Environment, Development and Sustainability

https://doi.org/10.1007/s10668-019-00405-5

Estimation of irrigation water quality index

with development of an optimum model: a case study

Sayiter Yıldız1 · Can Bülent Karakuş2

Received: 31 October 2018 / Accepted: 11 June 2019

© Springer Nature B.V. 2019

Abstract
Surface water quality parameters are important means for determination of water’s suitabil-
ity for irrigation. In this research, data from 32 irrigation stations were used to calculate the
sodium adsorption rate (SAR), sodium percentage (Na%), Kelly index (KI), permeability
index (PI) and irrigation water quality index (IWQI) for evaluation of surface water qual-
ity. The obtained SAR, KI and Na% values, respectively, varied between 0.10 and 9.43,
0.03–1.37 meq/l and 3.16–57.82%. The calculated PI values indicate that, 93.75% of the
water samples is in “suitable” category, and 6.25% is in “non-suitable” category. The IWQI
values obtained from the research area varied between 30.59 and 81.09. In terms of irriga-
tion water quality, 12.5% of the samples is of “good” quality, 15.62% is of “poor” qual-
ity, 68.75% is of “very poor” quality, and 3.12% is of “non-suitable” quality. Accordingly,
IWQI value was estimated on the basis of SAR, Na%, KI and PI values using multiple
regression and artificial neural network (ANN) model. The regression coefficient (R2) was
determined as 0.6 in multiple regression analysis, and a moderately significant relationship
(p < 0.05) was detected. As the calculated F value was higher than the tabulated F value, a
real relationship between the dependent and independent variables is inferred. Four differ-
ent models were built with ANN, and the statistical performance of the models was deter-
mined using statistical parameters such as average value (µ), standard error (SE), stand-
ard deviation (σ), R2, root mean square error (RMSE) and mean absolute percentage error
(MAPE). The training R2 value belonging to the best model was found to be significantly
high (0.99). The relation between the estimation results of ANN model and the experimen-
tal data (R2 = 0.92) verifies the model’s success. As a result, ANN proved to be a successful
means for IWQI estimation using different water quality parameters.

Keywords  Artificial neural network · Irrigation water quality index · Multiple regression

* Can Bülent Karakuş

[email protected]
Sayiter Yıldız
[email protected]
1
Environmental Department, Engineering Faculty, Sivas Cumhuriyet University, Sivas, Turkey
2
Urban and Regional Planning Department, Architecture Faculty, Sivas Cumhuriyet University,
Sivas, Turkey

13
Vol.:(0123456789)
 S. Yıldız, C. B. Karakuş

1 Introduction

The quality of their water is one of important factors, directly concerned with the health of
humans and other living beings (Bhuyan et al. 2018). The quality of surface water within
a region is governed by both natural processes and anthropogenic effects (Pejman et  al.
2009). Water quality monitoring is an important means for sustainable development and
provides critical data for water management (Jalali 2009). Due to spatial and temporal vari-
ations in water quality, which are often difficult to interpret, a monitoring program provid-
ing a representative and reliable estimation of the quality of surface waters is necessary
(Noori et al. 2012). Surface water quality is a delicate and critical issue in many countries.
The quality of surface waters in a region depends highly on the quality and scope of the
industrial, agricultural and other human-related activities in the basins of the specific area
(Singh et al. 2004).
The availability of water for irrigation purposes involves a number of issues such as
the quantity and quality of water. However, quality aspects are generally overlooked while
considering the quantity of water. Irrigation water quality is generally defined in terms of
total dissolved solids, major cations and anions. The three most common issues associated
with low water quality around the world are salinity, reduced permeability and increased
specific ion toxicity (Singh et al. 2018). However, whether water quality is good or bad is
not merely reflected in terms of physical and chemical parameters (Pham 2017).
Irrigation water quality is defined by the type and concentration of dissolved salts and
solids (Etteieb et al. 2017). Irrigation water quality information holds critical importance
for understanding the changes in the product quality, and the required modifications in
water management (Ramakrishnaiah et al. 2009). Product efficiency and the potential for
emergence of hazardous soil conditions should be taken into account during the evaluation
of water quality for irrigation. Evaluation of water quality is a prerequisite in planning,
design and operation of irrigation systems (Mirabbasi et al. 2008).
Water quality degradation through different sources as well as different monitor-
ing methods has been widely considered (Karbassi et  al. 2011). The monitoring process
includes programs and activities that check the quality of water resources and performance
of pollution reduction system (Tavakol et al. 2017). Agriculture is the major consumer of
water, as irrigation uses more than two-thirds of the World’s available freshwater resources
(Aliyu et al. 2017). Irrigation water quality should be continuously monitored for sustain-
able development in agricultural areas. Various methods have been applied to evaluate
the effect of irrigation water on plants and soils. Numerous researchers have used hydro-
chemical indices such as SAR, Na%, KI, PI and irrigation water coefficient (Li et al. 2013;
Fakhre 2014; Cieszynska et al. 2012; Brindha and Elango 2013). Instead of the use of a
single parameter, the combined use of chemical analyses of all ions is expected to yield
better results (Hem 1985). Additionally, IWQI uses a set of indicators and simplifies the
water quality to a single value, which is regarded as a better solution (Saeedi et al. 2010).
Many uses of water quality data cannot be met with an index. An index is most useful for
comparative purposes (what stations have particularly poor water quality?) and for general
questions (Semiromi et al. 2011).
Recently, artificial neural networks have found wide usage in various engineering applica-
tions (Gunaydin et  al. 2010; Das and Basudhar 2008; Yıldız and Değirmenci 2015; Yıldız
2017), water quality analyses (Rankovic et al. 2012; Wen et al. 2013; Baghapour et al. 2016)
and estimation of water quality index (Gazzaz et al. 2012). ANN models are generally used
to estimate and optimize qualitative parameters. As a data-driven approach, ANNs are quite

