N. Doerfliger (Y)
BRGM National geological survey, F 34000 Montpellier, France
P.-Y. Jeannin 7 F. Zwahlen
Center of Hydrogeology, Institute of Geology, University of
Neuchâtel, 11 rue E-Argand, CH-2007 Neuchâtel, Switzerland
Water vulnerability assessment in
karst environments: a new method
of defining protection areas using
a multi-attribute approach and GIS
tools (EPIK method)
N. Doerfliger 7 P.-Y. Jeannin 7 F. Zwahlen
Abstract Groundwater resources from karst aquif-
ers play a major role in the water supply in karst
areas in the world, such as in Switzerland. Defining
groundwater protection zones in karst environment
is frequently not founded on a solid hydrogeologi-
cal basis. Protection zones are often inadequate and
as a result they may be ineffective. In order to im-
prove this situation, the Federal Office for Environ-
ment, Forests and Landscape with the Swiss Na-
tional Hydrological and Geological Survey con-
tracted the Centre of Hydrogeology of the Neuchâ-
tel University to develop a new groundwater pro-
tection-zones strategy in karst environment. This
approach is based on the vulnerability mapping of
the catchment areas of water supplies provided by
springs or boreholes. Vulnerability is here defined
as the intrinsic geological and hydrogeological
characteristics which determine the sensitivity of
groundwater to contamination by human activities.
The EPIK method is a multi-attribute method for
vulnerability mapping which takes into considera-
tion the specific hydrogeological behaviour of karst
aquifers. EPIK is based on a conceptual model of
karst hydrological systems, which suggests consid-
ering four karst aquifer attributes: (1) Epikarst, (2)
Protective cover, (3) Infiltration conditions and (4)
Karst network development. Each of these four at-
tributes is subdivided into classes which are map-
ped over the whole water catchment. The attributes
and their classes are then weighted. Attribute maps
are overlain in order to obtain a final vulnerability
map. From the vulnerability map, the groundwater
protection zones are defined precisely. This method
was applied at several sites in Switzerland where
agriculture contamination problems have frequently
occurred. These applications resulted in recom-
mend new boundaries for the karst water supplies
protection-zones.
Key words Karst aquifer 7 Groundwater 7 Aquifer
protection 7 Vulnerability 7 GIS
General background
Introduction
Karst aquifers are generally considered to be particularly
vulnerable to pollution, because of their unique structure.
This structure is strongly heterogeneous. It can be con-
sidered as a network of conduits of high permeability
surrounded by large volumes of low permeability rock.
Recharge occurs by both dispersed and concentrated wa-
ter entry. This implies that a fair amount of the recharge
infiltrates directly into the conduit network so that atte-
nuation of contaminants does not occur effectively as in
porous aquifers. It can be pointed out that the percentage
of concentrated recharge increases with increasing re-
charge, therefore, the water quality can still be good at
low water level stage.
Applying the present Swiss regulations for defining water
supplies protection-areas in karst environments leads to
many protection areas which are too large and not perti-
nent. Porous media protection areas are outlined by a 10
days transit time limit (zone 2). Due to the heterogeneity
of groundwater velocity in karst environment and to the
impossibility for cost purpose to carry out many tracer
experiments, zone 2 should include the whole catchment.
As land use restrictions in zone 2 are too restrictive, in
most cases the whole catchment is assigned to zone 3.
Consequently zone 2 is either not present or too small.
As a result, water quality problems have occurred. For
this reason, the concept of vulnerability mapping using a
multi-attribute approach, the EPIK method, has been de-
veloped. The EPIK method was developed to assess the
Published in Environmental Geology 39, issue 2, 165-176,1999
which should be used for any reference to this work
1
intrinsic vulnerability of groundwater to surface contami-
nation and to provide a tool to define the protection
zones in karst environments for hydrogeological consult-
ing. (Intrinsic vulnerability represents the inherent hy-
drogeological and geological characteristics which deter-
mine the sensitivity of groundwater to contamination by
human activities. Intrinsic vulnerability refers essentially
to risk associated to non-point sources. Intrinsic vulnera-
bility as opposed to specific vulnerability considers all
kind of contaminants.) The most recent version is an up-
date of an earlier paper about EPIK (Doerfliger and
Zwahlen 1995). It now includes new values of weighting
and ranking. This EPIK method is an overlay weighting
and rating method similar to that of DRASTIC developed
by Aller and others (1987).
