension of water systems (Kjellen 2001, Lundqvist et al. 2003, Danert et al.
2014, Wescoatet al. 2007) or restricted boundaries (Aiken 1987, Johnson etal 2004,
MacDonald Gibson et al 2014). Residents then meet their water needs through connection to
other residences, non-piped sources, or from private domestic wells (e.g., Jepson and Brown
2014, U.S. GAO 1998). The U.S. Geologic Survey (USGS) estimated that 14% of the U.S.
population and 8% of Oklahomans provided their own domestic water in 2010 (Maupin et al.
2014), primarily through the use of private domestic wells (hereafter referred to as "wells").
In the United States, the Safe Drinking Water Act (SDWA) regulates public water systems and
uses source water protection, treatment, distribution integrity, and public information as
barriers between contamination and safe drinking water. Routine testing of public water
supplies is required for a list of natural and anthropogenic contaminants (U.S. EPA 2004).
However, private wells serving less than 25 persons are not regulated by the SDWA, and
routine monitoring is not required. Less frequent testing is mandated by some states,
commonly at installation and property transfers (see, e.g., Atherholt et al. 2009).
Numerous examples of private domestic well contamination demonstrate the potential risks for
people who drink from private wells. The contaminants include pathogens, nitrate, arsenic,
fluoride, radon, chromium VI, perchlorate, uranium, and organic compounds (including
pesticides, gasoline constituents, and chlorinated solvents) (e.g., Ander et al. 2016, DeSimone
8
-------
et al., 2009, Schaider et al. 2016, U.S. EPA 2002). Although each of the SDWA barriers to
drinking water contamination noted above is potentially involved with the safety of water from
wells, the lack of systematic monitoring leads to the potential for undetected exposure of a
large number of people (Levin et al. 2002), and demographic data on well users are currently
limited (Vanderslice 2011). A recent workshop on private wells recommended, among other
things, establishment of a standardized database of private well use, strategically incorporating
existing information (Fox et al. 2016), as there are no national data on the numbers and
locations of wells (Ridpath et al. 2016).
In addition, identification of users of private domestic wells is potentially useful for identifying
factors influencing cancer associations in epidemiologic studies (Patel et al., 2017), pediatric
disease diagnosis (CEHCID 2009), water treatment needs, strategies for protecting vulnerable
populations (Zheng and Ayotte 2015), and other public health concerns (U.S. DOC 1990a).
Identifying users of wells could improve emergency response to spills (MDCH 2013; NMED and
U.S. EPA 2015) and improve the evaluation of risk pathways for groundwater contaminant
remediation (ASTM 2015).
In the U.S., a common ground water contaminant source that potentially threatens well users is
underground storage tanks (USTs). Petroleum product releases have been reported from over
530,000 underground storage tanks, with almost 71,000 cleanups remaining to be completed
(US EPA 2016). One of the main potential pathways for exposure to petroleum hydrocarbons is
the consumption of water from private domestic wells. Studies of the length of contaminant
plumes indicate the expected extent of contamination from leaking underground storage tank
9
-------
sites (API, 1998, Connor et al., 2015). Although based on limited data, these studies indicate
that the maximum observed extent of contaminant plumes is on the order of 500 m.
Well logs reported to the states provide a source of information for developing a nationwide
estimate of well usage, but the availability of these data varies by state. Data may be
incomplete for various reasons including: reporting requirements being imposed relatively
recently (MDEQ2015; OACR 2015); variable compliance with reporting requirements (OWRB
2014); lack of physical location data (MEEA 2015); exclusion of "grandfathered" wells (NDER
2015); limited ability of state agencies to compile data (PGS 2015); or legal restrictions (CDWR
2015).
Indirect data on well use were developed from areas without public water by negative
inference from a state-wide dataset on water supply pipelines in New Hampshire (Hayes and
Horn, 2009). Indirect national data on well use were also developed after the U.S. Congress
authorized a housing survey in 1939. Beginning in 1960 and continuing through 1990, a
question on the source of water supply was added to the long form census asking if water was
obtained from a public water supply, individual well or other source (U.S. DOC 2009).
Respondents were instructed to indicate an individual well if it supplied four or fewer
residences (U.S. DOC 1993).
The source of water continued as a question on the American Community Survey, and was
subsequently transferred to the American Housing Survey. The current sample size of 55,000 in
rural areas and more than 5,000 in 21 selected metropolitan areas is not adequate to present
results on a county or smallerspatial basis after 1990 (Eggers 2009). Thus, Earle et al. (2011)
10
-------
inferred well usage from the 1990 U.S. census and developed a relationship to gas station
location data for estimating the potential for contamination of wells by leaking underground
storage tanks. Mashburn et al. (2013) based domestic groundwater use estimates on 1990
census data and the estimated population living outside the areas of public water supply.
Because these studies relied on the 1990 census data, a time and potentially spatial-resolution
gap exists in high-resolution estimates of well usage.
The purpose of this work was to develop a method to update 1990 estimates of well usage
from publicaIly-available data and to determine the relationship between underground storage
tanks (USTs) and wells in a pilot study. The study was designed so the method would be
extensible to the entire U.S.
11
-------
Data and Methods
Study Area
Because the state of Oklahoma freely distributes reported-well data and underground storage
tank data, Oklahoma was selected for the pilot study (Figure 1). Water resources in Oklahoma
include two major river systems (the Red and the Arkansas Rivers), ten major bedrock and
eleven major alluvial aquifers (OWRB 2012). Of interest to this study are major aquifers,
including the formations of the Hennessey Group, the Garber Sandstone and the Welling
formation, which yield small to moderate amounts of fairquality water (Bingham and Moore
2004) and are considered major aquifers (Figure 2). Cities adjoining Oklahoma City to the east
use public wells to tap the Garber-Wellington Aquifer (Figure 2 and Figure 3), as well as using
surface water supplies from reservoirs (Edmond 2009). In contrast, Tulsa is situated over shale,
sandstone, and thin coal beds of the Seminole formation, which yield small amounts of poor
quality water that are insufficient for public supplies (Engineering and News Record 1924;
Clinton 1945). For the most part, public supplies dominate in the areas around Tulsa, with the
exception of an area in Sand Springs, Oklahoma, which has no public supplies (OWRB 2017) and
is situated along the Arkansas River where terrace deposits supply good quality water (Marcher
and Bingham 1989, Figure 2). Similarly, Enid, Oklahoma is situated over the Enid Isolated
Terrace Deposit which supplies moderate amounts of fair to good quality water (Bingham and
Bergman 1980) and forms the supply for the city water system.
-------
Figure 1. Map of the United States showing the location of Oklahoma.
Figure 2. Major (green) and alluvial (brown) aquifers of Oklahoma (Oklahoma Water Resources Board,
1998).
13
-------
-JL s
[Sand%prinq"g
fe'-Tulsa*H
WKiahomalteiitVj
Wells per Km
0.00 - 0.01
0.02 - 0.1
o.ii
H 1 -10
Hi 10- 146
0 100 200 400 Km
Figure 3. Density of housing units using private domestic well use inferred from 1990 Census on
a census block group spatial basis, with locations of cities discussed in the text.
14
-------
Edmond
Lake Aluma
Forest Park Nicoma
_ , Park
Del
Choctaw
El Reno
Bethany
Oklahoma
City
Norman
10 20 40 Kilometers
J i i I i i I I
Figure 4, Locations of Oklahoma cities discussed as examples in the results section.
15
-------
Well Log Data
The Oklahoma Water Resources Board (OWRB) distributes compiled data from well logs
reported under the state's well driller registration requirements (OACR 2015, OWRB 2015a).
Well locations are either estimated by location within the U.S. public land survey system (PLSS)
township-section system or reported as latitude-longitude, which became a requirement after
2009.
Water supplies in Oklahoma
Many independent cities and unincorporated areas either border Oklahoma City or are
contained within the same county (Figure 4). These include communities with no water
systems (Forest Park, Nicoma Park, and Lake Aluma), areas of historical private domestic well
usage (Bethany, Oklahoma, see Jacobsen and Reed, 1949), and cities with water systems which
do not serve their entire populations (i.e., Choctaw, Del City, El Reno, Edmond, and Oklahoma
City (Edmond 2009, Layden 2013, OWRB 2017)), and allow use of private wells, although
sometimes limited to lots of a certain size or larger (e.g., Edmond, 2017).
Estimation Methods for Counties, Census Tracts and Census Block Groups
From U.S. Census, USGS, and Oklahoma Water Resources Board (OWRB) data, two approaches
were developed for estimating the density of private wells. To address scale and zoning issues,
16
-------
associated with administrative units (Salmivaara et al. 2015), the results are compared at the
county, census tract, and census block group administrative levels. The first method is based
on the number of reported wells (RW) and housing units lost during a specified time period:
, 11 _ (Ainit\ , a Nw _ r NHU-lost
\ •*-) rpdw—est rpdw—init I A j ' ^ A Jpdw A
v/1?tew/ ^new Anew
where ppwd-est is the well density estimate over an area of Anew, PpWd-init's the initial well
density over the area An/t, Nw is the number of wells, fpdw is the fraction of well use to total
water supply, and NHU~lost js the number of housing units lost per unit area. The initial well
^new
density and fpdw are inferred from the 1990 census results. The method is applied in two
increments corresponding to census years: 1990 to 2000, and 2000 to 2010. The quantity fpdw
is updated after each incremental calculation is made, allowing for changing spatial patterns of
well use. Including the loss of housing units accounts in part for the loss of wells, as the well
records may only indicate wells added.
The second method is based only on the net change in housing units (NHU):
(2) Ppdw-est ~ Ppdw-init (A ) fpdw
Anew
where A—^ is the net change in housing units per unit area. The fraction of private well use
^new
fpdw is determined from the 1990 census results. Variants on the NHU method allow for
updating of fpdw. First, for calculating the 2010 well density, fpdw was updated from the well logs
reported through 2000. Second, county-level water use data are available from USGS from
17
-------
1985 to 2010, which were used to update fPdw county-level estimates for 2000 and 2010 using
the NHU method.
Any estimates which produced negative well density for either method were replaced with a
value of zero. Tests of statistical significance were performed on results determined for each
spatial basis using Minitab 14 software. The Mann-Whitney non-parametric test was used
because the results were not normally distributed.