13
Estimation of irrigation water quality index with development…

suitable for solution of complex problems. Moreover, they are capable of determining the
underlying relation between input and output, which in turn yields a good generalization capa-
bility (Maier and Dandy 2000).
Several studies have addressed the estimation of water quality via artificial neural networks.
Chau (2006) investigated the recent developments in integration of artificial intelligence in
water quality modeling. Hatzikos et al. (2005) used neural networks for estimation of quality
indicators for sea water such as temperature, pH, dissolved oxygen and turbidity. SaralaTham-
bavani and Uma Mageswari (2014) also used artificial neural networks for effective estima-
tion of irrigation water quality. Gazzaz et al. (2012) used artificial neural network models for
estimation of WQI in River Kinta (Malaysia); and Alizadeh and Kavianpour (2015) used this
method for estimation of various ocean water quality parameters. Singh et al. (2009) investi-
gated the training, validity and applicability of ANN models to estimate the dissolved oxygen
and biochemical oxygen demand (BOD) levels in River Gomti (India). Alves et  al. (2018)
used ANN models to determine the water quality index, and Charulatha et  al. (2017) used
them to make an in-depth evaluation of groundwater pollutants.
Regression analysis is a statistical approach used to associate the relation variable, i.e., the
dependent variable, to a set of independent variables. Dependent and independent variables
are also referred to as response and explanatory variables, respectively. Regression model is
supposed to sufficiently define, forecast and control the dependent variable on the basis of
independent variables. Several researchers have addressed and applied regression analyses
(Charulatha et al. 2017; Yıldız and Değirmenci 2015; Noori et al. 2010).
In this research, SAR, KI, Na%, PI and IWQI were primarily calculated to evaluate surface
water quality for irrigation purpose. IWQI was estimated with ANN and multiple regression
models on the basis of SAR, KI, Na% and PI parameters. The optimum model for IWQI esti-
mation was thus introduced using different parameters that define water quality.

2 Materials and methods

Sivas province is located in the Yukarı Kızılırmak Basin of Central Anatolian Region, between
36°–39° eastern longitudes and 38°–41° northern latitudes with 28.488 km2 acreage, as the
second largest province of Turkey after Konya. Sivas province has the general geographical
view of a plateau, which consists of valleys between single mountains and mountain chains,
sunken plains and highlands. The research area is located within Sivas provincial boundaries
in Kızılırmak Basin. In this region, 32 irrigation water stations are managed by Sivas General
Directorate of State Hydraulic Works 19th Regional Directorate (Fig. 1). Parameters such as
SAR, KI, Na%, PI and IWQI, widely used in categorization of irrigation water quality, were
calculated to reveal the suitability of surface waters in the research area for irrigation through
estimation of IWQI via ANN and multiple regression models. During the research, the com-
puter aided software package MATLAB R2013 was used for ANN calculations, and Excel
2010 was used for regression analyses.

13
 S. Yıldız, C. B. Karakuş

Fig. 1  Location of research area and distribution of irrigation water stations

3 Results and discussion

3.1 Irrigation water quality

Parameters such as SAR, KI, Na% and PI were calculated to determine the suitability of
research area’s surface water quality for irrigation. In addition to these parameters, IWQI
values for all irrigation water stations were evaluated using EC, Na, Cl and ­HCO3 param-
eters (Table  1). Irrigation water quality of the research area was categorized in terms of
EC, SAR, KI, Na%, PI and IWQI parameters on the basis of the standard value intervals
given in Table 2 and previous irrigation water classifications (Shabbir and Ahmad 2015;
Seth et al. 2015; Meireles et al. 2010; Zahedi 2017; Omran 2012).

3.1.1 SAR

This parameter was initially proposed by Richards (1954), and it is used to evaluate Na
ions’ tendency for adsorption on soil, and dissolved cation levels’ tendency to enter into cat-
ion exchange regions in soil. High sodium concentrations affect soil permeability and have
direct effect on total salinity of water. This indicates that such concentrations can be toxic for