In many countries vulnerability maps are established on
a large scale, e.g. regional, county or state scale. To our
knowledge, EPIK is the first method to suggest vulnera-
bility mapping of karst systems at catchment scale, which
allows the determination of groundwater protection ar-
eas. Based on a hydrogeological conceptual model of
karst hydrological systems, EPIK is a method that takes
into account the most significant parameters considered
in this model. Some results of vulnerability mapping and
protection areas based on the EPIK approach on one test
site in Swiss Jura are presented in this paper.
Current Swiss groundwater protection legislation
The Swiss legislation to protect water – 1991 Federal Law
on Water Protection – requires defining protection zones
in the vicinity of public drinking water supplies. Accord-
ing to the regulations, three different zones have to be
delineated:
1. S1 zone: This zone has to protect the water supply
structure against damage and to prevent any direct
penetration of contaminants into the groundwater.
2. S2 zone: this zone has to provide protection against
microbiological and non-degradable contaminants and
to allow enough time for intervention in the case of an
accident.
3. S3 zone: this zone has to provide additional safety. It
could in many cases, correspond to the rest of the
catchment not covered by zones S1 and S2.
The three zones are subject to specific land-use restric-
tions according to the protection objectives. Today, the
definition of the protection zones for all water-catch-
ments in Switzerland is almost completed. But sadly, de-
spite this important effort, it is clear that the protection
of the water-catchments in karst environment still re-
mains insufficient. The protection zones are only partly
effective as water quality problems arise frequently due to
agricultural and industrial pollution (Doerfliger and oth-
ers 1997).
In most European countries three types of protection ar-
eas are distinguished, but their definition has no unifor-
mity. The immediate area is often a 10-m radius area
around the spring or well, sometimes including swallo-
wholes within the catchment. The inner protection zone
is often based on water transit time (10–100 days de-
pending on countries), it can also enclose some preferen-
tial infiltration areas. The outer protection area includes
the rest of the catchment or at least a 2 km or 400 days
transit time limit.
Definition of vulnerability mapping
The term vulnerability was used in the sixties in France,
introduced as a scientific term in the specific literature
by Albinet and Margat (1970). Since then, several defini-
tions of vulnerability have been presented in the techni-
cal literature. In the scope of the development of the
EPIK approach, the vulnerability term is defined, as fol-
lowing:
Intrinsic vulnerability represents the inherent hydrogeo-
logical and geological characteristics which determine the
sensitivity of groundwater to contamination by human
activities.
According to Foster and Hirata (1988), Adams and Foster
(1992) and Robins and others (1994), aquifer vulnerabili-
ty is a function of the natural properties of the overlying
soil and rock column or unsaturated zone of the aquifer.
The risk of groundwater pollution is dependent on both
the “natural” vulnerability according to the aquifer prop-
erties and to the subsurface contaminant load imposed
by human activity. This definition is in agreement with
the definition of Foster (1987) and Daly and Warren
(1994), as reported in the COST (European Scientific and
Technical Cooperation) action 65 final report (Hötzl and
others 1995).
Intrinsic vulnerability, as opposed to specific vulnerabili-
ty, considers all kind of contaminants. (Specific vulnera-
bility is defined for a given contaminant that is charac-
terized through particular properties, that could be differ-
ent from one contaminant to another. The assessment of
specific vulnerability requires consideration ot the char-
acteristics of the aquifer relative to the contaminant and
the contaminant itself, in addition to intrinsic hydrogeo-
logical and geological characteristics.) Even if the intrin-
sic vulnerability has only a general meaning in a pollu-
tion scenario where specific contaminant take place, ac-
cording to Andersen and Gosk (1987) and Adams and
Foster (1992), the concept of intrinsic vulnerability is
necessary to provide a maximum of unbiased informa-
tion, in order to define protection zones in karst environ-
ment. The degradation process are not taken into ac-
count in an intrinsic vulnerability approach.
