Accepted Manuscript

Regional Scale groundwater Modelling Study for Ganga River Basin

Maheswaran, R. Khosa, A.K. Gosain, S. Lahari, S.K. Sinha, B.R. Chahar, C.T.
Dhanya

PII: S0022-1694(16)30458-9
DOI: http://dx.doi.org/10.1016/j.jhydrol.2016.07.029
Reference: HYDROL 21415

To appear in: Journal of Hydrology

Received Date: 2 October 2015

Revised Date: 14 July 2016
Accepted Date: 17 July 2016

Please cite this article as: Maheswaran, Khosa, R., Gosain, A.K., Lahari, S., Sinha, S.K., Chahar, B.R., Dhanya,
C.T., Regional Scale groundwater Modelling Study for Ganga River Basin, Journal of Hydrology (2016), doi: http://
dx.doi.org/10.1016/j.jhydrol.2016.07.029

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Regional Scale groundwater Modelling Study for Ganga
River Basin
Maheswaran1,2#, R Khosa1,A K Gosain1, S Lahari1, S K Sinha3, B R Chahar1 C T Dhanya1,

1
Department of Civil Engineering

Indian Institute of Technology Delhi

New Delhi India

2
MVGR College of Engineering, Vizianagaram, India

3
Central Ground Water Board, New Delhi

#
Corresponding author: rmaheswaran@iitd.ac.in

Abstract

Subsurface movement of water within the alluvial formations of Ganga Basin Systemof
North and East India, extending over an area of 1 million km2,was simulated using
Visual MODFLOW based transient numerical model. The study incorporates historical
groundwater developments as recorded by various concerned agencies and also
accommodates the role of some of the major tributaries of River Ganga as geo-
hydrological boundaries. Geo-stratigraphic structures, along with corresponding
hydrological parameters,were obtained from Central Groundwater Board, India,and
used in the study which was carried out over a time horizon of 4.5 years. The model
parameters were fine tuned for calibration using Parameter Estimation (PEST)
simulations.
Analyses of the stream aquifer interaction using Zone Budget has allowed demarcation
of the losing and gaining stretches along the main stem of River Ganga as well as some
of its principal tributaries. From a management perspective,and entirely consistent with
general understanding, it is seen thatunabated long term groundwater extraction within
the study basin has induced a sharp decrease in critical dry weather base
flowcontributions. In view of a surge in demand for dry season irrigation water for
agriculture in the area, numerical models can be a useful tool to generate not only an
understanding of the underlying groundwater system but also facilitate development of
basin-wide detailed impact scenarios as inputs for management and policy action.
Keywords: Regional Groundwater Modelling, MODFLOW, Ganga River Basin.

1 INTRODUCTION

A river basin is an integrated system where interaction between surface water,

groundwater, water resources utilities and ecosystemare spatially widespread as well as
incessant and it is understood that in this dynamic complex, health of groundwater
systemshave an important bearing on water management plans in a given river basin.
Accordingly, in the latter context, regional or basin scale models have the potential to foster
betterunderstanding of the dynamics of the hydrological cycleand such an improved
understanding, in the process, translates into a more rational basis for water resources
management within the basin.Application of groundwater flow models to large scale aquifer
system simulation started in 1978 with the Regional Aquifer System Analysis (RASA)
program of U.S. Geological Survey (Sun and Johnson, 1994) in which typical regional
aquifer system models covered an area of the order of tens of thousands km 2. In the past,
regional models have also been used as decision support systems (Bauer et al., 2006; Carroll
et al., 2009; Refsgaard et al.,2010) as well as a tool to assess aquifer conditions under future
climate change scenarios (Scibek et al., 2007;Maxwell and Kollet, 2008). Software
engineering, computational speeds and memory infrastructure and wide use of Geographic
Information Systems (GIS) have been significant advances that has made use of 3D
groundwater flow models such as MODFLOW possible.On account of the latter’s flexible
modular structure, complete coverage of hydro-geological processes and its availability in the
public domain has made the platform a global industry standard for groundwater modelling.
Studies by Nishikawa (1998), Ting et al. (1998), Asghar et al. (2002), Faunt et al., (2004),
Rejani et al., (2007) , Umar (2008) and Wang (2009)aresome of the other successful
applications of MODFLOW for groundwater modelling studies.
Recently,Hou and Zhang, (2008) put together a modelling framework forthe Cretaceous
aquifer system in Ordos BasinwhileShao et al., 2009, developed a transient groundwater flow
model for the North China Plain having an area of 139,000 km2 with a uniform grid of 4 km
by 4 km. Wang et al., 2009, also developed a 3D transient groundwater flow model for the
Beijing P1ain in order to analyze groundwater flow systems within the integrated regional
water balance.
Ganga River basin isone of the largest alluvial systems in the world and over time, has
witnessed extreme demands on its water resources on account of its soaring population along
with its accompanying, and often competing, developmental aspirations. Additionally, Ganga
waters occupy an iconic status amongst the collective Indian psyche owing to its religious
and cultural importance and, understandably, environmental health of river Ganga has
emerged as a socio-political concern. Outfall of industrial wastes and sewage into the river,
over exploitation of its ground water resources, large scale dependence on irrigated
agriculture and possibly, as is widely perceived, imprudent water management policies have,
over the years, emerged as obdurate challenges.

