Connectivity in a Karst System Using Electrical

Resistivity Tomography and Network Theory

by Mariana Gómez-Nicolás1 , Mario Rebolledo-Vieyra2,3 , Efrain Canto-Lugo4 , Rodrigo Huerta-Quintanilla5 , and
Pablo Ochoa-Sandoval6

Abstract
The Ring of Cenotes (RC) is an alignment of numerous cenotes (sinkholes) in a semicircular form (with a radius of 100 km)
located in northwestern Yucatán, México. The formation roughly coincides with a concentric ring that corresponds to a buried
structure, which has been identified as the product of a meteor impact, known as the Chicxulub crater. Secondary permeability
generated by the fracturing and faulting of the sedimentary sequence in the Chicxulub crater has favored the karstification process
and therefore the development of underground rivers that transport water from the mainland to the sea.
This study implements the network theory to study the hydrological connectivity between a group of 11 cenotes within the
RC. Eight electrical resistivity tomography transects were used as an empirical basis. Each transect was acquired directly in the
field using the SuperSting R1/IP equipment with a dipole-dipole configuration. An adapted version of the reliability algorithm for
communication networks was used as a theoretical model.
We found evidence of the existence of water cavities in the study area. We made a network from the data and assigned
connection probabilities among cenotes as a function of the separation length and the number of water cavities, as well as their size.

Introduction out the southern boundary of the Chicxulub crater’s outer

The Yucatán Peninsula contains one of the most
extensive karst aquifer systems on the planet. The region
hosts the world’s longest underwater cave system and the
65-million-year-old multi-ring Chicxulub impact crater
(Gondwe et al. 2010; Bauer-Gottwein et al. 2011).
Using Landsat Thematic Mapper imagery, Pope iden-
tified in 1991 the Ring of Cenotes (RC), a semicircular
feature which extends over an arc of about 90 km tracing
1
Laboratorio de Ingeniería y Procesos Costeros, Instituto de
Ingeniería, Universidad Nacional Autónoma de México, Puerto de
Abrigo S/N, Sisal, Yucatán, 97355, México.
2 Centro de Investigación Científica de Yucatán, Unidad de
Ciencias del Agua, Mérida, Yucatán, México.
3 Currently at Independent Consultant, SM 312, Mza 7, Chipre
5, Cancún, Quintana Roo, 77500, México.
4 Departamento de Física Aplicada, Centro de Investigación y
de Estudios Avanzados del Instituto Politécnico Nacional, Mérida,
Yucatán, México.
5
Corresponding author: Departamento de Física Aplicada,
Centro de Investigación y de Estudios Avanzados del
Instituto Politécnico Nacional, Mérida, Yucatán 97310, México;
52+(999)9429431; rhuerta@cinvestav.mx
6
Facultad de Ingeniería y Ciencias, Universidad Autónoma de
Tamaulipas, Ciudad Victoria, Tamaulipas, 87149, México.
Article impact statement: Novel method combining electrical
resistivity tomography and network reliability theory to infer
preferential connectivity in a karst system.
Received March 2017, accepted October 2017.
© 2017, National Ground Water Association.
doi: 10.1111/gwat.12618
rim (Pope et al. 1991). Although the RC is clearly associ-
ated with the Chicxulub crater basin, several hypotheses
have been proposed to explain the formation of this
surface. Furthermore, new discussions could emerge in
coming years in the light of the findings derived from
the 364 drilling Expedition (Morgan et al. 2016).
Despite its uncertain origin, the RC plays a key role
in the regional hydrology of the Yucatán Peninsula; it
was a focus for Mayan settlements (Gilli et al. 2009)
and now, individual cenotes continue to be exploited
for human consumption or for tourism. As a result of
their importance, cenotes have become the subject of
an extensive variety of hydrogeological, chemical, and
geophysical studies (Pérez-Ceballos et al. 2012). These
studies have allowed large-scale patterns to be established
across the RC and have played a decisive role in the
formulation of laws to delimit protection zones such as
the geohydrological reserve (Decree117 2016). Despite
these efforts, there is a lack of integral management to
preserve the quality of their waters.
In cenotes which have a good hydraulic connection
with the rest of the aquifer, the water remains clear
during most of the year. These cenotes are primarily
fed by groundwater through fractures and dissolution
cavities (Schmitter-Soto et al. 2002). These conduits
serve as major pathways for groundwater flow in karst
aquifers and since a water-filled underground void has
lower resistivity than most of the surrounding rocks, in