To determine the initial density of private wells [ppaw-init) anc' fraction of private well use
{fpdw) 1990 census data were used. Well use data were collected from the "long form" which
was distributed to a sample of approximately 17% of the U.S. Population. Respondents were
asked if their water source was public, a drilled well, a dug well, or "other". As noted by
Maupin et al (2014), some water in the U.S. is self-supplied from surface water or cisterns. The
smallest unit used in the pilot study is the block group, because it is the smallest unit for which
the census bureau could supply sampled results (i.e., the well use data). The block groups were
designed to contain an optimum of 400 housing units (US. DOC, 1990b).
Census results were gathered for counties, census tracts, and census block groups from the U.S.
Census Bureau and the National Historical Geographical Information System (Minnesota
Population Center 2011). Shape files containing these data for Oklahoma were joined to
counties, census tracts, and census block groups to generate comparisons on three spatial
scales.
The number and size of census tracts and block groups ("geographies") can change from census
to census according to criteria established by the census bureau (FR 2008). The well estimates
18
-------
for these geographies were developed by beginning with 1990 and determining the areal
density for each quantity in equations 1 and 2. To provide common-sized geographies between
census years, and to account for the possibility of changing numbers of geographies, the
polygon to raster tool in ArcMap 10.1 was used to generate raster datasets with a cell size of 20
m. This size was smallest which allowed for practical computation, and also caused the term
An/t/Anew appearing in equations 1 and 2 to equal 1, allowing for direct comparison of quantities
between years. All outputs were normalized to a shapefile containing 2010 geographies, by
creating a zonal statistics table for each 2010 census tract and block group, which determined
the fraction of earliergeographies contained within the 2010 land division. The 1990 and 2000
data were then assigned using the weights from the zonal statistics table. OWRB data on
reported wells were joined to the 2010 shapefile in two groups covering the ten-year spans
between censuses. The well densities were then determined according to equations 1, 2, and
their variants.
Cities and Neighborhoods
Neighborhoods and cities that rely on solely on wells were identified by OWRB maps indicating
high well density. Residences were counted from Google maps available in July 2015 and the
number of wells counted from the OWRB map of reported wells (OWRB 2015b).
Positional Accuracy
Land in Oklahoma, and 34 other states is divided according to the United States Public Land
Survey System (PLSS), whose principle small-scale unit is the section (250 ha, 640 ac, 1 mi2)
19
-------
(Gates and Swenson, 1968). Common subdivisions are the quarter section (64.75 ha, 160 ac),
quarter-quarter section (16.19 ha, 40 ac), and quarter-quarter-quarter section (4.05 ha, 10 ac).
The Oklahoma well log dataset contained wells located by five methods: global positioning
system (GPS) corrected (8,103 wells), GPS uncorrected (10,060 wells), interpolation from PLSS
(37,065 wells), mathematical conversion program (11,201 wells), and unspecified (3,956 wells).
Well positions determined from PLSS units were assigned byOWRBtothe center of the unit,
and thus the maximum potential location error is equal to the distance from the center of the
unit to any corner. Because each well lies within a census block group, the significant
comparison was to a characteristic size of the census block groups. The potential error from
the estimates was assessed with a Monte Carlo procedure where a quarter-quarter-quarter
PLSS unit was randomly located within a representation of a census block group, which for
simplicity was taken as a square (Figure 5). The fraction of the PLSS unit that lay outside the
census block group was considered as the probability of error (see Figure 5, right). The process
was repeated 100,000 times for each of the 2965 census block groups and the statistics of the
results were determined.
20
-------
A
PLSS Unit
Census BlockGroup
Census BlockGroup
PLSS Unit
Census BlockGroup
Figure 5. Spatial relationships for public land survey system (PLSS) placement in census block
group error estimates. PLSS unit entirely contained within a census block group (left). PLSS
unit partly outside census block group (center). Illustration of error estimate which equals (B +
C+ D)/(A+ B + C+ D) (right).
Underground Storage Tank Data
The locations of regulated underground storage tanks (USTs) were obtained from a list
distributed by the Oklahoma Water Resources Board (OWRB. 2015b) and compared to reported
and estimated well locations. Each of the active 3033 USTs managed by the Oklahoma
Corporation Commission and the reported wells were located within a census block group. A
suite of potential impact distances was chosen (15, 30, 76, 150, 230, 300 and 1,610 m), based
on reported plume lengths (API, 1998, Connor et al., 2015) and knowledge of U.S. state agency
programs. Next, the neighboring census block groups of each census block group were
determined. The number of wells within each selected distance was determined for each UST
beginning with the census blockgroup containing the UST. Reported locations of wells were
21
-------
supplemented with estimated well locations in a Monte Carlo procedure. The latter were
randomly selected to match the estimated RW-method well density of the census block group.
The distance between the UST and each reported and estimated well was determined and the
distances were binned into categories based on the chosen impact distances. The same
calculations were performed on each neighboring census block group and neighbors-of-
neighbors, until no more UST-to-well distances fell within the potential impact distances.
Because estimated wells were included, the procedure was repeated 10,000 times, and the
statistical characteristics of the binned counts of wells were determined.
22
-------
Results and Discussion
Data Error and Method Evaluation
Error in reported well locations arose from three sources. First, a few wells were found to have
implausible coordinates as they plotted outside the state. Second, some wells were said to
occur in a county that differed from that in which they plotted. The majority of these were
found to be errors in designating the county rather than error in the well position perse. This
problem was particularly prevalent when the well was located near a county border, indicating
that the driller may not have known the precise county boundaries. For the analysis of
counties, the county designation was corrected to the county in which the well plotted. Third,
some wells designated for a specified land survey system unit/subunit plotted elsewhere or
were located only by reference to the PLSS unit. In total, of the 41,372 domestic wells reported
between the 1990 and 2010 censuses, 2.05% (847) were omitted when a correction could not
be made.
Of the wells located only by PLSS land units, the majority (99.52%) were located within a
quarter-quarter-quarter section which has a maximum positional uncertainty of 140 m (center
to corner). The minimum characteristic dimension of the census block groups was 236 m. Thus
the positional uncertainty in these well locations is on the same order as the smallest census
block group, but an order less than the median size (1,150 m), and two orders of magnitude less
than the maximum-sized census block group (40,400 m). The median estimated probability of a
well being placed in the wrong census block group was a maximum of 28% for the smallest
census block group and dropped steadily with increased block group size. The median
23
-------
estimated probability is zero for all census block groups above a characteristic dimension of 474
m (area of 0.45 km2) which accounts for 90.01% (2672) of the block groups. (Figure 6). Because
the approach is to estimate the spatial density of wells within administrative units (i.e., census
block group), the impact of the error in position is that the well could be assigned to an
incorrect areal unit. Because this possibility exists forall adjoining administrative units,
inaccurate well placement could both place wells outside a given administrative unit and, from
an adjacent unit, inside an administrative unit. The result could be shifted well densities on
both the maps and in calculated results, but any impacts are tempered by the positional
inaccuracy being less than 140 m for 99.52% of the wells.
24
-------
»«« A
0.1
i—i
>
0.01
Characteristic Census Block Group Dimension
• Bounded Error in Well Position
LO
(J)
CD
O
O
A 25% qua rtile through 100% qua rtiles
0.001
Median Probability of well Plotting in Adjacent Census Block
Group
0.1 1 10 100
Dimension (km)
Figure 6. Frequency distribution of characteristic dimension of census block groups and
relationship to public land survey system units. "Max Well Errors" follow from the size
distribution of reported well locations within PLSS units of which 99.5% are smaller than the
smallest characteristic dimension of the census block groups (blue line). The red line represents
the median probability of a well plotting in an adjacent census block group as determined by
Monte Carlo analysis.
As a check on the well use and housing unit census results for 1990, the fraction of well use was
calculated by dividing the number of well users by the number of housing units and also by
dividing by the sum of the four reported water uses (public, drilled well, dug well, and other). A
comparison of the two calculations found that the median absolute value of the difference was
9% with a range of 0 to 199%. A second check addressed sampling error. Because they are not
25
-------
derived from a 100% survey, the 1990 well use results are subject to sampling errors (US DOC,
1993). For each Oklahoma county, the census bureau provided the percent of housing units
sampled and a formula for calculating the sampling error (US DOC, 1993 and Table 1).
Following this approach, the composited sampling error for the entire state of Oklahoma was
0.60%. For the individual counties, the sampling error ranged from 1.9% to 32%, with a median
of 5.6% (Table 1). These results indicate increasing sampling error with increasing sparseness of
well use, as the maximum error occurred in a county with only 68 wells (Figure 7). The
associated error is only 22 wells or 0.01% of the 1990 estimate.
26
-------
Table 1. Error estimates forthe 1990 census inference of private domestic we 11 use.
