Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Geocarto International

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tgei20

Comparison of ANN model and GIS tools for


delineation of groundwater potential zones,
Fincha Catchment, Abay Basin, Ethiopia

Habtamu Tamiru & Meseret Wagari

To cite this article: Habtamu Tamiru & Meseret Wagari (2021): Comparison of ANN model and
GIS tools for delineation of groundwater potential zones, Fincha Catchment, Abay Basin, Ethiopia,
Geocarto International, DOI: 10.1080/10106049.2021.1946171

To link to this article: https://doi.org/10.1080/10106049.2021.1946171

Published online: 12 Jul 2021.

Submit your article to this journal

Article views: 11

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tgei20
GEOCARTO INTERNATIONAL
https://doi.org/10.1080/10106049.2021.1946171

Comparison of ANN model and GIS tools for delineation


of groundwater potential zones, Fincha Catchment, Abay
Basin, Ethiopia
Habtamu Tamirua and Meseret Wagarib
a
Department of Water Resources and Irrigation Engineering, Wollega University, Nekemte, Ethiopia;
b
Department of Natural Resources Management, Wollega University, Nekemte, Ethiopia

ABSTRACT ARTICLE HISTORY


In this article, the novelty of Artificial Neural Networks (ANN) Received 7 May 2021
model and GIS platform for the delineation of groundwater Accepted 14 June 2021
potential zones were compared in Fincha Catchment, Abay Basin,
KEYWORDS
Ethiopia. LULC, rainfall, soil, geology, drainage density, lineament
ANN model; comparison;
density and geomorphologic units were used as key factors in delineation; GIS tools;
both models. Weights were generated in ANN and Analytical groundwater
Hierarchy Process (AHP) to delineate the groundwater potential
zones. Groundwater potential zones with five and four categories
have been delineated in the ANN and GIS tools, respectively. The
potential zones were validated by overlapping the existing well
locations with an overall accuracy of 85% and 82.5% in ANN and
GIS tools, respectively. The ANN model revealed better perform-
ance in the delineation of groundwater potential zones in this
catchment when compared with GIS tools. Therefore, the
delineated groundwater potential zones will be valuable in solv-
ing the problem of drinking water in the catchment.

Introduction
Human life development began where water resources were available (Melesse and Abtew
2015). The quality and quantity issues of water for drinking are the major problems hap-
pening worldwide, and these problems are very visible in poor countries (Shomar et al.
2010; Rani et al. 2015). The importance of clean and potable drinking water is intensely
increasing due to population growth, and this is a common problem for sub-Saharan
countries, mainly in Ethiopia (Payne et al. 2013). Surface water sources can only be uti-
lized if the topography where the community is living and the source of water are easily
linked, financially affordable to treat the water and the source is available (Arefayne
Shishaye and Abdi 2015; Zhang et al. 2016). Groundwater source is an alternative source
of drinking water that can be utilized with requires less treatment and development cost.
Delineation of groundwater potential zones is the best compromising solution, where the
availability of surface sources (Dibaba et al. 2020) and financial affordability are very diffi-
cult. To utilize groundwater sources, first, the potential zones must be delineated. To
explore groundwater potential zones, different methods such as the areal method, surface

CONTACT Habtamu Tamiru habtemodel1345@gmail.com


ß 2021 Informa UK Limited, trading as Taylor & Francis Group
2 H. TAMIRU AND M. WAGARI

method, subsurface method and esoteric method are the mainly used methods globally
(Das and Pardeshi 2018). However, the integration of surface and subsurface methods are
the best and appropriate methods in which different geophysical factors are used to deter-
mine the subsurface phenomena. The geophysical factors such as slope, drainage density,
lineament density, geological characteristics, geomorphologic unit, land use land cover
and soil are an indicator for the occurrence of water in the aquifer (Li et al. 2014). The
majority of studies conducted in different parts of the world focused on the application of
Geographical Information System (GIS) and remotes sensing technology as an effective
tool to delineate the potential zones (Al-Adamat 2012; Nagaraju et al. 2016; Arulbalaji
et al. 2019). The application of GIS and remote sensing for the delineation of ground-
water potential zones is mainly focused on the remotely sensed data available online. The
accuracy of these data is relatively low when compared to the ground surveyed data
(Arefayne Shishaye and Abdi 2015). Commonly the data such as LANDSAT imagery and
Digital Elevation Models (DEM) are the primary data for exploration of groundwater
potential zones when GIS and Remote sensing are used as a method of potential zones
selection (Al-Nahmi et al. 2016; Akinluyi et al. 2018). However, these data are taken by
remote sensing technology in which cameras on satellites and airplanes take images of
large areas on the earth’s surface, and which doesn’t mean that the image can represent
the real world (Romilly and Gebremichael 2011). The uncertainty in the remote sensing
data will propagate throughout the data analysis and will disturb the result (Chang et al.
2007). Therefore, this article aims to improve this uncertainty by training the model with
the ground survey data. The novelty of Artificial Neural Networks (ANN) in this article is
presented by linking the real world on the ground with the remotely sensed data. In this
study, field-based data such as existing water sources (shallow and hand-dug well loca-
tions), the supervised LULC classification, slope developed from verified DEM, the soil
map prepared from the field surveyed data, Drainage density, Lineament density and geo-
morphologic unit were used as training data for identification of potential zones.
Therefore, two thematic maps were generated; the first one is the thematic maps gener-
ated using GIS and Remote sensing technology from the remotely sensed data, in this
case, data were used as it is without verifying with ground-based data. The second the-
matic maps were generated from the data prepared from the field survey. The weights in
ANN were assigned based on random initialization (optimized), while Analytical
Hierarchy Process (AHP) was applied for fixing weights in the delineation of groundwater
potential zones using GIS and Remote sensing.

