Professional Documents
Culture Documents
Pengantar Pemodelan Air Tanah-Heru Hendrayana 2012: Research
Pengantar Pemodelan Air Tanah-Heru Hendrayana 2012: Research
net/publication/275893905
CITATIONS READS
0 2,898
1 author:
Heru Hendrayana
Universitas Gadjah Mada
70 PUBLICATIONS 51 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Modeling Distribution of Rock Potential Acid Forming in Open Pit Coal Mining Areas View project
All content following this page was uploaded by Heru Hendrayana on 06 May 2015.
Based on Younger (2007), there are numerous reasons why the simulation of
aquifer behavior is often desirable. Mostly, the motives are entirely practical, for
instance:
Such applications of groundwater models are amongst the most valuable tools at the
disposal of aquifer managers and are probably the single greatest application of
advanced numerical analysis to the resolution of practical problems in any field of
engineering [Younger, 2007].
According to Kresic (2007), groundwater modeling area divided into two main
groups: (1) models of groundwater flow and (2) models of contaminant and solute
transport. Some of the more common questions that fully developed and calibrated
groundwater flow, and fate and transport models may help answer are:
Once of these questions area addressed by the models, many new ones may pop-up,
which is exactly what the purpose of a well-documented and calibrated groundwater
model should be answer all kinds of possible question related to groundwater flow,
and fate and transport of contaminates.
Based on Mandle (2002), groundwater flow models are used to calculate the rate
and direction of movement of groundwater through aquifers and confining units in the
subsurface. These calculations are referred to as simulations. The simulation of
groundwater flow requires a thorough understanding of the hydrogeologic
characteristics of the site. The hydrogeologic investigation should include complete
hydrogeological characteristics of aquifer such as:
Based on Kresic (2007), a model simulates the areal and temporal properties of a
system, or one of its parts, in either a physical (real) or mathematical (abstract) way.
A physical model in hydrogeology can be consider as a tank filled with sand and
saturated with water called “sandbox”, an equivalent to miniature aquifer of limited
extent. This aquifer can be subject to miniature stresses such as pumping from a
perforated tube placed into the sand thus representing a water well [Kresic, 2007].
Empirical (experimental)
Probabilistic
Deterministic
These models are derived from experimental data that fitted to some
mathematical function. Empirical models are limited in scope and are usually site or
problem-specify. However, these models can be an important part of a more complex
numeric modeling effort. For instant the behavior of a certain pollutant in porous
These models are based on laws of probability and statistics. They can have
various forms and complexity starting with a simple probability distribution of a
hydrogeological property of interest and ending with complicated stochastic, time-
dependent models. The main limitations for a wider use of probabilistic (stochastic)
models in hydrogeology are that: (i) they required large data sets needed for parameter
identification and (ii) they cannot be used to answer (predict) many of the most
common questions from hydrogeologic practice such as effects of a future pumping,
for example [Kresic, 2007].
These models assume that the stage or futures reactions of the system (aquifer)
studied are predetermined by physical laws governing groundwater flow. Most
problems in traditional hydrogeology are solved using deterministic models, which
can be as simple as the Theis equation or as complicated as a multiphase flow through
a multilayered, heterogeneous, anisotropic aquifer system. There are two large groups
of deterministic models depending upon the type of mathematical equation involved:
(a.) analytical methods and (b.) numerical methods [Kresic, 2007].
Referring to Kresic (2007), numerical methods described the entire flow field of
interest at the same time, providing solutions for as many data points as specified by
user. The area of interest is divided into many small areas (refer to as cells or
elements) and a basic groundwater flow equation is solved for each cell usually
considering its water balance (water input and output). The solution of a numerical
model is the distribution of hydraulic heads at points representing individual cells.
These points can be placed at the center of the cells, at interaction between adjacent
cells, or elsewhere. The basic differential flow equation for each cell is replaced
(approximated) by an algebraic equations so that the entire flow field is represented by
x equation with x unknowns, where x is the number of cells. This system of algebraic
equations is solved numerically, though an iterative process, thus the name numerical
models.
