EFFECT OF THE INDIAN OCEAN TSUNAMI ON GROUNDWATER QUALITY IN COASTAL AQUIFERS IN EASTERN SRI LANKA

EFFECT OF THE INDIAN OCEAN TSUNAMI ON GROUNDWATER QUALITY IN
COASTAL AQUIFERS IN EASTERN SRI LANKA

METHTHIKA VITHANAGE1,2,
KAREN G. VILLHOLTH3,
KUSHANI MAHATANTILA4,
PETER ENGESGAARD1 AND
KARSTEN H. JENSEN1

1Department of Geography and Geology, University of Copenhagen, Denmark,
2International Water Management Institute, Battaramulla, Sri Lanka,
3Geological Survey of Denmark and Greenland, Copenhagen, Denmark
4Department of Material Science, Interdisciplinary Faculty of Science and Engineering, Shimane
University, Matsue, Japan,

meththikavithanage@gmail.com
Tele: +94 112 833 101


ABSTRACT

Changes in water quality of a sand aquifer on the east coast of Sri Lanka due to the December 26,
2004 tsunami and subsequent disturbance due to well pumping and flushing by precipitation were
investigated. Two closely spaced tsunami affected transects, spanning the ocean and an interior
lagoon across a 2 km wide land strip were monitored from October, 2005 to September, 2006.
Water samples were collected from 15 dug wells and 20 piezometers, from the disturbed and
undisturbed sites respectively to evaluate the temporal and spatial trends in water quality.
The EC values observed from the undisturbed area showed a significant decrease (3000 to 1200
µS/cm) with the rain from November 2005 to March 2006, while the values in the disturbed area
appeared to have stabilized without further decline through the same period. The concentration
range of EC, Ca, K, Na, alkalinity, total hardness and sulphate were higher in the disturbed site than
in the undisturbed site. PHREEQC modeling showed that the mixed sea water fraction is higher in
the disturbed site than in the undisturbed site, and this is likely due to the movement of the
disturbed plume by water extraction through pumping and extensive well cleaning after the tsunami,
causing forced diffusion and dispersion. No arsenic contamination was observed as all observed
arsenic concentrations were below 10 µg/L. For the sites investigated, there are clear indications of
only a slow recovery of the aquifer with time in response to the onset of the monsoon.

Key words: Tsunami, water quality, groundwater, coastal aquifers, geochemical modelling, Sri
Lanka



Science of Tsunami Hazards, Vol. 28, No. 3, page 218 (2009)

1

---

INTRODUCTION

On December 26th 2004, tsunami waves devastated the coastal region of Sri Lanka. Coastal belt
of Sri Lanka comprises of shallow coastal sandy aquifers, which provides reliable water resources
throughout the year. The severely affected east coast of Sri Lanka consists of regosols overlying
directly unconfined sandy aquifers ranging from fine to moderately coarse structure-less sands
(Panabokke and Perera, 2004). These shallow groundwater aquifers are very susceptible to
contamination. When a dense fluid is situated above the less dense fluid it comprises of an unstable
density configuration and is referred to as an unstable-interface problem. Previous studies
(Schincariol and Schwartz, 1990; Koch and Zhang, 1992; Oostrom et al., 1992; Simmons et al.,
2001) have revealed that when density differences are significant, solute transport is a result not
only of forced (hydraulically driven) advection and dispersion or diffusion, but also of free
convection (buoyancy driven) resulting in gravitational instabilities. Therefore they have the
potential to contaminate larger regions of an aquifer and can mix naturally with the ambient water
to a larger degree due to mixed convection (Oostrom et al., 1992). The anthropogenic processes that
aggravate mixing (such as water abstraction) could result longer aquifer cleansing time compared to
the contamination by natural mixing.
People use shallow open dug wells for their domestic water supply as the surrounding sandy
aquifer. However, as a result of the tsunami the coastal belt was completely flooded and the shallow
groundwater aquifer was contaminated by saltwater. Open dug wells in the affected areas up to 1.5
km inland were filled with seawater (Villholth et al., 2006a, 2006b). It has been estimated that more
than 50,000 water supply wells were affected in Sri Lanka by the tsunami (UNEP, 2005). The
diffuse contamination of ground water and the point source contamination of wells along the
affected coastal zone were aggravated by the influx of wastes from septic materials, pit latrines,
chemical spills, dead bodies and surface debris. The waste and debris directly contaminated the
wells or accumulated in the surface depressions and infiltrated into to the aquifer for several days
following the event. This study investigates the change in water quality of the highly vulnerable
local sand aquifers due to tsunami and its “natural cleaning” response by precipitation over time and
the effect of well cleaning and pumping water for domestic uses.