13
Estimation of irrigation water quality index with development…

Table 1  Irrigation water quality parameters of surface water quality

Sample Na HCO3 Cl EC SAR KI Na% PI IWQI
number

1 0.58 3.09 0.55 569 0.34 0.10 10.30 37.47 49.80

2 0.39 6.24 0.26 636 0.21 0.06 6.25 41.32 47.10
3 0.98 4.36 0.43 725 0.53 0.15 13.68 39.69 46.57
4 0.86 4.86 0.63 586 0.53 0.16 14.54 49.83 44.84
5 0.1 2.99 0.16 315 0.08 0.03 3.17 52.87 55.19
6 0.21 3.81 0.12 382 0.15 0.05 6.04 52.73 53.61
7 0.46 3.26 0.21 381 0.34 0.13 12.08 55.26 52.89
8 0.31 3.98 0.14 402 0.22 0.08 7.97 52.99 52.83
9 0.15 4.18 0.17 434 0.10 0.03 4.04 47.09 52.81
10 0.41 3.4 0.47 862 0.19 0.04 4.40 23.65 49.35
11 0.3 4.55 0.5 1027 0.13 0.03 3.16 22.10 46.10
12 9.12 6.26 7.07 2410 3.11 0.53 34.99 44.24 56.84
13 0.17 4.84 0.32 520 0.10 0.03 3.52 41.95 49.18
14 32.52 2.99 29.69 5170 9.43 1.37 57.82 60.81 30.59
15 2.32 4.15 1.05 696 1.43 0.44 31.46 57.56 75.59
16 0.64 3.64 0.21 780 0.32 0.08 8.40 30.01 49.73
17 0.42 2.87 0.15 324 0.35 0.14 12.76 62.92 54.12
18 0.47 4.6 0.22 470 0.31 0.10 10.22 52.82 49.10
19 0.28 3.8 0.17 395 0.20 0.07 7.34 51.61 53.04
20 5.1 6.97 2.13 1760 2.16 0.46 35.99 47.78 75.08
21 0.44 6.22 0.28 630 0.25 0.07 7.21 43.47 46.89
22 0.77 5.73 0.52 627 0.44 0.13 12.05 45.85 44.92
23 0.58 3.73 0.28 436 0.42 0.15 13.53 54.68 51.20
24 0.71 3.32 0.71 608 0.41 0.12 11.24 38.19 48.06
25 4.04 6.02 1.98 2390 1.23 0.19 16.06 25.29 60.69
26 3.17 4.3 2.16 1322 1.38 0.30 23.57 38.25 68.49
27 0.2 2.71 0.25 464 0.13 0.04 4.19 36.92 53.69
28 0.34 3.6 0.15 448 0.22 0.07 10.53 45.38 52.80
29 0.56 5.11 0.28 538 0.34 0.11 10.02 48.13 47.75
30 10.11 2.79 10.11 2770 3.23 0.52 34.23 39.62 55.77
31 8.16 2.89 5.32 315 2.32 0.33 25.02 30.04 81.09
32 4.9 4.59 4.3 1243 2.37 0.58 36.84 52.48 78.59

*SAR and KI are unit-less; Na% and PI are percentage (%); RSC is meq/L; EC is defined in µS/cm

delicate products (González-Acevedo et al. 2016). The detection of salinity hazard depends on
electrical conductivity measurements. The concept of SAR is used for detection of a probable
sodium hazard (Almeida et al. 2008). SAR is calculated using Eq. 1 (Richards 1954).

Na
SAR = √
Ca + Mg (1)
2

13
 S. Yıldız, C. B. Karakuş

Table 2  Surface water quality- Parameters Value range Water classification Number of

based classification of water samples
samples
EC (µS/cm) < 250 Excellent 0
250–750 Good 22
750–2000 Permissible 6
2000–3000 Doubtful 3
> 3000 Unsuitable 1
SAR 0–6 Good 31
6–9 Doubtful 0
> 9 Unsuitable 1
KI < 1 Suitable 31
1 > Unsuitable 1
Na% < 20 Excellent 24
20–40 Good 7
40–60 Permissible 1
60–80 Doubtful 0
> 80 Unsuitable 0
PI% > 75% Good 0
25–75% Suitable 30
< 25% Unsuitable 2
RSC (meq/l) < 1.25 Good 32
1.25–2.5 Doubtful 0
> 2.5 Unsuitable 0
IWQI 85–100 Excellent 0
70–85 Good 4
55–70 Poor 5
40–55 Very poor 22
0–40 Unsuitable 1

The SAR values of irrigation water stations located in the research area varied between
0.10 and 9.43  meq/l (Table  1). The obtained SAR values showed that, almost all of the
samples received from the irrigation water stations are of “good” irrigation water quality
class (Table 2).

3.1.2 KI

Na, Ca and Mg concentrations in water represent alkali hazard (Dhembare 2012). For cal-
culation of KI parameter, Na concentration is measured against Ca and MG, and in most
waters Ca and Mg preserve their equilibrium state. KI is calculated using Eq.  2 (Kelley
1940).
Na
KI =
Ca + Mg (2)

The obtained KI values varied between 0.03 and 1.37 meq/l (Table 1). According to the
obtained KI values, the samples from 31 irrigation water stations are in “suitable” class,
and 1 irrigation water station is categorized as “non-suitable” (Table 2).

13
Estimation of irrigation water quality index with development…

3.1.3 Na%

The amount of sodium in irrigation water is generally defined as Na%. The sodium concen-
tration in water induces the exchange of ­Ca2+ and ­Mg2+ ions. This exchange process in turn
reduces soil permeability which results in poor internal drainage. As a result of its reactiv-
ity with soil, sodium is regarded as an important ion for classification of irrigation water,
and it reduces permeability (Vasanthavigar et al. 2010; Ishaku et al. 2011). Na% is used to
determine the water quality for agricultural purpose. Irrigation water with high Na% causes
growth retardation in plants (Mohammed et al. 2015). Na% is determined via calculation of
the relative proportion of all cations available in water (Eq. 3) (Ishaku et al. 2011).
Na + K
Na% =
Ca + Mg + Na + K
× 100 (3)

Na% values varied between 3.16 and 57.82  meq/l (Table  1). As indicated by the
obtained Na% values, 75% of the water samples received from irrigation water stations is
in “excellent” category, 21.87% is in “good” category, and 3.12% is in “acceptable” cat-
egory (Table 2).

3.1.4 PI

This index was developed by Doneen (1964) to determine the suitability of water for irriga-
tion and categorizes waters as Class 1 (PI > 75%), Class 2 (25% < PI < 75%) and Class 3
(PI < 25%). Class 1 and Class 2 waters are categorized as “good” and “suitable” with their
higher maximum permeability (Rahman et al. 2014). The permeability of soil is affected
by the amount of Na, Mg, Ca and H ­ CO3 ions in soil. PI is calculated using Eq. 4 (Ghaz-
aryan and Chen 2016).
√
Na + HCO3
PI = × 100 (4)
Ca + Mg + Na
The calculated PI values indicate that, 93.75% of the water samples is in “suitable” cat-
egory, and 6.25% is in “non-suitable” category (Table 4).