Deterministic Stochastic
- Monte Carlo simulation
- Random Walk Model
1D 2D 3D
Finite differences
Finite elements
Both types of methods above have their advantage and disadvantage and for
certain problems one way are more appropriate than the other. Both the finite
difference method and the finite element method are most widely used numerical
techniques for solving mathematical models.
Figure 4: The finite difference and finite element methods [Essink, 2000]
The computations result in values at nodal points. This means that the
piezometric head is determined for a certain area around these points. So the
piezometric head is an average for the so-called influence area. Figure 5 shows
influence areas for triangular and rectangular elements. In areas of interest a more
dense grid may be required. In the finite difference method, fluxes through a boundary
are inserted over the area of the elements, whereas in the finite element method,
boundary fluxes are inserted in the node [Essink, 2000].
h1 h2 h
i (I. 1)
L L
h1 h2 h
v k .i k . k. (I. 2)
L L
h1 h2 h
Q k .i. A k . .A k. .A (I. 3)
L L
h
q k . (I. 4)
l
Where
i = hydraulic gradient
k = proportionally factor or hydraulic conductivity/permeability (L/T)
h1 & h2 = heads (L)
v = velocity of flow through the aquifer (L/T)
A = cross-sectional area of diameter cylinder (L2)
Q = volume rate of flow of ground water (L3/T)
L = length of between hydraulic heard (T)
q = darcian specific discharge (L/T)
Darcy’s law is only valid in case of laminar flow: viz. at relative low velocities
when water particles move more or less parallel to each other. In quantified terms,
Darcy’s law is valid as long as the so-called Reynolds number Re does not exceed
some value between 1 and 10.
p.q.R
Re 1 10 (I. 5)
Where
p = mass density of water
μ = dynamic viscosity (M/ L/T)
R = hydraulic radius of the pore (L)
In case the atmospheric pressure equals zero (Figure 8), the relation between the
pressure and the so-called piezometric head is as follows:
VII.3. Permeability
One of the effective method to calculate the hydraulic conductivity is Slug Tests
Method. This method is applicable to determine the horizontal hydraulic conductivity
of distinct geologic horizons under in-situ conditions. Slug tests involve a short-term
introduction or removal of water via a well or boring into (or out of) a subsurface
interval of sediment, soil, or fractured rock. Monitoring the water level rise or fall as it
Likewise, one of the most applicable is Hvorslev Method, 1951 because this test
method is commonly use in the field since it is suitable for each type of aquifers and
tend to be more free and easy way of calculation. The parameters necessary recorded
at the observation wells such as radius of well casing, radius of well screen, length of
well screen, initial water level prior to removal of slug, and water level at time of
recorded equaled to zero.
Opening B,
1m wide and aquifer
height b
Opening A,
1m square
T K .b (I. 7)
It has unites of squared length over time (e.g.,m 2/d or ft2/d). In practice terms,
the transmissivity equals the horizontal groundwater flow rate through a vertical strip
where Vv = volume of all rock voids and V = total volume of rock (in geologic terms,
rock refers to all the following: soils, unconsolidated and consolidated sediments, and
any type of rock in general). Assuming the specific gravity of water equals unity, total
porosity, as a percentage, can be expressed in four different ways [Lohman, 1972].
Vi Vw V Vm V
n 1 m .100% (I. 9)
V V V V
where
n porosity, in percent per volume
VV volume of rock voids
V total volume
Vi volume of all int erstices voids
Vm aggregate volume of min eral solid particles
Vw volume of water in saturated sample
m d
n 1 d .100% (I. 10)
m m
A related parameter is termed the void ratio, designated as e, and stated as:
Vv
e (I. 11)
Vs
expressed as a fraction, where Vs is the solid volume. As total volume is the sum
of the void and solid volume, the following relationships can be derived:
n e
e or n (I. 12)
1 n 1 e
Porosity of sedimentary rock will depend not only on particle shape and
arrangement, but on a host of diagenetic features that have affected the rock since
deposition. Porosity can range from zero or near zero to more than 60% as shown in
Table 2.
Based on Domenico & Schwartz (1997), the specific storage was defined as a
proportionality constant relating the volumetric changes in fluid volume per unit
volume to the time rate of change in hydraulic head. The amount of water obtained per
unit volume drained is rather substantial and is obviously equal to the volume of pore
space actually drained. However, the physical mechanism that releases or stores water
in the storage is not the same for unconfined and confined aquifers (Figure 11).
