The Effects of the 2004 Tsunami on a Coastal

Aquifer in Sri Lanka

by Meththika Vithanage
1,2,3
, Peter Engesgaard
4
, Karen G. Villholth
2,5
, and Karsten H. Jensen
4
Abstract
On December 26, 2004, the earthquake off the southern coast of Sumatra in the Indian Ocean generated
far-reaching tsunami waves, resulting in severe disruption of the coastal aquifers in many countries of the region.
The objective of this study was to examine the impact of the tsunami on groundwater in coastal areas. Field
investigations on the east coast of Sri Lanka were carried out along a transect located perpendicular to the
coastline on a 2.4 km wide sand stretch bounded by the sea and a lagoon. Measurements of groundwater table
elevation and electrical conductivity (EC) of the groundwater were carried out monthly from October 2005 to
August 2007. The aquifer system and tsunami saltwater intrusion were modeled using the variable-density ow
and solute transport code HST3D to understand the tsunami plume behavior and estimate the aquifer recovery
time. EC values reduced as a result of the monsoonal rainfall following the tsunami with a decline in reduction
rate during the dry season. The upper part of the saturated zone (down to 2.5 m) returned to freshwater conditions
(EC < 1000 S/cm) 1 to 1.5 years after the tsunami, according to eld observations. On the basis of model
simulations, it may take more than 15 years for the entire aquifer (down to 28 m) to recover completely, although
the top 6 m of the aquifer may become fresh in about 5 years.
Introduction
The Sumatra-Andaman earthquake, which occurred on
December 26, 2004 in the Indian Ocean measured 9.0 on
the surface wave magnitude scale (Liu et al. 2005) and
generated a tsunami, which devastated coastal areas in
many southern Asian countries. In Sri Lanka, the wave
height recorded along the coast was approximately 5 to
1
Corresponding author: Department of Geography and
Geology, University of Copenhagen, Denmark; (94) 812 232 002;
fax: (94) 812 232 131; meththikavithanage@gmail.com
2
International Water Management Institute, Pelawatta,
Battaramulla, Sri Lanka.
3
Chemical and Environmental Systems Modelling Research
Group, Institute of Fundamental Studies, Kandy, Sri Lanka.
4
Department of Geography and Geology, University of
Copenhagen, Denmark.
5
Geological Survey of Denmark and Greenland, Copen-
hagen, Denmark.
Received May 2011, accepted October 2011.
2011, The Author(s)
Ground Water 2011, National Ground Water Association.
doi: 10.1111/j.1745-6584.2011.00893.x
12 m (Liu et al. 2005) and the average inundation distance
on the east coast was about 1 km (Villholth et al. 2005).
Groundwater is the only source of freshwater in
many parts of the eastern coastal belt of Sri Lanka. The
shallow unconned aquifers in these areas were affected
by the tsunami in several ways (Illangasekare et al. 2006).
During and after the tsunami, seawater inundated coastal
areas resulting in direct inltration of saline water from
ooded depressions as well as through the land surface
(Figure 1). In most of the affected areas, open dug wells
were instantly lled with seawater (Picture-2) which
inltrated into the freshwater below. Typically, the dug
wells are 1 to 2 m in diameter and 3 to 8 m deep.
Additionally, the subsurface saltwater wedge may have
pushed inland at the time of the tsunami wave due to the
increased head of the ocean (Figure 1).
A dense saltwater body located on top of fresh
phreatic groundwater is unstably stratied causing salt-
water to sink and mix convectively with water in deeper
parts of the aquifer (Oostrom et al. 1992). The density
contrast in the case of seawater from the Indian Ocean is
NGWA.org GROUND WATER 1
Figure 1. Saltwater contamination sources: (a) slow inltration/leakage from shallow depressions (see picture-1); (b) direct
inltration from the land surface; (c) inltration through open dug wells (see picture-2) and (d), inward movement of the
saltwater-freshwater interface. Photographs: National Science Foundation Tsunami Team, USA and Sri Lanka.
approximately 0.027 g/cm
3
, which is sufciently large to
create gravitational instabilities in the freshwater aquifer
(Schincariol and Schwartz 1990). Density effects have
been studied for storm surges (Andersen et al. 2005), but
similar effects have not been investigated previously in the
case of a tsunami, which occurs as relatively short pulses
of saltwater. Another important difference is the effect of
the signicant ponding of saltwater on the surface dur-
ing inundation (up to several meters), which promotes
inltration into the subsurface.
Another consideration in the case of a tsunami is the
length of time that is required to renew freshwater con-
ditions in the aquifer after the event. Assuming salts to
be conservative in the subsurface, the intrusion is consid-
ered transitory. In eastern Sri Lanka, the monsoon with
heavy rains (November to March) will recharge and poten-
tially renovate the shallow aquifers over some time period.
However, there have been very few studies specically
addressing this question. Sivakumar and Elango (2010)
carried out a three-dimensional (3D) modeling study to
understand tsunami-induced contamination in India and
found that about 4 to 5 years is required to restore fresh-
water conditions. Anderson and Lauer (2008) investigated
increases in total dissolved solids (TDS) in wells ooded
by a storm surge caused by a hurricane. Numerical mod-
eling predicted that it would take more than 10 years to
ush out the saltwater related to the overwash event.
The main objectives of this study are to under-
stand the behavior and the impact of the saltwater ooding
of the aquifer from the tsunami and the factors affecting
the aquifer recovery time. Field observations from eastern
Sri Lanka from 2005 to 2007 were combined with mod-
eling with the variable-density ow and solute transport
code, HST3D (Kipp 1987), to simulate the 2004 saltwater
intrusion related to the tsunami and to predict the time
needed to refresh the aquifer.
Field Investigations
Study Site
The study area includes parts of the eastern coast
of Sri Lanka that consists of elongated coastal lagoons
and associated land stretches oriented in a north-south
direction. The sandy near-shore setting is bordered by
the ocean to the east and brackish lagoons to the west. The
actual Kirankulum eld site is located south of the city
of Batticaloa in eastern Sri Lanka (Figure 2). The study
area is considered part of the dry zone of Sri Lanka with
an annual rainfall of 1200 to 1900 mm (Panabokke and
Perera 2005) with approximately 80% of the total rainfall
occurring from late October to the end of February.
Air temperatures are high year round (30