Study area

Location

The present study was carried out in a tsunami affected village located in the Batticaloa District of
eastern Sri Lanka (Fig. 1). The east coast is prevalent with elongated, north-south oriented coastal
lagoons with land stretches bordered by saline ocean water and brackish lagoon water on both sides.
Hydrogeologically, these land stretches act as oceanic islands and local, productive aquifers which
are created as freshwater lenses floating on the underlying denser saline water (Underwood et al.,
1992). The study area extends as an east west oriented narrow strip of land about two kilometers
wide. The ground surface is almost flat and the elevation in the middle part between the sea and the
lagoon is about 3 to 4 meters above mean sea level (msl). Freshwater at the site originates entirely
from local excess precipitation and the available quantity is thus limited.

Science of Tsunami Hazards, Vol. 28, No. 3, page 219 (2009)

2

---

Exploitation of the freshwater poses a risk of contamination by sea water due to over extraction
and hence the fresh water lens needs to be used in a sustainable manner as there are limited reliable
freshwater sources inland to supplement water for the coastal zone.


Fig. 1. Location map of the study area. Piezometer transect (undisturbed site) is in 2 km
demarcated by the sea (70 36' 24'' N and 810 47' 01") in the east and Lagoon (70 35' 59'' N and 810
45' 50") in the west. The village considered as the disturbed site.

Science of Tsunami Hazards, Vol. 28, No. 3, page 220 (2009)

3

---

Two transects are located in the Kirankulum village of the Batticaloa District. The two sites have
similar characteristics with the major difference being that one of them is only exposed to natural
rainfall conditions and no pumping activities present (hereafter referred to as the undisturbed site)
while the other site is subject to well pumping (referred to as the disturbed site). The undisturbed
site consists of thorny shrubs and cashew plantation, while the adjacent village, which consists of a
cluster of small one-story houses and relatively little paved areas with numerous dug wells for water
supply, is considered as the disturbed site.
Twenty piezometers were installed along a transect perpendicular to the coast and across the
entire land strip in the undisturbed site. Fifteen existing open dug wells from the village along a
second transect, covering a smaller section of the land strip, were selected for the detailed study
(Fig. 1). The two transects are 0.5 km apart. The disturbed site consists of a total of 57 dug wells
within an area of 0.25 km2 used by 113 families. Fifty wells in this area were affected by the
tsunami. Most of these wells are used for household purposes. In both sites, the tsunami run-up was
around 1 km from the coast line. In the village transect, only two dug wells were in the tsunami
non-affected area, while the undisturbed transect consists of 12 piezometers in the affected area and
8 piezometers in the non-affected area.

Climate

The study area belongs to the dry zone of Sri Lanka with an annual rainfall of 1200 to 1900 mm.
Eighty percent of the annual rainfall falls intensively during the north-east monsoon period from
November to February and during the inter monsoon period from March to April (Panabokke et al.,
2002). As the ambient temperature is high year round (30 – 35 0C) evaporation is estimated to be
high, around 30 – 50 % of the precipitation (Panabokke, 2001).

Geology and hydrogeology of the study area

In Sri Lanka, Miocene limestone are found along the northern and North West coast and
throughout the Jaffna Peninsula, while on the southeast, south and southwest coast, rocky headlands
are prevalent. The geology around the study area is dominated by Quaternary unconsolidated beach
deposits on the top of Precambrian metamorphic rocks Cooray (1985).
The two areas are assumed and expected to be similar in the lithological settings. The lithology
of a field site about 15 km north of the study area was reported by Wickremaratne (2004), first, a
medium to coarse sand layer of approximately 9 m depth with a small fraction of clay and shell
fragments is found.
A clay and silty clay layer of about 5-10 m in thickness is found below the sand layer, which is
then followed by a thin coarse sand layer of about 2 m in thickness. The thickness of the sandy
formations increases towards sea whereas the thickness of the clayey layers increases towards the
lagoon. It was reported that the bed rock is found around 15-25 m deep which is corroborated by
transient electromagnetic soundings carried out by Hoareau et al., (2006) in the particular study area.


