Hydrogeology Journal (2023) 31:571–587

https://doi.org/10.1007/s10040-023-02620-x

REPORT

Compilation simulation of surface water and groundwater resources

using the SWAT‑MODFLOW model for a karstic basin in Iran
Haniyeh Salmani1 · Saman Javadi2 · Mohammad Reza Eini2,3 · Golmar Golmohammadi4

Received: 17 December 2021 / Accepted: 16 February 2023 / Published online: 9 March 2023
© The Author(s), under exclusive licence to International Association of Hydrogeologists 2023

Abstract
Modeling of karstic basins can provide a better understanding of the interactions between surface water and groundwater, a
more accurate estimation of infiltrated water amount, and a more reliable water balance calculation. In this study, the hydro-
logical simulation of a karstic basin in a semiarid region in Iran was performed in three different stages. In the first stage,
the original SWAT model was used to simulate surface-water flow. Then, the SWAT-MODFLOW conjunctive model was
implemented according to the groundwater characteristics of the study area. Finally, due to the karstic characteristics of the
region and using the CrackFlow (CF) package, the SWAT-MODFLOW-CF conjunctive model was developed to improve
the simulation results. The coefficient of determination (R2) and the Nash-Sutcliffe efficiency coefficient (NSE) as error
evaluation criteria were calculated for the models, and their average values were 0.63 and 0.57 for SWAT, 0.68 and 0.61 for
SWAT-MODFLOW, 0.73 and 0.7 for SWAT-MODFLOW-CF, respectively. Moreover, the mean absolute error (MAE) and
root mean squared error (RMSE) of the calibration for groundwater simulation using the SWAT-MODFLOW model were
1.23 and 1.77 m, respectively. These values were 1.01 and 1.33 m after the calibration of the SWAT-MODFLOW-CF model.
After modifying the CF code and keeping the seams and cracks open in both dry and wet conditions, the amount of infiltrated
water increased and the aquifer water level rose. Therefore, the SWAT-MODFLOW-CF conjunctive model can be proposed
for use in karstic areas containing a considerable amount of both surface water and groundwater resources.

Keywords Karst · Iran · Conjunctive models

Introduction interaction between surface water and groundwater in these

areas can provide important information (Ghezelayagh et al.
Approximately one-fifth of the Earth’s surface is covered by 2020). Among the models currently available, the Soil and
various carbonate formations (Lamoreaux 1991). More than Water Assessment Tool (SWAT), developed by Arnold et al.
25% of the world’s population lives in these areas and con- (1998), is one of the most common and widely used for stud-
sumes water derived from carbonate aquifers (Afrasiabian ying basins (Bailey et al. 2016). This model has been used in
1998); therefore, using a modeling approach to determine the many parts of the world for hydrological simulation with dif-
ferent purposes such as surface-water and groundwater inter-
actions, the effects of climate change on water resources, and
* Saman Javadi evaluation of precipitation and land-use change (Eini et al.
[email protected]
2019). Being open-source and free access, this hydrologi-
1
Department of Civil Engineering, Ale Taha Institute cal modeling tool has led to much useful research (Yactayo
of Higher Education, Tehran, Iran 2009; Eshtawi et al. 2016; Amin et al. 2020; Nerantzaki
2
Department of Water Engineering, College of Agricultural 2015). Despite its numerous advantages, the original SWAT
Technology, University of Tehran, Tehran, Iran model suffers from several limitations, mainly in dealing
3
Department of Hydrology, Meteorology, and Water with karstic areas and groundwater resources, which have
Management, Institute of Environmental Engineering, been addressed in some studies—for example, Baffaut and
Warsaw University of Life Sciences, Warsaw, Poland Benson (2009) improved the SWAT code to investigate the
4
Department of Soil, Water and Ecosystem Sciences, time taken for water to enter the aquifer from the ground
University of Florida, Ranch Cattle REC, Gainesville, FL, surface through sinkholes, seams (channels) and cracks. By
USA

13
Vol.:(0123456789)
572 Hydrogeology Journal (2023) 31:571–587

making changes in the SWAT code, Nikolaidis et al. (2013), drains, is not considered precisely. Also, due to the wide
Palanisamy and Workman (2015), and Malagò et al. (2016) range of soil types and the presence of seams and cracks in
successfully simulated flow through seams and cracks in karstic areas, the amount of exchange between surface water
karst for better management of water resources in the areas and the aquifer requires more accurate measurements and
studied. Eini et al. (2020) used two approaches to simulate a adjusted calculations. Therefore, this study aimed to sim-
karstic basin in central Iran. They showed that the standard ulate the interaction of surface water and groundwater by
SWAT model had limitations in the accurate estimation of use of the SWAT-MODFLOW conjunctive model; thus, the
water balance parameters in karstic areas. SWAT-MODFLOW model was developed for investigating
During studies on modeling surface water in karstic areas, karstic areas, as well as investigating the infiltration values
changes have been applied in the SWAT code; however, the by combined use with the CrackFlow (CF) code. Hence,
SWAT model is limited in the simulation of groundwater dis- first, the SWAT model simulated surface-water flow, and
charge to surface streams and the infiltration of surface water then the SWAT-MODFLOW model was implemented for
into groundwater. For the first time in the United States, the Maharloo Lake basin, a dry basin in central Iran. Then
Sophocleous et al. (1999) simulated the interaction of sur- the SWAT-MODFLOW-CF model was developed to improve
face water and groundwater using the SWAT-MODFLOW the simulation results, taking account of the region’s karst
conjunctive model and estimated the runoff, groundwater characteristics.
level, and groundwater discharge rates. In another study,
Bailey et al. (2017) introduced a graphical user interface
based on the SWAT-MODFLOW code, called SWATMOD- Materials and methods
Prep, which facilitated simultaneous simulation. Kim et al.
(2017) used the SWAT-MODFLOW model to investigate Study area
the changes in groundwater level and the effects of the con-
struction of groundwater dams on the Oshipcheon River of The study area is located in Shiraz district, Fars Province,
South Korea. They considered the groundwater dam along- central Iran, and in the southwestern margin of Maharloo-
side the surface-water dam as an option to protect ground- Bakhtegan basin (29° 26′–29° 84′ N, 52° 13′–52° 48′ E),
water resources. Using SWAT, MODFLOW, MT3DMS, and with an area of 1,428 ­km2. Shiraz is located in the catchment
SWAT-MODFLOW-RT3D models, Eshtawi et al. (2016) area of the Central Plateau, which includes the basin of lakes
and Wei et al. (2019) determined the exchange of surface Tashk, Bakhtegan and Maharloo. The study area has two
water and groundwater in the Gaza aquifer and the Sprague seasonal rivers named Khoshkrood (which is formed by the
river basin (Middle East) in respect to the changes in quan- confluence of two tributary rivers, Nahr-e Azam River and
tity and quality. In another study, Park et al. (2019) intro- Khoshk River) and Chenar Rahdar River (Jamab Consulting
duced a new graphical user interface called QSWATMOD Engineers 2013).
in QGIS software to facilitate the link between the SWAT
and MODFLOW models. QSWATMOD can be a valuable
tool to create and manage SWAT-MODFLOW projects as Hydrology
SWAT-MODFLOW users increase worldwide. Aliyari et al.
(2019) implemented a modified SWAT-MODFLOW model Data from the Khoshk and Nahr-e Azam hydrometric sta-
for a study of the urban-agricultural basin of South Platte tions were selected for the modeling period 2003–2019.
River, Colorado, USA. They reported promising results for Also, daily precipitation data were obtained from one synop-
surface-water and groundwater simulation in this basin, tic weather station and two precipitation stations, while the
which show this model can be used for management pur- data from a synoptic and a climatology station in the region
poses (see Table 1 for a summary of previous research using were used for daily minimum and maximum temperatures.
the SWAT-MODFLOW model). The average precipitation and temperature recorded in the
Due to the great importance of groundwater resources in study area during the years 2003 to 2019 were 364 mm and
terms of quantity and quality as an essential and strategic 15.1 °C, respectively.
source of drinking and agricultural water in arid and semiarid The annual rainfall range in the study area is 400–600
regions, many studies have been conducted using SWAT- mm; the highest amount of rainfall occurs in the east and
MODFLOW to model the conjunctive use of surface water northwest regions at higher elevations. The temperature
and groundwater. However, to the authors’ knowledge, the ranges from 10 to 18 °C, whereby the lowest temperatures
combination of these two models has not been investigated are found in the highlands and the highest temperatures are
for use in karstic basins so far; moreover, in calculating the in the southeast of the plain. The lowest evaporation value
water balance, the amount of water exchanged between sur- is 1,000 mm/year from the western heights, and evaporation
face water and the aquifer, and vice versa, in rivers and in reaches its highest value of 3,000 mm/year at the center of