C to 38

C)
with evaporation estimated to be about 30% to 50%
of precipitation (Panabokke and Pathirana 2003). Fresh
groundwater at the site originates entirely from local
excess precipitation.
The Kirankulum site is natural with no nearby
housing or man-made wells. The land cover is thorny
shrubs and very few trees. The ground surface is almost
at, rising in the middle to an elevation of 3 to 4 m above
2 M. Vithanage et al. GROUND WATER NGWA.org
Figure 2. Map showing locations of the Kirankulum study
site and the piezometer transect used for water table and
water quality monitoring. The site is named after the village
near the site. The latitude and longitude of the site is N 7

36

03

and E 81

47

10

at the sea and in N 7

35

47

and
E 81

45 58 at the lagoon.
the mean sea level (msl). The width of this sand stretch
is 2.4 km oriented in an east-west direction. According to
eye witnesses, the December 2004 tsunami locally ooded
an area up to a distance of 1400 m from the sea, and the
run-up of the largest wave was about 5 to 7 m.
From previous studies, it is known that the upper part
of the coastal aquifers (top 10 m) consists of unlithied
sands with hydraulic conductivities ranging from 0.9
to 1.4 10
4
m/s (Nadarajah 2002). These aquifers are
underlain by impervious hornblende biotite gneiss or
granitic gneiss (Panabokke 2007). The depth to the
bedrock varies between 15 and 28 m below msl with
depressions close to the sea and at the center of the
land stretch (Wickramaratne 2004; Hoareau et al. 2006;
Panabokke 2007).
Measurements and Data
Groundwater levels and electrical conductivity (EC)
were measured monthly in 20 piezometers located along a
transect across the land stretch (Figure 2). The piezome-
ters were installed 9 months after the tsunami (October
2005) when the water table was at its annual low and
prior to the onset of the north-east monsoon. Note that the
tsunami took place in the middle of the rainy season in
2004/2005 and the 2005/2006 north-eastern monsoon was
therefore the rst full rainy season following the tsunami.
The average depth of the piezometers was 3.5 m,
ranging from 1.8 to 5.2 m below the ground surface
depending on the location of the groundwater table
at the time of installation. The screens were installed
1.5 m below the prevailing water table to avoid the
piezometers going dry. The piezometers were constructed
from standard PVC pipes with an inner diameter of 3.7 cm
and equipped with a 15 cm slotted screen and an end-
cap at the bottom. The screen was covered by a 1 mm
nylon mesh to minimize siltation. Hydraulic heads were
measured with a manual dip meter. EC on collected water
samples was measured in the eld using a Hanna hand-
held meter. Three well-casing volumes were purged before
sampling and immediate EC reading.
In addition to the piezometers, three deep wells were
drilled to the bedrock at 500, 1000, and 1600 m from
the sea using a jet boring method. Sediment samples
were extracted at 30 cm depth intervals to estimate
hydraulic conductivity (hydraulic conductivity) and to
assess the heterogeneity/layering of the subsurface. Grain
size distributions were determined from sieve analysis
of 25 representative samples, and hydraulic conductivity
values were estimated from the particle size distributions
using the Hazen method (Fetter 1988). Conventional
slug tests were also performed in 18 piezometers, and
the data were analyzed using the Hvorslev (1951)
method. The thickness of the aquifer and the bedrock
topography were estimated using borehole data and
transient electromagnetic (TEM) measurements (Hoareau
et al. 2006).
Field operations were hampered because of the
limited access and risks associated with the armed conict
in the region at the time. The monitoring program was
halted completely in late 2007 due to these security
increased risks.
Daily rainfall was measured on-site using a gauge
provided by the Meteorology Department of Sri Lanka.
Other meteorological data were collected from the Bat-
ticaloa Meteorological Station (20 km north of the
Kirankulum site). Reference evapotranspiration (ET
r
) for
the Kirankulum site was calculated using the Penman-
Monteith equation along with a temperature-based method
for estimating solar radiation (which was not available),
following the method of Allan et al. (1998). Reference
evapotranspiration was multiplied by an estimated crop
coefcient (1.1) representative of the crops at the site, to
obtain potential evapotranspiration (ET).
Numerical Modeling of Saltwater Intrusion
into the Aquifer
A two-dimensional (2D) cross-sectional model with
dimensions 2.4 km (length), 32 m (depth), and 2 m
(width) was set up with a quasi-rectangular simulation
domain. The lower boundary of the domain was stair-
stepped, reecting the bedrock topography (Figure 3).
The lateral boundary conditions for ow were prescribed
pressure boundaries at the two ends of the domain.
The top boundary was assumed to be a ux (recharge)
NGWA.org M. Vithanage et al. GROUND WATER 3
Figure 3. Two-dimensional simulation domain across the eld site. The layering is conceptualized in terms of seven layers
with different thicknesses (indicated by dotted horizontal lines). Boundary conditions for model setup are indicated.
boundary. The bottom was a no-ow boundary (Figure 3).
The boundaries at the ocean and lagoon sides were
assumed to be vertical. The ocean beach face slope is
rather steep and actually increased after the tsunami as
a result of erosion. On the lagoon side, the beach face
is atter, but because the tsunami wave did not affect
this area a vertical boundary was assumed for simplicity.
The constant hydrostatic pressures applied at these two
boundaries corresponded to mean sea level, assuming an
ocean water density of 1.025 g/cm
3
and a brackish water
density in the lagoon of 1.010 g/cm
3
. For the transport
simulations, the salinities at the boundaries, expressed
as scaled solute mass fractions, were assumed to be 1.0
and 0.04 for the ocean and the lagoon respectively. The
mass fraction for recharge at the top boundary was zero.
A transient rainfall recharge ux at the top boundary
was computed from the one-dimensional unsaturated zone
ow code, HYDRUS1D (