NGWA.org Groundwater 1
principle it can be detect using the geophysical method of
electrical resistivity tomography (ERT) (Zhu et al. 2011).
Electrical resistivity is based on the injection of an electric
current into the ground through electrodes. Bulk electrical
resistivities in the shallow subsurface are controlled
largely by electrolytic conduction in aqueous fluids that
are either distributed across grain boundaries or contained
in pores, fractures, and faults (Stummer et al. 2004).
We are interested in studying the connection between
sinkholes across conduits found by ERT. Since each
electrical resisitivity image is limited to a few meters in
depth and length, here we propose a new approach based
on a modified network reliability algorithm to integrate
the information provided by several ERT measurements
and characterize the connection between sinkholes.
By network, we refer to a collection of small circles
(nodes) interconnected by lines (links). For example, in
transportation, the circles might represent delivery points
of goods and the lines are the roads between them. The
study of reliability models starts from the knowledge of
the failure probabilities of the links and seeks to determine
an index that corresponds to the probability that the
network is operational; in other words, it determines the
probability that there is at least one path that connects
a pair of nodes. In this study, sinkholes comprising 11
cenotes within the RC in an area of 0.5 km2 and cavities
found by ERT are the nodes of the network. We establish
links between the nodes and then calculate the probability
that one cenote is connected to another by a specific
trajectory.
Site Description
This is a summary of the relevant geological features
of the Yucatán Peninsula as presented by Bauer-Gottwein
et al. (2011)). The current conceptual geological under-
standing of the Yucatán Peninsula was established by the
Pemex Exploration Drilling Program from the 1940s to
1960s, along with regional gravity and magnetic surveys,
and the Chicxulub Scientific Drilling Program carried
out by the National Autonomous University of México
(UNAM) in 2000 (Figure 1).
The Yucatán Peninsula is a low-lying platform con-
sisting of limestone, dolomites, and evaporates, reaching
thicknesses of greater than 1500 m. The Cretaceous, and
most of the Tertiary, lies upon a basement of red beds
from the Triassic-Jurassic and igneous rocks from the Tri-
assic and Paleozoic. The surface sedimentary rocks span
from the Upper Cretaceous to the Holocene, and are gen-
erally almost horizontally layered and off-lapping, with
gradually younger carbonates deposited toward the Penin-
sula margins. Lithological columns from the site UNAM-
2 (U2) drill site present this sequence from surface to
bottom: 1.7 m of soil and caliche followed by 85 m of
dolomitized crystalline limestone with 14 m of intercalated
dolomitized calcarenite, 95 m of fossiliferous dolomitized
crystalline limestone, 10 m of argillaceous limestone and
finally another 10 m of dolomitized crystalline limestone
(Rebolledo-Vieyra et al. 2000). Our study zone is located
2 M. Gómez-Nicolás et al. Groundwater
14 km from drill site U2, into the archeological site of
Mayapán (Figure 2), the last Maya capital of the Yucatán.
Bedrock samples, around Mayapán, show it as an impor-
tant geological boundary with contrasting 87 Sr/86 Sr sig-
natures on either side (Gilli et al. 2009).
In hydrogeological terms, the aquifer consists of a
mature karstic system with a low hydraulic gradient on
the order of 7 to 10 mm/km, suggesting very high perme-
abilities (Steinich and Marín 1996, 1997). Dissolution of
the carbonate rocks has resulted in a lack of aquitards. The
aquifer is unconfined except for a narrow band parallel to
the coast, possesses a thin lens of fresh water (45 m at
Mérida) underlaid by salt-water intrusion that penetrates
more than 40-km inland, with measured water depths of
up to 60 m (Perry et al. 1989, 1995). The aquifer is highly
vulnerable to contamination, as shown by previous studies
that reported that groundwater within the top 15 to 20 m
of the aquifer does not meet the standards for human con-
sumption (Marín et al. 2000; Pacheco-Castro et al. 2017).
Six kilometers to the east of our study site, Pérez-Ceballos
characterized a group of seven cenotes with low pH val-
ues, high NO− + −
3 , and high Ca2 and HCO3 and concluded
that these cenotes in the center of the RC are in an aquifer
recharge zone (Pérez-Ceballos et al. 2012).
At the end of the Cretaceous, a large asteroid
impacted the Yucatán Peninsula. The center of the
approximately 200 km impact crater, named the Chicxulub
impact crater, is located off the present Yucatán Peninsula
Gulf of México coastline, at approximately 21.4◦ N and
89.5◦ W. The outline of the impact crater is marked by
the RC, a fracture zone with a high density of sinkholes.
This zone plays an important role in the distribution of
water in the peninsula because of its high permeability
and subterranean-river behavior concentrates, transports,
and discharges groundwater where the ring intersects the
coast near Celestún and Dzilam de Bravo (Marín 1990;
Perry et al. 1995). In addition, due to its great biodiversity
and its hydrogeological properties, in 2016, the RC was
declared a Ramsar site and in the southwestern portion
the first hydrogeological State Reserve in the country was
created (Decree117 2016).
Mayapán was the largest Mayan political capital of
the Post-classic Period (A.D. 1150 to 1450). It has been
the subject of a comprehensive archeological study that
began in 1948 with the arrival of a team of researchers
from the Carnegie Institution (Hare et al. 2014). Nowa-
days, researchers focus their efforts on long-term studies
at the ancient city by documenting the regional system of
settlement and the environment in which it was developed.
As a result, they have created a bare earth digital eleva-
tion model of a 40 km2 area centered on the walled city
of Mayapán using LiDAR-generated data, ground survey,
surface collection, test excavations, artifact analysis, and
Geographic Information System–aided spatial analysis.
Underwater archeology presents some interesting
results 2 km far from our study site. Detailed underwater
mapping, photography, and video of the cenote Sac
Uayum shows two 15-m-deep chambers connected by a
narrow passage about 10 m long. The remains of human
NGWA.org
Figure 1. Compiled geologic map of Yucatán. The map shows the location of PEMEX exploratory boreholes (C1, S1, T1, Y1,
Y2, Y5A, and Y6) and UNAM scientific shallow-drilling program boreholes (U1 to U8). Q, Calcarenite and coquina; PCP,
Argillaceous limestone and coquinoide limestone; O, Marly calcarenite and coquinoide; EChCh, Fossilliferous crystalline
limestone; EChP, Crystalline limestone and argillaceous; Pal-E, Crystalline limestone, dolomitized with anhydritic facies. No
scale (modified from Rebolledo-Vieyra et al. 2010).