3.1.5 IWQI

The evaluation of IWQI is based on the principle of comparison of water quality param-
eters with specific standards, and it defines irrigation water quality by a single value, thus
avoiding water quality evaluations involving complex data intervals. (Ghazaryan and Chen
2016). Irrigation water quality index is based on the recommended limitations for continu-
ous water usage for all soil types (Stoner 1978).
In this research, the irrigation water quality index developed by Meireles et al. (2010)
was used to calculate IWQI. The specified model requires the use of the parameters such
as EC, N ­ a+, ­Cl−, SAR and ­HCO3− to calculate the index. The quality evaluation process
specified by this method differs from the WQI-based process used by WHO (Meireles
et al. 2010). The value of each parameter estimated on the basis of irrigation water quality
parameters recommended by University of California Committee of Consultants (UCCC)
and the criteria determined by Ayers and Westcot (1999) were used to evaluate the relative
value which provides basis for this method. Higher values of this water quality param-
eter (qi) represented by a dimensionless number are indicative of higher water qualities.

13
 S. Yıldız, C. B. Karakuş

Table 3  Parameter limiting values for quality measurements (qi). Source: Meireles et al. (2010)
qi EC (μS/cm) SAR (meq/L)1/2 Na+ (meq/L) Cl− (meq/L) HCO3− (meq/L)

85–100 200 ≤ EC < 750 2 ≤ SAR < 3 2 ≤ Na < 3 1 ≤ Cl < 4 1 ≤ HCO3 < 1.5

60–85 750 ≤ EC < 1500 3 ≤ SAR < 6 3 ≤ Na < 6 4 ≤ Cl < 7 1.5 ≤ HCO3 < 4.5
35–60 1500 ≤ EC < 3000 6 ≤ SAR < 12 6 ≤ Na < 9 7 ≤ Cl < 10 4.5 ≤ HCO3 < 8.5
0–35 EC < 200 or SAR < 2 or Na < 2 or Cl < 1 or HCO3 < 1 or
EC ≥ 3000 SAR ≥ 12 Na ≥ 9 Cl ≥ 10 HCO3 ≥ 8.5

Table 4  Relative weight wi of Parameters wi

each parameter in IWQI. Source:
Meireles et al. (2010)
[EC] 0.211
[Na +] 0.204
[HCO3−] 0.202
[Cl−] 0.194
[SAR] 0.189
Total 1

qi values were evaluated using Eq. 7 based on the tolerance limit values of the parameters
shown in Table 3.
[( ) ]
qi = qimax − xij − xinf xqiamp ∕qamp (5)
where qmax is the maximum value of qi for each class;  xij  is the observed value of each
parameter;  xinf is the corresponding value to the lower limit of the class to which the
parameter belongs;  qiamp is class amplitude;  xamp  is class amplitude to which the param-
eter belongs. In order to evaluate xamp of the last class of each parameter, the upper limit
was considered to be the highest value determined in the physical–chemical and chemical
analysis of the water samples. Then wi values were normalized such that their sum equals
one, according to Eq. (6) (Meireles et al. 2010):
k k n
∑ ∑ ∑
wi = Fj Aij ∕ Fj Aij (6)
j=1 j=1 i=1

where wi is the weight of the parameter for the WQI; F = component 1 autovalue; Aij is the
explainability of parameter i by factor j; i is the number of physical–chemical and chemical
parameters selected by the model, ranging from 1 to n; j is the number of factors selected
in the model, varying from 1 to k. Table 4 illustrates relative weight of each parameter. The
IWQI value can be calculated according to Eq. 7 taking into account the above-mentioned
values of qi and wi (Hallouche et al. 2017).
k
∑
IWQI = qi × wi (7)
j=1

The IWQI values obtained from the research area varied between 30.59 and 81.09
(Table  1). The evaluation made on the basis of irrigation water standard value intervals

13
Estimation of irrigation water quality index with development…

shows that, none of the samples from irrigation water stations is of “excellent” class. In
terms of irrigation water quality, 12.5% of the samples is of “good” quality, 15.62% is of
“poor” quality, 68.75% is of “very poor” quality, and 3.12% is of “non-suitable” quality
(Table 2).

3.2 Artificial neural network (ANN)

Artificial neural networks (ANN) use the modeling structure of a neural network which
is a very powerful technique for modeling complex nonlinear relations in particular cases
where the relationship between the variables is quite unclear (Smith 1994).
The structure of an ANN model typically consists of three layers. These are the input
layer where the data enter into the model and its weighted sum is calculated, intermedi-
ate hidden layer where the data are processed, and the output layer where ANN results are
generated (Fig. 2). Each layer consists of one or more basic elements termed as a neuron
or node. Each neuron represents an algebraic function which is assigned a parameter with
limit values. (Dryfus et al. 2002).
In this research, ANN was used for estimation of IWQI using SAR, KI, PI and Na%
among irrigation water quality parameters (Fig. 2).
The optimum number of neurons and training iterations in the hidden layer are impor-
tant indicators in ANN modeling. No specific algorithm is available for determination of
the optimum number of neurons in hidden layer, and this value should be determined on
a trial-and-error basis (Alizadeh and Kavianpour 2015). The number of hidden layers is
determined in accordance with the complexity of the problem, and in most cases a single
hidden layer suffices to model a problem (Rezvan et al. 2016).
The application of ANN was performed using an algorithm developed in MATLAB
platform for selection of the minimum number of principal components used as input, in
addition to the number of neurons in hidden layer that results in the optimum ANN model.
The proposed ANN structure is illustrated in Fig. 3.
The training, validation and test data of the ANN model with the optimum estimation
are shown in Fig. 4. The statistical performance of the models was evaluated using statisti-
cal parameters such as µ, SE, σ and R2. Additionally, RMSE and MAPE between the ANN
predicted data and the real data are used for evaluation of the quality of the developed
models (Nabavi-Pelesaraei et al. 2014). The statistical performance is given in Table 5.
RMSE, MAPE and R2 are often used as criteria in neural network studies to compare
and evaluate the errors and the received data (Alves et  al. 2018). RMSE and MAPE are
calculated using Eqs. 8 and 9.

