-
State of Delaware
DELAWARE GEOLOGICAL SURVEY
John H. Talley, State Geologist
REPORT OF INVESTIGATIONS NO. 73
ANALYSIS AND SUMMARY OF WATER-TABLE MAPSFOR THE DELAWARE COASTAL
PLAIN
By
Matthew J. Martin1 and A. Scott Andres2
University of Delaware
Newark, Delaware
2008
1Delta Development Group, Inc.2Delaware Geological Survey
RESEARCH
DELAWARE
GEOLOGICALSURVEY
EXPL
ORA
TIO
N
SERVICE
Dry Conditions Normal Conditions Wet Conditions
-
State of Delaware
DELAWARE GEOLOGICAL SURVEY
John H. Talley, State Geologist
REPORT OF INVESTIGATIONS NO. 73
ANALYSIS AND SUMMARY OFWATER-TABLE MAPS
FOR THE DELAWARE COASTAL PLAIN
By
Matthew J. Martin1 and A. Scott Andres2
University of Delaware
Newark, Delaware
2008
1Delta Development Group, Inc.2Delaware Geological Survey
RESEARCH
DELAWARE
GEOLOGICALSURVEY
EXPL
ORA
TIO
N
SERVICE
-
Use of trade, product, or firm names in this report is for
descriptive pur-poses only and does not imply endorsement by the
Delaware GeologicalSurvey.
-
Page
ABSTRACT
..............................................................................................................................................................................1
INTRODUCTION
....................................................................................................................................................................1
Purpose and Scope
..............................................................................................................................................................2
Acknowledgments
..............................................................................................................................................................2
METHODS
................................................................................................................................................................................2
Data Compilation and Statistical Evaluation
......................................................................................................................3Depth-to-water
data......................................................................................................................................................3Surface-water
features..................................................................................................................................................3
Model Development
............................................................................................................................................................3Hydrologic
conditions
..................................................................................................................................................3Multiple
linear regression and the initial water table
..................................................................................................4
Estimation of Water-Table Elevation
..................................................................................................................................5
Potential improvements to Water-Table DEM from LIDAR-Derived
DEM......................................................................5
RESULTS AND DISCUSSION
..............................................................................................................................................6
Sussex County Water-Table Elevation
................................................................................................................................7
Kent County Water-Table Elevation
..................................................................................................................................8
New Castle County Water-Table Elevation
........................................................................................................................8
Depth to Water
....................................................................................................................................................................8
Comparison of Existing DLG Hydrography to LIDAR-Derived DEM
............................................................................9
CONCLUSIONS
......................................................................................................................................................................9
REFERENCES
CITED..........................................................................................................................................................10
Page
1. Depth to water under normal conditions for the Inland Bays
watershed in Sussex County, Delaware.
...........................2
2. Hydrograph for well Qe44-01 showing monthly depth to water
below the land surface.
................................................4
3. Illustration of the initial water table including graphical
representation of water-table terms and vertical datums.
........5
(Figures 4A through 6C are located on Plate
1)........................................................................................................In
Pocket
4A. Water-table elevation under dry conditions for Sussex
County, Delaware.
4B. Water-table elevation under normal conditions for Sussex
County, Delaware.
4C. Water-table elevation under wet conditions for Sussex
County, Delaware.
5A Water-table elevation under dry conditions for Kent County,
Delaware.
5B. Water-table elevation under normal conditions for Kent
County, Delaware.
5C. Water-table elevation under wet conditions for Kent County,
Delaware.
6A. Water-table elevation under dry conditions for New Castle
County, Delaware
6B. Water-table elevation under normal conditions for New Castle
County, Delaware.
6C. Water-table elevation under wet conditions for New Castle
County, Delaware.
TABLE OF CONTENTS
ILLUSTRATIONS
-
Page
(Figures 7A through 9C are located on Plate 2 )
........................................................................................................In
Pocket
7A. Depth to water under dry conditions for Sussex County,
Delaware.
7B. Depth to water under normal conditions for Sussex County,
Delaware.
7C. Depth to water under wet conditions for Sussex County,
Delaware.
8A. Depth to water under dry conditions for Kent County,
Delaware.
8B. Depth to water under normal conditions for Kent County,
Delaware.
8C. Depth to water under wet conditions for Kent County,
Delaware.
9A. Depth to water under dry conditions for New Castle County,
Delaware.
9B. Depth to water under normal conditions for New Castle
County, Delaware.
9C. Depth to water under wet conditions for New Castle County,
Delaware.
10. Illustration showing stream segment elevation artifacts
caused by misalignment of 1:24,000 hydrography DLG with
LIDAR DEM and road crossings.
....................................................................................................................................10
TABLES
Page
Table 1. Long-period observation wells used for each county
..................................................................................................4
Table 2. Grid dimensions and numbers of ground-water points used
to estimate water-table elevation and depthto water table.
...............................................................................................................................................................6
Table 3. Coefficients for MLR and LR used to calculate the
water-table elevation for each county and/orcounty
section...............................................................................................................................................................7
Table 4. Statistical comparison of residuals from water-table
elevation estimations for normal conditions.
..........................8
Table 5. Comparisons of depth to water and percentage of land
area.......................................................................................9
ILLUSTRATIONS CONTINUED
-
INTRODUCTION
The water table is defined as the surface on which thewater
pressure in the pores of a porous medium is exactlyatmospheric
(Freeze and Cherry, 1979). In practice, the posi-tion of the water
table is measured in wells constructed withopenings along their
lengths and penetrating just deep enoughto encounter standing
water. Water located at or beneath thewater table is ground water.