Method and materials


Description of the study area
This study area is conducted in the Fincha catchment, Abay River basin and particularly
the study under consideration was delineated fixing the outlet at Fincha River. In this
study area (Figure 1), the delineated watershed is geographically found between
37 0.060 0000 E to 37 330 1800 E longitude and 09 210 1100 N to 10 010 0000 N latitude. The
delineated watershed has a total area of 82.11 km2.

GIS and remote sensing application in water resources


Geographical Information Systems (GIS) and Remote sensing techniques are being used by
many researchers for exploration of groundwater (Vijay et al. 2011; Rajaveni et al. 2017).
GEOCARTO INTERNATIONAL 3

Figure 1. River networks, Rainfall stations and Elevation in Fincha catchment.

The application of GIS and Remote sensing is a powerful tool for the delineation of
groundwater potential zones (Berhanu and Hatiye 2020; Tolche 2021). This method links
both the surface and subsurface significant factors such as slope, geology, rainfall, geomor-
phological units, lineament density, lithology and soil to delineate the potential zones
(Rajaveni et al. 2017; Berhanu and Hatiye 2020). The application of GIS and remote sens-
ing can minimize the cost and time desired during the exploration (Costache et al. 2020).
GIS and remote sensing technology are powerful method for a region where there is no
skilled manpower and financial affordability is very difficult for field-based investigation,
and this is a common problem poor country. The other importance of GIS and Remote
sensing is the scarcity of field-based data (Mideksa and Temesgen 2020). Exploration of
groundwater potential zones requires ground-based data such as geology, soil, lithology,
geomorphologic units and gathering cost and experienced person are available (Andualem
and Demeke 2019). The occurrence of the Groundwater potential zone is a function of
the aquifer characteristics, and its spatial extent. Therefore, the aquifer is the primary fac-
tor that determines the degree of occurrence of groundwater in a given watershed.
Rainfall, LULC, lineament density, drainage density, geology, slope, soil and geomorpho-
logic units are the deriving factors of groundwater potential zones considered in this
study (Figure 2).
The degree of significance of the driving factors is expressed by weights (Chuma et al.
2013; Atmaja et al. 2019). However, the driving factors such as LULLC, slope, soil, geol-
ogy, drainage density and lineament density derived from Remote sensing are the com-
monly used driving factors (Figure 3) for the delineation of groundwater potential zones
4 H. TAMIRU AND M. WAGARI

Figure 2. GIS and remote sensing-based delineation of groundwater potential zones.

Figure 3. Groundwater driving thematic map using GIS and remote sensing technique
GEOCARTO INTERNATIONAL 5

Table 1. Analytical hierarchy process scale and judgment.


Scale Judgment
1 Equal importance
3 Moderate importance one over the over
5 Essential or strong importance
7 Very strong or demonstrated importance
9 Extreme or absolute importance
2,4,6,8 Intermediate values between the two adjacent judgments

Table 2. Random Index (RI) for a number of attributes.


Attributes 3 4 5 6 7 8 9 10
RI 0.52 0.89 1.11 1.25 1.35 1.4 1.45 1.49

(Berhanu and Hatiye 2020). The integration of these significant factors using the weighted
overlay analysis and sensitivity analysis in AHP can appropriately be used in exploring
the potential zones. The Analytical Hierarchy Process (AHP) principle was applied to
assign weights through scale and judgment (Arefayne Shishaye and Abdi 2015). Then, set-
ting up priorities between the elements to be compared based on the scale, pair-wise com-
parison matrix for scale consists of the relative importance of the criterion from 1 (equal
importance) to 9 (extreme preference of one factor over the other) was applied (Table 1).
The consistency of the weights derived from the pair-wise matrix should be checked
and this improve the accuracy of the decision to be made in AHP method. The consist-
ency of the derived values of weights is checked by reducing the error in the estimation
and this can be achieved by the method called Consistency Index (CI) and Consistency
Ration (CR) as shown in Eqs. (1) and (2).
kmax n
ConsistencyIndexðCIÞ ¼ (1)
n1
CI
ConsistencyRatioðCRÞ ¼ (2)
RI
Where, kmax is the maximum eigen value of the pair-wise matrix, n is the number of
criteria used in the pairwise comparison, RI is a random Index for a number of an attrib-
utes used in the evaluation (Table 2). The CR > 0.1 indicates the correctness of weights
assigned in AHP.