Science of Tsunami Hazards, Vol. 28, No. 3, page 221 (2009)


4

---

Pre-tsunami hydrogeochemistry

Few studies have been carried out to investigate the pre-tsunami water quality in the shallow
ground water aquifers in eastern Sri Lanka (Mahendran et al., 2000; Vaheesar, 2002; Panabokke et
al., 2002; Jeyakumar et al., 2002; Sugirtharan et al., 2004; Wickramaratne, 2004). Some of the
recorded EC levels were relatively high in the pre-tsunami studies. According to Panabokke et al.,
(2002) only in two agro wells out of 25 from the Trincomalee District (located north of the study
area) had low EC values (<500 µS/cm) and most were above 1000 µS/cm while some were as high
as 2500 µS/cm). However, as agro wells have become larger in size and pumped more heavily to
supply water for irrigation, they may not be considered as representative of the domestic wells. As a
consequence, agro wells have shown consistently higher salinity levels than domestic dug wells
(Vaheesar et al., 2000; Jeyakumar et al., 2002). In the study by Wickramaratne (2004), EC levels
below 700 µS/cm were reported for shallow tube wells. Calcium and magnesium concentrations of
well water have been recorded between 250 – 1450 mg/L and 30 – 250 mg/L respectively
(Mahendran et al., 2000). Nitrate concentration in most areas in the Batticaloa District are above the
WHO standard limit (45 mg/L).

Post tsunami water quality

The shallow groundwater aquifers were affected in several ways by the tsunami: direct
infiltration from tsunami inundation, infiltration from pooled sea water left behind, permeation from
flooded wells, and salinization from flooding of lagoons and river mouths (Villholth et al., 2006a).
A few studies have documented the impact on the water quality in wells mainly in terms of
electrical conductivity (EC), total dissolved solids (TDS) and salinity (Villholth et al., 2006a;
Illangasekare et al., 2006). Water quality measurements carried out by Villholth et al., (2006b) have
documented a significant difference between the tsunami affected and non-affected well water.
Interestingly, Pilapitiya et al., (2005), have reported on elevated arsenic contamination due to
tsunami in Sri Lanka

METERIALS AND METHODS

Twenty piezometers were installed along a 2 km transect with 500 m mutual distance in the
undisturbed site. All these piezometers were installed just prior to the onset of the north east
monsoon when the water table was the lowest. The depth of piezometers varied along the transect
according to the level of the water table and they were installed 1.5 m below the water table to
avoid the possibility of being dry. The piezometers were made of standard pvc tubes with an inner
diameter of 3.7 cm and equipped with a 15 cm slotted screen at the bottom and with an end cap,
which was covered by a 1 mm mesh to protect the piezometer from siltation. They were installed in
8.7 cm diameter boreholes constructed by slug boring method at the proper level and securely
sealed. Fifteen existing open dug wells along a transect were selected from the village (disturbed
site) for detailed monitoring of water quality (Fig. 1). The wells had an average diameter of 1.3 m
and an average depth of 3.2 m. Rainfall at the site was recorded daily at 08:00 using a standard rain
gauge provided by the Meteorological Department of Sri Lanka.

Science of Tsunami Hazards, Vol. 28, No. 3, page 222 (2009)

5

---

Chemical Analysis

EC was measured in the field in both dug wells and piezometers using a Hanna hand held EC
meter, during the months of October, 2005, January, March, June and October, 2006. All 57 wells
in the village were monitored for EC, however, results given in this paper for the dug wells are all
from the 15 wells along the transect.
In January, March and October 2006, duplicate well water samples were collected for chemical
analysis. One sample of water was acidified to measure the metal ions and the other was frozen.
Nitrates, phosphates and sulphates were measured using a HACH 2400 UV-VIS Spectrophotometer
and the inorganic ions, Ca, Mg, Na, K, Fe an Mn were measured using an Atomic Absorption
Spectrophotometer (Perking Elmer, 2300).
In-situ arsenic analysis was carried out using the Arsenic measuring field kit (VISUAL
ARSENIC DETECTION KIT WAG-WE 10600). In this case only water in open dug wells, which
may have been filled with oceanic sediments by the tsunami were analyzed.