Given the climate and relativelypermeable subsurface materials in
Delaware, the water tableoften occurs at depths less than 10 ft
below land surface(Andres and Martin, 2005; Martin and Andres,
2005a, b, c).
The first efforts to map the water-table for the state
ofDelaware were undertaken in the 1950s and were a coopera-tive
effort between the United States Geological Survey(USGS), the
Delaware Division of Highways, and theDelaware Geological Survey
(DGS). Maps from this projectwere published as paper maps in the
Hydrologic Atlas seriesat a scale of 1:24,000 and depicted the
water table with con-tour lines at a 10-ft interval. These
water-table maps havebeen widely used by both the public and
private sectors(Andres and Martin, 2005). Despite the usefulness of
thesepaper maps, more data are now available and recent advancesin
computer technology and the expanding use of GeographicInformation
Systems (GIS) have made it necessary to updatewater-table maps into
a digital format.
The configuration of the water table is one of the majorfactors
that controls regional ground-water flow patterns(Freeze and
Witherspoon, 1967). Ground water moves slow-ly underground in the
down-gradient direction and eventual-ly discharges into streams,
lakes, and oceans (Perlman, 2005).Because ground water is such an
integral part of the watercycle, planners and developers often need
to have a strategicplan when dealing with water resources. Excess
pumping ofwells over extended periods of time can result in
lowering the
water table leading to an increase in the cost of pumping
fromgreater depths, depletion of the amount of water available
forimportant wetland habitats, and salt-water intrusion
intodomestic water supplies (Dunne and Leopold, 1998). In
addi-tion, the practice of well drilling to extract ground water
isdependent upon an understanding of the depth to the watertable.
Wells must be finished below the water table, and thedepth of the
water table determines the final specifications ofthe well.
Obtaining an accurate representation of the water table isalso
crucial to the success of many hydrologic modelingefforts (Williams
and Williamson, 1989). Estimated water-table elevation can be used
to specify heads in the surficialaquifer for a ground-water flow
model, to estimate depths toareas of potential ground-water
contamination, or to simulaterecharge and discharge rates of the
surficial aquifer(Sepulveda, 2003).
In many areas throughout Delaware, the depth to thewater table
has a direct effect on how people utilize the land.For instance,
based on depth to the water table, it can bedetermined whether or
not a site is suitable for a standard sub-surface
wastewater-disposal system. Water-table depth is akey facet in many
engineering, hydrogeologic, environmentalmanagement, and regulatory
decisions. Depth to water is animportant factor in risk
assessments, site assessments, evalu-ation of permit compliance
data, registration of pesticides anddetermining acceptable
application rates. Shallow depth toground water has been the
principal motive for constructingthe extensive ditch networks that
can be found in manywatersheds in Delaware. In many areas, the
water table is alsothe top of the aquifer that provides water for
potable, agricul-tural, commercial, and industrial uses. The
thickness of thisaquifer is one factor that controls the amount of
water that isavailable to wells (Andres and Martin, 2005).
Delaware Geological Survey • Report of Investigations No. 73
1
ANALYSIS AND SUMMARY OFWATER-TABLE MAPS
FOR THE DELAWARE COASTAL PLAIN
ABSTRACT
A multiple linear regression method was used to estimate
water-table elevations under dry, normal, and wet conditionsfor the
Coastal Plain of Delaware. The variables used in the regression are
elevation of an initial water table and depth to theinitial water
table from land surface. The initial water table is computed from a
local polynomial regression of elevations ofsurface-water features.
Correlation coefficients from the multiple linear regression
estimation account for more than 90 per-cent of the variability
observed in ground-water level data. The estimated water table is
presented in raster format as GIS-ready grids with 30-m horizontal
(~98 ft) and 0.305-m (1 ft) vertical resolutions.
Water-table elevation and depth are key facets in many
engineering, hydrogeologic, and environmental management
andregulatory decisions. Depth to water is an important factor in
risk assessments, site assessments, evaluation of permit
com-pliance data, registration of pesticides, and determining
acceptable pesticide application rates. Water-table elevations are
usedto compute ground-water flow directions and, along with
information about aquifer properties (e.g., hydraulic
conductivityand porosity), are used to compute ground-water flow
velocities. Therefore, obtaining an accurate representation of the
watertable is also crucial to the success of many hydrologic
modeling efforts.
Water-table elevations can also be estimated from simple linear
regression on elevations of either land surface or initialwater
table. The goodness-of-fits of elevations estimated from these
surfaces are similar to that of multiple linear regression.Visual
analysis of the distributions of the differences between observed
and estimated water elevations (residuals) shows thatthe multiple
linear regression-derived surfaces better fit observations than do
surfaces estimated by simple linear regression.
-
Depth to the water table is a prevailing factor in deter-mining
the ecological function of a landscape. For example,many wetlands
are found where the water table is at or nearland surface for
portions of the year. The duration of stand-ing water in large part
prescribes the plant and animal com-munities that can live at that
site. Under fair or “normal”weather conditions, the surfaces of
Coastal Plain streams andponds represent the intersection of the
water table with landsurface (Winter, 1999; Andres and Martin,
2005).
Purpose and Scope
The purpose of this report is to provide both a briefreview of
the pilot project, the Inland Bays Watershed Water-Table Mapping
Project, and a detailed summary and analysisof the results of
mapping the water table for the DelawareCoastal Plain. The goals of
the Delaware Coastal Plain pro-ject were to use the methodology and
procedures establishedduring the pilot project to map the water
table for the remain-der of Sussex County, as well as Kent County
and NewCastle County.