13
Hydrogeology Journal (2023) 31:571–587 573

Table 1  A summary of previous research

Study Study region Model used

Baffaut and Benson (2009) James River Basin, USA SWAT-B and B (karst)
Yactayo (2009) Virginia, USA SWAT (karst)
Nikolaidis et al. (2013) Crete, Greece SWAT (karst)
Palanisamy and Workman (2015) Kentucky, USA SWAT (karst)
Dowlatabadi and Ali Zomorodian (2016) FirouzAbad, Iran SWAT and MODFLOW
Izady et al. (2015) Neishaboor plain, Iran SWAT-MODFLOW
Guzman et al. (2015) Fort Cobb, USA SWAT-MODFLOW
Il-Moon Chung et al. (2015) Mihocheon watershed, South Korea SWAT-MODFLOW
Malagò et al. (2016) Crete, Greece Adapted SWAT (karst)
Ai-Di Huo et al. (2016) Heihe River basin, China SW-GW
Bailey et al. (2016) Sprague River Watershed, USA SWAT-MODFLOW
Surinaidu et al. (2016) Ramganga, India SWAT and MODFLOW
Eshtawi et al. (2016) Gaza Strip, Palestine SWAT and MODFLOW and MT3DMS
Kim et al. (2017) Oshipcheon River, Yeongdeok Gun, South Korea SWAT-MODFLOW
Bailey et al. (2017) Little River Experimental SWATMOD-PREP
Watershed (LREW), Georgia, USA
Kamali and Niksokhan (2017) Esfahan, Iran SWAT and MODFLOW and MT3DMS
Abbas et al. (2018) River Dee, UK SWAT, SWAT-MODFLOW
Ehtiat et al. (2018) Dehloran, Iran SWAT and MODFLOW and MT3DMS
Wei et al. (2019) Sprague River watershed, USA SWAT-MODFLOW-RT3D
Park et al. (2019) Middle Bosque River watershed, Texas, USA QGIS and SWAT-MODFLOW
Molina-Navarro et al. (2019) Odderkæk, Denmark SWAT, SWAT-MODFLOW
Chunn et al. (2019) Little Smoky river watershed, Alberta, Canada GW-SW(SWAT and MODFLOW)
Semiromi and Koch (2019) Ghareshoo River basin, Iran SWAT-MODFLOW
Sith et al. (2019) Todoroki, Japan SWATMOD-Prep
Aliyari et al. (2019) South Palatte, USA Modified SWAT-MODFLOW
Liu et al. (2019) Uggerby River basin, Denmark SWAT-MODFLOW
Eini et al. (2020) Shiraz, Iran Modified SWAT (karst)
Petpongpan et al. (2020) Yom and Nam River basin, Thailand SWATMOD-Prep
Sabzzadeh and Shourian (2020) Karkheh River, Iran SWAT and MODFLOW and PSO
Bailey et al. (2020) Middle Bosque River watershed, Texas, USA SWAT+, MODFLOW
Nguyen et al. (2020) Lower Saxony, Germany SWAT-IGF (karst)

the plain. Figure 1 depicts the Shiraz study area, waterway due to a large extraction from the limestone wells on the
network, and the locations of the climatic and hydrometric mountainside.
stations. Shallow groundwater-level contours, with depths
between 2 and 35 m, are associated with this alluvial
Hydrogeological properties aquifer. The water table is deeper in the western part and
shallower (to <2 m) in the eastern parts near Maharloo
The alluvial aquifer area of the Shiraz study region is Lake.
444.2 ­k m 2. Groundwater flows from the northwest and Areas with aquifer transmissivity in the range of
west highlands towards the southeast and Maharloo 100–3,000 ­m 2/day can be seen in the region, indicating
Lake. The iso-depth curves of the groundwater level in the good transmissivity of the aquifer in some parts of
the Shiraz plain vary from elevation 1,560 m above sea this plain. The transmissivity decreases from the north-
level (asl) in the west of the plain to 1,460 m asl near ern calcareous elevations to the south and southeast parts
Maharloo Lake located in the southeast of the plain. of the plain (outlet; Farsab Sanat Consulting Engineers
According to the spread of iso-depth curves in the north 2010). The water inflow to the aquifer enters from the
and northeast parts of the plain, groundwater flows from north and northwest, and the outflow exits from the
the alluvium towards the limestone highlands, possibly southeast part of the area, flowing into Maharloo Lake.

13
574 Hydrogeology Journal (2023) 31:571–587

Fig. 1  Study area, waterway network, and the locations of hydrometric and climatic stations

There are no other aquifers in the vicinity of this alluvial (MCM)/year, of which 112.90 MCM/year is from wells,
aquifer (Fig. 2). The total amount of discharge/withdrawal 69.13 MCM/year from aqueducts, and 7.09 MCM/year
from the alluvial aquifer is 189.12 million cubic meters from springs (Farsab Sanat Consulting Engineers 2010).

Fig. 2  Karstic areas and hydrogeology information for the case study

13
Hydrogeology Journal (2023) 31:571–587 575

Fig. 3  Active cells, river, and

exploitation and observation
wells in the SWAT and MOD-
FLOW models

Karst setting lime cement and weathering; despite its permeability due
to available gypsum and anhydrite, the Sachun formation
The basin of the Tashk-Maharloo and Bakhtegan lakes has causes a decrease in the quality of groundwater. Razak For-
regular folds along the northwest to southeast (Jamab Con- mation, which includes limestone, has a low transmissivity
sulting Engineers 2013). Northern and eastern parts of the due to the presence of marl.
study basin are lithologically composed of limestone and The Asmari-Jahrom, Bakhtiari conglomerate, Aghajari,
shale layers, with good permeability. The southwest part of Sachun, and Razak formations have infiltration rates of 47,
the basin contains red sediments of gypseous marl, dolomite, 45, 20, 35, and 10%, respectively, and a weighted average
and silty limestone, while the northwest includes a massive infiltration rate of 34.48% (Farsab Sanat Consulting Engi-
and resistant conglomerate structure (Fig. 2). Karstic areas neers 2010). Resulting from the suitable conditions of lithol-
of this region cover 37% of the basin, providing more than ogy, climate, tectonic activities, and the spread of soluble
50% of the drinking water of the city of Shiraz (Rahnemaei limestone and dolomite rocks, karst phenomena such as
et al. 2005). sinkholes, tower features, dry valleys, and caves have devel-
In the highlands of the study area, there are various oped (Khanlari and Momeni 2012). In this case study, there
formations such as Asmari-Jahrom, Bakhtiari, Aghajari, are 177 sinkholes in the highlands and karst formations and
Sachun, and Razak, with areas of 387.48, 109.82, 105.28, some faults, synclines, and anticlines. The size of the faults
42.93, and 183.89 k­ m2, respectively. Among these, the ranges from 1 to 24.5 km.
Asmari-Jahrom limestone formation can have good infiltra-
tion potential. The Asmari and Jahorm formations are actu- SWAT model
ally separate formations that are layered on top of each other;
therefore, it is very difficult to separate them in some areas SWAT is a conceptual, semidistributed, and catchment-scale
and they are considered integrated. The Bakhtiari conglom- model designed and established at the Agricultural Research
erate can also have a suitable infiltration potential if it has Service, United States Department of Agriculture (Arnold et al.