Sim unek et al. 2005) using daily

rainfall and potential ET from October 2005 to September
2006. The daily recharge values were averaged on a
monthly basis (repeated annually) and used as input to
the HST3D groundwater model for simulating saltwater
intrusion and migration in the sand aquifer.
Initial conditions for ow representing the pre-
tsunami situation in December 2004 were dened from
a steady-state simulation on the basis of mean annual
recharge. With this simulation, initially, the whole domain
contained freshwater. A transient simulation was run
to allow salt and brackish water to intrude from both
sides to reach yet another steady state. The subsequent
simulations utilized this result as the initial condition.
The model grid had a variable spacing in both hor-
izontal and vertical directions, with a ner discretization
near the ocean side up to 1500 m inland, near the lagoon
side up to 200 m inland, and in the top 10 m of the
aquifer. The grid blocks for the rest of the model domain
had x = 10 m and z = 1 m. Grid convergence was
tested for the nely discretized domain using several
other meshes (coarse: x = 5 m, z = 1 m; medium:
x = 2 m, z = 0.5 m; ne: x = 1 m, z = 0.5 m;
and nest: x = 1 m, z = 0.25 m). The tests showed
that accurate simulations for ow could be carried out with
the coarse mesh. However, for transport, a ner mesh was
required. The ne grid (206,720 nodes) resulted in a grid
Peclet number of 2 in the x-direction, which is generally
accepted as the limit for minimizing numerical dispersion
(Kipp 1987). The time steps were selected such that the
Courant criterion (Cr < 1) was fullled to reduce numer-
ical errors in the solution.
The calibration of HST3D was carried out in two
steps. First, a base case model with seven horizontal
stratigraphic layers of different thicknesses was calibrated
against transient observations of the water table levels
by multiplying the intrinsic permeability value (specied
based on measured hydraulic conductivity values) for
every layer by the same factor to obtain the best t to the
observed values. The effective porosity (specic yield)
was assumed to range from 0.3 to 0.4. The anisotropy
in permeability (k
z
/k
x
) for all layers was assumed to
be 0.2. In a second calibration step, longitudinal and
transverse dispersivities were adjusted, within reasonable
limits, to obtain a satisfactory t between observed and
simulated salinity distributions in the aquifer. The ow
and transport parameters obtained through calibration are
listed in Table 1.
Simulating the tsunami as a ood wave by imposing
a sudden change in head boundary at the top surface (i.e.,
ponding of saltwater over a certain time period over part
of the model surface) in HST3D is not possible. The code
is not capable of handling two different boundary types
(i.e., ux and head) along the same boundary. Inltration
(recharge) of saltwater due to the tsunami was therefore
approximated as a change in ux at the top of the model
domain. Calculation of the recharge during the tsunami
was carried out sequentially with HYDRUS1D assuming
the maximum wave height varied linearly from 7 m at the
coast to 0 m at the maximum inundation distance of
1400 m (Figure 4a). Initial wave heights were calculated
at 120 m intervals along the transect. At each point,
it was assumed that the wave height decreased from
its initial value to zero over 1 h, which corresponded
to the total time of inundation and wave recession.
HYDRUS1D was used to compute recharge variation
across the transect based on the variation in ponding
4 M. Vithanage et al. GROUND WATER NGWA.org
Table 1
Model Parameters of the Calibrated Base Case Model. Effective Intrinsic Permeability (Geometric Mean)
for All Layers Is 1.96 10
12
m
2
Model Input Parameters Value Units Data Source
Permeability 10
12
m
2
Grain size distribution data
Layer 1 1.58
Layer 2 1.41
Layer 3 6.92
Layer 4 1.20
Layer 5 0.69
Layer 6 1.61
Layer 7 5.26
Hydraulic conductivity m/s Calculated
Layer 1 1.51 10
5
Layer 2 1.38 10
5
Layer 3 6.78 10
5
Layer 4 1.18 10
5
Layer 5 6.76 10
5
Layer 6 1.58 10
5
Layer 7 5.15 10
5
Anisotropy ratio 0.2
Porosity 0.35 Fetter (1988) and sensitivity tests
Density kg/m
3
Freshwater 998.23 Fetter (1988)
Brackish water 1010.0 Measured
Saltwater 1025.0 Fetter (1988)
Viscosity 0.001 kg/ms Fetter (1988)
Diffusivity 1 10
9
m
2
/s Fetter (1988)
Dispersivities

L
0.5 m Calibration

T
0.0625
Recharge varies m/d Estimated from HYDRUS1D simulations
Permeability 10
12
m
2
Grain size distribution data
Layer 1 1.58
Layer 2 1.41
Layer 3 6.92
Layer 4 1.20
Layer 5 0.69
Layer 6 1.61
Layer 7 5.26
Anisotropy ratio 0.2
Porosity 0.35 Fetter (1988) and sensitivity tests
Density kg/m
3
Freshwater 998.23 Fetter (1988)
Brackish water 1010.0 Measured
Saltwater 1025.0 Fetter (1988)
Viscosity 0.001 kg/ms Fetter (1988)
Diffusivity 1 10
9
m
2
/s Fetter (1988)
Dispersivities