ANN model and GIS tools


Artificial Neural Networks (ANN) is a widely used machine-learning approach that is get-
ting international attention. This approach is intensively used in areas of water resources
management, utilization and modelling (Grimes et al. 2003; Chang et al. 2007;
Shamseldin 2010; Ahmed et al. 2016). A neural network is a machine-learning that
focuses on an information processing algorithm to solve a non-linear nature of the water
resources (Shamseldin and O’Connor 2001; Campolo et al. 2003; Ahmed et al. 2016;
Barbetta et al. 2016).
ANN are used for modelling non-linear problems and to predict the output values for
given input parameters from their training values (Dogan et al. 2007; Mustafa et al. 2012;
Asadi 2013; Ligaray et al. 2015). The selection of a specific model and its accuracy is gen-
erally governed by factors such as availability of input parameters, skilled manpower with
the watershed. In this study, the novelty of ANN model application for delineation of
6 H. TAMIRU AND M. WAGARI

Figure 4. Flowchart for the delineation of groundwater potential zones.

groundwater potential zones and the result was compared with the potential zones gener-
ated in GIS and Remote sensing technique. The general flowchart of the ANN model and
GIS tool used for the delineation of groundwater potential zones in the study area is pre-
sented in Figure 4.
An Artificial Neural Network (ANN) is one of the emerging methods of simulating the
non-linear and complex relationship between parameters and computes a multivariate
space of information to another in a given watershed (Lv et al. 2020). What makes differ-
ent the ANN model from other traditional models is that this approach can simply learns
from the existing situation and easily associates it with the sample data through training
processes (Chan and Chan 2020). The main objective of using the ANN model is to build
an algorithm of data-generating processes; therefore, the network can generalize and pre-
dict an output from sample input data. Artificial Neural Networks (ANN) are robust tools
for modelling many water resources such as groundwater modelling, surface water resour-
ces allocation and water quality simulation (Tsakiri et al. 2018). The significant factors
and other examples such as productive well sites were used as input to train the ANN
model. The neural Networks (NN) structure selected in this study is a multi-layer percep-
tron (MLP) consisting of an input layer with eight (8) input nodes, a hidden layer with
four (4) hidden nodes and an output layer with one (1) output node. In the ANN model,
the initial weights were assigned and the training process was started. The assigned
weights in the input nodes were multiplied with the pixel-based (12.5  12.5) point vector
data and the summation of these parameters sent to hidden nodes. Logically, the random
initialized values of weights and the nature of the input in the first instance are squashed
into between 0 and 1, which is called normalization. The normalization of input data in
the input nodes and the activation of the multiplied sum of weights and input were done
in the hidden layers (Simor et al. 2012). The normalization and activation equations for
these purposes were as given in (Eqs. (3) and (4)) as used in (Arun and Baskaran 2013).
XXmin
Normalization ¼ (3)
Xmax  Xmin
Where X is the pixel value in the vector point data, Xmin and Xmax the minimum and
maximum values in the point-based point values.
GEOCARTO INTERNATIONAL 7

Figure 5. Spatial Grids with Resolution of 50 m 3 50 m input parameters as point data.

1
Activation function ¼ (4)
1 þ ex
Where X is the total sum of the multiplied values of the normalized input and
assigned weights
The study under consideration has the total area of 82.11 km2.and converted into 30 m
 30 m spatial resolution. A pixel-to-pixel grid values (Figure 5) for factor such as slope,
rainfall, drainage density and lineament density were used whereas grid codes for LULC,
geology, Geomorphology and soil were used since these parameters have no specific pixel
value like slope or rainfall (Li et al. 2014).
In the feedforward propagation, the input parameters are pushed forward to get the
rough solution at the output node and it doesn’t take any account to minimize the error
between the result obtained from network and target output (Malmgren and Nordlund
1996; Tayebiyan et al. 2016; Dar 2017). The initial weight values assigned in the feed-
forward processes are just to start the modelling, and the accuracy of the model is very
low at this stage (Shamseldin and O’Connor 2003). The main importance of backpropaga-
tion (Figure 6) is to spread the error back into the networks to minimize the error
obtained in the feedforward process (Timbadiya et al. 2011; Sattari et al. 2017). The over-
all error obtained at the output layer starts to propagate back into the networks from the
output node to the entire networks (Mai and De Smedt 2017). Training in a sense mean-
ing that the network learns from the mistakes through the built-in learning algorithm in
ANN networks (Abhishek et al. 2012; Szegedy et al. 2014).