crania, bones, and ceramic fragments were clustered in
the southwest end of the first chamber (Russell 2017).
With all of the aforementioned, our main working
hypothesis is that the RC is originated and structurally
controlled by the faults and joints generated by the
gravitational accommodation of the Chicxulub Crater rim
(Gulick et al. 2008). Based on this, we developed a
conceptual model for the conduits and connections. The
model assumes that the vertical faults and joints are
continuous structures, no larger than 3 × 3 km. We also
hypothesize that this is a fault and joint system, where
the surface expression of them are nearly parallel. The
latter accounts for the connection along the RC, and, in
a lesser degree, to the connections perpendicular to the
main trace of the RC.
Research Method
Cenote Selection
We selected the study zone based on the database
provided by the Yucatán Government (SEDUMA 2016)
which satisfies the conditions of being in the RC and
presenting a high density of cenotes. The zone chosen
is 40 km southeast from Mérida, in the ancient city
of Mayapán. The original Carnegie map of Mayapán
(Jones 1962) and a map shared by Dr. Russell (Russell
2017) played an important role in our specific selection
of sinkholes. We used a Global Positioning System
®
Trimble (Sunnyvale, California) Recon 400X to aquire
coordinates for the 11 cenotes of interest (Figure 2). Table
S1, supporting information, shows the cenote labels and
their coordinates.
ERT Measurements
In this particular case, we used ERT to identify
macroscale porosities (> 3 m diameter). Our main goal
was to identify conduits larger than 3 m with a high
likelihood of connecting sinkholes. For this task, ERT is
a very effective tool. The spatial set up of the transects
(labeled T1, T2,...,T8) was designed considering the
accessibility to the field, using trails already available
and patches free of vegetation. The next consideration
was to position the transect in a way that would intercept
the imaginary line connecting one cenote to another. For
example, T7 intercepts an imaginary line from cenote
nine to cenote one and from cenote nine to cenote
seven. Transects: T1, T2, T5, T6, and T7 were laid
out on existing trails, while for T3, T4, and T8, it was
necessary to clear the vegetation. There was an overlap
among transects T1, T2, and T3 with the purpose of
reducing the edge effects. In Table S2, we show the
geographic coordinates for the first and the last electrodes
of each transect. The final setup of the transects and
cenotes is shown in Figure 3. The area in which the
transects and cenotes are located has an approximate size
of 600 m × 800 m. ERT imaging surveys were conducted
®
using the AGI SuperSting R1/IP (Advanced Geosciences
Inc., Austin, Texas) acquisition system. The 345-m
stationary cable consisted of 71 electrodes spaced 5 m

NGWA.org M. Gómez-Nicolás et al. Groundwater 3

Figure 2. Map of the Yucatán Peninsula. An aerial view of the RC, with Chicxulub Port at its center and two water discharge
locations (Celestún and Dzilam de Bravo). Study zone comprising a group of 11 cenotes, located in the Post-classic Mayan
political capital of Mayapán, within the only geohydrological Reserve of México. Cenote No. 11 (Chen Carro) belongs to the
Ramsar Site list no. 2043.
Figure 3. Array of the eight ERT transects (T1, T2, ..., T8)
located between the 11 cenotes. The geographical coordinates
of the cenotes and transects are given in Tables S1 and S2,
respectively.
apart along multi-node cable. A dipole-dipole array was
used because of its good lateral resolution (Stummer et al.
2004). A total number of 2175 measurements per transect
were collected.
Inversion Process
The apparent resistivity ρ a value is calculated from
the current (I ) and voltage (V ) values:
V
ρa = K (1)
I out using the software AGI EarthImager 2D v2.4.0
4 M. Gómez-Nicolás et al. Groundwater
where V is the potential difference and K is the
geometric factor of the dipole-dipole array (Loke 2013).
Prior to further processing, the data were filtered. We
discarded data with voltages lower than 0.2 mV and
resistivities smaller than 0.01 ohm-m and larger than
105 ohm-m. For the forward solution, we solved the
following two-dimensional partial differential equation in
the Fourier transform domain by the finite difference
method:
   