13
576 Hydrogeology Journal (2023) 31:571–587

Fig. 4  Schematic diagram of the modeling process in this study

1998). A basin is divided into several subbasins for simulation groundwater flow model (Varni and Usunoff 1999; Nastev
according to rivers. Each subbasin is further subdivided into et al. 2005; Palma and Bentley 2007; Sibanda et al. 2009;
smaller units known as hydrologic response units (HRUs), each Bedekar et al. 2011) that uses the groundwater flow rela-
with a unique slope, soil type, and land use. All available equa- tionships within an aquifer to simulate the flow (Todd and
tions in the model are calculated at the scale of each HRU. The Mays 2005)
Shiraz study area was divided into 41 subbasins and 188 HRUs Various code versions are used to solve the finite dif-
according to the threshold values and hydrometric stations. The ference equations associated with groundwater flow.
model was configured to extract waterways and subbasins using The MODFLOW-NWT model was used in this study.
a modified digital elevation model (DEM) map with a 15-m Introduced in 2011, MODFLOW-NWT is able to better
resolution. The land-use layer was derived from the Global Land model cell drying and rewetting (Niswonger et al. 2011).
Cover Characterisation (GLCC) categories of 2012 (USGS Eros For the MODFLOW model, the water level of the obser-
Archive 2012), while the soil layer, with a 1-km resolution, vation wells in the first month of the simulation, i.e.,
was provided by the Food and Agriculture Organization of the
United Nations (FAO 2008). A 17-year period, from January Table 2  Statistical evaluation criteria and their optimal values. See
2003 to December 2019, was considered for the modeling. The text for definitions of equation terms
surface-water modeling of the area in the period 2003–2016 Optimal Formula Coefficient
with a 3-year warm-up was considered for calibration, while the value
model was validated for the period 2017–2019. 1
�∑ � �� ��2
Coefficient of determination
i Qm −Qm Qs −Qs
R2 = ∑ � �2 ∑ � �2
MODFLOW i Qm −Qm i Qs −Qs

1 ∑
i (Qm −Qs )
2
Nash–Sutcliffe efficiency
NSE = 1 −
MODFLOW (McDonald and Harbaugh 1988) is a
∑ � �2
i Qm −Qm

computer code that simulates three-dimensional (3D) 0

�
∑N 2 Root mean square error
(Qs −Qm )
groundwater flow in a porous medium. Due to its exten- RMSE = 1
N
sive capability and validation in simulating various sys-
0 MAE = 1 ∑n
∣ Qs − Qm ∣ Mean absolute error
tems, MODFLOW is universally accepted as a standard n j=1

13
Hydrogeology Journal (2023) 31:571–587 577

Table 3  Sensitivity analysis of the parameters of the model and the range of their initial variations
Parameter No. Definition Variable name Default value Final value
Min Max Min Max

Surface runoff and channel processes

1 Initial SCS (The Soil Conservation Service) runoff curve number for mois- CN2.mgt 45 70 49.15 55
ture condition II
2 Effective hydraulic conductivity in main channel alluvium (mm/h) CH_K2.rte 0 500 59 120
3 Manning’s “n” value for the tributary channels CH_N1.sub 0.01 0.3 0.1 0.2
4 Manning’s “n” value for the main channel CH_N2.rte 0.01 0.3 0.09 0.2
5 Average slope length (m) SLSUBBSN.hru 10 130 69.25 75.55
6 Average slope steepness (m/m) HRU_SLP.hru 0 1 0.01 0.3
7 Temperature lapse rate (°C/km) TLAPS.sub –10 10 –5.7 7.1
8 Precipitation lapse rate (mm ­H2O/km) PLAPS.sub –1000 1000 –280 1000
9 Manning’s “n” value for overland flow OV_N.hru 0.01 30 15.22 25.98
10 Lateral flow travel time (days) LAT_TIME.hru 0 180 70.75 100.23
Evapotranspiration and plant water uptake
11 Soil evaporation compensation factor ESCO.hru 0 1 0.43 0.72
12 Threshold depth of water in the shallow aquifer for “revap” or percolation to REVAPMN.gw 0 500 50 220
the deep aquifer to occur (mm ­H2O)
13 Groundwater “revap” coefficient GW_REVAP.gw 0.02 0.2 0.02 0.1
Snowfall and snowmelt
14 Snowpack temperature lag factor TIMP.bsn 0 1 0.2 0.6
15 Melt factor for snow on 21st December (mm H ­ 2O/°C-day) SMFMN.bsn 0 10 0 2
Groundwater and crack flow processes
16 Baseflow alpha factor (1/day) ALPHA_BF.gw 0 1 0.11 0.2
17 Threshold depth of water in the shallow aquifer required for return flow to GWQMN.gw 0 5000 1000 3500
occur (mm H ­ 2O)
18 Groundwater delay time (days) GW_DELAY.gw 1 500 20 100
19 Available water capacity of the soil layer (mm ­H2O/mm soil) SOL_AWC(..).sol 0 1 0.4 0.6
20 Moist bulk density (Mg/m3 or g/cm3) SOL_BD(..).sol 0.9 2.5 1.5 2
21 Deep aquifer percolation fraction RCHRG_DP 0 1 0.2 0.8
22 Potential or maximum crack volume of the soil profile expressed as a fraction SOL_CRK 0 1 0.2 0.75
of the total soil volume

January 2003, was considered as the initial condition, there are exploitation wells, and (3) the subbasins and
and the bedrock was considered as the lower boundary HRUs that have agricultural irrigation, are entered into
of the aquifer. In this study, the MODFLOW network the swatmf_irrigate file, so that the model can estimate
was transformed into 28,557 cells with dimensions of the spatial conditions of exploitation wells for agricul-
300 × 300 m, each column of which had 167 cells, while tural, drinking, and industrial consumption. The amount
each row had 171 cells. The River package was used of discharge from each exploitation well in all simulation
to simulate the rivers in the model, and the river water time intervals is defined in the Well package. The active
levels and the hydraulic conductivity values of the river cells in the SWAT and MODFLOW models, as well as
were embedded in the model; moreover, 21 wells were river cells, are shown in Fig. 3. In the upstream and areas
selected for calibration, the groundwater levels of which that did not include the aquifer, the SWAT model calcu-
were recorded from January 2003 to December 2019. lated the infiltration and considered the groundwater and
Information about observation wells was applied using surface water from those upstream to enter the MOD-
the Observation package. Then, the data of 2009 was FLOW model. According to Fig 3, observation wells in
used in order to enter the groundwater discharge informa- the aquifer were generally located in the southern parts
tion. The number of exploitation wells was 1096. In this of the basin, reflecting that the amount of groundwater
model, the row and column numbers of cells (1) from exploitation is greater in the south and southeast due to
the MODFLOW network, (2) the subbasins in which the higher density of exploitation wells in these areas.