Identification of groundwater potential zones


The groundwater potential zones were delineated and mapped by overlapping the the-
matic layers of deriving factors. The potential zones were assessed based on the
Groundwater Potential Index (GWPI) and this index was widely used in Ethiopia
(Andualem and Demeke 2019; Berhanu and Hatiye 2020; Tolche 2021). Index of ground-
water potential (Eqn. 3) is a unitless parameter used to index the probability of ground-
water potential zones’ presence in a given catchment (Kumar and Srinivasan 2016). The
subsequent map is classified into quantitative classification such as good, moderate and
8 H. TAMIRU AND M. WAGARI

Figure 6. Flowchart of ANN model for the delineation of Groundwater Potential zones.

Table 3. GWPI and Analytical Hierarchy Process (AHP) with the description.
Class Potential zones Description
1 Very Low The chance of getting groundwater is very low
2 Low The chance of getting groundwater is low
3 Moderate The chance of getting groundwater is modest
4 High The chance of getting groundwater is better
5 Very High The chance of getting groundwater is best

poor zones. The explanation for the GWPI and AHP is presented in Table 3.
GWPI ¼ Rw Rr þ LULCw LULCr þ Slw Slr þ Ddw Ddr þ Sow Sor þ Gew Ger
þ Gmw Gmr þ Ldw Ldr (3)
Where Rr is the reclassified rainfall, LULCr is the reclassified land use/land cover, Slr is
the reclassified slope, Ddr is the reclassified drainage density, Sor is the reclassified soil,
Ger is the reclassified Geology, Gmr is the reclassified geomorphologic units and Ldr is the
reclassified lineament density. The indices Rw, LULCw, Slw, Ddw, Sow, Gew, Gmw and Ldw
are the normalized weights assigned for each key factor.

Result and discussions


For this particular study, eight groundwater deriving factors such as slope, rainfall, geol-
ogy, soil, LULC, Geomorphology, drainage density and lineament and the corresponding
thematic maps generated in the ANN model and GIS tools were presented in Figures
7–14. In this catchment, groundwater potential zones with five and four, a qualitative-
based categories have been identified in ANN model and GIS platform. As we can see
from the delineated groundwater potential zones presented in (Figure 15), the down-
stream parts of the catchment revealed high potential zones in both methods; however, a
very high prospect zones were only delineated in ANN model. The absence of the excel-
lent prospect zones (very high) category in GIS tool indicates that the interrelationship
between the groundwater significant factors is not properly captured when compared with
that of ANN model, which leads to a higher rate of delineating potential zones
(Figure 16). The probability of getting groundwater potential zones with five classifications
was obtained in the ANN model, whereas four classifications were obtained in GIS tools.
GEOCARTO INTERNATIONAL 9

Figure 7. Reclassified Rainfall thematic maps (a) GIS tools and (b) ANN model.

Figure 8. Reclassified LULC thematic maps a) GIS tools, b) ANN model.


10 H. TAMIRU AND M. WAGARI

Figure 9. Reclassified Drainage Density thematic maps a) GIS tools, b) ANN model.

Figure 10. Reclassified Lineament Density thematic maps a) GIS tools, b) ANN model.
GEOCARTO INTERNATIONAL 11

Figure 11. Reclassified Slope thematic maps (a) GIS tools and (b) ANN model.

Figure 12. Reclassified Slope thematic maps (a) GIS tools and (b) ANN model.
12 H. TAMIRU AND M. WAGARI

Figure 13. Reclassified Slope thematic maps (a) GIS tools and (b) ANN model.

Figure 14. Reclassified Geomorphologic units (a) GIS tools and (b) ANN model.
GEOCARTO INTERNATIONAL 13

Figure 15. Delineated Groundwater potential zones in (a) GIS tools and (b) ANN model.

Figure 16. The delineated groundwater potential zones and well locations (a) ANN model and (b) GIS tools.
14 H. TAMIRU AND M. WAGARI

Rainfall
Rainfall is the main source of groundwater. The intensity and duration of the precipita-
tion have a paramount role to increase the volume of the water stored in an aquifer
(Scanlon et al. 2005; Dibaba et al. 2020). For potential aquifer, much amount of water is
stored when the rain reached the water table. Therefore, if a sufficient duration and high
intensity of rainfall are there is a catchment, the possibility of getting potential zones for
groundwater exploration is very high. The result of reclassified rainfall maps generated in
the ANN model and GIS tools are presented in (Figure 7(b,a)), respectively. As we can
see from the results, the reclassified rainfall thematic layers are almost the same and this
indicates that both methods are able to generate the rainfall classifications equally.
The novelty of the ANN model in generating a thematic map from rainfall records is
seen as an alternative way of preparing significant factors in delineating groundwater
potential zones.

Land use/land cover


The reclassified LULC thematic layers created in the ANN model and GIS tools are pre-
sented in (Figure 8(b,a)). The reclassified thematic maps were categorized into five classifi-
cations and almost the same results were obtained both in ANN model and GIS tools. As
we can see from the generated thematic maps, five land use/land cover classifications
were dominantly identified. The qualitative classifications (Very poor ¼ 1, Poor ¼ 2,
Moderate ¼ 3, High ¼ 4 and Very High ¼ 5) based on rate of groundwater recharging
potential of the LULC were identified.