∂ ∂V ∂ ∂V
σ + σ − k 2 σ V = −I · δ (x) · δ (z)
∂x ∂x ∂z ∂z
(2)
where V is the potential, I is the electric current,
k is the wavenumber in the transform domain, and
σ is the electrical conductivity as a function of (x ,
z ) (Advanced Geosciences Inc. 2005). The results in
this study were produced using the finite difference
method solving Equation 2 (Dey and Morrison 1979).
For inversion, we used a smooth model inversion algo-
rithm (Constable et al. 1987) to minimize the objective
function
S (m) = (dobs − g (m))T Wd (dobs − g (m)) + α· mT Rm
(3)
where dobs is the measured data and g(m) is the calculated
data. Wd is a data weighting matrix, α is a Lagrange
multiplier, and R is a roughness operator (Advanced
Geosciences Inc. 2005). All these processes were carried
®
NGWA.org

Figure 4. Color scale associated with resistivity. The label
color is used to calculate the weighted area for the water
cavity in Equation 4.
(Advanced Geosciences Inc.). We applied a common scale
of colors to the eight inverted resistivity sections. In
Figure 4, the color scale and the corresponding resistivity
are displayed.
Water Cavity Determination
The ERTs were performed to obtain information
on possible underground water cavities which connect
cenotes. The water in the region studied has a resistivity
between 6 and 12 ohm-m (Pérez-Ceballos et al. 2012)
but since the inversion process introduces a degree of
uncertainty, we consider a larger range for the resistivity:
1 to 47 ohm-m. Later in the paper, we will see evidence
for 23 water cavities. The cross-sectional area for each
cavity was calculated as:

5
Z= Wi Ai (4)
i=1
where Ai is the area of the cavity with color ai (Figure 4)
and W i is the weight that represents the percentage of
water contained in a material with resistivity ai. We
use previous studies as a reference (Telford et al. 1990;
Bauer-Gottwein et al. 2011) to assign the following values
for W : W 1 = 0.8, since this is typical for saline water;
W 2 = 0.9, because in this resistivity range (3 to 6 ohm-
m), most of water is saturated with salt; W 3 = 1.0,
this is the resistivity for free water in the cenotes
(Pérez-Ceballos et al. 2012); W 4 = 0.5 and W 5 = 0.3,
correspond to water trapped in rocks with different
densities.
NGWA.org
Calculation of Probabilities
To calculate the connection probability between the
studied cenotes, we used an adapted version of the
network reliability algorithm (Fishman 1996, p. 55). This
algorithm provides us with the probability of any two
cenotes in a network, called s and t, being connected. To
frame the problem in terms of that algorithm we define
nodes, links and the reliability vector q as follows.
The nodes considered for the calculation consisted
of the 11 cenotes (Figure 3) and the 23 water cavities
identified by ERT analysis (Figure 5). In total, we obtain
a set of V = {1, ..., 34} nodes. For the other hand, there
is a link between two nodes s and t if there is a straight
line joining s and t and this line does not cross an ERT
transect. It is important to note that s and t cannot both be
cenotes, because we establish that the connection between
them is only across the cavities found by the ERT. With
this criteria we obtained a set of m = 138 links. Figure 6
shows the network with 34 nodes and 138 links. For
each link, we calculated q i with i ∈ 1, ..., m, where
1−q i means the probability that link q i “fails.” For the
calculation, we used the exponential functions D(X ) and
A(Z ), both with a range of (0, 1). The first one given in
terms of the nodes distance X = X (s, t) and expressed as
D (X) = e−αX (s,t) (5)
with α = γ /λ, where λ = 181.05 m is the average distance
between the nodes in the network; by assumption
D(3000 m) = 0.01, then, solving for γ we get γ = 0.277.
This assumption is based on geophysical observations:
Gulick et al. (2008), reported faults within the Chicxulub
crater rim, where the study area is located, up to 3 km
deep, from ground surface. From this, we assigned a
square shape to the fault’s geometry, with a 3 × 3 km side.
In Pope et al. (1991), we also found that the cenote density
and width of the ring vary from about three cenotes per
km2 along a 3 km-wide portion in the southwest, to a chain
of single cenotes 3 km apart in the southeast.
The second function is in terms of the weighted area
of the cavities Z = Z (t),
A (Z) = 1 − e−βZ (t) (6)
with β = η/τ , where τ = 81.34 m2 is the average weighted
area for all the cavities; by assumption A(56 m2 ) = 0.5,
then, solving for η we get η = 1. This assumption is based
on personal observations in the field that show the radius
of cenotes in this study area is approximately 4 m.
In this way, when we considered cenote (s)—water
cavity (t), q i = q(s, t) is calculated as
 
q (s, t) = e−αX (s,t) 1 − e−βZ (t) (7)
In the case water cavity (s)—water cavity (t). The
equation takes the form
  