Percent-in- Estimated
County
1990 Wells
1990 Housing
Units
Percent-in-
Sample
Standard
Error
sample Design
Error
Total Sample
Error
Error Fraction
Adair
3,258
7,124
24.6
94
1.2
113
0.035
Alfa Ifa
773
3,357
39.6
55
0.6
33
0.042
Atoka
1,274
5,110
18.5
69
1.2
83
0.065
Beaver
1,522
2,923
31.7
60
0.6
36
0.024
Beckham
1,448
9,117
17.2
78
1.2
94
0.065
Blaine
1,544
5,729
28.3
75
1.2
90
0.058
Bryan
2,958
14,875
19.8
109
1.2
131
0.044
Caddo
4,594
13,191
28.1
122
1.2
147
0.032
Canadian
3,597
28,560
14.4
125
1.3
163
0.045
Carter
2,557
19,201
16.3
105
1.2
126
0.049
Cherokee
4,182
15,935
13.5
124
1.3
161
0.039
Choctaw
2,363
6,844
18.9
88
1.2
106
0.045
Cimarron
539
1,690
36.6
43
0.6
26
0.048
Cleveland
10,928
71,038
13.9
215
1.3
280
0.026
Coal
348
2,725
31.7
39
0.6
23
0.067
Comanche
1,280
43,589
15.2
79
1.2
95
0.074
Cotton
376
3,152
24.4
41
1.2
49
0.130
Craig
582
6,041
19
51
1.2
62
0.106
Creek
4,765
25,143
18.5
139
1.2
167
0.035
Custer
1,276
11,636
18.5
75
1.2
90
0.071
Delaware
8,159
16,808
15.5
145
1.2
174
0.021
Dewey
787
2,733
40.6
53
0.6
32
0.040
Ellis
838
2,449
36.3
53
0.6
32
0.038
Garfield
2,793
26,502
16.2
112
1.2
134
0.048
Garvin
2,293
11,932
20.9
96
1.2
115
0.050
Grady
6,194
17,788
19.4
142
1.2
170
0.028
Grant
449
2,955
40.3
44
0.6
26
0.058
Greer
272
3,126
23.9
35
1.2
42
0.155
Harmon
68
1,793
17
18
1.2
22
0.319
Harper
457
2,077
41.3
42
0.6
25
0.055
H a s ke 11
1,557
5,138
28.7
74
1.2
88
0.057
Hughes
1,234
6,021
24.1
70
1.2
84
0.068
Jackson
657
12,125
19.4
56
1.2
67
0.102
Jefferson
493
3,522
41
46
0.6
28
0.056
Johnston
1,062
4,478
21.8
64
1.2
76
0.072
Kay
1,329
22,456
22
79
1.2
95
0.071
Kingfisher
1,225
5,791
22.7
69
1.2
83
0.068
Kiowa
436
5,645
28.4
45
1.2
54
0.123
Latimer
308
4,303
15.3
38
1.2
45
0.147
27
-------
Percent-in- Estimated
County
1990 Wells
1990 Housing
Units
Percent-in-
Sample
Standard
Error
sample Design
Error
Total Sample
Error
Error Fraction
Le Flore
2,185
18,029
21.7
98
1.2
118
0.054
Lincoln
6,138
12,302
22.6
124
1.2
149
0.024
Logan
5,288
12,277
19.2
123
1.2
147
0.028
Love
973
3,583
17.9
60
1.2
71
0.073
Major
1,520
3,855
20.2
68
1.2
81
0.054
Marshall
1,311
7,389
15.7
73
1.2
88
0.067
Mayes
2,728
15,470
14.8
106
1.3
138
0.051
McClain
3,700
9,300
21.7
106
1.2
127
0.034
McCurtain
4,545
13,828
18.9
124
1.2
148
0.033
Mcintosh
2,019
10,708
21.8
91
1.2
109
0.054
Murray
504
5,742
13.3
48
1.3
62
0.124
Muskogee
1,256
28,882
15.9
78
1.2
93
0.074
Noble
610
4,894
26.2
52
1.2
62
0.102
Nowata
281
4,534
18.8
36
1.2
44
0.155
Okfuskee
936
4,894
23
62
1.2
74
0.079
Oklahoma
21,092
279,340
14.4
312
1.3
406
0.019
Okmulgee
482
16,431
18.6
48
1.2
58
0.120
Osage
2,203
18,196
17.2
98
1.2
118
0.054
Ottawa
2,082
14,064
24
94
1.2
113
0.054
Pawnee
1,653
7,407
42
80
0.6
48
0.029
Payne
3,068
27,381
16
117
1.2
140
0.046
Pittsburg
1,426
19,433
21.8
81
1.2
98
0.068
Pontotoc
1,574
15,094
18.4
84
1.2
101
0.064
Pottawatom ie
8,237
24,528
18.4
165
1.2
198
0.024
Pushmataha
1,694
5,190
17
76
1.2
91
0.054
Roger Mills
880
2,048
33.3
50
0.6
30
0.034
Rogers
811
21,455
16.8
62
1.2
75
0.092
Seminole
2,921
11,404
19.9
104
1.2
125
0.043
Sequoyah
2,010
14,314
18.4
93
1.2
112
0.055
Stephens
3,593
19,675
16.6
121
1.2
145
0.040
Texas
1,558
7,328
23.9
78
1.2
94
0.060
Tillman
403
4,704
24.2
43
1.2
52
0.128
Tulsa
1,827
227,834
13.6
95
1.3
124
0.068
Wagoner
1,252
19,262
16.1
77
1.2
92
0.073
Washington
293
21,707
15.3
38
1.2
46
0.156
Washita
1,200
6,101
27.3
69
1.2
83
0.069
Woods
573
4,782
23.8
50
1.2
60
0.105
Woodward
1,473
8,512
19.8
78
1.2
94
0.064
28
-------
The standard sample error, SE(X), is calculated as SE(X) — F J5X (l — where X is the number of wells, W
is the number of housing units, and F is the percent-in-sample factor for the source of water census question (US
DOC, 1993, page C-10),
For sourceof water, the factor, F, is 1.3 for percent-in-sampleof 15% or less; 1.2 for percent-in-sample of 15%
to 30% and 0.6 otherwise (US DOC, 1993, page C-ll).
0.35
1990 Census Estimated Sampling Error (Oklahoma Counties)
<
~
0.30
a;
% °-25
ro
£
l/i
^ 0.20
o
1-
11 1
0.15
t
"5.
%
E
ro 0.10
&
to
~
~V
0.05
* *+ * * ~ ~
-
5,000 10,000 15,000 20,000 25,000
Number of Wells
Figure 7. Standard sampling error estimates for Oklahoma counties determined from equations
presented below Table 1.
29
-------
The Census Bureau attempts to control additional non-sampling errors that might result from
respondent, enumerator, and processing errors, as well as missed households and nonresponse
(US DOC, 1993). Non-sampling errors, if random, increase the variability and are reflected in
the standard sampling error (US DOC, 1993). Using the 9% error calculated from the first test
as a bound, the actual number of wells in use in Oklahoma in 1990 could have ranged from
161,000 to 193,000. Despite these errors, the estimates are an order-of-magnitude above the
number of reported wells for 1990 (15,042), so the inference from the census provides a
suitable basis forthe estimates updated to 2000 and 2010.
As a check on the net housing unit method, census data in ten-year increments were used to
estimate well use in Oklahoma counties for 1970, 1980, and 1990 using the net housing unit
method. The estimates were compared against the census-reported values (Figure 8 to Figure
13) as a means to assess the NHU method's viability. The highest statewide errors occurred for
the 1970 (26.8%, Figure 8 and Figure 9) and 1980 (27.6%, Figure 10 and Figure 11) estimates.
The error dropped to 5.7% for the 1990 estimate (Figure 12 and Figure 13). Given that the key
estimated factor is the ratio of well use to public supply, the lower 1990 estimate of 5.7%
indicates that the ratio remained relatively constant from 1980 to 1990. The result is explained
by data from OWRB (1998), which show ten or more rural water systems were added in
Oklahoma each year between 1962 and 1974, implying a shift from wells to public supplies
between 1960 and 1980. The decade between 1985 and 1995 typically showed less than five
rural water districts added in each year, implying a smallershift to public water supplies (Figure
14) and thus smallererror forthe 1990 result (Figure 12 and Figure 13).
30
-------
30.000
o
r-.
cr>
c 25,000
1970
!/>
?
-------
1970 Estimate
w
_D
-gO.5
c
-------
30.000
o
oo
2
c 25,000
19S0
t; 20.000
i_
Q_
O
i_
_q 10,000
y-G.9S49ji + &3D.24 ^
R* - D.92IG
y
,¦/
A 1
-------
1980 Estimate
H£=&
t
0.5
QJ
D
5
C
0J
>0.5
-------
30.000
1990
o
a*
CT>
c 25,000
/
o.smsi t- ia?.nj
i_
Q_
O
_q 10,000
T5
<0 5,000
E
r- -
o I *90 Estimate from 1 q =?n ft at a
1 lu 1 Ins
0 5,000 10,000 15,000 20,000 25,000 30,000
Private Domestic Weil Usage Inferred from 1990 Census Results
Figure 12. Comparison of county-level estimated private domestic well use and that inferred
from the U.S. Census for 1990.
35
-------
1990 Estimate
i
>0.5
County
Figure 13. Error estimates (Census Value - Estimate)/Census Value for county-level PDW
estimates given by county for 1990,
36
-------
Rural Water Systems Added
45
II 1
ml
II
II, II
1
ml
li
III lull. Ill
II
lllll
1
1
Figure 14. Rural water districts added in Oklahoma 1889-1995 (data from OWRB, 1998).
37
-------
State-Wide Inferences
In 1990, public or private water systems were reported as the source of water for over 1.2
million housing units in Oklahoma (Table 2), which were located both in urban and rural areas.
Individual drilled or dug wells were reported for about 177,000 housing units in Oklahoma, the
majority of which were located in rural areas. However, more than 30,000 housing units were
reported as using individual wells in urban areas. In total, well use constituted 12.6% of
Oklahoma water supply on a housing unit basis. When displayed on a census block group
spatial basis, the 1990 inferred densities of well usage showed high levels in areas surrounding
Oklahoma City and in northeastern Oklahoma (Figure 3) for reasons discussed below. The
census results illustrate the limitations of the well reports, because they began in the mid-
1980s (OACR, 2015), the reported wells undercount the total. A cumulative total of 15,042
were reported through 1990, which is 8.5% of the census inference of 177,074 (Figure 15).
Because reporting of well logs is now required, undercounting since the 1980s should be less
but this presumption requires testing (see section "City and Neighborhood Analysis" below).
Table 2. Oklahoma water sources from the 1990 census (U.S. DOC, 1993).
Water Supply Total Fraction of Urban Rural
Total
Public or private water system 1,223,121 0.8696 928,727 294,394
Total well use 177,074 0.1259 30,259 146,815
Individual drilled well 163,916 0.1165 28,026 135,890
Individual dug well 13,158 0.0094 2,233 10,925
Other Source 6,304 0.0045 556 5,748
38
-------
70,000
CO
cc
O 60,000
T3
o>
50,000
Q.
CD
cc
tst 40,000
~ 30,000
to
a 20,000
S 10,000
o
0
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
Year
Figure 15. Cumulative number of domestic wells reported to the Oklahoma Water Resources
Board.
County-level Estimates
Based on data aggregated to the county level, the population, number of housing units, and
estimated number of wells all increased in Oklahoma from 1990 to 2010 (Table 3). The
-------
reported-wells estimate for well usage in 2010 (215,806) was less than that determined from
simple addition of the 1990 Census inference and the raw number of wells reported between
1990 and 2010 (217,597), because of accounting for loss of housing units in the RW method.