This study was taken up as part of the Government of India initiative aimed at developing a
comprehensive basin wide Water Resources Management Plan for Ganga Basin and the
transient groundwater study was necessitated by the need to (i) study general ground water
flow patterns in the region, (ii) identify hot spots of groundwater depletion,(iii) understand
river- aquifer interaction to demarcate influent and effluent reaches and (iv)to investigate the
effects of further groundwater development. It merits mention that this study is the first such
attempt to model groundwater flow in the Ganga basin on such a scale ashas been carried out
and holds the promise of being a benchmark studyfor future studies.

2 Study Area

Ganga basinextends over an area of 1,086,000 km2and lies in Tibet, Nepal, Bangladesh and
India (at 8, 61,452 km2 constitutes nearly 26.2 percent of the total geographical area of India,
Figure 1). In India, Ganga basin also forms one of the largest multi-aquifer groundwater
reservoirs with depths of 2000 m and more. The river originates as Bhagirathi from the
Gangotri glaciers in the Himalayas at an elevation of about 7010m in Uttarkashi district of
Uttar Pradesh and flows for a total length of about 2,525 km up to its outfall into the Bay of
Bengal. Principal tributaries joining the river are the Yamuna, Ramganga, Ghaghra, Gandak,
Kosi, Mahananda and Sone.

2.1 Physiography

Ganga Basin is flanked by Himalayas in the north and north-west, Aravali Ranges and East
Rajasthan Uplands in west and south-west, Vindhyan Plateaux consisting Madhya Bharat
Pathar, Malwa Plateau, Baghelkhand Plateau and Chota Nagpur Plateau in south and alluvial
plains of Bangladesh in the East. Elevation levels range from 0 to 5 m amslat Sundarbans to
the high Himalayan peaks that include Mt. Everest, world’s tallest at 8848 m amsl.

Physiographic features of the western part of the basin falling in Rajasthan state can be
divided into three units: namelyAravalli hills, Eastern plains and theVidhyanScarpland and
Deccan Lava Plateau (CGWB,1996).

The elevation of Aravali Hills ranges from 600-900mamsl. The north eastern part of the hill
is composed of rocks ranging in age from Archaeani.e. 2500 million years to Proterozoic
i.e.740 million years old. The hill rangespresent an undulating topography with altitude of
plains east of Aravali ranging from 150m to 450m amsl. The central part of Ganga Basin in
the state of Uttar Pradesh can be divided into fourphysiographicunits namely Himalayas, sub-
Himalayas, alluvial plains and the southernplateaux.The Himalayan unit forms the northern
most part of the state covering the districts of Uttarkashi, Tehri, Pauri, Chamoli, Pithoragarh,
Almora and parts of Nainital and the zone is underlain by metamorphic and sedimentary
rocks and form hill ranges with high relief, deep gorges and narrow deep valleys.

The Sub-Himalayan unit forms the region between the aforementioned Himalayan unit in the
north and the alluvial plain to the south and comprises Doon Valley, river terraces and low
relief hilly tracts of theShivaliks.