q (s, t) = e−αX (s,t) 1 − e−βZ (s) 1 − e−βZ (t) (8)
M. Gómez-Nicolás et al. Groundwater 5
Figure 5. Electrical Resistivity Tomography Profiles. The eight transects with the 23 water cavities found, with resistivities
between 1 and 47 ohm-m and numbered from 12 to 34.
The 138 tuple of values of q i conform the compo-
nents of the reliability vector q. Once we obtain q, the net-
work is divided into six clusters or subnetworks (Figure 6)
and we applied the reliability algorithm to each of them to
calculate the final probability g(q) of connection. In sum-
mary, the network reliability algorithm indicates that the
probability of connection between two network nodes s
and t over one specific path is the product of the probabil-
ities that each node is connected with the next one in that
path. Afterward, it is necessary to sum these probabilities
over every possible path between s and t.
The two adjusted parameters γ and η can be
updated once more information is obtained from the
system. Figures S1 and S2 illustrate the variability of
the Euclidean norm of the reliability vector, as measured

by the difference ||q− q ||, under changes in our geological
suppositions.
Results
We perform a contact resistance test prior to each
ERT survey. Transects T1, T2, T3, T5, T7, and T8 passed
the test (<2000 ohm-m). However, 20% of the contact
resistance values in T4 and T6 were high. These two
transects were installed on exposed limestone bedrock.
We kept the data since they contribute significantly to the
analysis, but keeping in mind the possible introduction
6 M. Gómez-Nicolás et al. Groundwater
of uncertainties. Figure 5 shows the eight transects and
the positions of the 23 cavities that were found. The
comparison between the color hues (Figure 4) allowed us
to identify the non-saturation zone (red-yellow with resis-
tivities of 1529 to 100,000 ohm-m) and the saturated zone
(1 to 47 ohm-m). The range between 47 and 1529 ohm-m
is associated with fractured limestone (Telford et al.
1990). The water level was 14 m deep with respect to
the terrain level. This was based on measurements made
on cenotes with wide surface openings, using a portable
electronic water level meter. However, it was not possible
to do this with every cenote. With the others, we had to
substitute indirect estimates from ERT, uncorrected for
topography. Figure 5 shows that most of the underground
water lies at a depth range of 12 to 20 m.
The transects T1, T2, T3, T4, T5, and T7 indicate
possible water flux in the east to west direction (Figure 3);
this observation gives strength to the hypothesis of
circulation flux in the RC direction (Bauer-Gottwein et al.
2011).
To use Equation 5, coordinates of the cenotes were
used, as in Table S1. The coordinates of the geometrical
center of each water cavity was calculated from the
transect to which it belongs, using the first electrode of
the transect as a spatial reference. The results are given in
Table S3. For Equation 6, we calculated the weighted area
Z (Equation 4) of each water cavity (Table 1). Finally,
NGWA.org
Figure 6. The complex network with 34 nodes and 138 links. The colors show the six clusters analyzed. The nodes are labeled
in red from 1 to 11 for cenotes and in blue from 12 to 34 for water cavities. Black lines represent links between the nodes.

Table 1 Table 2
Weighted Area (Z ) of the Water Cavities Connecting Probabilities Between Cenotes that
Belong to Clusters 1, 2, and 3.
Water Cavity Z (m2 ) Water Cavity Z (m2 )
Cluster Cenotes g(q)
12 7 24 19
13 52 25 12 1 {1, 7} 0.875
14 92 26 125 {1, 8} 0.877
15 131 27 106 {1, 9} 0.878
16 7 28 61 {7, 8} 0.861
17 36 29 68 {7, 9} 0.873
18 50 30 9 {8, 9} 0.870
19 23 31 310 2 {2, 4} 0.594
20 62 32 28 {2, 5} 0.575
21 157 33 96 {2, 7} 0.003
22 296 34 107 {4, 5} 0.672
23 15 {4, 7} 0.004
{5, 7} 0.003
3 {2, 3} 0.581
{2, 6} 0.669
we used Equations 7 and 8 (depending on the case) to {2, 11} 0.615
determine the q values. {3, 6} 0.661
The 34 nodes, 138 links, and 6 clusters are presented {3, 11} 0.537
in Figure 6. Cenotes are the nodes in red (1 to 11) {6, 11} 0.688
and, water cavities are the nodes in blue (12 to 34).
Cluster 1 is composed of nine nodes of which nodes 1, 7, 8, and 9 are cenotes.
To reduce computational effort, and according to our Cluster 2 is composed of 10 nodes of which nodes 2, 4, 5, and 7 are cenotes.
principal interest, the network was divided into six clusters Cluster 3 is composed of 11 nodes of which nodes 2, 3, 6, and 11 are cenotes
(subgraphs); this does not modified our findings. (Figure 6).

In Tables 2 and 3, we applied network reliability

algorithm (Fishman 1996, p. 55) and we show the g(q)
probabilities of cenote-cenote connections. In particular,
we see a preferential pattern of connection in the west to
east orientation. As previous studies suggest, this could be
related to the fact that the preferential flux direction in this
part of the RC is in that direction. Cluster 6 oriented west-
east has 36% more connecting probability than cluster
5 (oriented north-south). Further analysis is needed to
confirm these findings. Another point that should be noted
is that for a pair of cenotes, these tables do not give equal