L
0.5 m Calibration

T
0.0625
Recharge varies m/d Estimated from HYDRUS1D simulations
height and thickness of unsaturated zone (Figure 4a and
4b), and using the same soil physical parameters as in
the calculations of freshwater recharge. It was further
assumed that the amount of saltwater entering the model
domain was constrained by the available pore space in the
soil column. The effect of density was not considered in
the HYDRUS1D simulations as this effect was assumed
secondary compared to the other assumptions.
Simulations with HST3D were continued until the
saltwater was completely ushed out of the system.
The criterion for complete ushing out was specied
as the time when concentrations everywhere were below
0.018 mass fraction of seawater, which corresponds to
an EC value of 1000 S/cm. Water with concentrations
below this value was considered as pre-tsunami ground-
water or freshwater inltrated after the tsunami.
NGWA.org M. Vithanage et al. GROUND WATER 5
(a)
(b)
Figure 4. (a) Schematic representation of the linear variation of maximum wave height used as the upper boundary condition
for HYDRUS1D simulations; (b) simulated seawater recharge rates during the tsunami as a function of distance from the
sea.
Results and Discussion
Shallow Variations in Hydraulic Conductivity
The grain size distribution results suggest up to 15
distinct layers with some of them showing signicant
contrasts in hydraulic conductivity values (50%). The
slug tests gave higher hydraulic conductivity values
(3 10
5
to 6 10
4
m/s) than those estimated from
the grain size distributions (7 10
6
to 5 10
4
m/s).
For modeling purposes, adjacent layers with only small
differences in hydraulic conductivity were aggregated
to reduce the number of layers in the model. This
simplication resulted in seven layers with hydraulic
conductivity varying from 1.2 10
5
to 7 10
6
m/s
(Table 1 and Figure 3).
Recharge Rates and Water Table Dynamics
From October 2005 to September 2006, recharge was
estimated to 654 mm/year, based on the HYDRUS1D
simulations, or approximately 50% of the 1214 mm
rainfall over this period (Figure 5). The calculated ET
r
ranged 0 to 6 mm/d and averaged 4.0 mm/d (E
p
= 1.1
ET
r
). The maximum daily recharge was 7 mm/d from 13
December, 2005 to 16 January, 2006. The water table
varied in time seasonally. Along the middle of the transect
at the end of the dry season, the water-table was nearly
at. It tended to rise with the onset of recharge, especially
from November to January (Figure 6). The extent of
seasonal uctuations of the water table was 1.75 to 2.0 m
(Figure 6 bottom).
Observed Saltwater Distribution
Background values of EC for water from the aquifer
(<400 S/cm) were estimated from water samples col-
lected from piezometers that were located beyond the
zone of tsunami inundation. Measurements of EC of
groundwater in the tsunami-affected area commenced
immediately after the installation of the piezometers,
some 9 months after the tsunami (i.e., October 2005).
At this time, the observed specic conductances were
already much lower than seawater concentrations. Typ-
ically values were less than 4000 S/cm, as compared
to approximately 50,000 S/cm for seawater. This obser-
vation indicates that convective mixing of seawater and
freshwater was relatively rapid or that saltwater by virtue
of its greater density had rapidly moved below the
piezometer screens by this time. In an area close to
the site, Villholth (2007) monitored the salinity of ground-
water from shallow wells beginning one month after the
tsunami. He found that salinity declined by about a factor
of 10 from levels half that of seawater (20,000 S/cm) to
about 2600 S/cm just 5 months after the tsunami (end of
the rainy season). A rapid decline in groundwater salinity
just 13 days after the tsunami was also reported along the
eastern coast of India (Ravisankar and Poongothai 2008).
6 M. Vithanage et al. GROUND WATER NGWA.org
Figure 5. Monthly rainfall, recharge, and potential evapotranspiration at the eld site from October 2005 to September 2006.