The alluvial unit can further be sub-divided into the following zones:

i) Bhabhar: It is highly porous dry zone and forms the southern limit of the sub-
Himalayan unit. Its width varies from10- 30 km along the foothills of Himalayas
and extends from Uttar Pradesh and through to West Bengal.
ii) Tarai: This Zone lies between Bhabhar in the north and central Ganga plain in the
south and has a variable width of 8-16kms and is characterised by change in
surface slope.
iii) Central alluvial tract: The Vast alluvial tract south of Tarai belt and extending
upto Yamuna river and Chhota Nagpur plateau, Palamu plateau in the east, covers
the largest, about 40-50% part of the basin and is highly cultivated throughout the
area.
iv) Marginal alluvial zone: It occupies the southern fringe area of the Ganga plains
lying south of Yamuna and close to the plateau region and slopes from south to
north towards Ganga River.
 v) Southern Plateaux:This unit occupies the extreme southern fringe of the basin
and is characterized by the generally plain tableland of Vindhyas and residual
conical hills in Bundelkhand region. The entire plateau region forms the cratonic
part of Ganga basin.

2.2Climate and Rainfall

The normal minimum temperature for the month of January ranges from 2°C in the north to
7.8°C in the south west though at altitude in Himalayas, the temperature dips to below
freezing during the winter months.

The water supply depends mainly on the rains brought by SW monsoons from July to
Septemberas well as the snowmelt beginning from March and lasting uptoJune and beyond.
The average annual rainfall varies from 30 inches (760 mm) at the western end of the basin to
more than 90 inches (2,290 mm) at the eastern end. The delta region at the western end of the
basin also experiences strong cyclonic storms both before the commencement of the monsoon
season, from March to May, and at the end of it, from September to October. Understandably,
rainfall is the main source of ground water recharge with nearly 90% of the annual totals
being contributed by SW monsoon from June to September.

2.3 Drainage

Drainage in Ganga basin is governed mainly by rainfall, physiography and lithology.

Drainage in mountains and hills are mainly dendritic and structurally controlled. On hill
slopes drainage is mostly characterized by parallel to sub-parallel drainage, which persists in
the Central Ganga plain and continues upto Bay of Bengal (CGWB,1996).

The Himalayan rivers are mainly rain-fed and snow-fed while rivers originating from
Aravalli and Vindhyan mountains mainly rain-fed and ground water fed. The major drainage
network of Ganga Basin is given in Figure 2.

The main perennial drainage lines in the Basin are: the Ganga, Yamuna, Ramganga, Gandak,
Kosi, Chambal, Sind, Betwa, Ken, Tons, and Son. In the northern and north-eastern parts of
Rajasthan, the Banganga, Banas, Sota, Sahibi and Kantli rivers are of inland nature. These
drainage lines become active only during the monsoon. The delta of the Ganga begins near
Gaur. The main branch of the river in the delta portion flows in the south-east direction and is
known as Padma. The Hooghly, an estuary of river Ganga is notorious for its sand banks.
2.4 Land-use and land cover

Agriculture is mainly restricted to the fertile plains, over 70 million acres in extent, of Ganga
that span from West Bengal in the east to western Uttar Pradesh and Haryana in the west
notwithstanding saline patches that are seen in parts of these latter states and Rajasthan(as
seen in Figure 3). Ganga basin is known for its extensive forest cover which is about 19% of
the total area of the basin in addition to supporting diverse land use practices such as urban
and arable developments, forestry andgrasslands, besides extensively distributed wasteland
patches, water bodies and marsh lands. Waste and grasslands are mainly restricted to the hilly
terrains of Aravalis, Vindhyan Plateau and in the plains of Rajasthan, Haryana, Uttar Pradesh
and Bihar while shrublands in plains of Uttar Pradesh are generally salt affected. The marshy
land and water bodies in the area are either natural depressions or manmade water bodies on
the rivers. The marshy and waterlogged areas have also developed along canals and in canal
command areas for which a dedicated live storage capacity of almost 38 billion cu.msupports
irrigation over a total of 23.41 million ha.