13
578 Hydrogeology Journal (2023) 31:571–587

Table 4  Results of calibration and validation of surface water using The model creates link files between subbasins and HRUs
the implemented models in the SWAT model and MODFLOW network cells to per-
SWAT-M0DFLOW SWAT-MODFLOW-CF form these exchanges. First, in the swatmf_dhru2hru file,
HRUs are separated according to the groundwater network
Discharge station Calibration Calibration Validation
and receive geographical coordinates to be distributed in the
Khoshk R2 0/66 0/70 0/76 groundwater model and are then called dhru. These sepa-
NSE 0/62 0/66 0/64 rated units are then linked to the MODFLOW groundwater
Nahre Azam R2 0/70 0/75 0/8 network in the swatmf_dhru2grid and swatmf_grid2dhru
NSE 0/60 0/73 0/71 files. Finally, the surface flow is linked to the subbasins in
the swatmf_river2grid file. The amount of infiltration from
surface water to groundwater in the SWAT-MODFLOW
model is calculated using these linking files. To estimate
the aquifer recharge in simulating groundwater flow, it is
SWAT‑MODFLOW model necessary to consider all factors such as infiltration through
the rainfall, river, and waterway. The SWAT model consid-
SWAT, as a semidistributed model, does not show the ers effective parameters such as topography, geology, soil,
heterogeneity of an aquifer, and therefore does not take climatic parameters, and plant management for each HRU to
into account the changes in hydraulic conductivity and estimate recharge. After matching, this value is assigned to
specific yield and the distribution of the groundwater the middle layers (dhru) and enters the MODFLOW network
level. Besides, to estimate discharge into the river, it com- cells in the SWAT-MODFLOW model. All simulation time
pares the depth of the shallow aquifer with the surface of intervals are defined in the Recharge file.
the river and does not consider the level of the riverbed
and the depth of the deep aquifer (Gassman et al. 2007; SWAT‑MODFLOW‑CF model
Kim et al. 2008; Arnold et al. 2012; Aliyari et al. 2019).
Although calculating groundwater recharge is one of the Since the depth, diameter, and amount of infiltration of the
main components of groundwater modeling, the MOD- sinkholes in karstic areas are different, it is not possible to
FLOW code does not consider some recharge components predict their behavior under different conditions, especially
such as river infiltration in groundwater recharge calcula- in response to precipitation. Therefore, modeling the karstic
tion. Therefore, due to the limitations of SWAT and MOD- components can provide a better understanding of a study
FLOW codes, the combination of these two models has area (Eini et al. 2020). Because infiltration in karstic areas
resulted in the development of efficient models in recent can be very quick, the CrackFlow approach in the SWAT
years—for example, Sophocleous et al. (1999) provided a model brings the values for surface infiltration closer to
link between SWAT and MODFLOW called SWATMOD, reality. CrackFlow or BypassFlow is one of the parts that
which could simulate surface-water flow, groundwater has recently entered the SWAT model code and was first
flow, and their interactions. The SWAT-MODFLOW used by Arnold and Fohrer (2005) to simulate some parts
integrated model (Bailey et al. 2016) allows simulation of Texas, United States. Later, successful applications of
of large basins with thousands of MODFLOW cells and the modified SWAT model for karstic conditions have been
tens of thousands of HRUs for SWAT. This model takes reported worldwide, including Nerantzaki et al. (2015,
account of withdrawal from resource exploitation wells by 2020), Amin et al. (2017), Wang et al. (2019), Delavar et al.
linking the exploitation cells in MODFLOW and SWAT (2020), Nguyen et al. (2020) and Eini et al. (2020). This
HRUs. Also, in a situation where the aquifer does not have approach is developed for regions with a soil texture of ver-
enough water, or there is no aquifer, the SWAT-MOD- tisol, and it is assumed that this type of soil behaves similarly
FLOW model supplies water for irrigation from a river in in karstic areas (Neitsch et al. 2005). In CrackFlow, factors
the subbasin, which is called the ‘auto-irrigation’ function. such as the Sol_CRK coefficient are essential to estimate
Figure 4 depicts a schematic diagram of the modelling the volume of seams and cracks by the model and consider
process in this study. The implementation period of this their maximum possible volume in the soil. The Sol_CRK
model for the study area was from 2003 to 2019. coefficient is one of the parameters used in the calibration
In order to perform surface-water and groundwater stage, which, among other factors, is effective in estimating
exchanges in the SWAT-MODFLOW model, the amount the volume of seams and cracks and is calibrated according
of surface infiltration is estimated by SWAT and enters to subbasins that have a karstic structure. Thus, according
the MODFLOW model automatically as input. Further- to this approach, seams and cracks are formed in the dry
more, the amount of drained water from the aquifer and the season, penetrating to a depth of more than 50 cm. These
MODFLOW network enters the river and SWAT model. seams and cracks, wider in the parts near the surface than the

13
Hydrogeology Journal (2023) 31:571–587 579

Table 5  Results of calibration and validation of groundwater level preparing the SWAT-MODFLOW-CF executable (.exe)
using the implemented models file.
SWAT-MODFLOW SWAT-MODFLOW-CF In order to more realistically represent karstic condi-
tions, the calculation of crack volume was modified so
Error coefficient Calibration Calibration Validation
that the cracks can also form in wet soils. Currently, the
RMSE 1/77 1/33 2/47 crack volume is a function of soil moisture. Although
MAE 1/23 1/01 2/39 volcr is zero for wet conditions in the standard SWAT
code, the CrackFlow module of standard SWAT was mod-
ified in SWAT-CF to remove the distinction of dry and
lower parts, are not unusual and are typically seen in nature wet soil conditions during the CrackFlow process
with a depth of about 6–15 cm.
SWAT calculates the crack volume of the soil matrix
volcrl,j = crlag × volcrl,j + (1 − crlag) × volcrnew (2)
for each day of simulation by layer. On days in which where volcr is the crack volume for the soil layer (mm),
precipitation events occur, infiltration and surface run- crlag is the lag factor, v­ olcr new is the crack volume for the
off are first calculated for the soil peds using the curve soil layer based on new moisture conditions (mm), j is
number or Green and Ampt method. If any surface runoff the HRU number, and l is a counter. The implementation
is generated, it is allowed to enter the cracks. A volume period for this model was 2003–2019, with a validation
of water equivalent to the total crack volume for the soil period between 2017 and 2019.
profile may enter the profile as crack flow. Surface runoff
in excess of the crack volume remains overland flow. The
water entering these seams and cracks first fills the lower Error evaluation criteria
layers of the soil and moves upwards to saturate the upper
layers as well, and then deep penetration conditions are Considering that the SWAT model is a physical model, its
provided for the infiltrated water. implementation and adaptation to the conditions of the region
Neitsch et al. (2011) presented Eq. (1) to calculate the initial to provide appropriate results require many input parameters.
(i) volume of the cracks (volcr) for one layer per day (mm): Each parameter needs to be calibrated due to the uncertainties
volcr(ly, i) = (vol(max, ly) × CF(crk) involved in its value. The SUFI2 method in the SWAT-CUP
×C(crk) − W(ly))∕(CF(crk) × C(ly)) (1) program was used to calibrate surface water in the SWAT,
SWAT-MODFLOW and SWAT-MODFLOW-CF models.
where vol(max,ly) is the maximum volume of the soil The models performed in the calibration and validation
layer (mm), CF(crk) is the adjustment coefficient of seams stages are compared with the observed values of monthly runoff
and cracks, which is 1, C(crk) is the amount of water in in the surface water and the groundwater level in observation
the soil according to the ground capacity (mm H ­ 2O), and wells using four statistical evaluation criteria shown in Table 2.
W(ly) is water content of the soil layer per day (mm H ­ 2O). The coefficient of determination (R2), Nash-Sutcliffe efficiency
In this study, to bring the conditions of the SWAT model (NSE), mean absolute error (MAE), and root mean squared
closer to the real conditions of karstic basins, the relevant error (RMSE) were calculated as error evaluation criteria. In
codes have been modified, the conditions of dry and wet soils these equations, Qm is the observed (measured) value, Q m ̅ is
have been equalized, and the closure of seams and cracks has the average observed value, Qs is the simulated value, Q s̅ is the
been prevented. Also, consider that temporal changes in this average simulated value, and N is the total number of samples.
part of the model are essential to increase surface runoff and
infiltration simulation performance. The significant differ-
ences in precipitation, temperature, evapotranspiration and Results
soil moisture in this study area caused a large difference in
surface runoff during various seasons, including winter and In this study, the SUFI2 method in the SWAT-CUP
summer. Therefore, the amount of soil moisture, the time program has been used to perform calibration steps in
needed for soil saturation, and the filling volume of seams SWAT-MODFLOW and SWAT-MODFLOW-CF models.
and cracks were different throughout the year. In order to perform effective calibration, it was neces-
In this study, using Microsoft Visual Studio, the karstic sary to analyze the sensitivity of the model parameters.
components and modification of CrackFlow were applied To perform sensitivity analysis, first, an initial selection
to the SWAT-MODFLOW codes, which were prepared in of parameters affecting the water flow was performed
the FORTRAN programming language. The changes were based on the available information on the basin and a
performed using the Intel Fortran Compiler, followed by review of previous studies (Eini et al. 2019). A total of

