GeoHydros was originally contracted by Ginnie Springs Outdoors, Inc. to evaluate a regional-scale steady-state equivalent porous media groundwater flow model of north Florida that was developed for the
Suwannee River Water Management District (SRWMD) in 2008 by the SDII Global Corporation (NFM-08)
and compare the results of that model to the results of a sub-regional scale steady-state hybrid
groundwater flow model of the western Santa Fe River basin that GeoHydros had previously completed for
Coca-Cola North America in 2008 (WSFM-08).
The stated purpose of the NFM-08 is for use in the evaluation of “the effects of existing and proposed
groundwater withdrawals on the aquifers … of the District” primarily to support the evaluation of impacts
associated with consumptive use permit applications, and the designation and management of Minimum
Flows and Levels. The goal of the WSFM-08 was to accurately simulate karstic groundwater flow patterns
to the 1st and 2nd magnitude springs on the Santa Fe River under both low water and high water
conditions.
The purpose of GeoHydros’ investigation was three-fold: 1) to identify any design issues that would be
reasonably expected to diminish the reliability of the model’s assessments of impacts to spring and river
flows associated with cumulative groundwater pumping in and surrounding the SRWMD; 2) to evaluate the
efficacy of the equivalent porous media approach through a comparison of the SDII model results in the
western Santa Fe River basin to results obtained from the hybrid model; and 3) to describe the key
groundwater modeling processes, reasonable expectations for model quality and disclosure, and the
degree to which the NFM-08 meets these expectations.
The NFM-08 was evaluated on the basis of three broadly accepted criteria for quality of a groundwater
flow model: 1) the degree to which simulated groundwater levels and flows match real-world conditions; 2)
the degree to which the framework of model parameters adheres to a reasonable conceptualization of the
hydrogeologic conditions being simulated; and 3) the appropriateness of the mathematical representation
of the flow processes. In addition, the supporting documentation was evaluated to determine the degree to
which it provides the reader with a complete and transparent understanding of the model development
process, all underpinning assumptions, and any limitations that have bearing on the model’s intended
applications.
The Upper Floridan Aquifer is represented in the NFM-08 as an equivalent porous media, homogenous
within 5,000 x 5,000 foot grid blocks, that does not contain conduits though conduits are known to be
ubiquitous throughout much of the model domain and to have evolved as a consequence of karstification.
Because the model does not address conduit flow, it relies on implausible parameter values to force the
porous media groundwater flow equations to simulate observed spring flows and river gains. As a result,
the simulated groundwater surface poorly represents observed groundwater levels and local hydraulic
gradients.
The NFM-08 was intended to calibrate to average groundwater levels and spring flows occurring between
June 1, 2001 and May 31, 2002, however it represents both poorly. SDII defined the model’s calibration
target as +/- 5% of the total change in observed Upper Floridan Aquifer groundwater levels as measured
in 676 wells across the model domain. The NFM-08 Model domain spans the width of the Florida
peninsula from South Georgia to southern Marion County. The calibration criterion of +/- 5 feet was only
applied to the average of the absolute differences between simulated and observed values at the 676
wells. The resulting criterion is broad relative to the observed variation in groundwater levels during the
calibration period, during which groundwater levels in more than 50% of the wells in the SRWMD having at
least monthly measurements varied by less than 3 feet.
Application of the chosen calibration criterion allowed widespread and large magnitude differences
between observed and simulated groundwater levels across the model domain. Differences at 147 of 534
wells (~28%) within the SRWMD were larger than the 5-foot criterion. Differences at more than 10% of those wells distributed throughout the central Suwannee River and Santa Fe River basins were greater
than 10 feet. Where such large-magnitude errors exist, the model cannot reliably predict groundwater level
fluctuations and the widespread distribution of large-magnitude errors significantly undermines the
reliability of the predictions throughout the model domain.
The model cannot simulate flow to discrete springs as is implied in the SDII report because the resolution
of the model is predicated on the use of 5,000 X 5,000 foot grid cells and many of the springs described as
correctly simulated fall within a single grid cell. Individual spring flows within a single grid cell were
accounted for through the use of multiple conductance terms associated with Drain and River assignments
to the grid cells. The conductance terms describe the ability of the streambed at the respective locations to
transmit water from the aquifer to the river thereby acting as confining material that separates the springs
and rivers from the Upper Floridan Aquifer.
The use of the streambed conductance terms is inconsistent with well-established unconfined conditions
of the Upper Floridan Aquifer existing at the majority of the simulated springs. The values assigned during
the calibration process are implausible because they equate to the presence of substantial confining
material where confining material does not exist. As a consequence and in almost all instances, the match
between simulated and observed spring and river flows was achieved at the expense of realistic
simulations of groundwater levels at the rivers. The simulated levels deviate from observed values by
more than 10 feet along much of the central Suwannee and western Santa Fe Rivers. These deviations
were not discussed or disclosed in the report accompanying the NFM-08.
Discrepancies between observed and simulated groundwater levels at the rivers exceeded the 5-foot
criterion for matching groundwater levels at more than 50% of the assignments, some exceeding 20 feet.