2.5 Hydrogeology

The hydro-geological conditions prevailing in the entire basin is highly diversified as shown
in Figure 4. It is possible to present the hydro-geological framework of the basin in three
broad categories as follows:

1. Area of Unconsolidated formations: The quaternary rocks comprising Recent

Alluvium, Older Alluvium and Costal Alluvium are by and large important
unconsolidated formations. These sediments are essentially composed of clays, silts,
sands, kankarand include some areas of Haryana and Rajasthan that carry Aeolian
cappings. Indo-Gangetic plains, Marusthali, Bengal basin and Foredeep regionfrom
western to eastern Himalayas are also occupied by these formations.
2. Area of Semi-consolidated formations: These belong to Palaeozoic-Mesozoic and
Cenozoic rocks extending from Corboniferous to Mid-Pliocene ages. They mainly
consist of shales, sandstone, and limestone and are also generally described as
Tertiaries and Mesozoics of UP and West Bengal. Further, the terrestrial fresh water
deposits belonging to Gondwana Super Groups of UP, MP, Bihar and West Bengal
are also included under this category.
3. Area of Consolidate formations: The consolidated formations which occupy almost
half of the basin have been classified into four broad lithological units(i) The
 Sedimentaries and meta-sedimentaries belong to Cainozoic, ii) the efflusives are
generally basaltic flows, chiefly represented by Dalma lava, iii) The intrusive
Cainozoic group and the iv) basal crystallines

2.6 Aquifer geometry

Fence diagram based on lithological logs (CGWB, 1996) reveals the vertical and lateral
disposition of aquifers, aquiclude and aquitard in the study area down to depth of 200 m bgl.
Supplement S1 shows the fence diagrams for Haryana, UP and Bihar & West
Bengalrespectively and were used for model conceptualization. While these figures suggest a
spatially variable lithology across the basin, it is also established that much of the basin of
interest is dominated by deep alluvial aquifers that extend to a depth of over 700m for much
of the middle basin. Also a reality is the presence of clay lenses, widely spattered and
extending up to 1-2m in some places. However, it is to be noted that the exact extent and
depth is highly uncertain and reportedly for North Bihar and West Bengal, the top clay layer
is persistent throughout the area varying in thickness from 3 to 4 m bgl.

The top clay bed is underlain by a granular zone that extends downward to depths varying up
to 400 m bgl and bears a texture that similarly varies from fine and medium to coarse sand
though occurrence of sub-regional clay beds and local clay lenses are also common
throughout the area. However, these features appear to merge with each other rendering the
aquifer to behave as a single consistent unit at least for depths up to 400 m.The inferred
aquifer geometry in Rajasthan and Madhya Pradesh suggests presence of water under
unconfined conditions and extending down to the basement rock. In contrast, presence of 4 to
5 aquifers, down to depths upto 750m, is suggested for Haryana, Uttar Pradesh, Bihar and
West Bengal and the conditions range from unconfined, semi-confined and, at places,
confined in nature.

Fissured formations, essentially composed of consolidated rock, are present in parts of the
Vindhyan range and the region in central Madhya Pradeshand occurrence and movement of
groundwater in them is essentially due to secondary porosity. It is therefore understood that
mechanics of groundwater movement in these secondary formations is distinct from that
which is observed in isotropic porous formations and cannot be modelled using approaches
developed for the latter medium.
3 GROUND WATER MODELDEVELOPMENT

Groundwater models are mathematical and digital tools of analyzing and predicting the
behaviour of aquifer systems on local and regional scale, under varying geological
environments (Balasubramanian, 2001) and help in establishing locations and characteristics
of aquifer boundaries besides facilitating assessment of the quantity of water available within
the system and the amount of recharge to the aquifer (Anderson and Woessner, 2002).

The general flow equation for unsteady flow in an unconfined aquifer underDupuit
assumptions (1) flow lines are horizontal and equipotential lines are vertical and (2) the
horizontal hydraulic gradient is equal to the slope of the free surface and is invariant with
depth) is given by Eq.(1).

 2h  2h  2h h
Kx  K  K  Sy R (1)
x y z t
2 y 2 z 2

Where Kx, Ky, and Kzare components of the hydraulic conductivity tensor. S yis the Specific
yield and R is general sink/source term that is intrinsically positive and defines the volume of
inflow to the system per unit volume of aquifer per unit of time.In this study, the Visual
MODFLOW with MODFLOW 2005 as the engine for simulating the groundwater dynamics
has been used. The 2005 version of MODFLOW family includes the simulation of saturated-
unsaturated flow process, groundwater simulation-optimization process, irrigation process,
density dependent flow process, parameter optimization process and solute transport process
(Zhou and Li, 2011).