An approach which would include land use, man’s activi-
ties and potential contamination from sources such as oil
spills, leakage from landfills, underground storage tanks,
is considered as “risk assessment”. It should be used to-
gether with the vulnerability mapping as a tool for deci-
sion makers and groundwater managers. This “risk map-
ping” concept is not included in the EPIK approach, pre-
sented in this paper.
Conceptual model of a karst aquifer
Our conceptual model of a karst aquifer includes a net-
work of conduits of high hydraulic conductivity (K val-
ues 1 10
–1
m/s) and of small volume – the karst network
2
Fig. 1
Conceptual model – scheme – of a karst aquifer
used to characterise the vulnerability mapping
(Doerfliger and Zwahlen 1995)
Fig. 2
Water recharge of “low
permeability volumes” and of
karst conduits by epikarst
zone (from Jeannin and
Grasso 1995 after Smart and
Friedrich 1986)
– connected to and discharging at an outlet. It is sur-
rounded by and connected with a large volume of low
permeability fractured and fissured rock (K values be-
tween 10
–3
and 10
–7
m/s). The karst network either drains
water out of the surrounding rock or recharges it, ac-
cording to the hydrodynamic state of the aquifer (Fig. 1).
We base our model on the facts that karst aquifers are
characterised by specific geomorphologic and hydraulic
phenomena: the absence of surface or near-surface drai-
nage, the existence of large springs, swallowholes and
dolines, the existence of networks of karst solution con-
duits, the typical spring hydrograph (rapid and violent
floods, rapid subsidence and slow tailing), rapid water
level variation in some wells, slow response in others and
the quick and strong variations in water chemistry as a
function of flow.
Recharge does not rapidly flow through the low permea-
bility volumes of the aquifer. This suggests that there are
some concentrated infiltration points such as sinkholes
directly connected to the karst network. The rest of the
quick recharge flows into the conduit network through
the epikarst. This epikarst, also called “subcutaneous
zone” is an immediate subsurface zone, highly fissured
due to the dissolution and pressure release of rock near
the ground surface. This zone is subject to extreme wea-
thering (Dodge 1982). Mangin (1973, 1975) defined it as a
possible temporary perched aquifer with a base that is es-
sentially a leaky capillary barrier – slow percolation in
tight fissures due to the fact that beneath the epikarst,
the rock mass has lower permeability – but also enclos-
ing connected pipes that provide effective drainage
(Fig. 2). The epikarst layer is not necessarily continuous,
its depth may be as thick as 10 m, even in tropical areas.
Water flow within the epikarst has a lateral component
moving through small conduits towards vertical pipes
and a vertical component with slow percolation into
small fissures (Williams 1983; Smart and Friederich 1986;
Ford and Williams 1989; Klimchouk 1995).
3
From conceptual model to vulnerability
Vulnerability of karst waters is a function of direct (mass
transport) and retarding (degradation and adsorption)
mass-transfer processes. These processes are governed by
mass-transfer physical parameters (molecular diffusion,
dispersion, sorption capacity, etc.), and also by flow pa-
rameters (mainly by the field of flow velocities). In karst
systems, the flow velocity field is highly variable due to
heterogeneous permeability field. The degradation and/or
adsorption processes take time to be effective and there-
fore the vulnerability of karst water depends essentially
on the residence time (or flow velocity) in (through) the
system.
Our conceptual karst aquifer model looks directly at the
heterogeneity of flow velocities within karst aquifers,
which can now be related to vulnerability. The following
statements related to the flow system are made:
1. During low water levels (base flow), spring outflow is
mainly fed by the water from the low permeability
volumes. This water has been resident in this part of
the aquifer for a long time. The vulnerability of the
spring water which flows during this part of the hy-
drologic cycle is thus relatively low.
2. During high water level (flood period) most of the wa-
ter from rainfall events infiltrates into conduits of the
epikarst and then flows into the conduit network
(Jeannin 1996); this water reaches the spring very
quickly. The filtration processes are thus less effective,
but compensated for by the dilution of any pollutants
in the large amount of water.