Appropriate methodologies and procedures for calculat-ing the
water table for the Coastal Plain of Delaware wereestablished in
the Inland Bays Watershed Water-TableMapping Project. Water-table
elevation maps were producedfor dry, normal, and wet conditions
using a variety of esti-mation methods, making qualitative
comparisons betweenthe different methods and pre-existing
water-table maps, anddetermining which of the estimation methods
could be usedto map the Coastal Plain of Delaware in a
cost-effective andtimely manner. One crucial constraint in choosing
a suitableestimation method for mapping the entire state was that
ithad to rely on existing data because available funding wasnot
sufficient to construct new wells or to support collectionof
additional water-level measurements.
The Inland Bays watershed (Fig. 1) was selected as thepilot
project by the Delaware Geological Survey and theDelaware
Department of Natural Resources andEnvironmental Control (DNREC)
Water Supply Section(WSS) because of the readily available
pre-existing water-level data from previous hydrologic studies
conducted in thisregion. In addition, the watershed was identified
as a highpriority area for a number of regulatory and
environmentalrestoration efforts that can use the resultant
information.
After creating and analyzing water-table maps for dry,normal,
and wet conditions using various statistical estima-tors, it was
determined that the method which produced themost desirable results
was an algorithm based on a multiplelinear regression (MLR)
equation to estimate the water table.Water-table elevation and
depth-to-water maps for theremainder of Sussex County, Kent County,
and New CastleCounty were then produced using this algorithm
(Martin andAndres, 2005a, b, c).
The map products created by this work are being uti-lized to
support various public environmental programs andprivate site
reviews that require hydrologic assessment.These map products will
be an important tool in the assess-ment process; however, they
depict estimates of water-tableelevation and are, therefore, not
intended to supplant on-sitedata collection efforts. The
water-table maps will not
be published paper maps; however, they are publishedas GIS-ready
products and are available fordownload from the Delaware Geological
Survey’s website(http://www.udel.edu/dgs).
Acknowledgments
This work was funded by the DNREC through grantsfrom the U.S.
Environmental Protection Agency Ground-Water Protection and Source
Water Protection Programs.Evan M. Costas, Cheryl A. Duffy, Bailey
L. Dugan, Scott V.Lynch, and Tamika K. Odrick assisted with the
work. RonaldGraeber, John Barndt, Blair Venables, Joshua Kasper,
andScott Strohmeier of the DNREC are thanked for makingmonitoring
data from numerous wastewater disposalfacilities available. Stacey
Chirnside of the University ofDelaware Department of Bioresources
Engineering isthanked for providing access to ground-water data
fromseveral research projects. Thomas E. McKenna (DGS), JohnT.
Barndt (DNREC), and Geoffrey C. Bohling (KansasGeological Survey)
critically reviewed the manuscript.
METHODS
This work has three primary components: data compila-tion,
statistical evaluation and model development, andestimation of the
water-table elevation. Water level and welldata were extracted from
a DGS, Oracle-based database.Spatial data management and processing
were done withdesktop and workstation components of ArcGIS v9.0
(ESRI,2003), ArcGIS v9.1 (ESRI, 2005) and Surfer v8
(GoldenSoftware, 2002) software. Horizontal coordinates of all
dataare in meters, using the Universal Transverse Mercator
2 Delaware Geological Survey • Report of Investigations No.
73
Figure 1. Depth to water under normal conditions for the
InlandBays watershed in Sussex County, Delaware.
http://www.udel.edu/dgs
-
Delaware Geological Survey • Report of Investigations No. 73
3
projection and North American Datum of 1983 (NAD83).Elevations
are reported relative to the NorthAmerican VerticalDatum of 1988
(NAVD88). Statistics were computed withfunctions and procedures
contained in Oracle, ArcGIS, andMicrosoft Excel.
Data Compilation and Statistical Evaluation
Land-surface elevation (LSE) data that were usedthroughout the
estimation process are from a 30-m digitalelevation model (DEM)
created by John Mackenzie of theUniversity of Delaware’s Spatial
Analysis Lab, and fromUSGS 1:24,000-scale topographic maps.
Depth-to-water data
Depth-to-water (DTW) and well data were acquired fromthe files
and electronic databases of the DGS, DNREC,USGS, and the University
of Delaware Department ofBioresources Engineering. DNREC data were
extracted fromthe files and electronic databases of the Site
Investigation andRestoration Branch, Water Supply Section,
Ground-WaterDischarges Section, Spray Irrigation Program, and
TankManagement Branch of the DNREC. Additional monitoring-well and
water-level data were obtained for New CastleCounty from the USGS
(1996). The data gathered were of twotypes: one type consisting of
time-series depth-to-water mea-surements from monitoring wells, the
other type from singlestatic depth-to-water measurements reported
by well drillerson well completion reports. Prior to analysis, all
depth-to-water data were converted to depth relative to
ground-surfacedatum. Depth-to-water data from monitoring wells
typicallyare reported to the nearest 0.01 ft; data from well
completionreports usually are reported to the nearest foot. The
accuracyof measurements from individual wells was evaluated
bycomparison to measurements in nearby wells and by convert-ing
depth-to-water to water-table elevation. Because the ele-vation of
the water table is above 0 ft under static conditions,water
elevations less than 0 ft and greater than LSE are gener-ally
considered to be non-representative of local conditionsor
inaccurate and were removed from the dataset.
Depth-to-water and well data were managed and ana-lyzed in a
relational database format. Structured QueryLanguage (SQL) queries
were assembled to create tables inOracle that contained well
locations, land surface elevations,water-level observations, and
computed statistics (mean, min-imum, maximum, standard deviation,
and number of observa-tions) of observations made in the months and
years of nor-mal, dry, and wet conditions. Each of Delaware’s
counties(Sussex, Kent and New Castle) had its own
well-informationdataset.