Based on the RW method results, the fraction of population relying on private wells (fpdw)
increased slightly from 0.1259 to 0.1297. The increase in reliance on private wells during this
time period was greater than the increase in population (10.4% versus 9.7% and 11.3% versus
8.7%, Table 3).
Table 3. Housing unit, population and private domestic well characteristics for the state of
Oklahoma
RW Direct Population %incin
Housing Method Well Using %incin Popluation
year Units Population Estimate Count Housing Units/km2 Wells/km2 fPdW Wells Population UsingWells AfPdw
1990 1,406,499 3,145,585 177,074 177,074 7.9189 0.9970 0.1259 396,020
2000 1,514,400 3,450,654 191,796 193,290 8.5265 1.0799 0.1266 437,019 9.7 10.4 0.0060
2010 1,664,378 3,751,351 215,806 217,597 9.3709 1.2150 0.1297 486,406 8.7 11.3 0.024
(°)1990 values are inferred from the 1990 U.S. Census.
High density of wells was found in the center of the state and along the border with Arkansas
and Missouri (Figure 16) using the RW method. The highest densities occurred in Oklahoma
County and adjoining Cleveland County, which also contains the extensive public water system
of Oklahoma City (OKC 2014). The third highest well density was found in Delaware County in
Northeastern Oklahoma, with a 2010 rural population of 29,103. Rural water districts covered
around 24% of the county in the latest year data are available, 1995 (OWRB 2017). These three
40
-------
counties share the characteristics of availability of high quality ground water (Marcher and
Bingham 1989, Figure 2), high or relatively high population, and limited coverage of public
water supplies (OWRB, 1998 and Figure 17).
Summary statistics (Table 4) show an increasing mean well density from 1.138 wells/km2 (1990
census inference) to values over 1.389 wells/km2for 2010, depending on the method used.
Similarly, the county-level maximum density in 1990 was 11.489 wells/km2, which increased to
over 12 or 13 wells/km2 for 2010 depending on method. As an example of the similarity of
their results, the RW and NHU (with fPdw updating to 2000) results for 2010 plotted close to a
1:1 line (Figure 18). The conceptual importance of updating the fraction of well usage, fPdw, is
illustrated by the changes in the mean from 0.1951 in 1990 to 0.2166 in 2010.
41
-------
Missouri
Figure 16. Reported-wells estimate of private domestic well density for 2010 on a county-wide
spatial scale. Oklahoma (top) and Cleveland (bottom) counties in the center of Oklahoma, and
Delaware County on the border with Missouri and Arkansas had the highest estimated well
density in the state.
42
-------
Amarillo
Wichita Falls
Lubbock
Denton
Piano,
Figure 17. Approximate service areas of public water systems in Oklahoma (Oklahoma Water
Resources Board, 1998).
43
-------
Table 4. Summary statistics for private domestic well density estimates (wells per km2) from
the 1990 census, reported wells, and net housing unit estimates for the 77 counties, 1,046
census tracts, and 2965 census block groups of Oklahoma.
Estimate Minimum Q1 Median Q3 Maximum IQR Mean Variance
County Scale (median size 2078 km2)
Dens 90
0.0489
0.301
0.592
1.313
11.489
1.013
1.138
2.763
RW 2000
0.0571
0.327
0.629
1.357
12.249
1.030
1.233
3.275
RW 2010
0.0636
0.368
0.700
1.461
14.232
1.094
1.388
4.423
NHU 2000
0.0449
0.296
0.594
1.453
12.134
1.157
1.254
3.509
NHU 2010
0.0421
0.295
0.631
1.617
13.154
1.322
1.389
4.584
NHU 2010 f2ooo
0.0396
0.292
0.629
1.583
13.170
1.290
1.372
4.472
NHUGS 2000
0.0447
0.316
0.604
1.398
11.674
1.082
1.254
3.328
NHUGS 2010
0.0435
0.289
0.616
1.519
12.463
1.231
1.327
3.934
Census Tract
Scale (median size
12.40 km2)
Dens 90
0.00
0.00
0.492
2.382
81.111
2.382
3.083
61.395
RW 2000
0.00
0.000830
0.618
2.556
81.691
2.382
3.196
62.491
RW 2010
0.00
0.0652
0.747
2.800
77.742
2.735
3.494
70.072
NHU 2000
0.00
0.00
0.535
2.704
114.178
2.704
3.901
105.340
NHU 2010
0.00
0.00
0.652
3.153
113.334
3.153
4.873
160.029
NHU 2010 (fzooo)
0.00
0.00
0.641
3.130
114.132
3.130
4.607
138.342
Census Block Group Scale
(median
size 2.66 km2)
Dens 90
0.00
0.00
0.276
2.082
143.454
2.082
3.465
106.146
RW 2000
0.00
0.00
0.376
2.322
145.071
2.322
3.575
105.966
RW 2010
0.00
0.00
0.478
2.739
145.313
2.739
3.887
113.597
NHU 2000
0.00
0.00
0.274
2.373
191.630
2.373
4.258
157.943
NHU 2010
0.00
0.00
0.282
2.762
401.921
2.762
5.403
295.804
NHU 2010 (f2ooo)
0.00
0.00
0.281
2.700
200.703
2.700
5.024
215.554
(a)IQR= interquartilerange, Q3 - Q1
44
-------
30,000
• Estimated private domestic wel Is for 2010
1:1 line
Linear (Estimated private domestic wel Is for 2010)
y = 0.9851x-9.1837
R2 = 0.9811
O 25,000
tH
O
CM
0) 20,000
15,000
00
c
10,000
~Z. 5,000
0
5,000
10,000
15,000
20,000
25,000
30,000
Reported-wells Estimate for 2010
Figure 18. County-level predicted private domestic well use for 2010 from housing unit
increase compared against predicted well use from well logs reported to the OWRB.
Using USGS domestic self-supplied water-use estimates to update fPdw (rather than the
reported well logs) gave results which were similarto the RW method results. The USGS
estimates were based on available data which differed among counties (Hutson 2007). The
estimates show fluctuating levels of well use over the period 1985 to 2010 (Figure 19 for five
selected counties and the Appendix for all Oklahoma counties). In particular, several counties
showed large increases followed by large decreases in public water usage (i.e., Adair, Hughes,
45
-------
Latimer, Roger Mills, and Washita), which suggests further refinement is needed for county
level estimates of well usage (Figure 19),
a;
Z)
"aj
+¦>
>
L.
CL
c
o
tJ
ra
1
0.8
0.6
0.4
0.2
-0.2
-0.4
Five Oklahoma Counties
•- -0.6
-------
Table 5. Statistical comparison of 1990 private domesticwell densityand estimates madethe RWand
NHU methods. Bold italic values differ statistically at 0.05 P value in the table. Quantity names are as
follows: Dens 90 = inference from 1990 census; RW 2000 = reported-welis result for 2000; RW 2010 =
reported wells result for 2010; NHU 2000 = nethousingunitresultfor2000; NHU 2010 = net housing
unit result for 2010; NHU 2010 (f2ooo) = net housing unit result for 2010 with the fraction of private
domesticwell use (fpdw) updatingfrom RW 2000 result; NHU USGS 2000 = net housingunit resultfor
2000 with/pdw updatingfrom USGS results; andNHU USGS 2010 = nethousingunitresultfor2010 with
fpdw updating from USGS results.
Test Quantity 1 Test Quantity2 P Value
Counties Census Tracts Census Block Groups
1990 Census Inference versus 2000 and 2010 estimates
Dens 90
RW 2000
0.6003
0.1409
0.0177
Dens 90
RW 2010
0.2799
0.0017
<0.0000
Dens 90
NHU 2000
0.8171
0.3688
0.2273
Dens 90
NHU 2010
0.5804
0.1082
0.0360
Dens 90
NHU 2010 (f2000)
0.8708
0.0752
0.2216
Dens 90
NHU 2000 (USGS)
0.6053
--
-
Dens 90
NHU 2010 (USGS)
0.6857
--
-
RW method versus NHU method
RW 2000 NHU 2000 0.8301 0.5963 0.2748
RW 2010 NHU 2010 0.6725 0.1872 0.0177
RW 2010 NHU2010 (f200o) 0.6359 0.2190 0.0164
NHU method with and withoutfraction well use updating
NHU 2010 NHU2010 (f200o) 0.9165 0.8871 0.9854
Fraction of Private Domestic Well (fpdw) Use
fpdw 1990 fPdw 2000 0.6489 0.9294 0.0005
fpdw 1990 fPdw 2010 0.5295 0.6774 <0.0001
fpdw 2000 fpdw 2010 0.8087 0.7255 0.1551
47
-------
Census Tract Scale Estimates
At the census tract scale, the mean well densities increased from 3.083 in 1990 to 4.873
wells/km2 for 2010 depending on method (Table 4). Differing from the counties, some census
tracts had no wells, thus densities of 0.0 wells/km2, but all counties had at least a few wells.
Reflecting the smallerspatial basis of the census tracts (170 km2 versus 2,307 km2), the
maximum density was as high as 114 wells/km2, which is a value much higher than seen in any
entire county. The counties can contain locales supplied by urban and rural public water, or
contain open land with no need for private wells, so they have both higher and lower extremes.
By design the census tracts are conceptually more homogeneous. As seen in the statistical
results, however, the only statistically significant difference is between the 2010 RW results and
the 1990 census inference. Although generally corresponding to the county-level result and on
the eastern boundary (Figure 20), the smaller size of the census tracts allows sub-county
heterogeneity to appear, in counties showing either high or low well density estimates (Figure
20, left).
48
-------
Colorado
Colorado
Colorado
Colorado
0 125 250 500 Km
1 i i i I i i i I
Figure 20. Net housing unit (NHU) method (top), and reported-wells (RW) method (bottom)
2010 results for census tracts(left) and census block groups (right) in Oklahoma,
Census Block Group Scale Estimates
At the smallercensus block group scale (mean 2010 land area of 60 km2), the mean well density
started at 3.465 wells/km2 in 1990 and increased to as much as 5.403 wells/km2 in 2010,
depending on method (Table 4). Likewise, the maximum density went from 143.5 wells/km2 in
1990 to as much as 401.9 wells/km2 in 2010. Similar to the census tract scale, there were
census block groups that contained no wells. At this spatial scale, the RW method estimates
49
-------
for 2000 and 2010, and the fraction of well use differed statisticallyfromthe 1990 census
inference (Table 5).