3.1 Model conceptualization and data acquisition

The purpose of building a conceptual model is to simplify the field problem and
organize the associated field data so that the system can be analyzed more readily (Anderson
and Woessner, 2002). The conceptualization includes synthesis and framing up of data
pertaining to geology, hydrogeology, hydrology, and meteorology. The model area and the
boundary conditions used in the present study are marked in Figure 4. It can be seen that
there isa no flow boundary along the north-western side of the Ganga basin. This is a
reasonable assumption as per CGWB (1996), there is a groundwater divide between the Indus
and Ganga Basin. The region adjoining the Bay of Bengal was taken as the specified head
with the head values being the tidal heights. Similarly, the northern and southern partsof the
basin is also designated as a specified flux boundary. Additionally, the mountainous region
was excluded from the study considering that these natural outcrops would act as a barrier to
groundwater movement and present itself as a natural boundary of the aquifer system.
However, these portions will contribute to the groundwater in the central alluvium part and
the recharge from this portion would have to be prescribed as a boundary condition.
Accordingly, the recharge from this area was indirectly derived as an output from theSWAT
(Soil and Water Assessment Tool, Arnold et al., 1998)hydrological modelthat was setup for
the entire Ganga system and incorporated in the groundwater model as a specified flux
boundary. It was also observed that the hard rock regions in the South had insignificant
overburden thicknesses and had secondary porous structure and, accordingly, the contribution
from this region was incorporated in the groundwater model as a specified flux.

3.2 Depiction of aquifer geometry

Geologic details of aquifers were combined with information on hydro geologic properties to
define hydrostratigraphic units for the conceptual model. It was shown throughfence
diagrams in Section 2 that the aquifer depth extends up to 750m below ground level in most
places of Ganga Basin, presence of occasional clay lenses notwithstanding. Noting that the
model setup was on a regional scale, the entire aquifer stratigraphy was combined together to
form a single layer unconfined aquifer of thickness 150 – 200m. Figure 5(a) shows the
snapshot from MODFLOW depicting the modelled area with the hypsometric colours
showing the water levels. The various boundary conditions used are also presented along the
boundary of the modelled region. Further, Figure 5b and 5c shows the topography and the
depth of aquifer across two different cross sections of the basin along with the water levels
and the pumps present in the region. The top profile shows the topography of the soil surface
as obtained from the DEM and the bottom profile shows the bottom of the aquifer surface.
The vertical dark blue lines in between the layer shows the positions of the rivers along the
cut sections AA and BB.

Aquifer Parameters

The entire (truncated in the manner as discussed above) study area was divided into 8
homogeneous and isotropic zones and hydraulic conductivity and specific yieldvalues were
assigned to each based on data and logs obtained from CGWB. The hydraulic conductivity
values assigned to the model ranged between 11.0 to 25.6 m/d and the particular values
assigned to each zone are as shown in Table 1.Similarly,region wisespecific yield values
(according to hydro geologic setting), ranging from 0.02 to 0.16, were assigned based CGWB
recommendations and are also depicted in Figure 7.

3.3 Recharge

Estimates of recharge due to external water applications, such as rainfall, irrigation and canal
seepage, were obtained from SWAToutputs and a synoptic presentation of the performance
of this model is shown in Figure 8. Similarly, recharge fromdesignated hard rock regions was
also prescribed to the GW model as a specified flux boundary in order to capture the lateral
contributions from these regions to the alluvium system. The monthly accumulated recharge
for the different regions of the model area from the SWAT model is provided as one of
inputsto the MODFLOW model.

However, it is to be noted that the time step for MODFLOW run is set to be at 10-daily scale.
In order to make the data interval of input recharge compatible with model time step, the
recharge value of a given month was disaggregated uniformly over 30days to yield the 10-
daily values of recharge using the inbuilt algorithm in MODFLOW.

