The following analysis is taken from:
Baker, A.E., Wood, A.R., and Cichon, J.R., 2008 (in review), The Citrus County Aquifer Vulnerability Assessment: Contract deliverable report prepared for the Florida Geological Survey - Florida Department of Environmental Protection.
THE CITRUS COUNTY AQUIFER VULNERABILITY ASSESSMENT
Part of Phase II of the Florida Aquifer Vulnerability Assessment (FAVA) Project, Florida
Department of Environmental Protection Contract No. RM059, July, 2008.
Prepared for the Florida Department of Environmental Protection by Advanced GeoSpatial Inc.
THE CITRUS COUNTY AQUIFER VULNERABILITY ASSESSMENT
NOTE: In discussion with the principal author of the Analysis, it was confirmed that no
consideration was given to the effect of fault lines or fracture sets which are prevalent in the area covered
by the analysis. [But see Fracture Sets Example]
Interpretation of Results in Context of FAVA
Results of the CAVA project have allowed delineation of new and unique zones of relative
vulnerability for the FAS in Citrus County, based on the county-specific model boundary used, use of numerous
well points for aquifer confinement characterization, incorporation of most recent soils data, and application
of recently-developed approaches for karst estimation in a GIS. These new results, though refined and highly
detailed, do not replace results of previous studies. In other words, the FDEP's regional FAVA results (Arthur
et al., 2005) for the FAS indicate that the Citrus County study area occurs in primarily a "more
vulnerable" zone relative to other areas in Florida (Figure 17); as a result the new CAVA model output
should be interpreted in the context of this major regional project. The new zones delineated in the CAVA
project are unique to the CAVA study area, and reveal more detailed information regarding aquifer
vulnerability within the regional "more vulnerable", and "vulnerable" zones identified in
the FAVA project.
|Figure 17. Results of the Florida Aquifer Vulnerability Assessment project (Arthur et al., 2005) for the
FAS in Citrus County. The CAVA model relative vulnerability zones, while based on more refined data than the
FAVA project, occur within the context of this regional model.
|Figure 1. Citrus County Aquifer Vulnerability Assessment project study area corresponds to the County’s
The political boundary of Citrus County was used as the CAVA model study area extent (shown in Figure 1).
Soil Hydraulic Conductivity and Soil Pedality Themes
The rate that water moves through soil is a critical component of any aquifer vulnerability analysis, as
soil is literally an aquifer system’s first line of defense against potential contamination (Arthur et
al., 2005). Two parameters of soils were evaluated for input into the CAVA model: soil hydraulic
conductivity, which is the “amount of water that would move vertically through a unit area of
saturated soil in unit time under unit hydraulic gradient” (U.S. Department of Agriculture, 2005); and soil
pedality, which is calculated based on soil type, soil grade, and soil pedon size, and is a unitless
parameter. Soil pedality is a relatively new concept used to estimate the hydrologic parameter of soil and is
generated for CAVA using the pedality point method developed by Lin et al. (1999).
In 2006, Citrus County soils data were expanded for the study area and made available by the Natural
Resources Conservation Service. This expansion included adaptation into ESRI geodatabase compatible format,
and specific soils values were updated (U.S. Department of Agriculture, 2006). As a result, more detailed
information is available for analysis for the CAVA project than during previous projects (e.g., Arthur et al.,
2005). Countywide data sets representing soil hydraulic conductivity and soil pedality were developed for use
as input into the CAVA model. This is completed for both hydraulic conductivity and soil pedality.
Weights calculated during sensitivity analysis for soil pedality were much stronger (i.e., had higher
absolute value) than weights calculated for soil hydraulic conductivity. As a result, soil pedality was chosen
as the better predictor of aquifer vulnerability because it shared the best association with training points.
Soil pedality, a unitless parameter, ranges from 0.0157 to 0.0533 across the study area. The analysis
indicated that areas underlain by 0.0429 to 0.0533 were more associated with the training points, and
therefore associated with higher aquifer vulnerability. Conversely, areas underlain by 0.0428 to 0.0157 were
less associated with the training points, and therefore lower aquifer vulnerability. Based on this analysis,
the evidential theme was generalized into two classes as displayed in Figure 10.
|Figure 10. Generalized soil pedality evidential theme; based on calculated weights analysis blue areas
share a weaker association with training points and thereby relatively lower aquifer vulnerability, whereas
red areas share a stronger association with training points.
In Copeland et al. (1991), the area of the Brooksville Ridge in central Florida is defined as having higher
recharge potential than adjacent areas. The Brooksville Ridge is chiefly composed of Undifferentiated Hawthorn
Group sediments which are poorly to moderately consolidated clayey sands and silty clays (Scott et al., 2001).
In Citrus County, these sediments reach a maximum calculated thickness of 199 feet, can be discontinuous,
deeply weathered and highly perforated by karst features.