Fig. 2  ANN architecture

13
 S. Yıldız, C. B. Karakuş

Fig. 3  ANN structure

Fig. 4  The training, validation 80.00

and testing data 70.00 y = 0.7939x + 9.9806
R² = 0.9234
60.00
Estimated IWQI

50.00
40.00
30.00
20.00
10.00
0.00
0 20 40 60 80 100
Calculated IWQI

Table 5  Statistical performance Model Structure R2 σ SE µ RMSE MAPE

of the ANNs models
I 4–3–1–1 0.764 0.0962 5.3410 1.0198 5.231 5.93
II 4–4–1–1 0.875 0.0579 3.8858 0.9936 3.479 2.96
III 4–5–1–1 0.855 0.0826 4.1941 1.0251 2.296 5.86
IV 4–6–1–1 0.923 0.0518 3.0473 1.0191 1.634 2.99

√ 1
1 ∑( )2
RMSE = ti − zi (8)
n n

( )|2
n |
1 ∑ || ti − zi || (9)
MAPE = × 100
n t=1 zi

where “ti” and “zi” are the predicted and actual output, and “n” is the number of points in
the data set.

13
Estimation of irrigation water quality index with development…

As shown in Table 5, a significant relationship is evident between the observed values

in the built models. However, 4–6–1–1 stands out as the best model in terms of R2 and SE.
Also, a low RMSE value is indicative of a good model performance. MAPE value is an
unbiased statistical value used to measure the estimation capability of a model. The lowest
MAPE value indicates the best model performance (Olyaie et al. 2015).
The correlation between the estimation results of the designed ANN model and the
experimental results were used to evaluate the ANN modeling performance which is used
as an effective means. The graph of calculated IWQI values versus those estimated with
ANN is shown in Fig. 5.
As indicated in Fig. 5, the estimated results of the designed ANN model were compared
with the calculated data and they were found to be in good agreement. ANN has proved to
be an effective method for IWQI modeling with high R2 (R2 = training (0.99), test (0.98)
and validation (0.95)) values. The effectiveness of ANN model is determined on the basis
of maximization of R2 and minimization of MSE value of the test set (neurons 1–20 cor-
respond to the hidden layer). As indicated by the number of epochs (Fig. 6) for optimum
ANN models in the minimum mean square error (MSE) graph, no significant change is
observed in the model’s performance after the 14th iteration. As inferred from Fig., the
network was successfully trained using a flexible back propagation algorithm.

Fig. 5  Comparison of calculated and estimated (ANN) IWQI results

13
 S. Yıldız, C. B. Karakuş

Fig. 6  MSE against the number of epochs for IWQI

3.3 Multiple regression analysis (MRA)

Multiple regression analysis defines two or more independent variables to describe the
variations of a dependent variable. In this research, SAR, KI, Na% and PI were used
as the independent variables and IWQI was used as the dependent variable. Multiple
regression analysis was performed to define the relationship between IWQI and four
independent variables. The statistical variables studied are given in Table 6.
The confirmation of the model was made taking into account F test, t test and cor-
relation coefficients. The statistical results are given in Table 7. The significance of R2
value can be determined via t test by assuming that all variables show a normal distribu-
tion and the observations are randomly selected. The test compares the t value calcu-
lated with null hypothesis and the tabulated t value. The critical t value was obtained as
± 1.66, as the reliability level was chosen as 95%. If the calculated t value is higher than
the tabulated t value, null hypothesis is then rejected. This indicates that R is significant.
Null hypothesis is not rejected in cases where the calculated t value is lower than the

Table 6  Statistics of the variables studied

SAR KI Na% PI IWQI

Mean 1.030312 0.209062 15.394375 44.46875 54.196875

Maximum 9.43 1.37 57.82 62.92 81.09
Minimum 0.08 0.03 3.16 22.1 30.59
Median 0.34 0.115 10.885 45.615 52.805
Variation 3.171532 0.071518 168.45485 110.49151 117.38204
Standard deviation 1.780879 0.267429 12.979016 10.511494 10.83429
Skewness 3.701276 2.970242 1.5910975 − 0.477615 1.041654
Kurtosis 16.291863 11.013439 2.381990 − 0.301733 1.417442

13
Estimation of irrigation water quality index with development…

Table 7  Statistical results
Inde- Dependent R2 Adjusted R2 Unstandard- Standard error Calcu- Tabu- Sign.
pendent value ized coef- lated F lated F
value ficients value value

SAR IWQI 0.6 0.539 − 5.46129 5.53739 10.06 0.55 0.000

KI − 43.65361 57.2527
PI 1.81852 0.51805
Na% − 0.13510 0.15536
Independent value Calculated t Tabulated t value Sig.