At the census blockgroup level, small-scale variability increased overthe census tract results
and (Figure 20, right). The RW method generated lower estimates than the NHU in 37% of the
census block groups with no fPdw updating and 38% with fPdw updated to 2000. Higher NHU
results were evident by both the visible higher densities and more widespread distribution of
high densities in the NHU results (Figure 20, right).
The estimated difference between the 1990 well density and the RW 2010 estimates showed
large areas with 10% to 100% increases (Figure 21). This result implies that the fraction of well
use increased in many areas of the state including the central and northeastern Oklahoma
(Figure 22), the area of highest well usage. At the census block group level the changes in well
use both in numbers and in the fraction of households using private wells were statistically
significant (Table 5). This result reflects that the changes are occurring on a small spatial scale
that is better represented by the block group estimates. This effect can be seen in the
Oklahoma City area, where some census block groups showed strong increases and others
showed strong decreases. These are associated with expansion of cities surrounding Oklahoma
City which have relied simultaneously on both public and private water supplies.
50
-------
+
Colorado
Missouri
400 Km
Percent Change
H -100% --75*54
o
¦¦ -74% - -50%
o
-49% - -25%
—
n -24% - -5%
X
-4% -0%
0)
I lo%
11%-25*
5
¦ 26% - 50%
0
¦1 51% -100%
2
¦¦ 101% ¦ 9996%
Increase from 0%
Texas
0 100 200
Figure 21. Percent change in estimated well density 1990 to 2010 using the reported well log
(RW) method on the census block group level.
51
-------
Colorado
o
o
X
Q>
$
Q>
f
Percent Change
!¦ -100% - -75% -10% - -5%
¦¦-71%--50% -5% - 0%
M -50% - -25% I 10%
H -25% - -10% 0% - 5%
5% -10% 1H 76% -100%
10% -25% B| > 100%
¦I 25% - 50%
¦i 50% - 75%
Kansas
Missouri
Texas
¦
400 Km
Figure 22. Estimated change in fraction of well to total water supply from 1990 to 2010 on the
census block group level.
52
-------
City and Neighborhood Scale Analysis
Example Cities without Public Water Supplies
Analysis at the city-and neigh borhood-level provides insight into the reasons forthe observed patterns
and a further check on the accuracy of reported data. Three communities we re identified in Oklahoma
County with no publicwatersupply systemsand with residents relianton wells: Forest Park, 2010
population 998; Lake Aluma, 2010 population 88; and Nicoma Park, 2010 population 2,393 (Figure 4
and Figure 17). Estimated densities of wells based on neighborhood counts of reported wellsand
existing residences resulted in higher we II density from the corresponding census block group 2010
estimatesforthe RW method fortwo reasons (Table 6.)
First, property records show that houses were built in Forest Park and Nicoma Park priorto the we li-
re porting requirement and few reported wells we re expected. Although the RW method is dependent
on the reported well data, it resulted in high well density because of its basis inthe 1990 census results
which had already identified these areas as relianton wells. Secondly, these cities include only small
undeveloped areas, sothe city estimates were higherwhen based onthe numberof housingunits,
ratherthan the area. Dividingthe neighborhood count by housingunits by the largerunit (county,
censustract, and census block group) well density givesan indication of how much higherthe well
densities withinthe cities with no public watersupplies can be compared withthe largerland unit;the
median result was 7.63 times larger for counties, 2.60 times larger for census tracts, and 2.19 times
largerfor census blockgroups (Table 6.)
Table 6. County, censustract, census blockgroup reported-wells method, and neighborhood-count
estimates of private domestic we II usage forthe cities of Lake Aluma, Forest Park, and Nicoma Park, all
of which have no publicwatersupply system.
53
-------
Location
Well density (wells/km2)
Count by Housing Units Divided by
Larger Unit Well Use
Wells Added Method 2010
Estimate
Neighborhood
Counts
County Census Census
Tract Block
Groups
by wells by
reported housing
units
County
Census
Tract
Census
Block
Groups
Whole
Lake Aluma Dr
(residences
only)
Lake Aluma Dr
(wholearea)
Whole
N. Bryant and
NE 50th ST
NE36TH and
N Coltrane
(1st
neighborhood)
NE36TH and
N Coltrane
(2nd
neighborhood)
N Bryant and
NE36th ST
Whole
NE23rd and N.
Westminster
14.3
14.28
14.28
14.3
14.28
14.28
14.28
14.28
14.3
14.28
41.9
41.9
41.9
41.9
41.9
41.9
41.9
41.9
77.3
77.3
36.3
36.3
36.3
36.3
97.9
28.8
36.3
36.3
97.9
145.6
75.6
147.7
42.6
77
147.7
42.6
77
Lake Aluma
14
Forest Park
6
13
25
12
Nicoma Park
15
98
48
63
109
184
141
213
6.86 2.34
3.36 1.15
4.41 1.50
7.63 2.60
12.89 4.39
9.87 3.37
14.92 2.76
2.70
1.32
2.19
3.00
5.07
1.44
1.44
Minimum for
AlI Cities
Medianforall
Cities
Maxi mumfor
Al I Cities
3.36 1.15
7.63 2.60
14.92 4.39
1.32
2.19
5.07
54
-------
Oklahoma City instituted an annexation policy to promote industrial development and to support Tinker
AirForce Base, amongotherobjectives(Oklahoman, 1959a, 1959b). Forest Park residents, however,
resisted annexation by the City of Oklahoma City (Oklahoman 1956,1957) and has remained
independent (Figure 23). Forest Park was originally 73 ha (180 ac) in size, and annexed surrounding land
to reach the size of 550 ha (1360 ac) (Everett, 2017a). Because the residences we re built before the
OWRB reporting requirements,fewwells have been reported to the state. Well use is presumed from
the 1990 census which indicated high density of private wells. Asimilarsituation exists in Nicoma Park.
Nicoma Park is another small city with no public water system (Figure 24). Nicoma Park was a planned
agricultural community whose purpose wasto develop a "poultry colony" where residents would raise
chickens to produce eggs on 1 to 5 acre lotsunderthe supervision of an expert (Everett. 2017b). The
projectended duringthe depression ofthe 1930s. The town persisted by residentsfindingemployment
in nearbyTinkerAirForce Base (sincethe 1940s), automobile manufacturing, and otherjobs inthe
Oklahoma City area (Everett, 2017b).
55
-------
Forest Park
L
0 0.5 1 Km
• OWRB Domestic Wells
Public Water Supply System
Forest Park
Figure 23. The City of Forest Park sits adjacent to Oklahoma City (Figure 4) and has no public
water supply system.
Nicoma Park
0 0.5 1 2 Km
1 i i i I i i i I
• OWRB Domestic Wells
X////\ Public Water Supply System
Nicoma Park
Figure 24. The City of Nicoma Park lies to the east of Oklahoma City (Figure 4) and has no
public water supply system.
56
-------
Individual Neighborhoods
Neighborhoods without public water supplies and recently-constructed homes were used
independently to estimate well usage and to evaluate the completeness of well reports (Table
7). The developments ranged in size from 0.15 to 1.32 km2, with median of 0.62 km2- The
median fraction of wells reported, as required, to OWRB from 22 developments was 0.55, and
the mean was 0.53, with range of 0.16 to 0.96. Adjacent subdivisions sometimes exhibited
large differences in reporting. OWRB (2014) recognized the failure among some drillers to file
required well completion reports. Similar to the communities discussed above, the estimates
of well density on a neighborhood basis generally gave higherdensity than the 2010 RW
method results for both the reported wells and the housing unit counts (Table 7). The higher
densities obtained from neighborhood counts reflects higherdensity of housing units within the
developments than in the undeveloped remainder of the areal unit (census blockgroups,
census tracts, and county). Similar to the cities, pockets of higher reliance on wells can exist
within larger units, and the median well densities can be 10.80, 11.24, and 9.36 times that of
the county, census tract, or census block group (Table 7). The maximum well density can be as
much as 500 times the census blockgroup density for certain neighborhoods.
57
-------
Table 7. Estimated private domestic well density for central Oklahoma neighborhoods, based
on county, census tract, census block group, and neighborhood counts.
County
Location
PD W density (wells/km2)
Estimated neighborhood count by housing
Fraction units/wells added estimate for
Reported counties, census tracts, and
census block groups
WellsAdded Method2010 Neighborhood Count wells<°>
Method
County Census Census By By
Tracts Block reported housing
Groups wells units
County Census Census
Tracts Block
Groups
Canadian
Canadian
Canadian
Cleveland
Cleveland
Logan
Logan
Logan
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
N Manning
Rd & E0980,
(SE 122 El
Reno)
SW59TH
and N. Czech
Hall Rd
NW
Expressway
and N. Frisco
Rd
E. Indian
Hills Rd &
168th Ave
NE
S. Harrah
Newalla Rd
and SE
104TH St
S. Penn & W.
Charter
NW 248th
(Waterloo) &
S. Portland
Rd (Hwy 74)
S. Douglas &
E. Waterloo
SE 74TH and
S. Choctaw
E. Danforth
and N.
Douglas Rd
E. Waterloo
& N.
Midwest
Blvd.
E. Covell and
N. Douglas
Blvd
SE 15th and
S. Dobbs Rd
SE 89TH St
and S. Indian
Meridian Rd
SE 44th & S
Choctaw Rd
E Hefner & N
Air Depot
Blvd
2.23 1.17 1.06
17
58
0.29 26.01 49.57
10.24 20.5 22.91
69
3.6 17.21 18.62 95
3.6 3.85 4.39 137
3.6 7.14 17.99 145
14.28 31.85 25.46 26
14.28 10.37 13.97 39
14.28 10.37 13.97 56
14.28 10.37 13.97 52
14.28 31.85 24.91 39
14.28 31.85 25.45 46
14.28 31.85 24.91 76
14.28 17.22 14.3 110
97
191
257
233
158
152
215
158
109
80
124
155
0.71 9.47
0.26
15.06
0.36 7.63
0.58 5.60
4.73
20.73
3.42
2.51
54.72
2.23 16.03 33.75 83 165 0.5 73.99 10.29 4.89
2.23 4.53 12.52 27 106 0.25 47.53 23.40 8.47
10.24 20.5 22.91 69 110 0.63 10.74 5.37 4.80
4.23
0.5 53.06 11.10 10.26
0.53 71.39 66.75 58.54
0.62 64.72 32.63 12.95
0.16 11.06 4.96 6.21
0.26 10.64 14.66 10.88
15.39
0.33 11.06 15.24 11.31
4.38
3.14
0.61 8.68 3.89 4.98
0.71 10.85 9.00 10.84
58
-------
Oklahoma E. Danforth 14.28 10.37 13.97 85 118 0.72 8.26 11.38 8.45
and N.