Drainage density
The length of the stream channel and the land surface drained across its length is an indi-
cator for groundwater potential zones. The potential of the aquifer is based on the density
of drainage and the climatic conditions of the watershed. Moisture holding capacity and
the degree of permeability of the geologic formation of the drainage initiates the infiltra-
tion capacity through high runoff (Andualem and Demeke 2019). The drainage density
prepared from DEM of (12.5 m  12.5) in the ANN model and GIS tools were presented
in (Figure 9(b,a)), respectively.

Lineament density
Lineaments features that distributed across the surface are considered as the key factor in
this study. The thematic maps of Lineament density generated in GIS tools and ANN
model was presented in (Figure10(a,b)). The data retrieved from the remote sensing tech-
nology (LANDSAT 8) were used to identify the satellite imagery by their relatively linear
alignments (Berhanu and Hatiye 2020), and these alignments express the surface topog-
raphy of the underlying structural features (Andualem and Demeke 2019).These features
can be attributed to the lineaments and artificial recharge structures such as fractures and
faults and also a large amount of sand which allows quick percolation of water down. The
significant factors such as slope (Figure 11(a,b)), soil (Figure 12(a,b)), Geology
(Figure 13(a,b)) and geomorphologic units (Figure 14(a,b)) are presented.
The pair-wise comparison matrix developed in AHP and the overall weights for the
evaluated key factors for the delineation of groundwater potential zones in this study is
GEOCARTO INTERNATIONAL 15

Table 4. Pair-wise comparison matrix of the key factors.


R LULC SL DD SO GE GM LD
R 1.00 3.00 0.25 5.00 2.00 0.33 0.17 3.00
LULC 0.33 1.00 0.50 0.33 5.00 4.00 0.13 4.00
SL 4.00 2.00 1.00 2.00 0.25 0.33 0.33 0.20
DD 0.20 3.00 0.50 1.00 0.33 0.50 2.00 3.00
SO 0.50 0.20 4.00 3.00 1.00 3.00 5.00 0.14
GE 0.08 0.19 0.07 0.21 0.38 1.00 0.19 0.03
GM 6.00 8.00 3.00 0.50 0.20 0.33 1.00 0.25
LD 0.33 0.25 5.00 0.33 7.00 0.20 4.00 1.00
Col. Total 12.44 17.64 14.32 12.37 16.16 9.69 12.82 11.62
Normalized Pair-Wise Comparison Matrix
R LULC SL DD SO GE GM LD
R 0.06 0.17 0.02 0.41 0.11 0.07 0.01 0.17
LULC 0.02 0.06 0.03 0.03 0.27 0.15 0.01 0.23
SL 0.23 0.11 0.07 0.16 0.01 0.31 0.02 0.01
DD 0.01 0.17 0.03 0.08 0.02 0.47 0.14 0.17
SO 0.03 0.01 0.28 0.24 0.05 0.35 0.34 0.01
GE 0.08 0.19 0.07 0.21 0.38 0.23 0.04 0.19
GM 0.35 0.45 0.21 0.04 0.01 0.01 0.07 0.01
LD 0.02 0.01 0.34 0.03 0.37 0.14 0.27 0.06
Normalized Sum of Rows Normalized average Rows Eigenvector
R 0.95 0.95/8 0.12
LULC 0.81 0.81/8 0.10
SL 0.85 0.85/8 0.11
DD 1.09 1.09/8 0.14
SO 0.98 0.98/8 0.12
GE 1.07 1.07/8 0.13
GM 1.17 1.17/8 0.15
LD 1.12 1.12/8 0.14
R ¼ rainfall, LULC ¼ land use/land cover, SL ¼ slope, DD ¼ drainage density, SO ¼ soil, GE ¼ geology,
GM ¼ geomorphologic units and LD ¼ lineament density.
k ¼ 8.125, n ¼ 8, CI (consistency index) ¼ 0.017, RI (random index) ¼ 1.41, CR ¼ 0.01.

Table 5. Updated weights after backpropagation in ANN model.


Factors Updated weights
Rainfall 0.35
LULC 0.19
Slope 0.27
Drainage density 0.28
Soil 0.07
Geology 0.31
Geomorphology 0.06
Lineament density 0.13

summarized in Table 2. The consistency of the AHP technique in capturing the explor-
ation of groundwater potential zones is evaluated by consistency Index (CI) and consist-
ency Ratio (CR) and the summary of these consistency evaluation is presented in Table 4.
As we can see from the table, the CI > 0.1, and this indicates that the values of weight
assigned for the individual key factor in AHP are correct (Nagaraju et al. 2016; Mandal
et al. 2018; Dinka 2020).
The updated weights used in the delineation of potential zones after training in the
ANN model are summarized in Table 5. The final updated weights are reallocated in the
neural networks. Once the training process is reached, the updated weights were fixed
and used to generate the thematic layers for the key factors and the thematic maps gener-
ated in this way were overlaid for groundwater potential zonation.
16 H. TAMIRU AND M. WAGARI

Generation of groundwater potential zones


The delineated Groundwater Potential zones using GIS tools and ANN model were pre-
sented in Figure 15. As we can see from Figure 15(a), delineated potential zones were
only classified into four qualitative-based classifications as High, Moderate, Low and Very
low, whereas, in Figure 14(b), five groundwater potential zones were identified including
very good qualitative classification. The probability of getting groundwater potential zones
with five classifications is obtained in the ANN model, whereas only four classifications
were obtained in GIS tools.