13
580 Hydrogeology Journal (2023) 31:571–587

Fig. 5  a Volumetric values of water infiltration obtained by the water infiltration obtained by the SWAT-MODFLOW-CF model; d
SWAT-MODFLOW model; b The calibrated values of the SolCrack Locations of faults, synclines, anticlines, and sinkholes
(SOL_CRK) parameter for each subbasin; c Volumetric values of

20 effective parameters were identified and selected for SWAT‑MODFLOW calibration

sensitivity analysis (Table 3). In the calibration stage,
soil-related parameters such as Sol_BD and Sol_AWC, In order to calibrate the surface water in the SWAT-MOD-
along with parameters such as surface-water delay, the FLOW and SWAT-MODFLOW-CF models by making
average slope of subbasin, and hydraulic conductivity of changes in the model executable file, the SWAT-CUP soft-
the waterway, were the most sensitive parameters. Also, ware was able to read the codes related to SWAT-MOD-
the parameters related to the base flow, which are related FLOW and execute them. Therefore, parameters related to
to groundwater, showed high sensitivity in the model. surface water, crack flow, and some groundwater parameters

13
Hydrogeology Journal (2023) 31:571–587 581

Fig. 6  Hydrology information,

names, and locations of the
piezometers (karstic areas are
defined in detail in Fig. 2)

were calibrated in SWAT-CUP, followed by a review in a for the Khoshk station and 0.70 and 0.60 for the Nahr-e Azam
geographic information system (GIS). Also, to calibrate the station (Table 4). According to Moriasi et al. (2007), these
hydraulic conductivity coefficient and specific yield in the results are considered acceptable for the monthly-scale data.
SWAT-MODFLOW model, the aquifer map was divided into On the other hand, the mean RMSE and MAE values
smaller parts and checked in ArcGIS. For each model run, for 21 observation wells in the study area were 1.77 and
files were created in ArcGIS, then converted to raster and 1.23, respectively (Table 5). These values are acceptable
ASCII files so they could be copied in the execution path for simulating groundwater in a 17-year period, and the
with .txt format. In order to simultaneously calibrate the sur- model has been able to efficiently simulate the observed
face runoff, groundwater level, hydraulic conductivity, and data in months with high and low amounts of rainfall.
specific yield, the model was run many times, and the output
files were checked in ArcGIS, Excel, and SWAT-CUP. The SWAT‑MODFLOW‑CF calibration
changes in the surface-water parameters affected the ground-
water level. Also, the changes in various parts of the map of By activating CrackFlow in the basin simulation, the seams
hydraulic conductivity and specific yield resulted in changes and cracks are considered open, and the infiltration amount
in the surface runoff. The output file swatmf_out_MF_obs increases. In model calibration, the SOL_CRK parameter
shows the amount of water level in the cell corresponding is essential and is calibrated according to the permeability
to each observation well on a daily basis. According to this of the formation and the area of the karstic basin. The ini-
file, the monthly average value of each observation well was tial values assigned to the sub-basins were entered into the
calculated, and the monthly hydrograph of each well was SWAT-CUP model as a numerical range according to the
obtained. geological map, soil type (Fig. 2), and the location of faults
The implementation of the SWAT-MODFLOW integrated and sinkholes (Fig. 5d).
model for the Shiraz study area to consider the effects of surface The greater the area, the permeability of the karstic for-
water and groundwater on each other, as well as their exchanges, mation in each subbasin and the number of sinkholes, the
resulted in the calibration R2 and NSE values of 0.66 and 0.62 larger the SOL_CRK parameter. The value of this parameter

13
582 Hydrogeology Journal (2023) 31:571–587

Fig. 7  Calibration results for surface water of Khoshk and Nahre-e Azam stations a using the SWAT-MODFLOW and b using the SWAT-MOD-
FLOW-CF, and c validation results for surface water using the SWAT-MODFLOW-CF

is considered zero for subbasins that do not have karstic higher in these areas. This amount of infiltration is entered
formations. into the MODFLOW model in the form of inflow water from
These values were calibrated by iterative executions of the highlands.
the model. The final values of the SOL_CRK parameter and As mentioned earlier, a period from 2003 to 2016 was
the output of the model are shown in Fig. 5b. The volu- considered for calibration, while the period 2017–2019
metric values of water infiltration obtained by the SWAT- was used to validate the SWAT-MODFLOW-CF model.
MODFLOW model are shown in Fig. 5b, where the map is Using this model, the R 2 and NSE values in the runoff
prepared from the output file of the model, showing the infil- calibration were 0.70 and 0.66, respectively, for the
tration values from SWAT as an annual average. This value, Khoshk station, while these values were 0.75 and 0.73,
after running the SWAT-MODFLOW-CF model, is shown in respectively, for the Nahr-e-Azam station (Table 4).
Fig. 5c. As can be seen in the figure, as the result of imple- Therefore, surface runoff simulation has been performed
menting the SWAT-MODFLOW-CF model, the amount of well using the SWAT-MODFLOW-CF model. In Fig. 6,
water infiltration has increased, reaching a realistic value the locations of hydrometric stations are shown, also
after implementing the CrackFlow code and considering the Fig. 7 shows the time-series data of simulated runoff
seams and cracks. along with the observational data at the hydrometric sta-
In the areas with karst formations where the dissolution tions considered in this study for the calibration period
ability of rocks was greater with a higher number of faults using the SWAT-MODFLOW and SWAT-MODFLOW-CF
and sinkholes, the calibrated value of the SOL_CRK param- models. Figure 7c shows the results after validation at
eter in the SWAT-CUP model was higher. Moreover, accord- both Khoshk and Nahr-e Azam stations. The similarity
ing to Fig. 5c, the amount of infiltration from SWAT was of the hydrograph of the observed and simulated data, as

13
Hydrogeology Journal (2023) 31:571–587 583

Fig. 8  Comparison of calibration results of groundwater levels using the SWAT-MODFLOW and SWAT-MODFLOW-CF versus time, with
time-steps of 1 month, at selected observation wells

well as their similar trends, shows the high performance Discussion

of this model in simulating base flow values, especially
from 2010 to 2016. In this study, first, surface-water modeling has been con-
Moreover, according to the information entered in ducted using the SWAT model. Then, surface-water and
the SWAT-MODFLOW-CF model, the average values groundwater modeling has been done by the SWAT-MOD-
of RMSE and MAE for the calibration processes were FLOW model, where the CrackFlow code has been modi-
1.33 and 1.01, respectively (Table 5). The groundwater fied to introduce a new model called SWAT-MODFLOW-
levels in some piezometers of the Shiraz study area after CF to simulate surface water and groundwater in the study
the calibration of SWAT-MODFLOW-CF are shown in area during a period between 2003 and 2019. According to
Fig. 8, and the location of these piezometers is shown in the results, by simulating both surface water and ground-
Fig. 6. The closeness of the hydrographs of the observed water as an integrated model, the simulated data were more
and simulated data, the uniformity of the trends of consistent with the observed data than with the SWAT
changes even in months with high and low rainfalls, and modeling. By considering groundwater for surface-water
the RMSE and MAE values after calibration show the modeling, the integrated simulation brought the simulated
remarkable performance of the SWAT-MODFLOW-CF water flows closer to the observed values. In the SWAT-
in the simulation process. MODFLOW modeling, the base flow in both study stations