Midwest Rd
Oklahoma Sorghum 14.28 10.37 11.65 94 122 0.77 8.54 11.76 10.47
Mill Rd and
N. Air Depot
Rd
Oklahoma S. Hiwassee 14.28 14.66 19.57 55 70 0.79 4.90 4.77 3.58
Rd and SE
89TH St
Oklahoma E.Waterloo 14.28 10.37 11.65 172 210 0.82 14.71 20.25 18.03
& N.
Midwest
Blvd.
Oklahoma SE74THand 14.28 0.39 0.27 133 138 0.96 9.66 353.85 511.11
S. Choctaw
Oklahoma Reduced 14.28 10.37 13.97 78 112 0.7 7.84 10.80 8.02
size: E
Danforth
and N.
Midwest Rd
Pottowatamie New Hope 5.22 13.26 12.95 14 50 0.28 9.58 3.77 3.86
Road and
WalkerRd
(NS 331)
Pottowatamie River Rd & 5.22 5.82 5.61 127 267 0.48 51.15 45.88 47.59
Rock Creek
Rd
Minimum forall 4.90 2.51 3.14
Neighborhoods
Median for all 10.80 11.24 9.36
Neighborhoods
Maximumfor 73.99 353.85 511.11
all
Neighborhoods
(a) The estimated fraction of reported wells is determined by dividing the number of reported wells by
the numberof residences.
59
-------
Coexistence of Public and Private Water Supplies
In some cities, there appears to be significant co-existence of public and private water supplies.
The city of Enid, Oklahoma (Figure 2 and Figure 25) is covered by an extensive public water
system, yet has a large number of shallow (15 to 18m deep) wells. Local information and the
observed response to a drought emergency declared in 2012, indicate that these wells are used
for landscape maintenance, rather than primary domestic supply (Enid News 2012a and 2012b).
OWRB sawa dramatic increase in well logs reported in August 2012, and, as the drilling backlog
was reduced, the amount of drilling returned to prior levels (Figure 26). This occurred after the
City of Enid imposed a ban on outdoor watering from public water supplies (Enid News 2012b).
In Bethany, OK (Figure 4 and Figure 27) historical use of wells has continued despite later
provision of public water (Jacobsen and Reed, 1949). Personal preference, cost of connection,
and cost of monthly water are typical reasons given for continued use of private supplies.
Persistence of private well use is also known in the expanding cities of Edmond (Figure 28) and
Choctaw (Figure 29). Edmond reached its current extent by the end of 1976. Water mains in
Edmond have largely been confined to the historic center of the city and surrounds (Figure 30),
although the mains have been extended to the eastern part of the city recently. If the patterns
exhibited in Bethany also exist in Edmond, however, use of private wells can be expected to
persist into the future.
60
-------
Enid n
Public Water Supply System
2.5
10 Km
| Enid
Figure 25. The City of Enid's water supply does not cover its entire territory, and contains areas
with large numbers of private wells.
61
-------
Number of Wells Drilled in Garfield County, OK for 2011-2012
70
€0
50
^ 40
I
*5
E
E
January February March April May June Juty1 August September October November December
¦ 2011
30
20
22
39
22
Figure 26. Wells reported to the Oklahoma Water Resources Board in 2011 and 2012 for
Garfield County (Enid). Drilling increased in August 2012 after imposition of an outdoor
watering ban.
62
-------
Bethany
1.5
3 Km
OWRB Domestic Wells
Public Water Supply System
] Bethany
Figure 27. City of Bethany, Oklahoma, which is contained within the city limits of Oklahoma
City (Figure 4 ).
Edmond
• OWRB Domestic Wells
Public Water Supply System
i Edmond
Figure 28. The City of Edmond public water supply has not extended throughout its entire
territory.
63
-------
Choctaw n
• OWRB Domestic Wells
Public Water Supply System
| Choctaw
2.5
5 Kilometers
Figure 29. The City of Choctaw public water supply extends only through a portion of its
territory.
64
-------
&Cf{pigm Mil Rfl
Cotlee Creek fid
Covtfl
Dnnfcrth Rd
Danfortt
Tower & Pump
1-35 Water Towers
.& Pump Station
Ia*<*
//h*l7VVrV
Arcadia Lake
Water Tower
.Irca&i
Jake
I.', ino' j| Rd
Water Tanks
Wells
Water Mains
Figure 30. The location of water mains, public wells, and water tanks in Edmond, Oklahoma
(2009).
65
-------
Distances between Underground Storage Tanks and Private Domestic Wells
The number of tanks with wells located within the selected trial distances (15, 30, 76, 150, 230
300, and 1,610 m) was determined for both reported wells only and total estimated and
reported wells for the entire state (Figure 31). Because there are appreciable numbers of
estimated wells for each distance (1.2 to 2.7 times the number of reported wells, Table 8), the
number of tanks estimated from the 10,000-run Monte Carlo estimates are higher than those
from the known wells only. The medians from the Monte Carlo estimates are indicative of the
proximity of USTs and wells and indicate that, for example, there are 9 (0.3%) USTs with wells
within 15 m (50 ft), and 823 (27.1%) USTs with wells located within 300 m (1,000 ft) (Table 8).
The latter distance is commonly used as a boundary by state environmental agencies, so almost
30% of tanks have a well within the distance of concern. The implications of this result are that
well owners within 300 m of a UST are potentially impacted if a release occurs from the tank.
Tank owners and state agency officials should have a relatively large expectation that a leaking
tank has the potential to impact a residential well. Thus site investigations should include
searches for private domestic wells.
66
-------
1 -1
0.9
0.8
0.7
&06
c
0)
3 0.5
CT
-------
Table 8. Results of 10,000 Monte Carlo simulations of the distance between USTs and reported
and estimated private domestic well locations. The distances were binned into categories and
the counts represent the number of USTs with at least one well within the specified distance.
Distance
m (ft)
Estimates of Numbers of USTs with Private Domestic Wei Is Within Specified Distances
Reported
well
Locations
Only
Reported and Estimated well locations
10,000 MonteCarioSimulation Results
Min 25th Percentile Median 75th
Percentile
Max
Ratio
median to
reported
only
result
15 (50)
5
5
8
9
11
20
1.8
30 (100)
10
14
24
27
30
46
2.7
76 (250)
53
99
126
131
137
171
2.5
150 (500)
181
328
368
377
385
422
2.1
230 (750)
284
542
583
593
603
655
2.1
300 (1,000)
438
765
813
823
834
879
1.9
1,610 (5,280)
1,652
1,998
2,028
2,034
2,039
2,063
1.2
(a) Ratio ofthe result forthe median number of USTs determined bythe Monte Carlo method to the results
determined from reported well locations only.
68
-------
Conclusions
Inferences from the 1990 census provide a baseline for developing current estimates of areas of
high density of well use and their relationship with underground storage tank sites. Oklahoma
population has increased since 1990 and its reliance on private domestic wells has also
increased slightly. County, census tract and census block group estimates produced consistent
qualitative results for well usage, with more discrimination of spatial patterns and numerically
higher estimates as the spatial size decreased. The lack of statistical significance forthe county
and census tract estimates suggest that the preferred analysis size is the census blockgroup.
The estimates developed by the RW and NHU methods follow established patterns evident
from 1990 census results. The estimates are best viewed as indicators of areas with high or low
well usage, as the data used for each method have limitations and the area associated with
spatial data is no smallerthan the census block group.
The reported wells method provided the most statistically significant results and is preferable
for conceptual reasons: the use of reported wells allows for updating of the magnitude and
spatial distribution of private domestic well use. This allows for the methodology to adapt as
time progresses, as the reliance on the 1990 census inference can be lessened with time.
The well-dependent city and neighborhood estimates have the potential to be the most
accurate well densities, because they include a minimum of undeveloped land, but data
collection or baseline information (i.e., neighborhoods known to lack public water) needs for
these methods are prohibitive on a large scale. Absent a future nationwide survey on source of
water, the use of accurate and complete well completion and abandonment data from state
69
-------
agencies has the most potential for indirectly determining shifts in the use of wells. Over the
long term, as communities expand their public water systems, however, data will be needed to
assess shifts in well use that aren't reflected by well reporting only.
The estimates are useful in understanding not only the usage of private wells for water supply,
but also in relation to sources of contamination that may potentially impact well users. In the
case of underground storage tanks, almost 30% of tanks had at least one well within the
possible extent of contamination. By identifying these tanks, environmental agencies can add
this information to their protocols for prioritizing leak prevention activities, and where leaks
have occurred, prioritize possibly scarce cleanup resources. Municipalities can also use this
information for decisions on optimizing the locations of new gas stations to minimize the
potential impact to wells.
Application of the methods developed in this paper requires the availability of widespread data,
which in general are collected by state and national governments. For countries without these
data, these general characteristics of areas with high reliance on wells were found:
• Expanding cities do not invest in infrastructure, either because of lack of resources or by
waiting for development to support the infrastructure costs.
• Wells were used historically and later provision of public water does not induce supply
change.
• Rural areas, except where
o Infeasible due to poor quality water
70
-------
o Public water is supplied by rural water districts
o Reliance on individual supply of surface water or cisterns
• Lawn or garden watering is taken from wells and is not the primary domestic supply.
These general characteristics are key indicators of high reliance on private domestic wells and a
guide to local investigation. These methods in estimating well density can be used to assess the
potential impact to private well users from potential contaminant sources, including
underground storage tanks, confined animal feeding operations, industrial and hazardous
waste sites, and landfills. In addition to Oklahoma, private domestic wells are an important
component of overall water supply in the U.S., and understanding the geospatial relationships
between these wells and potential contaminant sources is a fundamental aspect of resource
management and human health protection.