Surface-water features
In the Coastal Plain of Delaware, topographic relief issmall and
aquifers consist of unconsolidated sediments. In thistype of
hydrogeologic setting, the surfaces of streams, ponds,and swamps
can be assumed to be the water table under fairweather conditions
(Freeze and Cherry, 1979). This assump-tion also was used in the
production of the 1960s hydrologicatlases and other regional
evaluations of the water table inDelaware (Johnston, 1973,
1976).
Streams, ponds, and swamps have a direct correspon-dence to the
water table; therefore, acquiring the elevationsand maintaining the
spatial configuration of these surface-water features is an
important part of modeling the water table.Locations of
surface-water features are from the 1992 USGS1:24,000 hydrography
digital line graph (DLG) datasetobtained from DataMIL
(datamil.delaware.gov). These dataare stored in an ArcGIS personal
geodatabase. DLG hydro-graphic data were converted into 30-m
gridded raster datasetswith each grid node set to a value of zero.
The grid geometrieswere set to correspond to the 30-m land surface
DEM. Two30-m grids were created: one for shorelines and fringing
tidalmarshes, and one for fresh-water streams. Two 90-m gridswere
created from DLG hydrographic polygons: one for fresh-water ponds
and swamps, and one for tidal marshes and theocean (Andres and
Martin, 2005) For the grids representingfresh-water features, the
elevation of each grid node was setequal to the elevation from the
corresponding land-surfaceDEM. The raster calculator was also used
to set elevations ofnodes representing salt-water marshes to 1 ft,
and to set theelevations of nodes representing the shorelines to 0
ft. Thesegrids were converted to point datasets and merged
(Andresand Martin, 2005).
Surface-water feature point data were modified toreduce noise in
the dataset. Areas of steep land slope nearstreams produced some
data points with anomalous eleva-tion values. The number of these
anomalous values was min-imized by removing points occurring more
than 15 m fromsurface-water features.
Model Development
Hydrologic conditions
For this work, dry, normal, and wet conditions weredetermined
from time-series measurements of depth-to-water.Depth-to-water
measurements have been collected at approx-imately monthly
intervals for more than 30 years in a numberof observation wells
located throughout the state. A set ofobservation wells was chosen
for each county to define thehydrologic conditions for that
particular area (Table 1). Amulti-step procedure was used to
identify dry, normal, and wetconditions from observations made in
those wells.
Ideally, comparison of long-term water-level observa-tions made
at different locations should use data measured onthe same days and
at regular intervals (e.g., monthly measure-ments should be made on
the same day of the month in thewells being compared). To correct
for the fact that this did notoccur, the observed water levels were
used to interpolate waterlevels on the 15τη of each month for each
month that a waterlevel was measured. For some months when water
levels werenot measured, levels were interpolated from
measurementsmade within 25 days of the 15τη day of the unmeasured
month.Interpolation was done by on-screen digitizing of
hydro-graphs. No estimates were made if water levels were
notobserved within 25 days of the 15τη day of the
unmeasuredmonth
Statistical measures of the water-level observations
werecomputed and the corresponding dates that those water
levelsoccurred were identified. From these statistics (Table 1),
dry,
datamil.delaware.gov
-
4 Delaware Geological Survey • Report of Investigations No.
73
normal, and wet hydrologic conditions were defined.Normal
conditions were defined as the months with DTWlevels falling
between the 40τη and 60τη percentiles (Fig. 2) inthe wells that
were compared. Dry conditions (lowest waterlevels) were defined as
months where the DTW levels fellbetween the 75τη and 95τη
percentiles (Fig. 2) and wet condi-tions (highest water levels)
were defined as months wherethe DTW levels fell between the 5τη and
25τη percentiles(Fig. 2). These percentile values were chosen as a
balancebetween having an adequate number of dates to identifywells
for estimating the water table and minimizing thedifferences in
water levels within a particular groupcompared to differences
between dry, normal, and wetgroups. Extreme values (< 5τη and
> 95τη percentiles) wereexcluded from the analysis.
Multiple linear regression and the initial water table
Sepulveda (2003) reported that estimation of the water-table
elevation by linear regression (LR) on LSE could beimproved by a
multiple linear regression (MLR) procedure
that used a “minimum water table” along with land
surfaceelevation to estimate the water-table elevation. A
keyassumption used in many water-table estimation projects isthat
streams, ponds, and swamps represent the intersection ofthe water
table with land surface (Winter, 1999) (Fig. 3).Thus, the
land-surface DEM was used to assign elevations tothe surface-water
features.
The minimum water table was then estimated by com-puting a grid
from the elevations of surface-water features(Sepulveda, 2003). In
this process, the minimum and maxi-mum elevations of the minimum
water table are 0 ft (e.g.,tidal-water elevation) and land-surface
elevation, respective-ly. For clarity, Sepulveda’s term “minimum
water table” isreplaced by “initial water table” (INITWT) for
application toDelaware. For Sussex County and Kent County,
estimates ofthe INITWT were created by a 5th-order local
polynomialregression method.A kriging algorithm was used to
calculatethe INITWT for New Castle County. Three separate
initialwater-table grids were created for Sussex, three for
Kent,and two for New Castle Counties.
The second variable in the MLR equation is a depth tothe initial
water table, which was calculated by subtractingthe initial
water-table elevation from the land surface eleva-tion DEM. Thus,
the general form of the multiple linearregression equation is:
Est WTi = β1 * INITWTi + β2 * (LSEi-INITWTi) (1)
where:
Est WTi = estimated water-table elevation at point iβ1 =
regression coefficient 1INITWTi = initial water-table at point iβ2
= regression coefficient 2LSEi = land-surface elevation at point
i(LSEi-INITWTi) = depth to the initial water table atpoint i
The regression coefficients, β1 and β2, were calculatedfrom the
depth-to-water and well datasets. INITWT anddepth to INITWT were
converted into point feature class for-mat in ArcGIS and then
exported into Microsoft Excel forthe regression analysis. The dry,
normal, and wet welldatasets for each county produce their own
unique sets ofregression coefficients. The effectiveness of MLR was
com-pared to simple LRs on LSE and the INITWT by comparing
Table 1. Long-period observation wells used for each county.