In Citrus County these sediments are generally highly weathered, leaky, thin and intensely breached by
karst features. These factors combine to increase the recharge potential to the FAS in the study area.
Recharge potential values were calculated for the study area by subtracting the USGS 2000 potentiometric surfacean imaginary surface defined by the level to which water in an aquifer would rise in a well due to the natural pressure in the rocks. of the FAS (USGS, 2000) from land surface elevation derived from USGS 7.5” quadrangles.
Resulting recharge potential values range from -18 ft to greater than 150 ft (relative to mean sea level).
Negative values generally correspond to areas where the aquifer is estimated to be discharging while higher
positive values are restricted to the more substantial hills located on the Brooksville Ridge.
Categories of recharge potential were derived from records of the potentiometric surface. A preliminary
weights of evidence indicated a very strong relationship between training points and recharge potential.
Categories of recharge potential were ranked as displayed in Figure 5. Use of recharge potential via this
approach is restricted to areas of Florida where the FAS is not well confined (e.g., this layer may not be
usable in areas which are also underlain by thicker, contiguous Intermediate confining unit sediments), and
where there is not a laterally contiguous Surficial aquifer system present.
Intermediate Confining Unit / Overburden Thickness Themes
Aquifer confinement – either in the form of overburden overlying the FAS, or the Intermediate
confining unit (ICU) – is another critical layer in determining aquifer vulnerability. Where aquifer
confinement is thick and the FAS is deeply buried, aquifer vulnerability is generally lower, whereas in areas
of thin to absent confinement, the vulnerability of the FAS is generally higher.
To allow the calculation of aquifer confinement thickness in various study areas, surface models were
developed using a dataset of borehole records supplemented with well gamma logs that contain descriptions of
subsurface materials. AGI (the consultants) used these surfaces to calculate thickness of the ICU (Figure 6)
and thickness of overburden overlying the FAS (Figure 7) in the study area. These two layers were tested for
input in the model as described in Sensitivity Analysis.
|Figure 6. Thickness of the ICU calculated by subtracting predicted surface of ICU (Figure 6) from
predicted surface of FAS as generated by FGS/FDEP.
||Figure 7. Thickness of sediments overlying the FAS calculated by subtracting digital elevation data from
predicted surface of FAS as generated by FGS/FDEP.
Potential Karst Feature Theme
Karst features, or sinkholes and depressions, can provide preferential pathways for movement of surface
water into the underlying aquifer system and enhance an area’s aquifer vulnerability where present. The
closer an area is to a karst feature, the more vulnerable it may be considered. Closed topographic depressions
extracted from U.S. Geological Survey 7.5-minute quadrangle maps served as the initial dataset from which to
estimate potential karst features in the study area. To supplement these data, the FGS/FDEP sinkhole database
was included to identify karst features possibly not represented on USGS maps. These two data sources
displayed in Figure 8 were combined and analyzed to develop a potential karst features evidential theme. See
Figure 13 below.
|Figure 13. Potential karst features evidential theme buffered into 100-ft zones for proximity analysis in
the weights of evidence analysis.
Potential Karst Features
As mentioned above, areas closer to a potential karst feature are normally associated with higher aquifer
vulnerability. Based on this, features were buffered into 100-ft zones to allow for a proximity analysis (Figure
13). The analysis indicated that areas within 3,346 feet of a karst feature were more associated with the
training points, and therefore with higher aquifer vulnerability. Conversely, areas greater than 3,346 feet
from a karst feature were less associated with the training points, and therefore lower aquifer vulnerability.
Based on this analysis, the evidential theme was generalized into two classes as displayed in Figure 14.
|Figure 14. Generalized potential karst feature evidential theme; based on calculated weights analysis blue
areas share a weaker association with training points and thereby relatively lower aquifer vulnerability,
whereas red areas share a stronger association with training points.
Using evidential themes representing soil pedality, recharge potential, ICU thickness, and potential karst,
weights of evidence was applied to generate a response theme, which is a GIS raster consisting of posterior
probability values ranging from 0.0002697 to 0.0343457 across the study area (Figure 15). These
probability values describe the relative probability that a unit area of the model will contain a training
point – i.e., a point of aquifer vulnerability as defined above in Training Points – with
respect to the prior probability value of 0.01043. Prior probability is the probability that a training
|Figure 15. Relative vulnerability map for the Citrus County Aquifer Vulnerability Assessment project.
Classes of vulnerability are based on calculated probabilities of a unit area containing a training point, or
a monitor well with water quality sample results indicative of vulnerability.
|Figure 18. Confidence map for the CAVA model calculated by dividing the posterior probability values by
the total uncertainty for each class to give an estimate of how well specific areas of the model are predicted.