NGWA.org M. Gómez-Nicolás et al. Groundwater 7

 Table 3 flow; this variable was not measured. Nevertheless, we
Connecting Probabilities Between Cenotes that can infer the preferential direction of flow from the
Belong to Clusters 4, 5, and 6. conduits found and their connections. This methodology
can complement archeological underwater studies such
Cluster Cenotes g(q) as those of Dr. Bradley Russell’s group. Despite the
arbitrariness of the scale factors γ and η, we have found
4 {6, 8} 0.230 that the connectivity hierarchy is preserved even when the
{6, 10} 0.728 factors are modified. If, for instance, in cluster 6 the two
{8, 10} 0.182
most connected cenotes are {1,8} and the least connected
5 {1, 4} 0.004
{1, 8} 0.278 are {3,9}, this would remain true even if the parameters
{1, 10} 0.095 are modified.
{4, 8} 0.004 We hope that future work in this area could support
{4, 10} 0.001 our findings. For instance, we expect cenotes 1, 7, 8, and
{8, 10} 0.097 9 in cluster 1, which are the most connected, to share
6 {1, 3} 0.005 geochemical characteristics, for example, clear waters,
{1, 6} 0.098
clean, sandy or rocky bottoms, and a homogeneous, well-
{1, 8} 0.670
{1, 9} 0.625 oxygenated water column. In any case, in cenote 1, a great
{3, 6} 0.035 sound of flowing water was detected that was big enough
{3, 8} 0.005 to generate an air column which was expelled from the
{3, 9} 0.004 cenote.
{6, 8} 0.097 Cenote 11 is very probably connected to cenotes 6, 2,
{6, 9} 0.077 and 3, and because of this, any action to preserve cenote
{8, 9} 0.609 11 as a RAMSAR site, would require this connection to
Cluster 4 is composed of 10 nodes of which nodes 6, 8, 10 are cenotes. be considered in the preservation scheme. With regards
Cluster 5 is composed of 11 nodes of which nodes 1, 4, 8 and 10 are cenotes. to cenotes 4 and 10, which in the calculations appear to
Cluster 6 is composed of 8 nodes of which nodes 1, 3, 6, 8 and 9 are cenotes be poorly connected, possibly due to the collapse of a
(Figure. 6).
roof or wall and sedimentation, we expect turbid water,
and a thermally stratified, alkaline surface water layer
values, that is, the probability of 3 and 6 being connected which is oversaturated with oxygen, while water near the
is 0.661 in cluster 3 and is not the same as the probability bottom is likely to be acidic, devoid of O2 , and contain
that 3 being connected to 6 in cluster 6: 0.035. These H2 S. It is beyond the scope of this paper to corroborate
probabilities need not to be the same since they depend the aforementioned, however, we are sure that it would
on the transects that have been included in the calculation. be easy for the community to test it.
Even though we have split the network into clusters to
calculate the connecting probabilities, we also calculated
them using the entire network. As an example we obtain Conclusion
0.662 for the connecting probability between cenotes {2, The methodology proposed in this study is poten-
11}. Inside cluster 3, this probability is 0.615. tially a very powerful tool for investigating aquifers where
traditional hydrogeological methods (e.g., pumping tests)
are not available or difficult to perform. The combina-
Discussion tion of a proved geophysical technique to investigate
Many attempts have been made to describe the the underground geological structure, with a powerful,
network structure of underground karstic cavities (Borghi yet flexible mathematical algorithm such as the network
et al. 2012). Geophysical methods have been successfully theory, allows us to assign a number to the connection
used to locate preferential flow paths and structures probability between different cenotes based on the dis-
in karst aquifers (Gondwe et al. 2010). In the Yucatán tance between them and the empirical evidence obtained
Peninsula, electrical resistivity methods have never been using ERT. In this particular case, we applied our pro-
applied to the study of connectivity between sinkholes, posed methodology to an area where we have very reliable
the only antecedent is the work of (Steinich and Marín macroscale geological information, such as information
1996). These authors detected the main azimuth angles of on the RC, where extensive, regional, geophysical studies
subsurface high-permeability structures at a regional scale have been performed (Gulick et al. 2008). These provide
using vertical electrical sounding. us with regional information, predicting connectivity in
In the present study, we are not aiming to determine the southern portion of the RC, roughly oriented W-E,
the complete structure of a karstic aquifer, such as the one which concurs with our results.
we have in the Yucatán Peninsula, since the representation These probabilities depend on the setup we used to
of sinkholes and conduits in terms of nodes and links measure the resistivities, which means they are sensitive
lacks important information. However, we have gained to the trajectory. The uncertainties can only be reduced
an insight into the system’s pattern of connections. Our by increasing the number of transects. With regards to
method does not provide quantitative information on water computation, it was possible to program in parallel to