Considering that river stage is known to match the groundwater level in the unconfined portion of the
Upper Floridan Aquifer where the rivers flow directly on Upper Floridan Aquifer limestones, these
discrepancies raise the average difference between observed and simulated groundwater levels in the
SRWMD to 5.6 feet, which violates SDII’s criterion for model calibration.
The absence of conduits from the model design required the calibration effort to rely on implausible
hydraulic conductivity, recharge, and streambed conductance assignments in order to force the model to
approximate hydrogeologic conditions that the underlying mathematical equations were not intended to
represent. Hydraulic conductivities deviate from values derived from aquifer performance tests and
reported by the US Geological Survey (USGS) by 0.5 to 2.6 orders of magnitude across much of the
model domain. Assigned recharge distributions fail to correlate to precipitation or documented land use.
The magnitude of assigned recharge results in simulated groundwater discharge to rivers and streams that
flow to the Gulf of Mexico that exceeds measured values by between 300 and 950 cfs, or when compared
to sub-watershed scale discharge, exceeds measured values by between 231 and 750 cfs. The streambed
conductance terms imply the existence of confining material in the unconfined part of the aquifer that does
not exist. As a consequence of implausible parameter values, the model violates the assumptions
underpinning the groundwater flow equations with which it was constructed throughout approximately half
of the model domain including much of the Suwannee River basin.
The model under-estimates the measured impacts to Upper Floridan Aquifer groundwater levels from
municipal groundwater pumping at two locations evaluated, the City of Gainesville and Fernandina Beach,
by more than 30 feet in both cases. The model under-estimates the capture zone for City of Gainesville’s
well field by more than 100 square miles, and it fails to accurately simulate documented groundwater flow
paths to the Santa Fe and Ichetucknee Rivers.
Model boundaries were not designed or assigned according to standard practices that focus on limiting the
degree to which simulated pumping is derived directly from external model boundaries. Approximately
38% of the simulated flow through the UFA is to external model boundaries (24% to the general head
nodes defining the southern model boundary, and 14% to the constant head nodes defining the Gulf of Mexico boundary). Removing the assigned pumping (not including wells used to represent river siphons)
revealed that the boundary conditions permit more than 40% of the simulated well extractions to intercept
flow that would otherwise be to the external boundaries (35.5% to the general head nodes representing
the southern boundary, and 5.1% to the constant head nodes representing the Gulf of Mexico). These
boundary condition effects are not disclosed in the NFM-08 report and the associated limitations on the
model’s ability to reliably predict impacts to groundwater levels or flows have therefore not been disclosed
to readers or model users.
These problems reveal that the NFM-08 is poorly constructed and not reliable for its stated purpose.
Furthermore, the model report is misleading because it does not disclose the necessary information for
readers or model users to indentify the degree to which the model fails to meet these criteria.
With respect to the technical practicability of improving on these model limitations, comparisons of the
NFM-08 (equivalent porous media model) to the WSFM-08 (hybrid model that includes conduits) reveal
substantial differences that are consequential to groundwater resource management decisions. The hybrid
model, in which the UFA was simulated as a dual-permeability framework consisting of conduits
embedded in a porous media, achieved substantially better matches to observed groundwater levels and
spring flows under both low-water and high-water conditions where the improvement stemmed from
significantly different simulations of groundwater flow patterns and velocities. Where the NFM-08 failed to
simulate tracer defined groundwater flow paths, the hybrid model accurately did so. Where the NFM-08
failed to match tracer-defined groundwater velocities, the hybrid model accurately did so. And, where the
NFM-08 used unrealistically high hydraulic conductivities, resulting in an inability to simulate observed
impacts to groundwater levels derived from municipal groundwater pumping in areas such as Fernandina
Beach and the Gainesville municipal well field, the substantially lower hydraulic conductivity values used in
the hybrid model support the simulation of much larger simulated drawdowns in the aquifer matrix that are
more consistent with observed conditions.
These discrepancies demonstrate that the equivalent porous media approach is incapable of adequately
simulating the patterns of groundwater flow to springs and therefore the impacts of groundwater pumping
on those flow patterns. Moreover, the fact that the hybrid model was constructed with commercially
available, widely used software as well as publically available datasets demonstrates that the decision to
use and rely on equivalent porous media assumptions and methods cannot be argued to be based on
technological impracticability.
In summary, the flaws in the NFM-08 and the manner in which it is being used by the SRWMD identified
through this investigation impart substantial limitations on the model’s assessments of the magnitude and
spatial distribution of impacts to spring and river flows associated with current and future groundwater
extractions. The limitations have direct bearing on water management district consumptive use permit
application review processes and Minimum Flows and Levels programs. The most relevant conclusions in
this regard are: 1) the NFM-08 is poorly constructed and fails to meet broadly accepted measures of
quality, and therefore cannot be reliably used to simulate or predict impacts to groundwater flows and
levels created by groundwater extractions within or surrounding the SRWMD; 2) the approach and
software used for the NFM-08 do not represent the best available technology; 3) alternative methods and
software could be, and could have been leveraged to build a better model that provides substantially more
reliable predictions; and 4) by using the NFM-08, the SRWMD is not pursuing a reasonably conservative
approach to the characterization and mitigation of impacts to spring and river flows associated with
groundwater withdrawals.