3.4 Groundwater extraction by pumping

As per available records based on Minor Irrigation census, a total of pumping wells in excess
of 10,000,000 are indicated and a district wise extraction rate of 1500-2200 m3/dhas been
extrapolated by CGWB (CGWB, 2004 and 2009). Using these enormous number of pumps
would certainly increase the complexity of the model. Therefore, we tried to vary the
pumping rate from 2000m3/day to 80000m3/day in such a way there is a trade-off between
model complexity and reality situation. Based on a sensitivity analysis it was observed that
for a more manageable count of the number of wells, aggregation to yield 150,000extraction
points and corresponding to an extraction rate ranging from 40000-60000m3/d was used
purpose of modelling. This is reasonable considering the spatial extent of the model solution
domainand the focus of the study essentially being the long term dynamics of the system. As
an alternative approach, simulations were attempted by prescribing net recharge rates
(recharge rates– pumping rates) for each cell but a preponderance of negative net recharge
rates often led to serious computational difficulty.
3.5 Boundary conditions

3.5.1 River Boundary

These boundary conditions(in terms of observed water levels and riverbed elevations)were
applied along the main stem of river Ganga as well as along the main tributaries (see Figure
9). The data was obtained from Central Water Commission for the period starting from June
2000.River bed conductance values ranging from 30 to 150 m2/d were adopted for the
different rivers.

3.5.2 General Head boundaries (GHB)

GHB were assigned at the eastern fringes that are in close proximity to the sea based on
observed tidal levels.

3.5.3 Specified flux boundaries

As already reported, the model solution domain has been restricted to the alluvial system of
the Ganga plains and would receive contributions from beyond this zone but do come under
the purview of modelling focus. For emphasis, it may be noted that these excluded areas
include, as stated earlier, the high ranges in the north as well as the hard rock areas. In order
to accommodate these contributions, appropriate fluxboundaries have been prescribed as
shown in Figure 4. The specified flux from the contributing regions were estimated from the
SWAT model as groundwater recharge.

3.6 Conceptualization of flow regime and model setup

Choice of spatial grid size, time steps, prescription of boundary and initial conditions across
the model solution domain, model parameter initialization and other assumptions are some of
the important considerations for model setup design. The details, pertaining to model
formulation, are as presented below:

1. The entire aquifer system is homogeneous and isotropic and without discontinuities.
2. The single layer, model solution domain comprises of a 2500m x 2500m, uniformly
discretized, cells distributed over 500 rows and 500 columns.
3. The entire model solution domain has been sub-divided into eight zones which were
individually assigned hydraulic conductivity values ranging from 11 m/d to 26.6 m/d.
4. The study assumes that the aquifer has a uniform thickness of 200m.
 5. Natural recharge from monsoon rainfall, recharge from irrigation application,
contributions from hard rock areas and contributions from the high altitude ranges
constitute the main inputs to the model.
6. Pumping rates vary from 40000-60000m3/d.
7. River boundary conditions, in the form of observed water levels, were applied along
the main stem of river Ganga as well as the main tributaries.

4 RESULTS

4.1 Calibration of the model- Transient State

In view of widespread mining of groundwater resources, steady state description of transport

phenomenon in aquifers is indeed unrealistic. In this study, therefore, transient state
calibration of the aquifer system was carried out for a four yearperiod from 2000-2004 using
(i)a time step of 10days, (ii) 30 day length of each stress period, and (iii) recharge flux and
extraction prescribed at intervals of 30days with the provision for estimating intermediate
values by interpolation.Calibration was initiated with initial head values as recorded for
August 2000 and similarly repeated with other years.It was observed that model simulations
rendered similar water levels for all years, suggesting that the model calibration was weakly
sensitive to initial conditions. To establish a benchmark for comparison, water levels in a
total of 100 CGWB observation wells for the months of January, May, August and November
were used for calibration and validation. The model simulated water levels compared well
with the corresponding observations and Figure 9shows the location of these wells in
MODFLOW screenshot.