13
584 Hydrogeology Journal (2023) 31:571–587

Table 6  Comparison of Water budget Source

groundwater balance (million
cubic meters, MCM) Ministry of Energy SWAT-MODFLOW SWAT-MODFLOW-CF

Input
Groundwater 17.45 45.93 57.51
Infiltration of rain 41.75
Infiltration of surface water 6.00 4.5 5.2
Infiltration of sewage 40.21 35.4 39
Infiltration of drinking water 92.23 87.57 91.32
and from industry
Total input 197.64 173.4 193.03
Output
Withdrawals 189.12 172.67 185.43
Drainage 4.1 4.2 4.2
Groundwater output 0.86 1.3 0.95
Evaporation from aquifer 4.2 6.7 5.1
Total output 198.28 184.87 195.68
Aquifer storage change –0.64 –11.47 –2.65

showed a higher value, especially in the years 2010 to balances and is equal to the excessive withdrawal from
2015, which significantly impacted the improvement of the aquifer storage in negative balances. The downstream
statistical evaluation criteria. However, the amount of base observation wells showed a more realistic value as well
flow at these stations is still higher than the observed data, as a greater infiltration value for each of the balance
and a relatively accurate value in peak and base flows has parameters using the SWAT-MODFLOW-CF model due
not been obtained. Also, although the groundwater level in to the rainfall and seams and cracks in the highlands.
piezometers has been well simulated in time intervals with The implementation of this model reduced the uncertain-
low and high rainfall amounts using this modeling tech- ties in calculating the water balance and also brought the
nique, the observed and simulated data are not completely modeled values of the balance parameters closer to the
similar to each other. At some points, they differ greatly, observed values.
and the water level in the piezometers is lower than the
observed data. Considering the infiltration values in the
karstic areas and modifying the CrackFlow code brought Conclusions
about an increase in the infiltration value and raised the
water level in most piezometers. Also, modifying the Integrated simulation of surface water and groundwater led
CrackFlow code had a significant effect on reducing the to a significant understanding of the groundwater and sur-
base flow in the surface water, and the volume of base and face-water resources in the study area as well as its karstic
peak flows was less than the modeled values before acti- formations. Identifying the karstic areas of the study region
vating the CrackFlow parameter, resulting in more accu- and estimating the infiltration values of these karstic forma-
rate surface runoff simulation. These changes indicated tions resulted in a better simulation of surface water and
better performance of the SWAT-MODFLOW-CF model groundwater in this study.
than the SWAT-MODFLOW model. It should be noted that The SWAT-MODFLOW and SWAT-MODFLOW-CF
after the implementation of the CrackFlow code, Eini et al. models were calibrated for the 2003–2016 period and vali-
(2020) also obtained better results in surface-water simula- dated for the 2017–2019 period. After implementing the
tion of Maharloo and Tashk Bakhtegan studies compared SWAT-MODFLOW integrated model and simultaneous
to the original SWAT model. calibration of surface water and groundwater at the Khoshk
The comparison of groundwater balance obtained by hydrometric station, the values of R2 and NSE were obtained
the SWAT-MODFLOW and SWAT-MODFLOW-CF mod- at 0.66 and 0.62, respectively. R2 and NSE values of 0.70
els is presented in Table 6. According to area balance cal- and 0.60, respectively, were obtained for the Nahr-e Azam
culations from a Shiraz study in 2009, the changes in the hydrometric station. Also, the RMSE and MAE values
storage volume of the study aquifer were equal to –0.67 obtained were 1.77 and 1.23, respectively, in calibrating the
MCM. This value is almost equal to zero in equivalent groundwater level.

13
Hydrogeology Journal (2023) 31:571–587 585

The CrackFlow code has been modified in this inte- Afrasiabian A (1998) The importance of studies and researches on
grated model, and the seams and cracks in the karst areas water resources in Iran. Articles of the second international
symposium in water resources on karstic area of Tehran-Ker-
are reconsidered when the soil in the area is dry or wet. In manshah. Western Regional Water Organization, Kerman-
the simultaneous calibration of surface water and ground- sha, Iran, pp 126–137
water of the SWAT-MODFLOW-CF model, the R 2 and Aliyari F, Bailey RT, Tasdighi A, Dozier A, Arabi M, Zeiler K
NSE values were obtained at 0.70 and 0.66, respectively, (2019) Coupled SWAT-MODFLOW model for large-scale
mixed agro-urban river basins. Environ Model Softw 115:200–
for the Khoshk station, and 0.75 and 0.73, respectively, 210. https://​doi.​org/​10.​1016/j.​envso​ft.​2019.​02.​014
for the Nahr-e Azam station. Also, in the validation stage, Amin MGM, Veith TL, Collick AS, Karsten HD, Buda AR (2017)
the values of R2 and NSE were 0.76 and 0.64 for the Kho- Simulating hydrological and nonpoint source pollution pro-
shk hydrometric station and 0.80 and 0.71 for the Nahr-e cesses in a karst watershed: a variable source area hydrology
model evaluation. Agric Water Manag 180:212–223. https://​doi.​
Azam hydrometric station, respectively. In the ground- org/​10.​1016/j.​agwat.​2016.​07.​011
water calibration, RMSE and MAE values were obtained Amin MGM, Veith TL, Shortle JS, Karsten HD, Kleinman PJA
at 1.33 and 1.01, respectively, and these values were (2020) Addressing the spatial disconnect between national-scale
obtained as 2.47 and 2.39 in the validation stage. total maximum daily loads and localized land management deci-
sions. J Environ Qual 49(3):613–627. https://​doi.​org/​10.​1002/​
In the SWAT-MODFLOW-CF model, by modifying the jeq2.​20051
CrackFlow code and reconsidering the seams and cracks Arnold JG, Fohrer N (2005) SWAT2000: current capabilities and
in both dry and wet conditions, the amount of infiltrated research opportunities in applied watershed modelling. Hydrol
water increased and better modeling results were obtained Process 19(3):563–572
Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area
by reducing the difference between the observed and simu- hydrologic modeling and assessment: part 1, model development.
lated values. Furthermore, recharge, which is one of the J Am Water Resour Assoc 34:73–89
critical and influential factors of groundwater flow, is not Arnold JG, Moriasi DN, Gassman PW, Abbaspour KC, White MJ,
precisely determined in most models, so it should be cali- Srinivasan R, Santhi C, Harmel RD, Van Griensven A, Van Liew
MW, Kannan N, Jha MK (2012) SWAT: model use, calibration,
brated in the calibration stage. However, in this study, the and validation. Trans ASABE 55(4):1491–1508
amount of infiltration was automatically transferred from Baffaut C, Benson VW (2009) Modeling flow and pollutant transport
the HRUs of the SWAT model to the MODFLOW cells; in a karst watershed with SWAT. Trans ASABE 52(2):469–479.
hence, the infiltrated water enters groundwater almost https://​doi.​org/​10.​13031/​2013.​26840
Bailey RT, Wible TC, Arabi M, Records RM, Ditty J (2016) Assessing
accurately and does not need to be calibrated, thus reduc- regional-scale spatio-temporal patterns of groundwater–surface
ing the number of parameters to be calibrated. water interactions using a coupled SWAT-MODFLOW model.
The calibration and validation results in the observation Hydrol Process 30(23):4420–4433. https://​doi.​org/​10.​1002/​hyp.​
wells of the aquifer during a 17-year period were acceptable, 10933
Bailey RT, Rathjens H, Bieger K, Chaubey I, Arnold J (2017) SWAT-
and the SWAT-MODFLOW-CF model has simulated the MOD-Prep: graphical user interface for preparing coupled SWAT-
groundwater level of the Shiraz aquifer well, which confirms MODFLOW simulation. J Am Water Resour Assoc 53(2):400–
that the amount of recharge used in the model has been cal- 410. https://​doi.​org/​10.​1111/​1752-​1688.​12502
culated correctly. It should also be noted that the CrackFlow Bailey RT, Park S, Bieger K, Arnold JG, Allen PM (2020) Enhancing
SWAT + simulation of groundwater flow and groundwater-surface
modification had a significant effect on the simulation of the water interactions using MODFLOW routines. Environ Model
aquifer level as well as the better simulation of the surface Softw 126. https://​doi.​org/​10.​1016/j.​envso​ft.​2020.​104660
runoff. It is recommended to use the CrackFlow approach in Bedekar V, Niswonger RG, Kipp K, Panday S, Tonkin M (2011)
the SWAT+, MODFLOW, and QSWAT-MODFLOW inte- Approaches to the simulation of unconfined flow and perched
groundwater flow in MODFLOW. Ground Water 50(2):187–198.
grated models in other karstic areas. https://​doi.​org/​10.​1111/j.​1745-​6584.​2011.​00829.x
Chung IM, Lee J, Kim NW, Na H, Chang SW, Kim Y, Kim GB (2015)
Estimating exploitable amount of groundwater abstraction using
Declarations an integrated surface water–groundwater model: Mihocheon
watershed, South Korea. Hydrol Sci J 60(5):863–872. https://​doi.​
Conflicts of interest The Authors declare that they have no conflict of org/​10.​1080/​02626​667.​2014.​980261
interest. Chunn D, Faramarzi M, Smerdon B, Alessi DS (2019) Application
of an integrated SWAT-MODFLOW model to evaluate potential
impacts of climate change and water withdrawals on ground-
water–surface water interactions in west-central Alberta. Water
11(1):110–138. https://​doi.​org/​10.​3390/​w1101​0110
References Delavar M, Morid S, Morid R, Farokhnia A, Babaeian F, Srinivasan R,
Karimi P (2020) Basin-wide water accounting based on modified
Abbas S, Xuan Y, Bailey R (2018) Improving river flow simulation SWAT model and WA+ framework for better policy making. J
using a coupled surface-groundwater model for integrated water Hydrol 585:124762
resources management. EPiC Series Eng 3:1–9. https://​doi.​org/​ Dowlatabadi S, Ali Zomorodian SM (2016) Conjunctive simulation of
10.​29007/​6ft7 surface water and groundwater using SWAT and MODFLOW in