Validation of the potential zones


The accuracy of the delineated groundwater potential zones further evaluated with exist-
ing water sources (Duan et al. 2016; Das and Pardeshi 2018; Berhanu and Hatiye 2020).
The total number of 51 well locations were collected from the local government office
and out of this, only 36 well have potential yield and the rest are dried. As we can see
from the validated potential zones presented in Figure 15, the majority of existing water
sources such as shallow and hand-dug wells are almost overlapped up to 86% and 82.5%
with the delineated potential zones in ANN model and GIS tools, respectively.

Conclusion
In this article, the novelty of the ANN model and GIS tools for delineation of
Groundwater Potential Zones in Fincha Catchment, Abay Basin, in Ethiopia is presented.
ANN model and GIS tools were applied for the delineation of Groundwater potential
zones using the parameters such as rainfall, slope, LULC, soil, drainage density, lineament
density, geology and Geomorphology as influencing factors. The training was done in the
ANN model using the information obtained from significant factors and mimicking with
the existing groundwater sources as training inputs to train the model to capture the
potential zones, Whereas, weighted overlay analysis for the significant factors using AHP
was implemented in GIS tools. Thematic maps were generated in both methods and the
final groundwater potential zones were developed using the updated weights in the ANN
model and the weighted sum overlay analysis in GIS was used, respectively. The delineated
groundwater potential zones obtained in the ANN model revealed that a better chance of get-
ting potential zones classified under five qualitative classifications as Very High, High,
Moderate, Low and Very low were identified whereas the delineated groundwater potential
zones obtained in GIS tools revealed only the chance of getting High, Moderate, Low and
Very low potential zones. The delineated groundwater potential zones obtained on both meth-
ods were validated based on the existing water supply sources (wells location). According to
the comparison made between the results obtained in both models and the existing water
sources, the ANN model showed better agreement based on the overlapping percentages of
potential well and the qualitative classifications. The accuracy and high probability of getting
groundwater potential zones in the case of the ANN model is better than GIS tools.
Therefore, it is logically acceptable that the ANN model is more powerful and accurate in the
delineation of groundwater potential zones than GIS tools where the data are scarce. Finally, it
was concluded that the application of the Artificial Neural Networks (ANN) model in water
resources development is a powerful tool for the delineation of groundwater potential zones in
the region where skilled manpower and financial affordability are very limited.
GEOCARTO INTERNATIONAL 17

Disclosure statement
No potential conflict of interest was reported by the author(s).