Geochemical Modeling

The water quality data sets obtained from all the sites were analyzed using the geochemical
model, PHREEQC (Parkhurst and Appelo, 1999), The PHREEQC code simulates chemical
reactions and transport processes in water using observed field data. The model has an inverse
modeling capability, which attempts to find sets of mineral and gas mole transfers that account for
differences in composition between waters while accounting for specified uncertainties (Parkhurst
and Appelo, 1999).
Our intention was to use inverse modeling to back calculate and reproduce the data obtained
from the water quality analysis, and determine the composition of the water samples. In inverse
modeling, one aqueous solution is assumed to be mixed with the other aqueous solutions to react
with mineral and gas phases to produce the observed composition of the tertiary solution and all the
mixing fractions and mole transfers of the phases are calculated (Parkhurst and Appelo, 1999).
Assumed here are three water analyses, sea water (from the PHREEQC data base), fresh water
(from none affected dug wells) and water in the affected area (water quality observations) subject to
mixing and ending with a certain water composition along the flow path.
The numerical method of inverse modeling of PHREEQC is based on the mole balance
equations, which are included for (a) each element, (b) alkalinity, (c) electrons, (d) water and (e)
each isotope. Furthermore, it includes inequalities to constrain the uncertainty (Parkhurst and
Appelo, 1999).
The uncertainty represents potential analytical error and spatial or temporal variability in
concentration of each component in each aqueous solution and in mineral compositions.
Uncertainty in the aqueous species in the models was set in between 0.1-0.2 to make the model
converge with reasonable results. To consider uncertainties in mineral compositions, three mineral
phases (calcite, dolomite and anhydrite) and one gas phase (CO2) were added to the models.


Science of Tsunami Hazards, Vol. 28, No. 3, page 223 (2009)

6

---

RESULTS AND DISCUSSION

Precipitation

The precipitation data recorded at the field site is show (Vithanage et al., 2009) that the rainfall
due to the onset of the north-east monsoon period in October with the highest rainfall in November.
Nearly 90% of total annual rainfall precipitates during the north-east monsoon period from
November to February.

Electrical Conductivity (EC)

The tsunami impact on chemical groundwater characteristics was significantly manifested in the
salinity levels. Because there were very few records of pre-tsunami EC levels and chemical
parameters of shallow ground water (Mahendran et al., 2000; Vaheesar, 2002; Wickaramaratne,
2004) the data from the present study of the areas not flooded by the tsunami, basically the 8 and 2
most inland wells for the undisturbed and disturbed site, respectively (Fig. 2), were considered
representative of the original water quality of the local aquifer system. The original EC levels of the
Kirankulum study area, except for the well closest to the lagoon, which could have been affected by
the tsunami by the seawater flooding directly into the lagoon mouths (Villholth et al., 2006b), were
relatively lower than the data from previous reports, generally < 300 µS/cm.
It is noted that the EC levels showed more spatial variability across the transect in the disturbed
site than in the undisturbed site (Fig. 2).

Fig. 2. Change in EC levels both in undisturbed (solid line) and disturbed sites (circles) in October,
2005, January, February, March, May and September, 2006. Red circles represent dug wells that
are pumped for cleaning purposes and green circles represent abandoned dug wells.

Science of Tsunami Hazards, Vol. 28, No. 3, page 224 (2009)