SAR − 0.986257 ± 1.66 0.332763

KI − 0.762471 0.452388
PI 3.510281 0.001590
Na% − 0.869633 0.392166

tabulated t value. In such cases, R is considered insignificant (Yıldız and Değirmenci

2015).
As indicated in Table 7, calculated t values are higher than tabulated t values; thus,
R is considered significant. R2 is determined as 0.6, and there is a medium signifi-
cance level (p ˂ 0.05). Also, the calculated F value is higher than the tabulated F value.
Accordingly, null hypothesis is rejected; thus, it can be inferred that there is a real rela-
tionship between the dependent and independent variables.

4 Conclusions

This research aims to determine irrigation water quality using irrigation water quality
parameters of surface water stations, and estimation of IWQI using ANN and multiple
regression methods. SAR, Na%, KI, PI and IWQI were evaluated for evaluation of sur-
face water in the research area for irrigation purposes. The majority of samples received
from surface water irrigation stations were found to be “suitable” on the basis of a SAR,
Na%, KI and PI-based evaluation. It was concluded as a result of this research that,
estimation of IWQI value on the basis of different water quality parameters is possi-
ble through the use of ANN and multiple regression analysis. R2 was found as 0.6 by
multiple regression, and the optimum R2 was found as 0.92 by ANN model. Also, the
calculated F value was found to be higher than the tabulated F value; thus, null hypoth-
esis is rejected and a real relationship between the dependent and independent variables
is inferred. The values estimated using ANN model with statistical parameters and the
calculated IWQI values were found to be consistent (R2 = 0.92). R2 values (0.98) were
found to be high in different studies (Latha 2019). These results show that, multi-regres-
sion and ANN analyses can be used for estimation of IWQI, which is an important qual-
ity parameter, with the parameters which are used for determination of irrigation water
quality. However, the best estimation results are obtained by use of ANN. A similar
result has been demonstrated in different studies (Kadam et al. 2019).

13
 S. Yıldız, C. B. Karakuş

References
Aliyu, T., Balogun, O., Namani, C., Olatinwo, L., & Aliyu, A. (2017). Assessment of the presence of
metals and quality of water used for irrigation in Kwara State, Nigeria. Pollution, 3(3), 461–470.
Alizadeh, M. J., & Kavianpour, M. R. (2015). Development of wavelet-ANN models to predict water
quality parameters in Hilo Bay, Pacific Ocean. Marine Pollution Bulletin, 98(1–2), 171–178.
Almeida, C., Quintar, S., González, P., & Mallea, M. (2008). Assessment of irrigation water quality. A
proposal of a quality profile. Environmental Monitoring and Assessment, 142(1–3), 149–152.
Alves, E. M., Rodrigues, R. J., Dos Santos Corrêa, C., Fidemann, T., Rocha, J. C., Buzzo, J. L. L., et al.
(2018). Use of ultraviolet–visible spectrophotometry associated with artificial neural networks as
an alternative for determining the water quality index. Environmental Monitoring and Assessment,
190(6), 319.
Ayers, R. S., & Westcot, D. W. (1999). Water quality for agriculture, FAO Irrigation and Drainage Paper
No. 29, Rev. 1, U. N. Food and Agriculture Organization, Rome.
Baghapour, M. A., Nobandegani, A. F., Talebbeydokhti, N., Bagherzadeh, S., Nadiri, A. A., Gharekhani,
M., et  al. (2016). Optimization of DRASTIC method by artificial neural network, nitrate vulner-
ability index, and composite DRASTIC models to assess groundwater vulnerability for unconfined
aquifer of Shiraz Plain, Iran. Journal of Environmental Health Science and Engineering, 14(1), 13.
Bhuyan, M., Bakar, M., Sharif, A. S. M., Hasan, M., & Islam, M. (2018). Water quality assessment
using water quality indicators and multivariate analyses of the old Brahmaputra River. Pollution,
4(3), 481–493.
Brindha, K., & Elango, L. (2013). Environmental assessment of water quality in Nagarjuna Sagar reser-
voir, India. Earth Resources, 1(1), 33–36.
Charulatha, G., Srinivasalu, S., Maheswari, O. U., Venugopal, T., & Giridharan, L. (2017). Evaluation
of ground water quality contaminants using linear regression and artificial neural network models.
Arabian Journal of Geosciences, 10(6), 128.
Chau, K. W. (2006). Areview on integration of artificial intelligence into water quality modeling. Marine
Pollution Bulletin, 52, 726–733.
Cieszynska, M., et  al. (2012). Application of physicochemical data for water-quality assessment of
watercourses in the Gdansk Municipality (South Baltic coast). Environmental Monitoring and
Assessment, 184, 2017–2029.
Das, S. K., & Basudhar, P. (2008). Prediction of residual friction angle of clay artificial neural network.
Engineering Geology, 100, 142–145.
Dhembare, A. J. (2012). Assessment of water quality indices for irrigation of Dynaneshwar Dam Water,
Ahmednagar, Maharashtra, India. Archives of Applied Science Research, 4(1), 348–352.
Doneen, L. D. (1964). Notes on water quality in agriculture. Published as a water science and engineer-
ing, Paper 4001, Department of Water Sciences and Engineering, University of California.
Dryfus, G., Martinez, J. M., Samuelides, M., Gordon, M. B., Badran, F., Thiria, S., et  al. (2002).
Reseaux de Neurones: Methodologie et Applications. Paris: Editions Eyrolles.
Etteieb, S., Cherif, S., & Tarhouni, J. (2017). Hydrochemical assessment of water quality for irrigation:
A case study of the Medjerda River in Tunisia. Applied Water Science, 7(1), 469–480.
Fakhre, A. (2014). Evaluation of hydrogeochemical parameters of groundwater for suitability of domes-
tic and irrigational purposes: A case study from central Ganga Plain, India. Arabian Journal of
Geosciences, 7, 4121–4131. https​://doi.org/10.1007/s1251​7-013-1055-6.
Gazzaz, N. M., Yusoff, M. K., Aris, A. Z., Juahir, H., & Ramli, M. F. (2012). Artificial neural network
modeling of the water quality index for Kinta River (Malaysia) using water quality variables as pre-
dictors. Marine Pollution Bulletin, 64(11), 2409–2420.
Ghazaryan, K., & Chen, Y. (2016). Hydrochemical assessment of surface water for irrigation purposes
and its influence on soil salinity in Tikanlik oasis, China. Environmental Earth Sciences, 75(5),
383.
González-Acevedo, Z. I., Padilla-Reyes, D. A., & Ramos-Leal, J. A. (2016). Quality assessment of irri-
gation water related to soil salinization in Tierra Nueva, San Luis Potosí, Mexico. Revista Mexi-
cana de Ciencias Geológicas, 33(3), 271–285.
Gunaydin, O., Gokoglu, A., & Fener, M. (2010). Prediction of artificial soil’s unconfined compression
strength test using statistical analyses and artificial neural networks. Advances in Engineering Soft-
ware, 41, 1115–1123.
Hallouche, B., Hadji, F., Marok, A., & Benaabidate, L. (2017). Spatial mapping of irrigation ground-
water quality of the High Mekerra watershed (Northern Algeria). Arabian Journal of Geosciences,
10(11), 233.