Values are depth to water measured in feet below land surface.
Figure 2. Hydrograph for well Qe44-01 showing monthly depth
towater (DTW) below the land surface. Data points are estimated
forthe 15th of each month. Statistics derived from estimated data.
Linesrepresent 25τη, 40τη, 60τη, and 75τη percentiles of data
distribution.
-
Delaware Geological Survey • Report of Investigations No. 73
5
statistical measures of observed and predicted WTEs.
Estimation of Water-Table Elevation
The elevation of the water table is the distance of
thewater-table surface from a vertical datum, in this case
theNAVD88, which is approximately sea level. The two variables(the
initial water table and the depth to the initial water table)and
the two coefficients (coefficient β1 and coefficient β2)were
computed and applied to the multiple linear regressionequation. The
resultant water-table elevation maps for eachcounty are continuous
surfaces; however, observations of thesurface exist at irregularly
spaced locations. The multiple lin-ear regression equation
interpolates water-table elevationsbetween these surface
observations to produce the continuoussurface. The water-table
elevation grids are in the form of GISgrids with 30-m horizontal
and 1-ft vertical resolution.
The water-table grids were completed as a series ofsub-grids
that were subsequently merged into single county-wide grids. For
example, a water-table DEM for normalconditions for eastern Sussex
County was merged with awater-table DEM for normal conditions for
western SussexCounty.
There are several different methods that can be used tomosaic
raster datasets, and because the grids overlapped insome areas, a
weighted average algorithm was used followedby a filtering step.
The weight-based algorithm is dependenton the distance from the
pixel to the edge within the over-lapping area. As an example of a
filter, the different sectionsof Sussex County were separated based
upon hydrography,so the merged Sussex County WTE grids were put
through a3x3 Gaussian low-pass filter that removes higher
frequencyvariations in grid values and smooths the artifacts along
theseams.
Merging the county DEMs into a statewide grid wasexplored;
however, the large size of the resultant grid severe-ly taxes the
performance of even high-end PC workstations.In addition, the merge
process resulted in unwanted grid arti-facts because the DEMs are
regular grids and the county
boundaries are in part formed by meandering streams.
Theseartifacts are a problem because when the statewide grid iscut
into county grids, there are no-value nodes located in theinterior
of the resultant county grids. To work around theseissues, the
grids are completed by county and include anoverlap of 200 m into
the adjacent county. If a user needsa simple map covering more than
one county for displaypurposes, then the county grids are adequate.
Any analyticalwork (i.e., slope and aspect, hillshade, etc.) that
requiresa seamless grid across county boundaries will require
theuser to merge the grids and develop the appropriate smooth-ing
procedures most suited to the location and scale
ofinvestigation.
Potential Improvements to Water-Table DEM from
LIDAR-Derived DEM
DEMs of land surface produced from aircraft-borneLIDAR (Light
Detection And Ranging) data offer the poten-tial for increasing the
horizontal and vertical resolutions ofthe elevations of
surface-water features and water-tableDEMs. Compared to the 30-m
horizontal and 1-ft verticalresolution DEMs derived from 1:24,000
DLG data, experi-mental DEMs produced from LIDAR data collected by
air-craft-borne sensors in the past few years typically result
inDEMs with 2-m horizontal and approximately 0.328-ftvertical
resolutions.
A simple experiment was conducted with experimentalLIDAR-derived
DEMs produced by the USGS for two ran-domly selected small
watersheds located in Sussex County. Inthe same way that the
elevations of surface-water featureswere determined from existing
DLG hydrography data(USGS, 1992) and 30-m DEMs (Mackenzie, 1999),
LIDAR-derived DEMs and the USGS (1992) hydrography DLG datawere
used to determine elevations of surface-water features.The
resulting point elevation data were visually compared tothe
LIDAR-derived DEM to determine if the DLG-derivedpoints were
aligned with the local elevation minima on the
Figure 3. Illustration showing the initial water table including
graphical representation of water-table terms and vertical
datums.Illustration modified from Sepulveda (2003).
baxterTypewritten Text
baxterText BoxLIDAR-derived DEM.
-
RESULTS AND DISCUSSION
Water-table elevations were estimated as a series ofoverlapping
grids (Table 2). For each grid area, a set of long-term observation
wells was used to define dry, normal, andwet periods (Table 1).
This enabled collection of water-levelobservations made in
additional wells during those periods.Separate INITWT and MLR
estimations (Tables 3 and 4)were run for each grid area.
Two interesting observations can be made regarding
thecoefficients of the MLR equations and of the coefficient forthe
regression on INITWT. First, β1 values (weighting factorfor
INITWT), except for one grid, are slightly less than 1.
InSepulveda’s analysis of the water-table elevation in
Florida(Sepulveda, 2003), the regression coefficients of the
initialwater table were all ≥ 1 in all but one of his study
groups.This indicates that the data (i.e., elevations of
surface-waterfeatures) and methods (polynomial surface fit) used to
esti-mate the INITWT in Delaware slightly overestimateobserved
water-table elevation (WTE) rather than underesti-mate WTE as in
Florida. Spatially, the INITWT elevations,regardless of hydrologic
condition, are less than the estimat-ed WTEs in low-lying areas
along the shorelines and aroundthe streams and bays. In addition,
INITWT elevations foreastern Sussex County are less than the
estimated WTEs inall areas under wet conditions. These conditions
also can bepartially due to artifacts from estimating elevations of
sur-face-water features from DLGs and DEMs. Second, themagnitude of
β2 (weighting factor for depth to INITWT) islargest in New Castle
County, and lowest in Sussex County.This is likely due to deeper
incision of streams and greatertopographic relief in New Castle and
Kent counties, andresultant greater depth to the INITWT.