7

---

This could indicate that different patterns of use of individual wells after the tsunami had meant
a different degree of residual tsunami imprint. This pattern of larger small-scale variability within
the disturbed transect than the undisturbed one was also evident in most of the other chemical
variables, supporting the hypothesis that the degree of reminiscence of the tsunami impact locally is
a function of the aquifer disturbance in terms of well pumping and cleaning going on in the wells
since the tsunami. Intensive and frequent well cleaning through emptying of wells and chlorination
had been carried out since shortly after the tsunami. Furthermore, the most prevalent cleaning
procedure was simple and entailed discharging the well water next to the well and hence basically
recycling the contaminated water into the system and increasing the overall mixing and retarding
the downward and lateral movement of the tsunami pulse. Hence, well cleaning, abstraction of
water and degradation of debris may have delayed the natural recession of the tsunami effect at the
disturbed site. In the undisturbed site, natural flow process had taken place, encompassing rapid
sinking of the denser saltwater immediately after the tsunami (Illangasekare et al., 2006) and hence,
the tsunami effect, after about one and half years, was less and more uniform across the transect
compared to the disturbed site. Further supporting this hypothesis is the fact that well no. 1, 4, and
11 in the disturbed site were abandoned since the tsunami and these wells showed lower values of
the chemical parameters.

Water Quality Analysis

The values obtained from the water quality analysis were normalized according to the sea water
and plotted against time. The results of water quality analysis in January, 2006 showed that Ca
concentrations in dug wells are slightly higher than in the piezometers (Fig. 3).

Fig. 3. Normalized calcium concentrations in the undisturbed site (P), dug wells in use (DWU) and
abandoned dug wells (DWA) for January, March and October, 2006

Science of Tsunami Hazards, Vol. 28, No. 3, page 225 (2009)

8

---

Sodium concentrations for undisturbed site were almost same (>100 mg/L) as for the disturbed
site in January. In March the concentrations decreased for both sites, most notably for the
undisturbed site (Fig. 4). The original Na concentration in the aquifer was below 10 mg/L, and still
most of the dug wells had much higher Na concentrations in March, 2006 (several above 100 mg/L)
which is a clear imprint of the tsunami.

Fig. 4. Normalized sodium concentrations in the undisturbed site (P), dug wells in use (DWU) and
abandoned dug wells (DWA) for January, March and October, 2006

The behavior of potassium concentrations was similar to Na (Fig. 5). The original K levels were
less than 4 mg/L, and as shown by the figure many of the dug wells were higher in concentrations.
Similar to sodium the concentrations are generally higher than for the piezometer measurements
except for the two piezometers adjacent to the sea and lagoon respectively.

Fig. 5. Normalized potassium concentrations in the undisturbed site (P), dug wells in use (DWU)
and abandoned dug wells (DWA) for January, March and October, 2006

Science of Tsunami Hazards, Vol. 28, No. 3, page 226 (2009)

9

---

The measured concentrations for alkalinity corroborate the findings for the other chemical
parameters that the concentration levels for the disturbed site are marginally higher than for the
undisturbed site (Fig. 6). Manganese also shown a similar behavior, and conversely, iron
concentrations were high in undisturbed environment. Manganese from the sea sediments, which
has deposited in the dug wells may have played a major role here causing the difference in between
affected and non-affected area of the village transect. Though Pilapitiya et al., (2005) has measured
elevated arsenic contaminations in groundwater due to tsunami there were no sign of arsenic in the
water samples tested onsite. All the samples showed concentrations below 10 µg/L (WHO
recommended guidelines).

Fig. 6. Alkalinity concentrations in the undisturbed site (P), dug wells in use (DWU) and
abandoned dug wells (DWA) for January, March and October, 2006

Hydrogeochemical Modeling of data

PHREEQC was used to model the chemical datasets from both undisturbed and disturbed sites to
validate the analytical work and obtain the mixed sea water and fresh water fractions. Different
mineral and gas phases were added according to the inequalities obtained in the output. Some
samples showed that they are strongly undersaturated with solid sulphates and iron, and saturated
with respect to calcite and dolomite. Therefore, possible precipitation of calcite or dolomite and
dissolution of anhydrite were set as model constraints. For those solutions, it was necessary to add
redox phases such as goethite or pyrite. A possible sink for CH2O and a CO2 source were set for
equilibrate the reactions with organic matter. Since the upper most layer of the aquifer is
geologically characterized as coarse to fine sand and no clay minerals are present, cation exchange
was not considered. Sea water composition was obtained from the PHREEQC data base. The
observed values of pH, alkalinity, C, Cl, Mg, Na, S and Ca were used as input to include mole
balance equations for elements not contained in any of the phases.

Science of Tsunami Hazards, Vol. 28, No. 3, page 227 (2009)

10

---