During the next rainy season, November 2005 to
January 2006, a further decline in salinity was evident.
However, the rate of decline slowed during the subsequent
dry season (Figure 7). Water samples by piezometers
located along the middle of the transect exhibited
a slower reduction of salinity compared to samples
collected from piezometers closer to the ocean. Likely
groundwater ow near the mound (middle) was slower
than near the margins. By March 2006, samples from
many of the piezometers still showed EC values greater
than 1000 S/cm. Although the background EC values are
around 400 S/cm, we use 1000 S/cm (or 0.7 ppt) as the
upper limit for potable freshwater, which is close to the Sri
Lankan standard for drinking water quality (750 S/cm)
(Villholth and Rajasooriyar 2010).
The observed recovery of water quality in the shal-
low aquifer corresponds well with results of monitoring
of about 150 shallow drinking water wells (average depth
3.4 m) in neighboring areas. Water samples indicated that
the return to pre-tsunami salinities occurred in approxi-
mately 1.5 years (Villholth 2007). After the onset of the
second full monsoon (October 2006 to February 2007),
rainfall continued to dilute and/or displace saltwater fur-
ther down, leading to EC values below 1000 S/cm.
Modeling of Flow and Transport
Two sets of data are utilized to develop condence
in the results of the simulation. First, the base case
was calibrated to match the observed water table data.
As shown in Figure 6 there is excellent agreement
between observations and simulations (R
2
> 0.93). The
calibrated permeability value in every layer was reduced
by approximately 50% compared to the permeability
values obtained from the grain size distributions (Table 1).
However, such a reduction is not unreasonable given the
uncertainty of estimating k from grain size distributions.
Second, salinity data were used to calibrate the trans-
port model. The best t was obtained with a longitudinal
dispersivity of 0.5 m and a transverse dispersivity of
0.0625 m (Figure 7). The dispersivity values are consis-
tent with values reported from eld experiments in similar
aquifer materials and at a similar scale (Sudicky et al.
1983; Jensen et al. 1993). The simulated salinity was
slightly higher than the observations during some months
(October and December 2005 and August 2007. This dif-
ference could be caused by inaccuracies in specifying
saltwater recharge, instabilities, and unknown heterogene-
ity not included in the conceptual model. However, in
comparison with the uncertainty of the pattern of ini-
tial tsunami inundation, the discrepancies are considered
small. It is interesting to observe an abrupt drop in simu-
lated and observed salinity from 1400 to 1500 m, which
demarcates the tsunami inundation distance. The simu-
lated and observed increased salinity close to the lagoon
could be due to brackish water intrusion (Figure 7).
Simulated patterns of salinity are plotted for at
various times after the tsunami in Figure 8. Note that
resulting instabilities appear vertically elongated due to
the scale exaggeration. Clearly, the spatial density and
depth of observations wells are not sufciently resolved
to verify these types of detailed concentration patterns,
but they are useful for observing the development in
salinity in the shallow groundwater after the tsunami.
Modeling results show that instabilities throughout the
affected area decline throughout the simulation period.
The Peclet criteria (Pe
x
= 2, Pe
z
= 4) were satised with
the ne grid (Voss 1984; Kipp 1987) suggesting that the
instabilities were not a result of numerical instability, but
rather due to density effects.
Owing to freshwater recharge from rainfall and mix-
ing, the salinity of the groundwater declined signicantly
with time. By mid-April, 2005, the simulated maximum
salt concentrations in the groundwater are around 30%
of the seawater salinity (not shown). Although concentra-