Unit area
Unit declines
in heads
Confined
aquifer Confined
aquifer
Volume of
water released
from storage
The specific retention Sr of a soil or rock is the ratio of the volume of water it
will retain after saturation against the force of gravity to its own volume. Thus,
wr
Sr (II. 13)
Vt
where wr is the volume occupied by retained water and Vt is the bulk volume of
the soil or rock.
Based on Todd (1990), a specific yield Sy of a soil or rock is the ratio of the
volume of water that, after saturation, can be drained by gravity to its on volume.
Therefore,
wy
Sy (I. 14)
Vt
where wy is the volume of water drained. Values of Sr and Sy can also be expressed as
percentages.
Regarding to Kresic (2007), specific yield has been called the “effective
porosity” or “practical porosity” because it represents the pore space that will
surrender water to wells and is there effective in furnishing water supplies. Therefore,
S r S y (I. 15)
where α presented the effective porosity. Figure 12 below illustrated the effective
porosity of unconfined aquifer in function to specific retention and specific yield.
Regarding to Todd (1980), the values of specific yield depend on grain size,
shape and distribution of pores, compaction of the stratum, and time of drainage.
Representative specific yield for various geologic materials are listed in Table II. 3;
individual values for a soil or rock can very considerably from these value. It should
be noted that fine-grained materials yields little water, whereas coarse-grained
materials permit a substantial release of water and hence serve as aquifers. In general,
specific yields for thick unconsolidated formations tend to fall in the range of 7 to 15
percent, because the mixture of grain sizes presented in the various strata; furthermore,
thy normally decrease with depth due to compaction.
Table 3: Values of Specific Yields for Various Geologic Materials (Johnson, 1967)
- Natural outflows occur when water leaves the aquifer at springs or into rivers.
Other natural outflows include low-lying areas which act as a sink to
groundwater systems; this form of outflow may be associated with areas of
evapotranspiration especially from deep-rooting vegetation. These low-lying
areas often form wetlands which have a high ecological value. One further
natural outflow occurs when water flows into other aquifers.
- Man-made outflows occur when water is pumped from wells and boreholes.
Since the velocities in the vicinity of the pumped borehole are far higher than
the natural groundwater velocities, there is a risk of deterioration of the aquifer
in the vicinity of the well or borehole and a deterioration of the borehole
structure. Horizontal wells or adits provide alternative means of collecting
water from an aquifer; this approach is especially suitable for shallow aquifers
or for aquifers with thin lenses of good quality water.
d
KP m
(I. 16)
d
1 K m
pm
Where
d = total thickness
Based on Essink (2000), beefore going on, it is essential to identify the purpose
of the modeling effort. Therefore, some of the following questions are needed.
The responses of these questions will lead the modeler in determining the
modeling effort: analytical or numerical, lumped or distributed, transient or steady
state, etc [Essink, 2000].
Regarding to Essink (2000), since a very wide range of computer codes exists
for application of different problems, it is needed to select the best code which is the
most available to the interest area. When choosing a code from the selection available,
the following points should be considered:
Figure 14: Orientation is dependent on aquifer’s shape and degree of anisotropy [Kresic, 2007]
o Choosing the Model-Grid Cell Size
Stated by Essink (2000), Selecting the model grid is comparable to the selecting a
net for fishing; the openings in the net must match the size of the “fish”
(heterogeneities and predictive details) to be captured. In choosing a model grid
size, the following factors should be considered:
Based on Essink (2000), a model boundary is the interface between the model
calculation domain and the surrounding environment. Boundaries occur at the edges of
the model domain and at other points where external influences are represented, such
as rivers, wells, leaky impoundments, or chemical spills and so forth. Boundary
conditions are expressions of the effect of the external world on the model domain,
and they required to complete the description of a flow or transport problem. The
mathematical expression of boundary condition is required for a well-posed problem.
Likewise, boundary conditions are mathematical statements at the boundary of the
problem domain. A correct selection of boundary conditions is a critical step in the
model design, as a wrong boundary may lead to serious errors in the results.