Vulnerability depends on the residence time in the differ-
ent parts of the aquifer. Three main parts can be distin-
guished:
1. The endokarst (conduit network and low permeability
volumes) where the flow velocity is high in karst con-
duits and low in low permeability volumes. A well de-
veloped conduit network suggests high vulnerability.
2. The epikarst where part of the water is stored and
slowly released (low vulnerability) and the rest quickly
concentrated into the endokarst conduit network (high
vulnerability). The more directly the epikarst is con-
nected to the conduit network, the higher the vulnera-
bility.
3. The protective cover (sediments overlaying the limes-
tone) where the residence time essentially depends on
the permeability of the cover and of its thickness
(among other transport parameters). The permeability
of the cover is a function of the cover water saturation
range.
The distinction between the three parts presented above
is meaningful when the recharge is diffuse (spread off)
over the catchment. In some cases, recharge is concen-
trated in features such as swallowholes and the water in-
filtrates more or less directly into the endokarst conduit
network.
Thus, the characteristics and hydraulic behaviour of the
three parts of a karst aquifer allow us to define four attri-
butes that may be used in a multi-attribute method of
vulnerability assessment. They are the epikarst, the pro-
tective cover, the infiltration conditions and the karst
network development.
The four attributes do not include the depth to the water
table. This attribute is inappropriate to karst aquifer be-
cause of the potential and common immediate and direct
recharge of runoff into the karst aquifer via swallowholes
and the epikarst conduits without filtration through the
unsaturated zone (Ray and O’dell 1993).
EPIK method
The method used to assess the vulnerability of waters of
a karst spring over the catchment area is based on our
conceptual model. This method is called EPIK, an acro-
nym for Epikarst (E), Protective cover (P), Infiltration
conditions (I) and Karst network development (K).
It is a multi-attribute weighting-rating method (overlay
and index method) that assesses the groundwater sensi-
tivity of karst terrain in a strict manner. A multiplier, re-
flecting a relative importance weighting, is assigned to
each attribute. The ratings for each class of given attri-
bute are multiplied by the weight related to the attribute
and then the products are added up to arrive at a final
score. The higher the score, the greater the protection of
the area is, i.e. the less vulnerable the area is. At the end,
the final numerical score range is assigned to classes of
different degrees of vulnerability. One of the first point-
count system models called DRASTIC was developed by
Aller and others (1987). This point-count system model
was chosen for the EPIK method as explained in the next
section. The vulnerability assessment with EPIK is made
at a scale range from 1:10000 to 1:5000.
Four major steps have to be carried out :
1. The boundaries of the water catchment basin from the
spring or well have to be defined on the basis of geo-
logy, hydrogeology and tracer tests.
2. The four attributes are assessed, measured (if possible)
and mapped. The evaluation is semi-quantitative, by
means of classes assigned numbered values.
3. The resulting maps for attributes E, P and I are digi-
tised and integrated into a GIS. The attribute K may
globally be assessed for the whole catchment (one
overall value), but may also be put on a regional scale
according to the geological and tectonic context, and
to the speleological knowledge. The GIS can then cal-
culate the vulnerability values for each raster (cell of
the map that has been defined) of the catchment ba-
sin. Thus, we obtain a map in a format, such that the
numbered values of the attribute classes can be com-
bined with an additive procedure. The result is a com-
pilation of the values which effectively is summarised
in a vulnerability map.
4. On the basis of this map showing the spatial distribu-
tion of the vulnerability, one can determine the differ-
ent vulnerability classes, respectively the different pro-
tection zones.
4
The most difficult step is the second one, which repre-
sents the originality of the EPIK method. In this step, the
attributes are combined into a ranking by means of as-
signing values to the classes. This assessment is made
with the help of several methods – direct or indirect –
such as tracer tests, geophysics, geomorphologic studies,
hydrographs analysis, shallow subsurface probes, inter-
pretation of aerial photographs and so forth. The various
indices assigned to each attribute, E, P, I and K are de-
scribed in the following.
Definition of the attributes and their classes and
their characterisation methods
Attribute E : Epikarst
The epikarst zone is located under any consolidated soil.