13
586 Hydrogeology Journal (2023) 31:571–587

Firoozabad watershed. KSCE J Civ Eng 20(1):485–496. https://​ for irrigation and drinking water. Hydrol Earth Syst Sci Discuss.
doi.​org/​10.​1007/​s12205-​015-​0354-8 https://​doi.​org/​10.​5194/​hess-​2019-​232
Ehtiat M, Mousavi SM, Srinivasan R (2018) Groundwater modeling Malagò A, Efstathiou D, Bouraoui F, Nikolaos N, Franchini M,
under variable operating conditions using SWAT, MODFLOWand Bidoglio G, Kritsotakis M (2016) Regional scale hydrologic
MT3DMS: a catchment scale approach to water resources man- modeling of a karst-dominant geomorphology: the case study
agement. Water Resour Manag 32:1631–1649. https://​doi.​org/​10.​ of the Island of Crete. J Hydrol 540:64–81. https://​doi.​org/​10.​
1007/​s11269-​017-​1895-z 1016/j.​jhydr​ol.​2016.​05.​061
Eini MR, Saman Javadi S, Delavar M, Monteiro JAF, Darand M (2019) McDonald MG, Harbaugh AW (1988) A modular three dimensional
High accuracy of precipitation reanalyses resulted in good river finite-difference ground-water flow model. US Geol Surv Tech
discharge simulations in a semi-arid basin. Ecol Eng 131:107–119 Water Resour Invest 06-A1
Eini MR, Javadi S, Delavar M, Gassman PW, Jarihani B (2020) Molina-Navarro E, Bailey RT, Andersen HE, Thodsen H, Nielsen
Development of alternative SWAT-based models for simulat- A, Park S, Jensen JS, Jensen JB, Trolle D (2019) Comparison
ing water budget components and streamflow for a karstic- of abstraction scenarios simulated by SWAT and SWAT-MOD-
influenced watershed. Catena 195:104801 FLOW. Hydrol Sci J 64:434–454
Eshtawi T, Evers M, Tischbein B, Diekkrüger B (2016) Integrated Moriasi DN, Arnold JG, Van Liew MW, Binger RL, Harmel RD,
hydrologic modeling as a key for sustainable urban water Veith T (2007) Model evaluation guidelines for systematic
resources planning. Water Res 101:411–428. https://​doi.​org/​ quantification of accuracy in watershed simulations. Soil Water
10.​1016/j.​watres.​2016.​05.​061 Division ASABE 50(3):885–900
Food and Agriculture Organization of the United Nations (2008) The Nastev M, Rivera A, Lefebvre R, Martel R, Savard M (2005) Numer-
digital soil map of the world version 1.2. FAO, Rome. http://​ ical simulation of groundwater flow in regional rock aquifers,
www.​fao.​org/​soils-​portal/​soil-​survey/​soil-​maps-​and-​datab​ases/​ southwestern Quebec, Canada. Hydrogeol J 13:835–848. https://​
harmo​n ized-​world​s oil-​d atab​a se-​v 12/​e n. Accessed February doi.​org/​10.​1007/​s10040-​005-​0445-6
2023 Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2005) Assessment
Gassman PW, Reyes MR, Green CH, Arnold JG (2007) The soil tool soil and water documentation theoretical version 2005.
and water assessment tool: historical development, applications, Report no. 76502, Grassland, Soil and Water Research Labora-
and future research directions. Trans ASABE 50(4):1211–1250 tory, Temple TX
Ghezelayagh P, Javadi S, Kavousi A (2020) Assessment of groundwa- Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2011) Soil and water
ter recharge in carbonate aquifers based on a modified KARST- assessment tool theoretical documentation version 2009. Report
LOP–AHP method (case study: Dorfak region, Iran). Environ no. 406, Texas Water Resources Institute, College Station, TX
Earth Sci 79:92. https://​doi.​org/​10.​1007/​s12665-​020-​8829-4 Nerantzaki SD, Giannakis GV, Efstathiou D, Nikolaidis NP, Sibetheros
Huo AD, Dang J, Song JX, Chen XH, Mao HR (2016) Simulation IA, Karatzas GP, Zacharias I (2015) Modeling suspended sedi-
modeling for water governance in basins based on surfacewater ment transport and assessing the impacts of climate change in a
and groundwater. Agric Water Manag 174:22–29. https://​doi.​ karstic Mediterranean watershed. Sci Total Environ 538:288–297.
org/​10.​1016/j.​agwat.​2016.​02.​027 https://​doi.​org/​10.​1016/j.​scito​tenv.​2015.​07.​092
Iran’s Ministry of Energy (2013) Comprehensive water plan update Nerantzaki SD, Hristopulos DT, Nikolaidis NP (2020) Estimation of
studies, geology and studies of soil resources, Iran (in Persian). the uncertainty of hydrologic predictions in a karstic Mediter-
Jamab Consult. Eng., Tehran ranean watershed. Sci Total Environ 717:137131
Izady A, Davary K, Alizadeh A, Ziaei NA, Akhavan S, Alipoor A, Nguyen VT, Dietrich J, Uniyal B (2020) Modeling interbasin ground-
Joodavi A, Brusseau ML (2015) Groundwater conceptualiza- water flow in karst areas: model development, application, and
tion and modeling using distributed SWAT based recharge for calibration strategy. J Env Soft 124:104606. https://​doi.​org/​10.​
the semi-arid agricultural Neishaboor plain, Iran. Hydrogeol J 1016/j.​envso​ft.​2019.​104606
23:47–68. https://​doi.​org/​10.​1007/​s10040-​014-​1219-9 Nikolaidis NP, Bouraoui F, Bidoglio G (2013) Hydrologic and geo-
Kamali A, Niksokhan MH (2017) Multi-objective optimization for chemical modeling of a karstic Mediterranean watershed. J Hydrol
sustainable groundwater management by developing of coupled 477:129–138. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2012.​11.​018
quantity-quality simulation-optimization model. J Hydroinformat- Niswonger RG, Panday S, Ibaraki M (2011) MODFLOW-NWT, a
ics 19(6):973–992. https://​doi.​org/​10.​2166/​HYDRO.​2017.​007 Newton formulation for MODFLOW-2005. US Geol Surv Tech
Khanlari GH, Momeni A (2012) Geomorphology, hydrogeology and Methods 6-A37
the study of factors affecting to karst development in Garin Palanisamy B, Workman SR (2015) Hydrologic modeling of flow
area, west of Iran. Geography Territorial Spatial Arrangement through sinkholes located in streambeds of Cane Run Stream,
2(3):61–73 Kentucky. J Hydrol Eng 20(5). https://​doi.​org/​10.​1061/​(asce)​he.​
Kim N, Chung I, Yoo S, Arnold JG (2008) Development and appli- 1943-​5584.​00010​60
cation of the integrated SWAT-MODFLOW model. J Hydrol Palma HC, Bentley LR (2007) A regional-scale groundwater flow
356(1):1–16 model for the Leon-Chinandega aquifer, Nicaragua. Hydrogeol J
Kim JT, Choo CO, Kim MI, Jeong GC (2017) Validity evaluation of 15:1457–1472. https://​doi.​org/​10.​1007/​s10040-​007-​0197-6
a groundwater dam in Oshipcheon River, eastern Korea using a Park S, Nielsen A, Bailey RT, Trolle D, Bieger K (2019) A QGIS-based
SWAT-MODFLOW model. Environ Earth Sci 76(22):769–781. graphical user interface for application and evaluation of SWAT-
https://​doi.​org/​10.​1007/​s12665-​017-​7085-8 MODFLOW models. Environ Model Softw 111:493–497. https://​
Lamoreaux PE (1991) History of karst hydrogeological studies. doi.​org/​10.​1016/j.​envso​ft.​2018.​10.​017
Quaderni del dipartimento di gegrafia no. 13. In: IGU Proceed- Petpongpan C, Ekkawatpanit C, Kositgittiwong D (2020) Climate
ings of the International Conference on Environmental Changes change impact on surface water and groundwater recharge in north-
in Karst Areas, IGU, Universita di Padova, Italy, September ern Thailand. Water 12:1029. https://d​ oi.o​ rg/1​ 0.3​ 390/w
​ 12041​ 029
1991, pp 215–229 Rahnemaei M, Zare M, Nematollahi AR, Sedghi H (2005) Applica-
Liu W, Park S, Bailey RT, Molina-Navarro E, Andersen HE, Thodsen tion of spectral analysis of daily water level and spring discharge
H, Nielsen A, Jeppesen E, Jensen JS, Jensen JB et al (2019) hydrographs data for comparing physical characteristics of karstic
Comparing SWAT with SWAT-MODFLOW hydrological simu- aquifers. J Hydrol 311:106–116. https://​doi.​org/​10.​1016/j.​jhydr​
lations when assessing the impacts of groundwater abstractions ol.​2005.​01.​011