71
-------
Acknowledgements
The U.S. Environmental Protection Agency through its Office of Research and Development
funded the research described herein. Weaver and Kremer were directly supported, Murray
through Interagency Agreement Number DW-89-92433001. This document has been reviewed
by the U.S. Environmental Protection Agency, Office of Research and Development, and
approved for publication. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use. We appreciate the assistance of Joel Hennessey and
Richard Lowrance, U.S. EPA, Chris Luttrel, Environmental Research Apprenticeship Program,
U.S. EPA, Ashley McElmurry and Kevin Blackwood, Student Services Contractors, U.S. EPA, Kay
Pinleyand Pat Bush, Senior Environmental Enrollees, U.S. EPA; Steve Beyeu, Oklahoma
Department of Libraries, Jack Keeley, U.S. EPA and OWRB retired, Kent Wilkins, OWRB.
72
-------
References
Aiken, C.S. 1987, Race as a factor in municipal underbounding, Annals of the Association of
American Geographers, 77(4) 564-579.
Ander, E.L., Watts, M.J., Smedley, P.L., Hamilton, E.M., Close, R., Crabbe, H., Fletcher, T., Rimell,
A., Studden. M., Leornardi, G. 2016, Variability in the chemistry of private drinking water
supplies and the impact of domestic treatment systems on water quality, Environmental
Geochemistry and Health, 38:1313-1332.
API. 1998. Characteristics of Dissolved Petroleum Hydrocarbon Plumes: Results from Four
Studies. American Petroleum Institute.
http://www.api.0rg/~/media/Files/EHS/Clean_Water/Bulletins/O8_Bull.pdf
ASTM. 2015. Standard Guide for Risk-Based Corrective Action Applied at Petroleum Release
Sites. E1739-95 (reapproved 2015). West Conshochocken: PA ASTM International.
Atherholt, T.B., J.B. Louis, J. Shevlin, K. Fell, and S. Krietzman. 2009. The New Jersey Private Well
Testing Act: An Overview. New Jersey Department of Environmental Protection, Division of
Science, Research and Technology. 8pp.
Bingham, R.H., and D.L. Bergman. 1980. Reconnaissance of the water resources of the Enid
quadrangle, north-central Oklahoma. United States Geologic Survey.
Bingham, R.H. and R.L. Moore. 2004. Reconnaissance of the water resources of the Oklahoma
City quadrangle, central Oklahoma. 4th printing. United States Geologic Survey.
CDWR. 2015. Well Completion Reports, California Department of Water Resources.
http://www.water.ca.gov/groundwater/wells/well completion reports.cfm, accessed 7-31-
2015.
Clinton, F.S. 1945. Tulsa's water resources, Chronicles of Oklahoma, 23 no. 2: 59-70.
CEHCID.2009. Policy Statement, Drinking Water from Private Wells and Risks to Children.
Committee on Environmental Health and Committee on Infectious Diseases. Pediatrics. 123 no.
6: 1599-1605.
Connor J.A., Kamath, R., Walker, K.L., and McHugh, T.E. 2015. Review of quantitative surveys of
the length and stability of MTBE, TBA, and benzene plumes in groundwater at UST sites.
Groundwater, 53(2), 195-206.
Danert, K., D. Adelike, and J. Gesti Canuto. 2014. Manually Drilled Boreholes: Providing Water
in Nigeria's Megacity of Lagos and Beyond. St. Gallen, Switzerland: Skat
Foundation/UNICEF/Rural Water Supply Network, http://www.rural-water-
supply.net/en/resources/details/618.
73
-------
DeSimone, L.A., P.A., Hamilton, and R.J. Gilliom. 2009. The quality of our nation's waters-
Quality of water from domestic wells in principal aquifers of the United States, 1991-2004—
Overview of major findings. Circular 1332. U.S. Geological Survey, 48 p.
Earle, R. C., J. T. Wilson, F. Kremer, and J. W. Weaver. 2011. GIS Analysis to Assess where
Shallow Ground Water Supplies in the United States Are Vulnerable to Contamination by
Releases of Motor Fuel from Underground Storage Tanks. EPA/600/R-11/108. Washington, DC:
U.S. Environmental Protection Agency.
Edmond. 2009. 50 Year Water Supply Plan-Final, CityofEdmond OK:
http://edmondok.com/DocumentCenter/Home/View/735, accessed 1-11-2017.
Edmond. 2017. Title 21-Subdivisions, Code of Ordinances, Edmond, OK, Minicode Library,
http://www.municode.com/library/ok/edmond, accessed 1-11-2017.
Eggers, F.J. 2009. Streamlining the American Housing Survey. Washington, DC: U.S.
Department of Housing and Urban Development, June, 23pp.
Enid News. 2012a. Water well, well, well. EnidNews.com. August 16.
http://www.enidnews.com/news/local_news/water-well-well-well/article_ac8513de-7a01-
57bl-9ff2-cbbe5790aac9.html, accessed 8-4-2015.
Enid News. 2012b. Effective Immediately: Phase 2 of Water Conservation Ordinance, Posted
8/1/2012. Assessed 9-11-2014.
Engineering News-Record. 1924. Spavinaw Water Supply Project for Tulsa, Oklahoma.
Engineering News-Record. 92, no. 19: 816-818.
Everett, D. 2017a. "Forest Park," The Encyclopedia of Oklahoma History and Culture,
www.okhistory.org (accessed June 19, 2017).
Everett, D. 2017b. "Nicoma Park," The Encyclopedia of Oklahoma History and Culture.
www.okhistory.org (accessed June 19, 2017).
Fox, M.A., Nachman, K.E., Anderson, B., Lam, J., Resnick, B. 2016 Meeting the public health
challenge of protecting private wells: Proceedings and recommendations from an expert panel
workshop, Science of the Total Environment, 554-555, 113-116.
FR. 2008. DOC, Bureau of the Census docket number 070126022-8027-02, Census blockgroup
program for the 2010 decennial census—Final criteria. Federal Register. 73 no .51: 13829-
13836.
Gates, P.W., and Swenson, R.W. 1968. History of Public Land Law Development. Washington,
D.C.: Public Land Law Review Commission. 828pp.
Hayes, Laura, and Horn, M.A., 2009, Methods for estimating withdrawal and return flow by
census block for 2005 and 2020 for New Hampshire: U.S. Geological Survey Open-File Report
2009-1168, 32 p., available at http://pubs.usgs.gov/of/2009/1168, assessed 6-12-2017.
74
-------
Hutson, S.S. 2007. Guidelines for Preparation of State Water-Use Estimates for 2005,
Techniques and Methods Book 4, Chapter El, Reston, Virginia: U.S. Geological Survey.
Jacobsen, C.L., and E.W. Reed. 1949. Ground-Water Supplies in the Oklahoma City Area,
Oklahoma. Mineral Report No. 20. Norman, Oklahoma: Oklahoma Geological Survey.
Jepson, W., Brown, H.L. 2014, 'If no gasoline, no water': Privatizing drinking water quality in
South Texas colonias, Environment and Planning A, 46, 1032-1048.
Johnson Jr, J.H., Parnell, A., Joyner, A.M., Christman, C.J., Marsh, B. 2004, The Review of Black
Political Economy, Spring, 89-107.
Kjellen, M. 2001, Water provisioning in Dar Es Salaam, Tanzania: The public-private interface,
Frontiers in Urban Water Management: Deadlock or Hope?, J.A. Tejada-Guibert and C.
Maksimovic, eds., UNESCO, Paris, Technical Documents in Hydrology, No. 45. 337-343.
Layden, L. 2013. Why Oklahoma City won't tap water from the aquifer under its own feet, State
Impact Oklahoma. August 13. https://stateimpact.npr.org/oklahoma/2013/10/31/why-
oklahoma-city-wont-tap-water-from-the-aquifer-under-its-own-feet/ (accessed 10-5-2015).
Levin, R.B., Epstein, P.R., Ford, T.E., Harrington, W., Olson, E., Reichard, E.G. 2002. U.S. Drinking
Water Challenges in the Twenty-First Century, Environmental Health Perspectives, 110(1), 43-
52.
Lundqvist, J. Appasamy, P. Nelliyat, P. 2003, Dimensions and approaches for Third World city
water security, Phil. Trans. R. Soc. Lond. B 358, 1985-1996.
MacDonald Gibson, J., DeFelice, N., Sebastian, D. Leker H. 2014, Racial disparities in access of
community water supply service in Wake County, North Carolina, Frontiers in Public Health
Services and Systems Research, 3(3) DOI: 10.13023/FPHSSR.0303.06.
Marcher, M.V., and R.H.Bingham, R.H. 1989. Reconnaissance of the Water Resources of the
Tulsa Quadrangle, Northeastern Oklahoma, 2nd printing. United States Geologic Survey.
Mashburn, S.L., D.W. Ryter, C.R. Neel, S.J. Smith, S.J., and J.S. Magers. 2013. Hydrogeology and
Simulation of Ground-Water Flow in the Central Oklahoma (Garber-Wellington) Aquifer,
Oklahoma, 1987 to 2009, and Simulation of Available Water in Storage, 2010-2059. U.S.
Geological Survey Scientific Investigations Report 2013-5219. 92 p.
Maupin, M.A., J.F. Kenny, S.S. Hutson. J.K Lovelace, N.L. Barber, and K.S. Linsey. 2014. Estimated
Use of Water in the United States in 2010. Circular 1405. Reston, Virginia: United States
Geological Survey. 56 p.
MDCH. 2013. Public Health Assessment, Final Release, Kalamazoo River/Enbridge Spill:
Evaluation of Crude Oil Release toTalmadge Creek and Kalamazoo River on Residential Drinking
Water Wells in Nearby Communities (Calhoun and Kalamazoo Counties, Michigan). Michigan
Department of Community Health. February 27.
75
-------
MDEQ. 2015. MDEQ Scanned Well Record Retrieval System. Michigan Department of
Environmental Quality, http://www.deq.state.mi.us/well-logs/, accessed 8-2-2015.
MEEA. 2015. Well Driller Program -SearchWell. Massachusetts Executive Office of Energy and
Environmental Affairs, http://www.mass.gov/eea/agencies/massdep/water/drinking/well-
driller-program-searchwell.html. accessed 7-31-2015.