13
Estimation of irrigation water quality index with development…

Hatzikos, E., Anastasakis, L., Bassiliades, N., & Vlahavas, I. (2005). Simultaneous prediction of multi-
ple chemical parameters of river water quality with tide. In Proceedings of 2nd International Sci-
ence Conference on Computer Science, IEEE Computer Society, May 11–13, Varna, Bulgaria.
Hem, J. D. (1985). Study and interpretation of the chemical characteristics of natural water USGS. Water
Supply Paper, 2254, 117–120.
Ishaku, J. M., Ahmed, A. S., & Abubakar, M. A. (2011). Assessment of groundwater quality using
chemical indices and GIS mapping in Jada area, Northeastern Nigeria. Journal of Earth Sciences
and Geotechnical Engineering, 1(1), 35–60.
Jalali, M. (2009). Groundwater geochemistry in the Alisadr, Hamadan, Western Iran. Environmental
Monitoring and Assessment, 166, 359–369.
Kadam, A. K., Wagh, V. M., Muley, A. A., Umrikar, B. N., & Sankhua, R. N. (2019). Prediction of water
quality index using artificial neural network and multiple linear regression modelling approach in
Shivganga River basin, India. Modeling Earth Systems and Environment. https​://doi.org/10.1007/
s4080​8-019-00581​-3.
Karbassi, A. R., Mir Mohammad Hosseini, F., Baghvand, A., & Nazariha, M. (2011). Development of
water quality index (WQI) for Gorganrood River. International Journal of Environmental Research,
5(4), 1041–1046.
Kelley, W. P. (1940). Permissible composition and concentration of irrigated waters. Proceedings of
ASCF, 66, 607.
Latha, M. (2019). Prediction of irrigation water quality index using artificial neural network. Interna-
tional Journal of Applied Engineering Research, 14(4), 952–956.
Li, P., Wu, J., & Qian, H. (2013). Assessment of groundwater quality for irrigation purposes and iden-
tification of hydrogeochemical evolution mechanisms in Pengyang County, China. Environmental
Earth Sciences, 69, 2211–2225.
Maier, H. R., & Dandy, G. C. (2000). Neural networks for the prediction and forecasting of water
resource variables: A review of modelling issues and applications. Environmental Modelling and
Software, 15, 101–124.
Meireles, A., Andrade, E. M., Chaves, L., Frischkorn, H., & Crisostomo, L. A. (2010). A new proposal
of the classification of irrigation water. Revista Ciencia Agronomica, 413, 349–357.
Mirabbasi, R., Mazloumzadeh, S. M., & Rahnama, M. B. (2008). Evaluation of irrigation water quality
using fuzzy logic. Research Journal of Environmental Sciences, 2(5), 340–352.
Mohammed, I. U., Ndahi, A. K., & Adamu, I. C. (2015). Rapid assessment of reservoir water quality
and suitability indices for irrigation purpose: A case study of Ero and Ele Reservoirs in Ekiti State
Nigeria. International Journal of Multidisciplinary and Current Research, 3, 215–219.
Nabavi-Pelesaraei, A., Kouchaki-Penchah, H., & Amid, S. (2014). Modeling and optimization of C ­ O2
emissions for tangerine production using artificial neural networks and data envelopment analysis.
International Journal of Biosciences, 4(7), 148–158.
Noori, R., Karbassi, A., Khakpour, A., Shahbazbegian, M., Badam, H. M. K., & Vesali-Naseh, M.
(2012). Chemometric analysis of surface water quality data: Case study of the Gorganrud River
Basin, Iran. Environmental Modeling and Assessment, 17(4), 411–420.
Noori, R., Sabahi, M. S., Karbassi, A. R., Baghvand, A., & Zadeh, H. T. (2010). Multivariate statistical
analysis of surface water quality based on correlations and variations in the data set. Desalination,
260(1–3), 129–136.
Olyaie, E., Banejad, H., Chau, K. W., & Melesse, A. M. (2015). A comparison of various artificial intel-
ligence approaches performance for estimating suspended sediment load of river systems: A case
study in United States. Environmental Monitoring and Assessment, 187(4), 189.
Omran, E. S. E. (2012). A proposed model to assess and map irrigation water well suitability using geo-
spatial analysis. Water, 4, 545–567.
Pejman, A. H., et al. (2009). Evaluation of spatial and seasonal variations in surface water quality using
multivariate statistical techniques. International Journal of Environmental Science and Technology,
6(3), 467–476.
Pham, L. (2017). Comparison between Water Quality Index (WQI) and biological indices, based on
planktonic diatom for water quality assessment in the Dong Nai River, Vietnam. Pollution, 3(2),
311–323.
Rahman, M., Das, R., Hassan, N., Roy, K., Haque, F., & Akber, M. A. (2014). Environmental study on
water quality of Mayur River with reference to suitability for irrigation. International Journal of
Environmental Sciences, 4(6), 1150.
Ramakrishnaiah, C. R., Adashiv, C., & Ranganna, G. (2009). Assessment of water quality index for the
groundwater in Tumkur Taluk, Karnataka State, India. Egyptian Journal of Chemistry, 6, 523–530.