tions are quite quickly reduced just after the tsunami, it
takes a long time (15 years) to ush the saltwater from
the entire aquifer and to recover the ambient salinity of
freshwater (EC < 1000 S/cm). However, the upper 6 m
of the aquifer becomes fresh in 5 years (Figure 8). Note
that this estimate is obtained for non-pumping conditions.
Pumping may prolong the recovery time because it dis-
turbs the natural downward movement of the saltwater
NGWA.org M. Vithanage et al. GROUND WATER 7
Figure 6. Observed (symbols) and simulated (solid line) water table variations across the land stretch for different times
(top). Seasonal uctuations of the water table in four piezometers located at 260, 620, 1100, and 1580 m away from the sea
coast (bottom).
8 M. Vithanage et al. GROUND WATER NGWA.org
Figure 7. Comparison of observed and simulated salinity across the transect for selected times. Red symbols indicate salinity
values in piezometers 2 to 20, while blue line represents simulated data.
body due to density effects and enhances mixing par-
ticularly near pumping wells (Vithanage et al. 2009). As
suggested by Oostrom et al. (1992), an unstably stratied
plume can undergo signicant convective mixing, sub-
stantially increasing the volume of contaminated water.
Natural recovery can therefore tend to be slower. In our
case, the salty water also ows downward to the deep-
est parts of the aquifer, rather than migrating laterally
toward the sea. This effect likely contributes to increased
times required to recover the water quality. The piezome-
ters sample water only from the upper part (<4.5 m) of
the aquifer (Figure 6). The plume migrated down quickly
due to its high density and, apparently, only the tail-end
of effects related to the tsunami was captured with the
eld observations.
Vithanage (2009) showed simulation results from
a number of other possible conceptual models assum-
ing homogeneous conditions, different thicknesses of
the aquifer, and different anisotropy ratios. The refresh-
ing time varied from 12 to 20 years with the shortest
recovery time for the same layered system, but with a
lower anisotropy ratio. In that system, the higher per-
meability in the horizontal ow direction forced quicker
ushing toward the coast. The longest refreshing time was
also observed with the layered system, same anisotropy
as the base case, but with a thicker aquifer, where the
bottom has been assumed to be at at 28 m msl. The
homogeneous cases fall in between these two systems with
refreshing times between 13 and 18 years.
Conclusions
A monitoring program in a shallow sandy aquifer in
a rural area on the east coast of Sri Lanka showed that
shallow groundwater was affected 1 to 1.5 years after
the 2004 tsunami up to 1.5 km inland from the coast.
Subsequent recharge to the aquifer from the monsoon has
been effective in ushing the saltwater and recovering the
water quality of the aquifer.
2D variable-density ow and transport modeling was
used to evaluate the time required for complete recovery to
pre-tsunami quality conditions. The model was calibrated
by matching the observed dynamics of the water table and
observations of the salinity in the shallow groundwater
from 2005 to 2007. Unfortunately, the armed conict
in the region stopped the monitoring program in 2007.
The results of the calibration show good agreement
with observed data despite lack of observations in the
NGWA.org M. Vithanage et al. GROUND WATER 9
Figure 8. Simulated saltwater distribution from October 2005 to December 2015. The layered stratigraphy and bedrock are
not shown in the gures. The blue zones show EC levels higher than 1000 S/cm (>0.018 mass fraction). Symbols represent
the screen depths of the piezometers.
months immediately following the tsunami and limited
knowledge of specic details of saltwater loading to the
aquifer. The simulations show a rapid initial decrease in
salinity just after the tsunami, which are in agreement
with observations due to the ow of the denser uid