Simple linear regressions were performed with the nor-mal
condition water-level data on both LSE and the INITWTfor
statistical comparison to the multiple linear regressionmethod
(Tables 3 and 4). The coefficients of determination(R2), which show
the proportions of sample variancesaccounted for by the regression
equations, are very similar
between the MLR and both LR models. Root mean squareerror (RMS),
a statistical measure of the magnitude of thetotal estimation
error, were also calculated for the MLR andboth LR methods. The RMS
value for the MLR method issmaller than the RMS values produced
from the LSE LR andthe INITWT LR analyses. These statistical
measures indicatethat the MLR method is a slightly more accurate
predictor ofwater-table elevation than a simple LR on LSE or
INITWT.It is important to note that a statewide analysis by simple
LRon LSE fairly accurately predicts the normal WTE, and thatWTE is
approximately 80 percent of LSE (Table 4).
A second way of assessing the goodness of fit betweenthe
different estimation methods is to evaluate the
individualresiduals, or observed minus predicted WTE values. In
gen-eral, differences between the 2νδ and 3ρδ quartiles and 5τη
and95τη percentiles of residuals from the MLR method are lessthan
similar differences from both of the LR methods. Thisindicates that
the MLR method better estimates 90 percent ofwater-level
observations than do the LR methods. However,the MLR method did
produce a higher range (maximum-minimum) of residuals in Kent
County and New CastleCounty than did the LSE LR; this is likely a
result of increas-ing LSE values and ranges in these two
counties.
On closer inspection, many of the largest residuals arelocated
near areas of steepest topography, and some arelocated near bodies
of tidal surface water. In the cases ofsteep topography, errors in
coordinates of measurement loca-tion can result in significant
changes in LSE and observedWTE. In cases of measurements made in
northern NewCastle County, where topographic contours have 10 ft
inter-vals, an error in horizontal position of just 60 m can
easilyresult in a change in LSE and of the observed WTE of 10 to20
ft. Larger residuals associated with measurement pointslocated near
bodies of tidal surface water indicate that thosemeasurement points
may not be indicative of WTE of thewater-table aquifer. It is also
possible that some of the largeresiduals are artifacts from
estimating the elevations of sur-face-water features from DLGs and
DEMs.
Table 2. Grid dimensions and numbers of ground-water points used
to estimate water-table elevation and depth to water table.
Surfacewater points were used to estimate the initial water table.
Ground-water points were used in the multiple linear regression
estimationprocedure.
6 Delaware Geological Survey • Report of Investigations No.
73
-
Sussex County Water-Table Elevation
Creating the water-table elevation maps for SussexCounty,
Delaware, was a multi-step process that involveddividing the county
into three separate geographic sections(east, west, and north). In
large part, the geographic sectionsof Sussex County were delineated
based on watershedboundaries and DLG lines representing hydrography
(e.g.,streams) in the area. Each geographic region of SussexCounty
has its own unique well data set, and thus its ownunique set of
regression coefficients for dry, normal, and wetconditions.
Water-table elevation grids for each hydrologiccondition were
created for eastern, western, and northernSussex County and were
then merged to create a unified,county-wide water-table elevation
grid for dry (Fig. 4A),normal (Fig. 4B), and wet (Fig. 4C)
conditions (Martin andAndres, 2005a).
Hydrologic conditions for the area identified as easternSussex
County were determined by comparing the long-termwater-level
measurements in monitoring wells Ng11-01 andQe44-01. The
observation well dataset used to compute theregression coefficients
in this area included water-levelsfrom 1,320 wells (Table 2).
Locations of these measurementsare not spread evenly across the
study area.
Long-term water-level measurements from monitoringwells Nc45-01
and Qe44-01 were compared in order todefine the time periods for
dry, normal, and wet conditionsfor western Sussex County. The
water-level observationdatasets for this area included data from
728 water-levelobservation points (Table 2) that were unevenly
distributedacross the study area.
Dry, normal, and wet conditions for the area defined asnorthern
Sussex County were determined from the compari-son of long-term
water-level measurements in monitoringwells Nc45-01 and Ng11-01. A
total of 1,114 wells wasincluded in the water-level observation
data set (Table 2).
In all three areas, MLR equations and weighing factors(Table 3)
were used to calculate water-table elevation grids.The final
water-table elevation maps for Sussex County hadelevations ranging
from 0 to 66 ft. Because land-surface ele-vation was a component of
the multiple linear regression, thewater-table elevation maps
resemble the land surface eleva-tion DEM maps. Water-table
elevations, in general, increasewith increasing land surface
elevations. This is true for Kentand New Castle counties as
well.
Delaware Geological Survey • Report of Investigations No. 73
7
Table 3. Coefficients for MLR and LR used to calculate the
water-table elevation for each county and/or county section. The
first regres-sion coefficient (β1) is multiplied by the initial
water table (INITWT). The second regression coefficient (β2) is
multiplied by the regressor(LSE-MINWT).
-
Table 4. Statistical comparison of residuals from water-table
elevation (WTE) estimation residuals for normal conditions.
Residuals werecalculated using multiple linear regression (MLR),
linear regression on land surface elevation (LSE), and linear
regression on the initialwater table (INITWT).