8 M. Gómez-Nicolás et al. Groundwater NGWA.org

reduce the time taken to perform the analyses, which
allowed for more possible links in the network. We
are sure that other teams that have been working in
Mayapán for many years have something to say about
our results, and with their collaboration the results of this
study could provide valuable information. We hope that
people working in underwater anthropology or isotopic
techniques can verify our results, since it could be a
technique to help make conservation decisions such as
the restriction of human activities, the selection of waste
disposal areas, and the protection of recharge areas to
promote the sustainable use of the water resources in this
geographical area.
Acknowledgments
M.G.N. acknowledges a master grant from CONA-
CyT, under which this contribution first started. Also we
thank Acrelio Pacheco, Archeologists Carlos Peraza and
Eunice Uc, for providing free access to the lands where
we carried out the survey. The authors wish to thank
CICY for providing the vehicles and field equipment for
conducting ERT surveys. We also thank the project “Geo-
physical survey of the Yucatán Aquifer,” where M.R.V.
is responsible. Also would like to thank three anonymous
reviewers for their insightful comments, which greatly
improved this manuscript. To Dr. Adrián Flores-Orozco
for the useful suggestions in the ERT inversion process.
To Dr. Bradley Russell for sharing geographic locations
of Mayapan cenotes. To Daniel P.B. Smith for editorial
assistance.
Supporting Information
Additional Supporting Information may be found in the
online version of this article. Supporting Information is
generally not peer reviewed.
Table S1. Cenote labels and their locations.
Table S2. Location of each of the transects. In this table,
we give the position of the first (E1) and last (E71)
electrodes.
Table S3. Location of the geometrical center of each water
cavity.
Figure S1. Variability of the reliability vector q under
changes in the geological suppositions in Equation 5.
Here, q is the reliability vector when we assume D(X *)
= 0.01 with X * = 3000m, and q is the reliability vector
obtained when X* is perturbed.
Figure S2. Variability of the reliability vector q under
changes in the geological suppositions in Equation 6.
Here, q is the reliability vector when we assume A(Z *)
= 0.5 with Z* = 56 m2 , and q is the reliability vector
obtained when Z* is perturbed.
References
Advanced Geosciences Inc. 2005. Instruction Manual for Earth-
Imager 2D. Resistivity and IP Inversion Software. Copyright
(2002–2009). Austin, Texas: Advanced Geosciences Inc.
NGWA.org
Bauer-Gottwein, P., B. Gondwe, G. Charvet, L.E. Marín, M.
Rebolledo-Vieyra, and G. Merediz-Alonso. 2011. Review:
The Yucatán Peninsula karst aquifer, México. Hydrogeol-
ogy Journal 19, no. 3: 507–524.
Borghi, A., P. Renard, and S. Jenni. 2012. A pseudo-genetic
stochastic model to generate karstic networks. Journal of
Hydrology 414–415: 516–529.
Constable, S.C., R.L. Parker, and C.G. Constable. 1987.
Occam’s inversion: A practical algorithm for generating
smooth models from electromagnetic sounding data. Geo-
physics 52, no. 3: 289–300.
Decree117. 2016. Decreto 117, decreto que establece el
área natural protegida denominada reserva estatal geo-
hidrológica del anillo de cenotes. Gobierno del estado
de Yucatán (spanish) [decree 117, decree establishing the
protected natural area called the geohydrological state
reserve of the ring of cenotes. Government of the state
of Yucatán]. http://www.yucatan.gob.mx/gobierno/diario_
oficial.php?f=2013-10-28 (accessed March 2016).
Dey, A., and H.F. Morrison. 1979. Resistivity modelling for
arbitrarily shaped two-dimensional structures. Geophysical
Prospecting 27, no. 1: 106–136.
Fishman, G.S. 1996. Monte Carlo: Concepts, Algorithms, and
Applications. New York: Springer.
Gilli, A., D.A. Hodell, G.D. Kamenov, and M. Brenner. 2009.
Geological and archaeological implications of strontium
isotope analysis of exposed bedrock in the Chicxulub crater
basin, northwestern Yucatán, México. Geology 37, no. 8:
723–726.
Gondwe, B.R., S. Lerer, S. Stisen, L. Marín, M. Rebolledo-
Vieyra, G. Merediz-Alonso, and P. Bauer-Gottwein. 2010.
Hydrogeology of the south-eastern Yucatán Peninsula: New
insights from water level measurements, geochemistry,
geophysics and remote sensing. Journal of Hydrology 389,
no. 1: 1–17.
Gulick, S.P., P.J. Barton, G.L. Christeson, J.V. Morgan,
M. McDonald, K. Mendoza-Cervantes, Z.F. Pearson, A.
Surendra, J. Urrutia-Fucugauchi, P.M. Vermeesch, et al.
2008. Importance of pre-impact crustal structure for