Model Limitations and Scale of Use
When implementing the CAVA model results, it is vital to remember that aquifer systems in Florida are
vulnerable to contamination; an invulnerable aquifer does not exist. Model results are based on features of
the natural system that have significant association with the location of training points and thereby aquifer
vulnerability. The CAVA project results provide a probability map that identifies zones of relative
vulnerability in the study area based on input data; as a result the CAVA model output is an estimation of
natural aquifer vulnerability and the results do not account for activities at land surface, contaminant type,
groundwater flow paths or fate/transport of chemical constituents.
Derivative Products: Protection Zones
Relative vulnerability zones defined in this project may be applied to develop derivative maps, such as a
protection-zone map (Cichon et al., 2005). Ideally, data layers not included as input in the aquifer
vulnerability model would be considered to help in defining such protection zones and may include groundwater
flow modeling, stream-sink features, induced drawdown areas from large well fields, and distribution of
drainage wells. These layers, while important to aquifer vulnerability, do not form usable input into this
aquifer vulnerability assessment project.
As mentioned above, a confidence map of the model’s posterior probability values can be calculated by
dividing the posterior probability by its total uncertainty. This essentially applies an informal student T-test
(as in Table 2) to the posterior probability values. The higher the confidence values, the greater the
certainty is with regard to the posterior probability. This map essentially indicates the degree of confidence
to which the posterior probabilities are meaningful and should be referenced when interpreting and
implementing the model results. In other words, the confidence map should be used to help guide implementation
of the vulnerability map as it reveals the confidence level associated with each vulnerability class (Mihalasky
and Moyer, 2004).
Surface Water Areas
In addition to large surface-water bodies omitted from the analysis, there are many other surface-water
features which were not removed. Many of these features may represent areas of groundwater discharge; however,
these discharging surface waters are not part of the aquifer, although they originate from it. Accordingly,
the CAVA model is not intended to be used to assess contamination potential of surface waters, though the
discharging surface waters are highly vulnerable to contamination.
Recommendations on Scale of Use
Use of highly detailed evidential theme data as model input results in highly resolute model output as can
be seen in the model response theme. These resolute features are reflections of real data used as input;
however, the final maps should not be applied to very large scales such as to compare adjacent small parcels.
CAVA model output is, in a sense, as accurate as the most detailed input layer, and as inaccurate as the
least detailed layer. For example, wells used to define confinement thickness represent an area up to 25
square miles (mi 2 ), the potentiometric surfacean imaginary surface defined by the level to which water in an aquifer would rise in a well due to the natural pressure in the rocks. map used in the development of the recharge potential
evidential theme was mapped at 1:500,000, and soils polygonal data represent an area as small as 19,375 ft 2 .
Every raster cell of the model output coverage has significance per the model input as discussed above.
However, it is important to note that aquifer vulnerability assessments are predictive models and no
assumptions are made that all input layers are accurate, precise or complete at a single-raster cell scale. As
mentioned above, the confidence map, because it is an indicator of the meaningfulness of the vulnerability
classes, should be used to help guide implementation of the vulnerability map. For example, in the CAVA
confidence map (Figure 18), local-scale land-use decisions might be more defensible in with the higher
vulnerability classes (more vulnerable and most vulnerable) as these areas are associated with highest
Ultimately, accuracy of the maps does not allow for evaluation of aquifer vulnerability at a specific
parcel or site location. It is the responsibility of the end-users of the CAVA model output to determine
specific and appropriate applications of these maps. In no instance should use of aquifer vulnerability
assessment results substitute for a detailed, site-specific hydrogeological analysis.
As demands for fresh groundwater from the Floridan aquifer system underlying Citrus County increase
resulting from continued population growth, identification of zones of relative vulnerability becomes an
increasingly important tool for implementation of a successful groundwater protection and management program.
The results of the CAVA project provide a science-based, water-resource management tool allowing for a pro-active
approach to protection of the FAS, and, as a result, have the potential to increase the value of protection
Model results will enable improved decisions to be made about aquifer vulnerability based on the input
selected, including focused protection of sensitive areas such as springsheds and ground-water recharge areas.
The results of the CAVA vulnerability model are useful for development and implementation of groundwater
protection measures; however, the vulnerability output map included in this report should not be viewed as a
static evaluation of the vulnerability of the FAS. Because the assessments are based on snapshots of best-available
data, the results are static representations; however, a benefit of this methodology is the flexibility to
easily update the response themes as more refined or new data becomes available. In other words, As the
scientific body of knowledge grows regarding hydrogeologic systems, this methodology allows the ongoing
incorporation and update of datasets to modernize vulnerability assessments thereby enabling end users to
better meet their objectives of protecting these sensitive resources. The weights of evidence modeling
approach to aquifer vulnerability is a highly adaptable and useful tool for implementing ongoing protection of
Florida’s vulnerable groundwater resources.
NOTE: In discussion with the principal author of the Analysis, it was confirmed that no consideration was
given to the effect of fault lines or fracture sets which are prevalent in the area covered by the analysis. [See Fracture Sets Example]