and subsequent recharge by freshwater. The simulations
suggest that it takes about 5 years to recover the ambient
water quality in the upper 6 m of the aquifer and that the
aquifer will recover totally through ushing of saltwater
approximately within 15 years.
Acknowledgments
The authors wish to thank J. Hoareau at Pierre
and Marie Curie University (France), P. Jeyakumar, N.
Amalraj, and S. Prathepaan at the Eastern University of
Sri Lanka and L. Chandrarajah at the village close to the
eld site for their generous support. Franklin W. Schwartz
is acknowledged for his kind review and his interest in this
work. We are also grateful for helpful comments from the
journal reviewers.
References
Allan, G.R., L.S. Pereira, D. Raes, and M. Smith. 1998. Crop
Evapotranspiration; Guidelines for Computing Crop Water
Requirements. FAO Irrigation and Drainage Paper 56.
Rome: FAO.
Andersen, M.S., V.N.R. Jakobsen, and D.K. Postma. 2005.
Geochemical processes and solute transport at the seawater/
freshwater interface of a sandy aquifer. Geochimica et
Cosmochimica Acta 69, no. 16: 39793994.
Anderson, W.P. Jr., and R.M. Lauer. 2008. The role of overwash
in the evolution of mixing zone morphology within barrier
islands. Hydrogeology Journal 16, no. 8: 14831495.
Fetter, C.W. 1988. Contaminant Hydrogeology. Upper Saddle
River, New Jersey: Prentice-Hall Publishing Company.
Hoareau, J., R. Guerin, J.M. Vouillamoz, and M. Vithanage.
2006. Sustainable development of water resources, water
supply and environmental sanitation. 32nd WEDC Interna-
tional Conference, Colombo, Sri Lanka, 2006.
Hvorslev, M.J. 1951. Time lag and soil permeability, in ground-
water observations. U.S. Corps Engineers Waterways
Experiment Station, Vicksburg, Mississippi, Bulletin no.
36, 50.
Illangasekare, T., S.W. Tyler, T.P. Clement, K.G. Villholth,
A.P.G.R.L. Perera, J. Obeysekera, A. Gunatilaka, C.R.
Panabokke, D.W. Hyndman, K.J. Cunningham, J.J. Kalu-
arachchi, W. W-G. Yeh, M.R. Van Genuchten, and K.H.
Jensen. 2006. Impacts of the 2004 tsunami on groundwater
resources in Sri Lanka. Water Resources Research W05201.
DOI: 10.1029/2006WR004876.
Jensen, K.H., K. Bitch, and P. L. Bjerg. 1993. Large-scale
dispersion experiments in a sandy aquifer in Denmark:
Observed tracer movements and numerical analyses, Water
Resources Research 29, no 3: 673696.
Kipp, K.L. 1987. HST3D: A computer code for simulation of
heat and solute transport in three dimensional groundwater
ow systems. Water resources investigation report 86-4095,
USGS, Denver, Colorado.
Liu, P.L., P. Lynett, H. Fernando, B. Jaffe, H. Fritz, B. Higman,
R. Morton, J. Goff, and C. Synolakis. 2005. Observations
by the International Tsunami Survey Team in Sri Lanka,
Science 308: 1595.
Nadarajah, K. 2002. Well inventory in the Nilaveli-Kuchchaweli
area. Technical Paper 17. Integrated Agricultural Food
Security Program. Trincomalee, Sri Lanka.
Oostrom, M., J.S. Hayworth, J.H. Dane, and O. G uven. 1992.
Behavior of dense aqueous phase leachate plumes in
homogeneous porous media. Water Resources Research 28,
no. 8: 21232134.
10 M. Vithanage et al. GROUND WATER NGWA.org
Panabokke, C.R. 2007. Groundwater Conditions in Sri Lanka;
A Geomorphic Perspective. Colombo, Sri Lanka: National
Science Foundation of Sri Lanka.
Panabokke, C.R., and A.P.G.R.L. Perera. 2005. Groundwater
resources of Sri Lanka. World Water Assessment Program.
Sri Lanka.
Panabokke, C.R., and S.R.K. Pathirana. 2003. Water quality
monitoring in the coastal area of Tincomalee District.
Technical Paper 16. Integrated Agricultural Food Security
Program. Trincomalee, Sri Lanka. Rep. 84-4369.
Ravisankar, N., and S. Poongothai. 2008. A study of ground-
water quality in tsunami affected areas of Sirkazhi
Taluk, Nagapattinum District, Tamilnadu, India. Science of
Tsunami Hazards Journal 27, no. 1: 4755.
Schincariol, R.A., and F.W. Schwartz. 1990. An experimental
investigation of variable density ow and mixing in
homogeneous and heterogeneous media. Water Resources
Research 26, no. 10: 23172329.