Minnesota Population Center. 2011. National Historical Geographic Information System:
Version 2.0. Minneapolis, Minnesota: University of Minnesota.
NDER. 2015. Groundwater. Nebraska Department of Natural Resources.
http://www.dnr.ne.gov/gwr, accessed 7-31-2015.
NMED and U.S. EPA. 2005. Free Private Domestic Well Testing San Juan County—Animas River
Valley. New Mexico Environment Department and U.S. Environmental Protection Agency.
https://www.env.nm.gov/ ri ve r waters a fety/, accessed 8-10-2015.
OACR. 2015. Oklahoma Administrative Code and Register, Title 785 Oklahoma Water Resources
Board, Chapter 25 Well Driller and Pump Installer Licensing, http://www.oar.state.ok.us.
accessed 7-30-2015.
OHC. 2017. Forest Park, Oklahoma History Center,
http://www.okhistory.org/publications/enc/entry.php?entrv=FO023, assessed 6-19-2017.
OKC. 2014. City of Oklahoma City 2014 Consumer Confidence Report.
http://www.okc.gov/CCR 2014.pdf. accessed 7-28-2015.
Oklahoman. 1956. City is Larger by 1,012 Acres, May 23.
Oklahoman. 1957. City Annexation is Ruled Illegal, October30.
Oklahoman. 1959a. A City Must Grow, February 14.
Oklahoman. 1959b. City Spells Out Annexing Policy, October 9.
Oregon DEQ. 2015. Groundwater Quality Protection in Oregon, State of Oregon Department of
Environmental Quality.
http://www.deq.state.or.us/pubs/legislativepubs/2015/GoundwaterReport.pdf. (accessed 7-6-
2015).
OWRB. 1998. Rural Water Systems in Oklahoma, publication 138. Oklahoma Water Resources
Board, Planning and Management Division, Oklahoma City.
OWRB. 2012. 2012 Oklahoma Comprehensive Water Plan Executive Report, Oklahoma Water
Resources Board
OWRB. 2014. Coordinator's Corner, Well Driller's Log. Oklahoma Water Resources Board. 14 no.
2: 1.
76
-------
OWRB. 2015a. Well Drilling and Pump Installation. Oklahoma Water Resources Board.
Oklahoma City, http://www.owrb.ok.gov/supply/wd/drillers.php. accessed 1-11-2017.
OWRB. 2015b. Groundwater Data & Maps.
http://www.owrb.ok.gov/maps/PMG/owrbdata GW.html, accessed 6-01-2015.
OWRB. 2017. Rural Water Systems. Oklahoma Water Resources Board. Oklahoma City.
http://www.owrb.ok.gov/ma ps/ma ps2/rws/Delaware.pdf,
http://www.owrb.ok.gov/ma ps/ma ps2/rws/ca nadian.pdf,
http://www.owrb.ok.gov/ma ps/ma ps2/rws/cleveland.pdf,
http://www.owrb.ok.gov/maps/maps2/rws/oklahoma.pdf ,accessed 1-11-2017
Patel, A.P., Jacobs, E.J., Dudas, D.M., Briggs, P.J., Lichtman, C.J., Bain, E.B., Stevens, V.L.,
McCullough, M.L., Teras, L.R., Campbell, P.T., Caudet, M.M., Kirkland, E.G., Rittase, M.H., Joiner,
N., Diver, W.R., Hildebrand, J.S., Yaw, N.C., Gapsture, S.M. 2017. The American Cancer Society's
Cancer Prevention Study 3 (CPS-3): Recruitment, study design, and baseline characteristics.
Cancer, January, DOI: 10.1002/cncr.30561.
PGS. 2015. Pennsylvania groundwater information system (PaGWIS). Pennsylvania Geological
Survey. 4th ser. SQL database.
http://dcnr.state.pa.us/topogeo/groundwater/pagwis/records/index.htm. accessed 7-31-2015.
Ridpath, A., Taylor, E., Greenstreet, C., Martens, M., Wicke H., Martin, C. 2016, Description of
calls from private well owners to a national well water hotline, 2013, Science of the Total
Environment, 544, 601-605
Salmivaara, A., Porkka, M. Kummu, M. Keskinen, J.H.A. Guillaume, and 0. Varis. 2015. Exploring
the modifiable areal unit problem in spatial water asessments: A case of water shortage in
monsoon Asia. Water. 7: 898-917.
Schaider, L.A., Ackerman, J.M., Rudel, R.A. 2016, Septic systems as sources of organic
wastewater compounds in domestic drinking water wells in a shallow sand and gravel aquifer.
Science ofthe Total Environment, 547, 470-481.
U.S. DOC. 1990a. 1990 Census of Population and Housing Guide, Part A. Text. 1990 CPH-R-1A.
Washington, DC: U.S. Department of Commerce. Economics and Statistics Administration,
Bureau of the Census.
U.S. DOC. 1990b. 1990 Census of Population and Housing—History,
https://www.census.gov/history/pdf/1990proceduralhistory.pdf., accessed 5-18-2017.
U.S. DOC. 1993. 1990 Census of Housing Detailed Housing Characteristics Oklahoma. 1990 CH-
2-38. Washington, DC: U.S. Department of Commerce, Economics and Statistics Administration,
Bureau of the Census.
77
-------
U.S. DOC, 2009, History, 2000 Census of Population and Housing, Volume 1. PHC-R-VI.
Washington, DC: U.S. Department of Commerce, Economics and Statistics Administration, U.S.
Census Bureau.
U.S. EPA. 2002. Drinking Water from Household Wells, EPA816-K-02-003. Washington, D.C:
United States Environmental Protection Agency.
U.S. EPA. 2004. Understanding the Safe Drinking Water Act. EPA 816-F-04-030. Washington,
DC: U.S. Environmental Protection Agency, Office of Water.
U.S. EPA. 2016. Semiannual Report of UST Performance Measures End of Fiscal Year 2016
(October 1, 2015 - September 30, 2016). https://www.epa.gov/sites/production/files/2016-
ll/documents/ca-16-34.pdf. assessed 12-23-2016.
U.S. GAO 1998. Drinking Water: Some Households Rely on Untreated Water from Irrigation
Systems, U.S. General Accounting Office, Washington DC, GAO/RCED-98-244.
VanDerslice, J. 2011. Drinking water infrastructure and environmental disparities: Evidence and
methodological considerations, American Journal of Public Health 101 (si), S109-S114.
Wescoat, Jr, J.L., Headington, L., Eheobald, R. 2007. Water and poverty in the United States,
Geoforum (38), 801-804.
Wright, M.H. 1926. First Oklahoma oil was produced in 1859. Chronicles of Oklahoma. 4 no. 4:
322-328.
Zheng Y., and Ayotte, J.D. 2015, At the crossroads: Hazard assessment and reduction of health
risks from arsenic in private well waters of the northeastern United States and Atlantic Canada,
Science ofthe Total Environment, 505, 1237-1247.
78
-------
Appendix: Fluctuation of Private Domestic Well Use in Oklahoma
Counties.
USGS data were used to estimate the fraction of private domestic we 11 use irs each county of Oklahoma.
'k.
Cl
C
o
ts
ra
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
Adair - Cherokee Counties
•- -0.6
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
tj
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Choctaw - Ellis Counties
•= -0.6
<1>
<5 -0.8
-C
u
- Choctaw
-Cimarron
-Cleveland
-Coal
-Comanche
- Cotton
-Craig
-Creek
-Custer
Delaware
-Dewey
"Ellis
1985
1990
1995
2000
2005
2010
Year
Figure 33. Changes in fraction of private domestic well use from USGS water supply data (1985-
2010) for Choctaw to Ellis counties.
80
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
tj
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Garfield - Jefferson Counties
•= -0.6
<1>
<5 -0.8
-C
u
-Garfield
-Garvin
-Grady
- Grant
-Greer
- Harmon
Harper
- Haskell
Hughes
-Jackson
Jefferson
1985
1990
1995
2000
2005
2010
Year
Figure 34. Changes infraction of private domestic well use from USGS water supply data (1985-
2010) for Garfield to Jefferson counties.
81
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
"ts
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Johnston - McCurtain Counties
•- -0.6 -
0)
&©
<5 -0.8
JZ
u
Johnston
Kingfisher
; c5
Latimer
Le F ore
Lincoln
Love
McClain
McCurtain
1985
1990
1995
2000
2005
2010
Year
Figure 35. Changes in fraction of private domestic well use from USGS water supply data (1985-
2010) for Johnston to McCurtain counties.
82
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
"ts
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Mcintosh - Okmulgee Counties
•- -0.6 -
0)
&©
<5 -0.8
-C
u
- Mcintosh
- Major
- Marshall
- Mayes
- Murray
- Muskogee
Noble
- Nowata
Okfuskee
Oklahoma
Okmulgee
1985
1990
1995 2000
Year
2005
2010
Figure 36. Changes infraction of private domestic well use from USGS water supply data (1985-
2010) for Mcintosh to Okmulgee counties.
83
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
tj
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Osage - Seminole Counties
•= -0.6
<1>
<5 -0.8
-C
u
-Osage
-Ottawa
-Pawnee
-Payne
-Pittsburg
-Pontotoc
-Pottawatomie
-Pushmataha
Roger Mills
-Rogers
Seminole
1985
1990
1995
2000
2005
2010
Year
Figure 37. Changes infraction of private domestic well use from USGS water supply data (1985-
2010) for Osage to Seminole counties.
84
-------
0)
(/>
Z)
ID
u
+->
Q.
C
o
tj
1
0.8
0.6
0.4
0.2 -
0
-0.2
-0.4
Sequoyah - Woodward Counties
•= -0.6
<1>
<5 -0.8
-C
u
-Sequoyah
-Stephens
-Texas
-Tillman
-Tulsa
-Wagoner
- Washington
- Washita
Woods
- Woodward
1985
1990
1995
2000
2005
2010
Year
Figure 38, Changes infraction of private domestic well use from USGS water supply data (1985-
2010) for Sequoyah to Woodward counties.
85
-------
&EPA
United States
Environmental Protection
Agency
Office of Research arid Development (8101R)
1200 Pennsylvania Ave. NW
Washington, DC 20460
epa.gov/research
EPA/600/R-17/209
September 2017
------- |