4.2 Parameter Estimation (PEST)run

In order to obtain improved match between simulated water levels and their
corresponding observations and suggestive of an apparently improved parameter estimation,
the groundwater model setup was auto calibrated using the PEST module of MODFLOW.
Hydraulic conductivities of the entire sub basin were made as target decision variables while
theobjective was to minimize the root mean square error between the observed and simulated
water levels. The calibrated conductivity and the initial conductivity values are shown in
Table 1 and these yielded simulations that were comparatively better in comparison to the
manual calibration simulations as seen from the plots shown in Figure 10. Figure 10ashows
the scatter plot between the observed and the simulated water levels (w.r.t the msl datum) at
all the observation wells corresponding to simulation run of 723 days, or approximately at the
end of two years. It can be seen that simulated water levels during June 2002match
reasonably well with the observed water levels during that period. The model results
inreference to the local ground water level is shown in Figure 10b and the scatter plot shows
that barring a few exceptions which report a relatively high RMSE equal to 5.23m, most of
the points lie around the 45Oline.
For a better insight, it is instructive to examine the model results pertaining to three
observation wells as depicted in Figures 11which shows time series plots of simulated and
observed water levels for these wells. It can be seen that the model is able to closely capture
the real observation and the trend in the observed water levels.Further, Figure 12 shows the
scatter plot of the observed and the simulated water levels for all observation wells at
different time periods and these results also underscore the point that the proposed model
setup can simulate the geo-hydrological reality of the Ganga basin with a reasonable level of
accuracy. Similar conclusions are indeed in order as RMSE and other error statistics are
examined for different periods andshown tabulated in Table 2.
The surface plot of the depth of water levels for both pre and post monsoon is shown in
Figure 13 and it is seen that with the exception of a few green patches, for much of the
region, the water table is in between 11-20m which is a greater concern. Also, the red patches
in the figure indicate the hard rock terrain and therefore these areas can be ignored from the
perspective of water management.

Figure 14 shows the primary directions of the water flow as obtained from the model time
t=730 days from May 2000.The majority of the flow direction is according to the laws of
groundwater physics and was verified using the water level contour map produced by
CGWB. Thus, the developed model is able to mimic the reality to a satisfactory level. It can
be seen that at majority of the locations the groundwater is contributing to the river in the
form of base flow which is true in reality. It is also observed that at head waters of the rivers,
at some locations, the river is recharging the aquifers.
4.3 Mass Balance

Figure 15(a) and (b)shows the derived water balance of study region end of 365 days and
1560 days from the start of the simulation, respectively. The blue colour bars indicate the
mass that goes into the system and the red indicates abstractions out of it. It can be observed
that the overall mass balance has been established is indeed in order. Further, from Figure
15(a) it is also observed that during the given period of analysis, the groundwater abstraction
by pumping aquifer storages is disproportionately higher - five times higher - than the natural
recharge from rainfall. The overall mass can be interpreted using the following equation,
(Storage)in + (General Head)in + (River Leakage)in + (Recharge)in = (Storage)out +(General Head)out+ +
(Wells)out + (River Leakage)out

(2)

where,

Storage is the amount of water going into ground water storage;

General Head is the boundary condition for such places where required;

River Leakage is the leakage from and to the groundwater that occurs across river boundaries
as modelled;

Recharge is the boundary condition to account for zone wise recharge of the groundwater
system.

Figure 15(b) shows the total water balance at the end of simulation (after 4 years). It can be
seen that the groundwater extraction in terms of pumping is much greater when compared to
the available storage and recharge which leads to the over-exploitation of aquifers. Overall it
can be seen that the storage in the aquifers is depleting over time due to the over exploitation
by excessive pumping.

In order to understand the river aquifer interaction, the exchange between the river and the
adjoining aquifer cells were estimated separately and was analysed. In order to compare the
present status of the river-aquifer interaction with virgin (pristine conditions) a separate
model scenario was simulated without any pumping. Figure 16shows the pictorial
representation of the flux between the aquifer and the river for different stretches along the
River Ganga. The results are presented as long-term pre-monsoon and post-monsoon
averages. It can be seen that during the virgin conditions where there is no pumping, all the
stretches are gaining stretches(effluent) except for the one between Chapra and Varanasi.
This might be attributed to the topographical nature of this stretch and the adjoining areas.

It can be observed that the upper reaches of the Ganga and Yamuna are becoming loosing
streams owing to the over-exploitation of adjacent groundwater aquifers. Similarly, the
middle stretches of Yamuna and Ganga are gaining much lesser water in comparison with the
virgin conditions. Therefore, with the over exploitation of the groundwater under the present
conditions, most of the river stretches have become either loosing or are gaining much lesser
amount of water from the groundwater. For example, some of the stretches under the virgin
condition that were blue have become orange or red indicating the over exploitation of
aquifer in these areas. It has also been noted that there is a seasonal behavior in the influent /
effluent nature of the river stretches. Since, the range of the flux is high, this seasonal
difference is not being captured in the figures.