the asymmetry of the Chicxulub impact crater. Nature
Geoscience 1, no. 2: 131–135.
Hare, T., M. Masson, and B. Russell. 2014. High-density lidar
mapping of the ancient city of Mayapán. Remote Sensing
6, no. 9: 9064–9085.
Jones, M. 1962. Map of Mayapán. In Mayapán, Yucatán,
México, ed. H.E. Pollock, R.L. Roys, T. Proskouriakoff,
and A.L. Smith, 619. Washington, DC: Carnegie Institute
of Washington.
Loke, M.H. 2013. Tutorial: 2-D and 3-D Electrical Imaging
Surveys. Copyright (1996–2004). Malaysia: Geotomo Soft-
ware.
Marín, L.E., 1990. Field Investigations and Numerical Simula-
tion of Groundwater Flow in the Karstic Aquifer of North-
western Yucatán, México. DeKalb, Illinois: Northern Illi-
nois University.
Marín, L.E., B. Steinich, J. Pacheco, and O.A. Escolero. 2000.
Hydrogeology of a contaminated sole-source karst aquifer,
Mérida, Yucatán, México. Geofísica Internacional 39, no.
4: 359–365.
Morgan, J.V., S.P.S. Gulick, T. Bralower, E. Chenot, G.
Christeson, P. Claeys, C. Cockell, G.S. Collins, M.J.L.
Coolen, L. Ferrière, C. Gebhardt, K. Goto, H. Jones,
D.A. Kring, E. Le Ber, J. Lofi, X. Long, C. Lowery, C.
Mellett, R. Ocampo-Torres, G.R. Osinski, L. Perez-Cruz,
A. Pickersgill, M. Poelchau, A. Rae, C. Rasmussen, M.
Rebolledo-Vieyra, U. Riller, H. Sato, D.R. Schmitt, J. Smit,
S. Tikoo, N. Tomioka, J. Urrutia-Fucugauchi, M. Whalen,
A. Wittmann, K.E. Yamaguchi, and W. Zylberman. 2016.
The formation of peak rings in large impact craters. Science
354, no. 6314: 878–882.
M. Gómez-Nicolás et al. Groundwater 9
Pacheco-Castro, R., J. Pacheco-Ávila, M. Ye, and A. Cabrera-
Sansores. 2017. Groundwater quality: Analysis of its
temporal and spatial variability in a karst aquifer. Ground-
water. https://doi.org/10.1111/gwat.12546
Pérez-Ceballos, R., J. Pacheco-Ávila, J.I. Euán-Ávila, and H.
Hernández-Arana. 2012. Regionalization based on water
chemistry and physicochemical traits in the ring of cenotes,
Yucatán, México. Journal of Cave and Karst Studies 74, no.
1: 90–102.
Perry, E., J. Swift, J. Gamboa, A. Reeve, R. Sanborn, L.
Marín, and M. Villasuso. 1989. Geologic and environmental
aspects of surface cementation, north coast, Yucatán,
México. Geology 17, no. 9: 818–821.
Perry, E., L. Marín, J. McClain, and G. Velazquez. 1995.
Ring of cenotes (sinkholes), northwest Yucatán, México: Its
hydrogeologic characteristics and possible association with
the Chicxulub impact crater. Geology 23, no. 1: 17–20.
Pope, K.O., A.C. Ocampo, and C.E. Duller. 1991. Mexican site
for k/t impact crater? Nature 351: 105.
Rebolledo-Vieyra, M., J. Urrutia-Fucugauchi, L.E. Marín, A.
Trejo-García, V.L. Sharpton, and A.M. Soler-Arechalde.
2000. Unam scientific shallow-drilling program of the
Chicxulub impact crater. International Geology Review 42,
no. 10: 928–940.
Rebolledo-Vieyra, M., J. Urrutia-Fucugauchi, and H. López-
Loera. 2010. Aeromagnetic anomalies and structural model
of the Chicxulub multiring impact crater, Yucatán, México.
Revista Mexicana de Ciencias Geológicas 27, no. 1:
185–195.
10 M. Gómez-Nicolás et al. Groundwater
Russell, B. 2017. Dr. Bradley Russell personal page. https://
strose.academia.edu/BradleyRussell (accessed July 2017).
Schmitter-Soto, J.J., F.A. Comín, E. Escobar-Briones, J.
Herrera-Silveira, J. Alcocer, E. Suárez-Morales, M. Elías-
Gutiérrez, V. Díaz-Arce, L.E. Marín, and B. Steinich. 2002.
Hydrogeochemical and biological characteristics of cenotes
in the Yucatán peninsula (SE México). Hydrobiologia 467,
no. 1–3: 215–228.
SEDUMA. 2016. Secretaría de desarrollo urbano y medio
ambiente de Yucatán (Spanish) [Secretariat of urban
development and environment of Yucatán]. http://
www.seduma.yucatan.gob.mx/ (accessed March 2016).
Steinich, B., and L.E. Marín. 1996. Hydrogeological investi-
gations in northwestern Yucatán, México, using resistivity
surveys. Groundwater 34, no. 4: 640–646.
Steinich, B., and L.E. Marín. 1997. Determination of flow
characteristics in the aquifer of the northwestern peninsula
of Yucatán, México. Journal of Hydrology 191, no. 1:
315–331.
Stummer, P., H. Maurer, and A.G. Green. 2004. Experimental
design: electrical resistivity data sets that provide optimum
subsurface information. Geophysics 69, no. 1: 120–139.
Telford, W.M., L.P. Geldart, and R.E. Sheriff. 1990. Applied
Geophysics, 2nd ed. Cambridge, UK: Cambridge Univer-
sity Press.
Zhu, J., J.C. Currens, and J.S. Dinger. 2011. Challenges of using
electrical resistivity method to locate karst conduits—A
field case in the inner bluegrass region, Kentucky. Journal
of Applied Geophysics 75, no. 3: 523–530.
NGWA.org