Kent County Water-Table Elevation
Initially, the process for calculating the water-table
ele-vation maps for Kent County was going to be consistent withthat
for Sussex County. Long-term water-level measure-ments from
monitoring wells Mc51-01, Md22-01, and Jd42-03 were compared in
order to define the time periods for dry,normal, and wet hydrologic
conditions. Kent County wasdivided into three geographical sections
(south, central, andnorth) based on watershed boundaries and
hydrography.However, the water-level observation point data sets
for cen-tral and northern Kent County were not adequate
calculateaccurate water-table elevation grids. The regression
analysisproduced poor R2 values for each condition due to an
insuf-ficient number of water-level observation points.
Therefore,the water-level points were incorporated together to form
asingle data set and the water table was estimated for theentire
county as a whole entity. As a result, the initial water-table
grids for southern, central, and northern Kent Countywere also
merged together to form a unified Kent Countyinitial water-table
grid. The water-level observation pointdata set for Kent County
consisted of 2,176 wells (Table 2).The regression analysis
performed on these data sets yieldedthe regression equations (Table
3) for each hydrologic con-dition and the resulting water-table
elevation grids for dry(Fig. 5A), normal (Fig. 5B), and wet (Fig.
5C) conditions inKent County (Martin and Andres, 2005b).
New Castle County Water-Table Elevation
Calculating the water-table elevation maps for NewCastle County
also involved applying the same concepts thatwere established in
Sussex County by dividing New CastleCounty into two separate
sections (north and south) with theC&D Canal acting as the
hydrologic boundary. Long-term
water-level measurements from monitoring wells Jd42-03,Hb14-01,
and Db24-10 were compared in order to define thetime periods for
dry, normal, and wet conditions. However,as was the case in Kent
County, the regression analysis per-formed on the water-level point
data sets for these areasfailed to produce useable correlation
coefficients; therefore,the water-level points for the north and
south sections werejoined to produce a single data set for the
entire county.INITWT grids for New Castle County were created with
anordinary kriging algorithm because grid elevations comput-ed by
local polynomial regression were too high at low LSEand too low at
high LSE. It is likely that local polynomialregression could not
adequately reproduce the greater reliefof land surface and the
water table in New Castle County.The INITWT grids for southern, and
northern New CastleCounty were also merged together to form a
unified NewCastle County INITWT grid. The water-level point data
setfor New Castle County contained an uneven distribution of812
wells (Table 2). The regression analysis performed onthese wells
produced the regression equations (Table 3) thatwere used to create
the water-table elevation maps for dry(Fig. 6A), normal (Fig. 6B),
and wet (Fig. 6C) conditions inthe Coastal Plain of New Castle
County (Martin and Andres,2005c). When using the water-table
elevation and subse-quent depth-to-water maps for New Castle County
it isimportant to note that the Piedmont region of Delaware
wasexcluded from this work due to the sparse availability
andinaccuracy of water-level data for this area.
Depth to Water
The water-table DEMs for dry, normal, and wet condi-tions for
Sussex (Figs. 7A, B, and C), Kent (Figs. 8A, B, andC), and New
Castle (Figs. 9A, B, and C) counties were
8 Delaware Geological Survey • Report of Investigations No.
73
-
subtracted from the land surface DEM to produce depth-to-water
grids. When comparing depth to water to the percentageof land area
(Table 5) it becomes apparent that a significantportion of the
Coastal Plain of Delaware can be classified ashaving a shallow
water table. Under normal conditions, 71percent of the land area
has a depth to water of less than 10 ftand 21 percent of the land
area has a depth to water of less than5 ft, with these percentages
being significantly higher inSussex and Kent counties.
When dealing with depths to water of less than 10 ft therewill
likely be significant environmental issues with largerwastewater
disposal facilities such as rapid infiltration basinsand community
disposal systems (USEPA, 1999, 2003).When dealing with depths to
water of less than 5 ft, sitesbecome high risk for individual
standard domestic subsurfacewastewater disposal systems (DNREC,
2005) and for anyexcavations, building foundations, and
basements.
Comparison of Existing DLG Hydrography toLIDAR-Derived DEM
Utilizing existing DEMs, DLGs, water level data, andGIS tools to
estimate the water-table elevation was cost effi-cient and
effective; however, the potential still exists forgreater precision
and accuracy through the use of LIDAR-derived DEMs of land surface
and elevations of surface-waterfeatures. It is not unreasonable to
expect that water-table gridresolutions could be increased to the
2-m horizontal and0.328-ft levels of the LIDAR-derived DEMs.
However,improvements to the water-table DEMs from LIDAR DEMswill
require significant additional efforts as visual comparisonindicate
that there are inaccuracies in the DLG locations ofsurface-water
features. There also are data processing artifacts
that result from the procedures used to estimate elevations
ofsurface features from land surface DEMs.
Locational inaccuracies in the 1:24,000 hydrographyDLGs were
evident where DLG locations of surface-waterfeatures did not align
with the local minimum land surface ele-vations on the 30-m DEMs;
data processing artifacts are evi-dent where the elevations of
stream features do not decrease inthe downstream direction (Andres
and Martin, 2005). Theseissues become even more apparent when
comparing the1:24,000 hydrography DLGs with LIDAR-derived DEMS(Fig.
10). Reducing the effects of these problems will requirework to
locate the streams within the areas of local topo-graphic minima
and to mitigate any other artifacts in theLIDAR DEMs that result
from bridges, culverts, channelobstructions, and/or data processing
problems.