Mathematically, the boundaries are divided in three types:
Specific head boundary describing specified head boundaries for which a head is
given. Examples of specified head boundaries are: the water level at a lake or at the
sea. A specified head boundary represents an inexhaustible supply of water. For
example, water is pulled from or discharged in the boundary without changing the
head at the boundary. In some situations, this is probably an unrealistic approximation
of the response of the system. The equation of specific head boundary expressed by:
This boundary describing the head-dependent flow for which flux across a
boundary is calculated, given a value of the boundary head. The condition of heard-
dependent flow boundary is:
. cons tan t (I. 19)
x
Based on Essink (2000), the initial condition of groundwater flow can be given
in three features (Figure15): (1) the static steady state condition in which the head is
constant throughout the problem domain and in which there is no flow is the system
(e.g. used for drawdown simulations in response to pumping); (2) the dynamic average
steady state condition in which the head varies spatially and flow into the system
equals flow out the system (this condition is used most frequently); and (3) the
dynamic cyclic steady state condition in which the head varies in both space and time
(a set of heads represent cyclic water level fluctuations, e.g. monthly head fluctuations
or monthly average recharge rates).
Trial-and-error calibration
In this alternative is the input all the parameters based on physical observation, and
provides estimates of the unknown parameters as a first trial. As such, the adjustment
of parameters is manual. The model is run and the computed output is compared to the
measured output from the prototype (Figure 16). The comparison is done by means of
visual pattern recognition of the measured and computed flow hydrographs or solute
distributions, or it is based on some mathematical criterion. Based on this comparison,
adjustments are made to one or more of the trial parameters to improve the fit between
In this alternative, also called the inverse problem, the model itself contains internal
programming which will adjust the trial parameters in a systematic step by step
manner until the goodness of fit criterion is satisfied. In this way, the model will
automatically calibrate itself and carry out the necessary number of trial runs until the
best set of parameters is achieved [Essink, 2000].
Traditionally, two methods are used to evaluate the calibration: (a) qualitatively,
by comparison of contour maps of measured and computed parameters, which
provides only a qualitative measure of the similarity between the patterns; and (b)
quantitatively, by a scatterplot of measured and computed parameters, where the
deviation of points from the straight line should be randomly distributed. Three ways
of expressing the average difference between measured and computed parameters are
normally used to quantify the average error in the calibration:
a. The mean error (ME): which is the mean difference between measured
(pmeasured) and computed (pcomputed) parameters, such as piezometric heads:
1 n
ME . ( Pmeasured Pcomputed )i (I. 21)
n i 1
Where
As both negative and positive differences are incorporated in the calculation, they may
cancel out the error. As such, a small error may not indicate a good calibration, and
this way of quantifying the error should be used with care.
b. The mean absolute errors (MAE): which is the mean of the absolute value of
the difference between measured and computed parameters:
1 n
MAE . Pmeasured Pcomputed )i (I. 22)
n i 1
0.5
1 n
RMS . ( Pmeasured Pcomputed )i2 (I. 23)
n i 1
This error is usually thought to be the best measure of error if errors are
normally distributed.
In general, the making of the first error-free output will probably require many
additional hours, at least, more than expected. The whole phase of the first model
execution includes the preparation of the input data file, the entry of input data file into
computer lines, the execution of the model (the so-called run), and the interpretation of
the results [Essink, 2000].
In order to improve the confidence in the calibration of the model, the model has
to be tested by using a second independent set of data. This is called the model
verification or model validation. A model is verified when the verification targets are
matched without changing the calibrated parameters. If it is necessary to adjust
parameters during the model verification because the verification targets are not
matched, the verification becomes a second calibration and other independent data sets
should be needed until the verification of the model is performed [Essink, 2000].
VIII.7. Simulation
In this category can often rely on the laws of fluid mechanics and hydraulics. At
the most, interpolation and short extrapolation occurs. The availability of test data
provides a relatively good safeguard against misconceptions [Essink, 2000].
VIII.9. Postaudit
Code verification No
Verified?
CODE
SELECTION Yes
Presentation of result
Figure 17: Steps of groundwater model application [adapted from Anderson & Woessner,
1992]