13
Hydrogeology Journal (2023) 31:571–587 587

Sabzzadeh I, Shourian M (2020) Maximizing crops yield net benefit in a Todd DK, Mays L (2005) Groundwater hydrology, 3rd edn. Wiley,
groundwater-irrigated plain constrained to aquifer stable depletion Hoboken, NJ
using a coupled PSO-SWAT-MODFLOW hydro-agronomic model. USGS EROS archive (2012) Land cover products: global land use
J Clean Prod 262:121349 land-cover characterization. https://​lta.​cr.​usgs.​gov/​glcc/​globe_​
Semiromi MT, Koch M (2019) Analysis of spatio-temporal variability int. Accessed February 2023
of surface–groundwater interactions in the Gharehsoo river basin, Guzman JA, Moriasi DN, Gowda PH, Steiner JL, Starks PJ, Arnold
Iran, using a coupled SWAT-MODFLOW model. Environ Earth JG, Srinivasan R (2015) A model integration framework for link-
Sci 78:201–222 ing SWAT and MODFLOW. Environ Model Softw 73:103–116.
Sibanda T, Nonner JC, Uhlenbrook S (2009) Comparison of groundwa- https://​doi.​org/​10.​1016/j.​envso​ft.​2015.​08.​011
ter recharge estimation methods for the semi-arid Nyamandhlovu Varni MR, Usunoff EJ (1999) Simulation of regional-scale groundwater
area, Zimbabwe. Hydrogeol J 17:1427–1441. https://​doi.​org/​10.​ flow in the Azul River basin, Buenos Aires Province, Argentina.
1007/​s10040-​009-​0445-z Hydrogeol J 7:180–187. https://​doi.​org/​10.​1007/​s1004​00050​190
Sith R, Watanabe A, Nakamura T, Yamamoto T, Nadaoka K (2019) Wang C, Wu J, Zeng W, Zhu Y, Huang J (2019) Five-year experimental
Assessment of water quality and evaluation of best management study on effectiveness and sustainability of a dry drainage system
practices in a small agricultural watershed adjacent to Coral Reef for controlling soil salinity. Water 11:111–126. https://​doi.​org/​
area in Japan. Agric Water Manag 213:659–673. https://​doi.​org/​ 10.​3390/​w1101​0111
10.​1016/j.​agwat.​2018.​11.​014 Wei X, Bailey RT, Records RM, Wible TC, Arabi M (2019) Compre-
Sophocleous MA, Koelliker JK, Govindaraju RS, Birdie T, Ramired- hensive simulation of nitrate transport in coupled surface-sub-
dygari SR, Perkins SP (1999) Integrated numerical modeling for surface hydrologic systems using the linked SWAT-MODFLOW-
basin-wide water management: the case of the Rattlesnake Creek RT3D model. Environ Model Softw 122. https://d​ oi.o​ rg/1​ 0.1​ 016/j.​
basin in south-central Kansas. J Hydrol 214:179–196. https://​doi.​ envso​ft.​2018.​06.​012
org/​10.​1016/​S0022-​1694(98)​00289-3 Yactayo GA (2009) Modification of the SWAT model to simulate
Surinaidu L, Muthuwatta L, Amarasinghe UA, Jain SK, Ghosh NC, hydrologic processes in a karst-influenced watershed. MSc Thesis,
Kumar S, Singh S (2016) Reviving the Ganges water machine: Virginia Polytechnic Institute, Blacksburg, VA
accelerating surface water and groundwater interactions in the Ram-
ganga sub-basin. J Hydrol 540:207–219. https://​doi.​org/​10.​1016/j.​ Publisher’s note Springer Nature remains neutral with regard to
jhydr​ol.​2016.​06.​025 jurisdictional claims in published maps and institutional affiliations.
The Regional Water Company of Fars (2010) Water budget report (in
Persian). Report no. 10090-RPT-WB-SH-001-02, Farsab Sanat
Consult. Eng., Fars, Iran

13