References
Abhishek K, Kumar A, Ranjan R, Kumar S. 2012. A rainfall prediction model using artificial neural net-
work. In Proceedings – 2012 IEEE Control and System Graduate Research Colloquium, ICSGRC 2012;
February 2016; Amsterdam: Elsevier Ltd; p. 82–87.
Ahmed S, Ghumman AR, Ahmad I, Ahsan M, Ahmad S, Ghumman AR, Ahmad I, Khan KA, Ahsan M.
2016. Data Driven Modelling for Real-time Flood Forecasting. In 2nd International Multi-Disciplinary
Conference, Gujrat, Pakistan, December. University of Lahore; p. 1–8.
Akinluyi FO, Olorunfemi MO, Bayowa OG. 2018. Investigation of the influence of lineaments, lineament
intersections and geology on groundwater yield in the basement complex terrain of Ondo State. Appl
Water Sci. 8(1):1–13.
Al-Adamat R. 2012. The use of GIS and Google Earth for preliminary site selection of groundwater
recharge in the Azraq Oasis Area—Jordan. JWARP. 04(06):395–399.
Al-Nahmi F, Alami OB, Baidder L, Khanbari K, Rhinane H, Hilali A. 2016. Using remote sensing for lin-
eament extraction in Al Maghrabah area - Hajjah, Yemen. Int Arch Photogramm Remote Sens Spatial
Inf Sci. XLII-2/W1(2W1):137–142.
Andualem TG, Demeke GG. 2019. Groundwater potential assessment using GIS and remote sensing: a
case study of Guna tana landscape, upper blue Nile Basin, Ethiopia. J Hydrol: Reg Stud. 24:100610.
Arefayne Shishaye H, Abdi S. 2015. Groundwater exploration for water well site locations using geophys-
ical survey methods. Hydrol Current Res. 07(01):1–7.
Arulbalaji P, Padmalal D, Sreelash K. 2019. GIS and AHP techniques based delineation of groundwater
potential zones: a case study from Southern Western Ghats. Sci Rep. 9(1):1–17.
Arun BS, Baskaran K. 2013. Design and Development of Artificial Neural Networking (ANN) system using
sigmoid activation function to predict annual rice production in Tamilnadu. IJCSEIT. 3(1):13–31.
Asadi A. 2013. Application of HEC-HMS for Flood Forecasting in Kabkian Basin and Delibajak Subbasin
in Iran. IOSRJEN. 03(09):10–16.
Atmaja RRS, Putra DPE, Setijadji LD. 2019. Delineation of groundwater potential zones using remote
sensing, GIS, and AHP techniques in southern region of Banjarnegara, Central Java, Indonesia. In
Sixth Geoinformation Science Symposium, Indonesia, November; Stuart Phinn, Ammar A. Aziz, Proc.
of SPIE; p. 23.
Barbetta S, Coccia G, Moramarco T, Todini E. 2016. Case study: a real-time flood forecasting system with
predictive uncertainty estimation for the Godavari River, India. Water (Switzerland). 8(10):463.
Berhanu KG, Hatiye SD. 2020. Identification of groundwater potential zones using proxy data: case study
of Megech watershed, Ethiopia. J Hydrol: Regional Stud. 28(January):100676.
Campolo M, Soldati A, Andreussi P. 2003. Artificial neural network approach to flood forecasting in the
River Arno. Hydrol Sci J. 48(3):381–398.
Chan VKH, Chan CW. 2020. Towards explicit representation of an artificial neural network model: com-
parison of two artificial neural network rule extraction approaches. Petroleum. 6(4):329–339.
Chang FJ, Chiang YM, Chang LC. 2007. Multi-step-ahead neural networks for flood forecasting. Hydrol
Sci J. 52(1):114–130.
Chuma C, Orimoogunje OOI, Hlatywayo DJ, Akinyede JO. 2013. Application of remote sensing and geo-
graphical information systems in determining the groundwater potential in the crystalline basement of
Bulawayo Metropolitan Area, Zimbabwe. ARS. 02(02):149–161.
Costache R, Bao Pham Q, Corodescu-Roşca E, C^ımpianu C, Hong H, Thi Thuy Linh N, Ming Fai C,
Najah Ahmed A, Vojtek M, Muhammed Pandhiani S, et al. 2020. Using GIS, remote sensing, and
machine learning to highlight the correlation between the land-use/land-cover changes and flash-flood
potential. Remote Sens. 12(9):1422.
Dar LA. 2017. Rainfall-runoff modeling using artificial neural network rainfall-runoff modeling using arti-
ficial neural network technique. Int Res J Eng Tech. 4(11).
Das S, Pardeshi SD. 2018. Integration of different influencing factors in GIS to delineate groundwater
potential areas using IF and FR techniques: a study of Pravara basin, Maharashtra. Appl Water Sci.
8(7):1–16.
Dibaba WT, Demissie TA, Miegel K. 2020. Watershed hydrological response to combined land use/land
cover and climate change in highland ethiopia: Finchaa catchment. Water (Switzerland). 12(6):1801.
18 H. TAMIRU AND M. WAGARI