13
 S. Yıldız, C. B. Karakuş

Rankovic, V., Radulovic, J., Radojevic, I., Ostojic, A., & Comic, L. (2012). Prediction of dissolved oxy-
gen in reservoirs using adaptive network-based fuzzy inference system. Journal of Hydroinformatics,
14(1), 167–179.
Rezvan, K., Fakhri, Y., Mehrorang, G., & Kheibar, D. (2016). Back propagation artificial neural network
and central composite design modeling of operational parameter impact for sunset yellow and azur (II)
adsorption onto MWCNT and MWCNT-Pd-NPs: Isotherm and kinetic study. Chemometrics and Intel-
ligent Laboratory Systems, 159, 127–137. https​://doi.org/10.1016/j.chemo​lab.2016.10.012.
Richards, L. A. (1954). Diagnosis and improvement of saline and alkaline soils (p. 60). Washington: US
Department of Agriculture Hand Book.
Saeedi, M., Abessi, O., Sharifi, F., & Meraji, H. (2010). Development of groundwater quality index. Envi-
ronmental Monitoring and Assessment, 163, 327–335.
SaralaThambavani, D., & Uma Mageswari, T. S. R. (2014). Modeling of irrigation water quality using mul-
tilayer perceptron back propagation neural network (MLBP-NN). International Journal of ChemTech
Research, 6(5), 3053–3061.
Semiromi, F. B., Hassani, A. H., Torabian, A., Karbassi, A. R., & Lotfi, F. H. (2011). Water quality index
development using fuzzy logic: A case study of the Karoon River of Iran. African Journal of Biotech-
nology, 10(50), 10125–10133.
Seth, R., Mohan, M., Singh, P., et al. (2015). Assessment of seasonal variations in surface water quality of
Bageshwar District, Uttarakhand, India for drinking and irrigation purposes. The Proceedings of the
National Academy of Sciences, India, Section A: Physical Sciences, 85(2), 283–293.
Shabbir, R., & Ahmad, S. S. (2015). Use of geographic information system and water quality index to assess
groundwater quality in Rawalpindi and Islamabad. Arabian Journal for Science and Engineering,
40(7), 2033–2047.
Singh, K. P., Basant, A., Malik, A., & Jain, G. (2009). Artificial neural network modeling of the river water
quality—A case study. Ecological Modelling, 220(6), 888–895.
Singh, S., Ghosh, N. C., Gurjar, S., Krishan, G., Kumar, S., & Berwal, P. (2018). Index-based assessment of
suitability of water quality for irrigation purpose under Indian conditions. Environmental Monitoring
and Assessment, 190(1), 29.
Singh, K. P., Malik, A., Mohan, D., & Sinha, S. (2004). Multivariate statistical techniques for the evalua-
tion of spatial and temporal variations in water quality of Gomti River (India)—A case study. Water
Research, 38, 3980–3992.
Smith, M. (1994). Neural networks for statistical modelling (p. 235). NY: Van Nostrand Reinhold.
Stoner, J. D. (1978). Water-quality indices for specific water uses. Department of the Interior, Geological
Survey.
Tavakol, M., Arjmandi, R., Shayeghi, M., Monavari, S. M., & Karbassi, A. (2017). Application of multivari-
ate statistical methods to optimize water quality monitoring network with emphasis on the pollution
caused by fish farms. Iranian Journal of Public Health, 46(1), 83.
Vasanthavigar, M., Srinivasamoorthy, K., Gantha, R. R., Vijayaraghavan, K., & Sarma, V. S. (2010). Char-
acterization and quality assessment of groundwater with special emphasis on irrigation utility: Thiru-
manimuttar sub-basin, Tamil Nadu, India. Arabian Journal of Geosciences, 5(2), 245–258.
Wen, X., Fang, J., Diao, M., & Zhang, C. (2013). Artificial neural network modeling of dissolved oxygen in
the Heihe River, Northwestern China. Environmental Monitoring and Assessment, 185(5), 4361–4371.
Yıldız, S. (2017). Artificial Neural Network (ANN) methods for modeling of Zn(II) adsorption in batch pro-
cess. Korean Journal of Chemical Engineering, 34(9), 2423–2434.
Yıldız, S., & Değirmenci, M. (2015). Estimation of oxygen exchange during treatment sludge compost-
ing through multiple regression and artificial neural networks. International Journal of Environmental
Research, 9(4), 1173–1182.
Zahedi, S. (2017). Modification of expected conflicts between drinking water quality index and irrigation
water quality index in water quality ranking of shared extraction wells using multi criteria decision
making techniques. Ecological Indicators, 83, 368–379.

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13