Sim unek, J., M.T. Van Genuchten, and M.

Sejna. 2005. The
HYDRUS-1D Software Package for Simulating the Move-
ment of Water, Heat, and Multiple Solutes in Variably
Saturated Media, Version 3.0, HYDRUS Software Series 1,
Department of Environmental Sciences, University of Cal-
ifornia Riverside, Riverside, California.
Sivakumar, C., and L. Elango. 2010. Application of solute
transport modelling to study tsunami induced aquifer
salinity in India. Journal of Environmental Informatics 15,
no. 1: 3341
Sudicky, E.A., J.A. Cherry, and E.O. Frind. 1983. Migration of
contaminants in groundwater at a landll: A case study, 4,
A natural gradient dispersion test. Journal of Hydrology 63,
no. 12: 81108.
Villholth, K.G. 2007. Tsunami impacts on groundwater and
water supply in eastern Sri Lanka. Waterlines 26, no. 1:
811.
Villholth, K.G., and L.D. Rajasooriyar. 2010. Groundwater
resources and management challenges in Sri Lanka -
An overview. Water Resources Management 24, no. 8:
14891513. DOI: 10.1007/s11269-009-9510-6.
Villholth, K.G., P.H. Amerasinghe, P. Jeyakumar, C.R. Pana-
bokke, O. Woolley, M.D. Weerasinghe, N. Amalraj, S.
Prathepaan, N. B urgi, D.M.D.S. Lionelrathne, N.G. Indra-
jith, and S.R.K. Pathirana. 2005. Tsunami Impacts on Shal-
low Groundwater and Associated Water Supply on the East
Coast of Sri Lanka. Colombo, Sri Lanka.: International
Water Management Institute (IWMI). ISBN 92-9090-622-7,
68.
Vithanage, M. 2009. Effect of tsunami on coastal aquifers:
Field studies and tank experiments, PhD thesis, Faculty of
Science, University of Copenhagen, Denmark.
Vithanage, M., K.G. Villholth, K. Mahatantila, P. Engesgaard,
and K.H. Jensen. 2009. Effect of well cleaning and
pumping on groundwater quality of a tsunami-affected
coastal aquifer in eastern Sri Lanka. Water Resources
Research 45: W07501. DOI: 10.1029/2008WR007509.
Voss, C.I. 1984. Sutra: Saturated-unsaturated transport, a nite
element simulation model for saturated-unsaturated uid-
density-dependent groundwater ow with energy transport
or chemically-reactive single-species solute transport. U.S.
Geological Survey Water Resources Investigation Report
84-4369.
Wickramaratne, H.U. 2004. Use of groundwater for coastal water
supply schemes, minimizing the salinity problem; A case
study at Kattankudy in Batticaloa District, Sri Lanka. MSc
thesis. Post Graduate Institute of Science, University of
Peradeniya, Peradeniya, Sri Lanka.
NGWA.org M. Vithanage et al. GROUND WATER 11