Dinka MO. 2020. Quantification of soil erosion and sediment yield for ungauged catchment using the
RUSLE model: case study for Lake Basaka catchment in Ethiopia. Lakes Reserv. 25(2):183–195.
Dogan E, Isik S, Toluk T, Sandalci M. 2007. Daily streamflow forecasting using artificial neural networks.
Int Congress River Basin Manage. 5(12):449–459.
Duan H, Deng Z, Deng F, Wang D. 2016. Assessment of groundwater potential based on multicriteria
decision making model and decision tree algorithms. Math Prob Eng. 2016:1–12.
Grimes DIF, Coppola E, Verdecchia M, Visconti G. 2003. A neural network approach to real-time rainfall
estimation for Africa using satellite data. J Hydrometeor. 4(6):1119–1133. https://doi.org/10.1175/1525-
7541(2003)004 < 1119:annatr > 2.0.co;2.
Kumar G, Srinivasan D. 2016. Evaluation of groundwater potential index (GWPI) using geophysical sur-
vey in Kallar Watershed, Tamil Nadu, India. Int J Earth Sci Eng. 9(5):1902–1906.
Li X, Li G, Zhang Y. 2014. Identifying major factors affecting groundwater change in the North China
plain with grey relational analysis. Water (Switzerland). 6(6):1581–1600.
Ligaray M, Kim H, Sthiannopkao S, Lee S, Cho KH, Kim JH. 2015. Assessment on hydrologic response
by climate change in the Chao Phraya River basin, Thailand. Water (Switzerland). 7(12):6892–6909.
Lv Z, Zuo J, Rodriguez D. 2020. Predicting of runoff using an optimized SWAT-ANN: a case study. J
Hydrol: Reg Stud. 29:100688.
Mai DT, De Smedt F. 2017. A combined hydrological and hydraulic model for flood prediction in
Vietnam applied to the Huong river basin as a test case study. Water (Switzerland). 9(11):879.
Malmgren BA, Nordlund U. 1996. Application of artificial neural networks to chemostratigraphy.
Paleoceanography. 11(4):505–512.
Mandal B, Dolui G, Satpathy S. 2018. Land suitability assessment for potential surface irrigation of river
catchment for irrigation development in Kansai watershed, Purulia, West Bengal, India. Sustain Water
Resour Manag. 4(4):699–714.
Melesse AM, Abtew W. 2015. Landscape dynamics, soils and hydrological processes in varied climates,
v–vi. Basel, Switzerland: MDPI.
Mideksa G, Temesgen T. 2020. Irrigation water potential and land suitability assessment in Kurfa Chele.
Turkish JAF Scitech. 8(1):139–146.
Mustafa MR, Isa MH, Rezaur RB. 2012. Artificial neural networks modeling in water resources engineer-
ing : infrastructure and applications. World Acad Sci, Eng Technol. 6(2):341–349.
Nagaraju A, Sreedhar Y, Thejaswi A, Dash P. 2016. Integrated approach using remote sensing and GIS
for assessment of groundwater quality and hydrogeomorphology in certain parts of Tummalapalle
Area, Cuddapah District, Andhra Pradesh, South India. ARS. 05(02):83–92.
Payne SM, Magruder IA, Woessner WW. 2013. Application of a groundwater classification system and
GIS mapping system for the lower Ruby Valley Watershed, Southwest Montana. JWARP. 05(08):
775–791.
Rajaveni SP, Brindha K, Elango L. 2017. Geological and geomorphological controls on groundwater
occurrence in a hard rock region. Appl Water Sci. 7(3):1377–1389.
Rani VR, Pandalai HS, Sajinkumar KS, Pradeepkumar AP. 2015. Geomorphology and its implication in
urban groundwater environment: case study from Mumbai. Appl Water Sci. 5(2):137–151.
Romilly TG, Gebremichael M. 2011. Evaluation of satellite rainfall estimates over Ethiopian river basins.
Hydrol Earth Syst Sci. 15(5):1505–1514.
Sattari M-T, Rezazadeh-Joudi A, Kusiak A. 2017. Assessment of different methods for estimation of miss-
ing data in precipitation studies. Hydrol Res. 48(4):1032–1044.
Scanlon BR, Reedy RC, Stonestrom DA, Prudic DE, Dennehy KF. 2005. Impact of land use and land
cover change on groundwater recharge and quality in the southwestern US. Global Change Biol.
11(10):1577–1593.
Shamseldin AY, O’Connor KM. 2003. A “consensus” real-time river flow forecasting model for the Blue
Nile River. Vol. 281. IAHS-AISH Publication; p. 82–89.
Shamseldin AY. 2010. Artificial neural network model for river flow forecasting in a developing country. J
Hydroinformatics. 12(1):22–35.
Shamseldin AY, O’Connor KM. 2001. A non-linear neural network technique for updating of river flow
forecasts. Hydrol Earth Syst Sci. 5(4):577–598.
Shomar B, Fakher SA, Yahya A. 2010. Assessment of groundwater quality in the Gaza Strip, Palestine
using GIS mapping. JWARP. 02(02):93–104.
Simor V, Hlavcova K, Kohnova S, Szolgay J. 2012. Application of Artificial Neural Networks for estimat-
ing index floods. Contributions GeophysGeodesy. 42(4):295–311.
Szegedy C, Zaremba W, Sutskever I, Bruna J, Erhan D, Goodfellow I, Fergus R. 2014. Intriguing proper-
ties of neural networks. arXiv. preprint. https://doi.org/978-0131471399.
GEOCARTO INTERNATIONAL 19

Tayebiyan A, Mohammad TA, Ghazali AH, Mashohor S. 2016. Artificial neural network for modelling
rainfall-runoff. Pertanika J Sci Technol. 24(2):319–330.
Timbadiya PV, Patel PL, Porey PD. 2011. Calibration of HEC-RAS model on prediction of flood for
Lower Tapi River. JWARP. 03(11):805–811.
Tolche AD. 2021. Groundwater potential mapping using geospatial techniques: a case study of Dhungeta-
Ramis sub-basin, Ethiopia. Geol, Ecol Landscapes. 5(1):16–65.
Tsakiri K, Marsellos A, Kapetanakis S. 2018. Artificial neural network and multiple linear regression for
flood prediction in Mohawk River, New York. Water (Switzerland). 10(9):1158.
Vijay R, Samal D, Mohapatra PK. 2011. GIS based identification and assessment of groundwater quality
potential zones in Puri City. JWARP. 03(06):440–447.
Zhang Y, Zheng H, Chiew FHS, Arancibia JP, Zhou X. 2016. Evaluating regional and global hydrological
models against streamflow and evapotranspiration measurements. J Hydrometeorol. 17(3):995–1010.

You might also like