If there is no soil, the morphological features associated
with the epikarst are essentially similar to the Karren-
fields.
So far, epikarst features have not been studied in great
detail. Geomorphologic as well as hydrological character-
istics are poorly known. There are no real classification
nor investigation tools which allow recognition and map-
ping of different types of epikarst zones. Further, the epi-
karst can be heterogeneous over the aerial distribution of
a karst aquifer. Therefore, it is difficult to assign detailed
classes to different epikarst zones (for example well de-
veloped epikarst connected to the karst network, or de-
veloped but non-connected to the network, or absence of
epikarst, see also Doerfliger 1996).
For this reason and as a first step to address these diffi-
culties, the epikarst zone was characterised indirectly,
based on geomorphologic features which can easily be
mapped. Three indices have been established over the
range of vulnerability:
Table 1
Attribute classes of the Epikarst
Epikarst Karst morphological features
Highly developed E
1
Shafts, sinkholes or dolines (from all
kinds of genesis), karrenfields, cuesta,
outcrops with high fracturing (along
roads and railways, quarries)
Moderately
developed
E
2
Intermediate zones in the alignment of
dolines, dry valleys. Outcrops with
medium fracturing.
Small or absent E
3
No karst morphological phenomena.
Low fracture density.
Characterisation of the Epikarst. Mapping the three
classes E
1
–E
3
is equivalent to mapping the concerned
geomorphologic features. In order to do this, topographic
maps (1:25000, 1:10000 or 1 :5000) are first used to iden-
tify and outline these features. Interpretation of aerial
photographs allows confirmation and definition of the
delineated objects from the topographic map. Observed
intersections of lineaments from the aerial photographs
or from remote sensing analysis (Landsat photo analysed
with GIS) may correspond to highly fractured zones. If
no typical geomorphologic features are associated to
these zones, they could be mapped under E
2
instead of E
3
to be conservative (Doerfliger 1996).
Attribute P: Protective cover
For this attribute, we include both the soil and other geo-
logical overburden such as Quaternary deposits (glacial
till, silt, loess, rocks debris), and other non-karst layers,
for example, clay and sandstone.
The upper unconsolidated zone overlying the aquifer is
commonly regarded as one of the most important attri-
butes in the assessment of groundwater vulnerability. Soil
and other geological layers potentially have an important
attenuation capacity (Zaporozec 1985) due to specific pa-
rameters such as texture/structure, thickness, content of
organic matter and clay minerals, degree of water satura-
tion and hydraulic conductivity, in general to various
kinds of contaminants. These soil parameters are related
to physical, chemical and biological properties that allow
attenuation.
The thickness of a soil is strongly related to water resi-
dence time. It is an important property when assessing
groundwater vulnerability. The thinner the soil, the
greater the vulnerability.
As a first approach, in order to be able to assess the vul-
nerability of a karst water catchment basin, we consider
the thickness of the protective cover as basic parameter
and its hydraulic conductivity (Doerfliger 1996). We dis-
tinguished two cases, both according to the presence of
geological layers overlying the limestone and their hy-
draulic conductivity. To be conservative, we identified
four classes with boundary ranges of 20 cm, 100 cm,
200 cm and 1 200 cm (Table 2).
Characterisation of the Protective cover. This attribute
requires field verification using for example boreholes, or
auger methods for soil probes.
From published sources such as geological, pedological
and topographic maps, geological and regional studies,
we can define the areas of the watercatchment with or
without overlying geological layers. Aerial photographs
and satellite imagery are helpful to determine the soil
presence, and probably the thickness (just a scale of
sizes) if supplemented with field verification.
With a hand auger, the thickness can be measured direct-
ly in the field. If the catchment is not large, it may be
cost effective to perform auger holes according on a reg-
ular grid. If the catchment is large, the mesh size has to
be bigger and it may be necessary to assume similar at-
tributes for similar topography or morphology. This
would mean that for a measured thickness at one point,
the same thickness is given inside a square of 100-m side,
if the point is surrounded by the same morphology
(Doerfliger 1996). However, the spatial heterogeneity has
to be kept in mind.
5