CONCLUSIONS
Water-table depth is a key facet in many
engineering,hydrogeologic, and environmental management and
regulato-ry decisions. Depth to water is an important factor in
riskassessments, site assessments, evaluation of permit compli-ance
data, and registration of pesticides and determiningacceptable
application rates. Obtaining an accurate represen-tation of the
water table is also crucial to the success of manyhydrologic
modeling efforts.
An extensive cooperative effort to produce readily avail-able
water-table elevation maps for the state of Delaware wasundertaken
in the 1950s. However, despite the usefulness ofthese paper contour
maps, contemporary advances in comput-er technology and the
expanding utilization of GIS has madeit necessary to update these
maps and has brought about thedemand to have them published in a
suitable digital format.
Table 5. Comparisons of depth to water (∆ΤΩ) and percentage of
land area.
Delaware Geological Survey • Report of Investigations No. 73.
9
-
Mapping the water-table elevation of the DelawareCoastal Plain
was accomplished by using pre-existing datasuch as long-term
hydrographs to determine dry, normal, andwet hydrologic conditions,
a 30-m DEM to assign elevationsto surface-water features, and well
completion reports used toobtain static water levels of shallow
domestic wells to producethe regression coefficients that were
inserted into the multiplelinear regression equation. The resultant
products are GISready grids with a horizontal spacing of 30 m and a
verticalresolution of 1 ft.
Existing DEMs, DLGs, water-level data, and GIS toolsprovided a
cost efficient and relatively accurate means to esti-mate the
water-table elevation; however newer technologyoffers potential for
greater precision and accuracy. LIDARmeasured DEMS offer the
potential for increasing the hori-zontal and vertical resolutions
of the water-table grids. Use ofLIDAR DEM data to estimate higher
resolution grids ofwater-table elevation will require more accurate
locationaldata for surface-water features as well as more powerful
andefficient computers and software.
REFERENCES CITED
Andres, A. S., and Martin, M. J., 2005, Estimation of thewater
table for the Inland Bays watershed, Delaware:Delaware Geological
Survey Report of Investigations No.68, 20p.
Delaware Department of Natural Resources andEnvironmental
Control (DNREC), 2005, Regulationsgoverning the design,
installation and operation of on-sitewastewater treatment and
disposal systems: DNRECDocument No. 40-08-05/04/07/01, 75p.
Dunne, T., and Leopold, L. B., 1998, Water in
environmentalplanning: New York, New York, W. H. Freeman
andCompany, 818 p.
ESRI, Inc., 2003, ARCGIS software v. 9.0:
Redlands,California.
____ 2005, ARCGIS software v. 9.1: Redlands, California.Freeze,
A. R., and Cherry, J. A., 1979, Groundwater:
Englewood Cliffs, New Jersey, Prentice Hall, Inc., 604 p.Freeze,
A. R., and Witherspoon, P. A., 1967, Theoretical
analysis of regional ground-water flow: Water ResourcesResearch,
vol. 3, p. 623-634.
Johnston, R.H., 1973, Hydrology of the Columbia(Pleistocene)
deposits of Delaware: An appraisal of aregional water-table
aquifer: Delaware Geological SurveyBulletin No. 14, 78p.
____ 1976, Relation of ground water to surface water in
foursmall basins of the Delaware coastal plain: DelawareGeological
Survey Report of Investigations No. 24, 56p.
Golden Software, Inc., 2002, Surfer software v. 8,
Golden,Colorado.
Mackenzie, J., 1999, Watershed delineation project, alphadata
release: www.udel.edu/FREC/spatlab/basins.
Martin, M. J., and Andres, A. S., 2005a, Digital water-tabledata
for Sussex County, Delaware: Delaware GeologicalSurvey Digital
Product 05-01, ESRI grid format.
____ 2005b, Digital water-table data for Kent County,Delaware:
Delaware Geological Survey Digital Product05-01, ESRI grid
format.
____ 2005c, Digital water-table data for New Castle
County,Delaware: Delaware Geological Survey Digital Product05-01,
ESRI grid format.
Perlman, H., 2005, Earth’s water: Ground water, UnitedStates
Geological Survey, United States Department ofthe Interior,
http://ga.water.usgs.gov/edu/earthgw.html.
Sepulveda, N., 2003, A statistical estimator of the spatial
dis-tribution of the water-table altitude: Ground Water, vol.41, p.
66-71.
U.S. Environmental Protection Agency, 1999, The Class
Vunderground injection control study, large-capacity septicsystems:
U.S. Environmental Protection Agency,EPA/816-R-99-014e, 120p.
____ 2003, Rapid infiltration land treatment, U.S.Environmental
Protection Agency wastewater technologyfact sheet: EPA
832-F-03-025, 6p.
U.S. Geological Survey 1992, Hydrography digital line graphfor
Delaware: U.S. Geological Survey.
____ 1996, Water-level data for the industrial area northwestof
Delaware City, Delaware, 1993-94, USGS Open FileReport 96-125: U.S.
Environmental Protection Agency,Towson, Maryland, 23 p.
Williams, T. A., and Williamson, A. K., 1989,
Estimatingwater-table altitudes for regional ground-water
flowmodeling, U.S. Gulf Coast: Ground Water, vol. 27,
p.333-340.
Winter, T. C., 1999, Relation of streams, lakes, and wetlandsto
groundwater flow systems: Hydrogeology Journalvol. 7, no. 1, p.
28-45.
10 Delaware Geological Survey • Report of Investigations No.
73
Figure 10. Illustration showing stream segment elevation
artifactscaused by misalignment of 1:24,000 hydrography DLG
withLIDAR DEM and road crossings.
-
RESEARCH
DELAWARE
GEOLOGICALSURVEY
EXPL
ORA
TIO
N
SERVICE
Delaware Geological SurveyUniversity of DelawareNewark, Delaware
19716