5 DISCUSSIONS

Analysis of derivedmodel simulations leads to the reasoned inference that the developed
regional groundwater model is able to simulate the actual groundwater reality reasonably
satisfactorily and, therefore, is suggestive of the assurance that the model conceptualization,
as hypothesized in the study, together with all the accompanying assumptions about boundary
fluxes and other conditions as prescribed in the model setup is a reasonably accurate
depiction of the physical system. Further, it is also observed that the model is able to capture
satisfactorily both the pattern of observed variability in groundwater levels as well as the
overall flow pathways along which groundwater movement may be taking place. It would,
therefore, be reasonable to aver that the groundwater model as proposed in the study is able
to mimic the abstract reality of the physical system and may be used with confidence to put
together a more rational framework for policies on groundwater management within the study
basin.
The modelling based study of Ganga basin, as described in the paper was further based on
two profound assumptions. With regard to the first, it is seen that the approximations made in
terms of the pumping wells does not seem to have compromised the model’s capability to
capture the overall extraction rates and the second assumption also appears to be reasonable
that each zone wise groundwater system was homogeneous and isotropic and individually,
the groundwater storage system functions as a single, undivided entity without any embedded
hydraulic or physical discontinuity.

From even a cursory analysis of the modelling results, the stark reality that emerges is
suggestive of the widely held perception that groundwater resources in Ganga basin are
depleting and much of the recorded abstraction can safely be categorized as groundwater
mining as the rate of these abstractions by groundwater pumping is significantly higher than
rate at which these alluvial aquifersare actually recharging.
While overall, the results appear to be reasonably satisfactory, however, a mismatch between
model simulations and actual observations on groundwater levels at various well locations is
also observed. This may be attributed to the following likely factors:
1. Error in DEM gets propagated into the water levels and thereby creating errors in the
water level values.
2. The errors may also be due to the assumption ofhomogeneityand isotropy of the
groundwater transport medium. Therefore, refinement of model parameters based on a
more rigorous field study is likely to improve the model performance.
3. In the study, flow domain is assumed to be a single homogeneous aquiferand the
reality of transport mechanics in stratified multilayered systems has been ignored. The
proposed groundwater model can be improved once the CGWB initiative on detailed,
nationwide aquifer mapping concludes.
4. As a further improvement, uncertainty analysis can be made part of any future studies.

6 CONCLUSIONS AND SUGGESTIONS

In this study, a three-dimensional groundwater flow model of Ganga basin was developed.
Based on the simulation results, following conclusions are in order:
1. The MODFLOW model was successfully run and calibrated for 4 years in transient
mode. The model results were found to be satisfactory and were seen to matchthe
observed water levels to a reasonable level of accuracy.
2. The model results show that the assumptions regarding boundary conditions and
aquifer geometry are reasonableand therefore the can be used for future scenario
generation and as aid for developing groundwater managementpolicies.
3. It was observed that with the over exploitation of the groundwater under the present
conditions, most of the river stretches have become either loosing or are gaining much
lesser amount of water from the groundwater
4. As a further scope for this study, the hydraulic conductivities of the river bed and the
aquifer can be more spatially refined. Also, sensitivity analysis can be performed to
identify the key parameters.In this regard, an elaborate discussion on the river-aquifer
interaction is provided in the second part of this study.

Acknowledgement
The authors acknowledge the Ministry of Environment and Forest for providing the funding
for work through the Ganga River Basin Management Plan project. The authors also
acknowledge the contribution from Dr.Sharma, Dr.Sanjay, Dr.Sandhya Rao, INRM and other
faculty involved in the GRBMP project through scientific discussion and expert advises. We
are also grateful for the Associate Editor Dr.Han and other two anonymous reviewers.

REFERENCES

Anderson MP, Woessner WW 1992. Applied groundwater modeling: simulation of flow and
advective transport. Academic, California, p 381

Arnold J.G, R. Srinivasan, R. S. MuttiahandJ. R. Williams, 1998. Large area hydrologic modeling and
assessment part I: model development.JAWRA Journal of the American Water Resources
AssociationVolume 34, Issue 1, pages 73–89.

Asghar MN, Prathapar SA, Shafique MS 2002.Extracting relatively fresh groundwater from
aquifersunderlain by salty groundwater. Agric Water Manag, 52:119–13

Bauer P, Held RJ, Zimmermann S, Linn F, Kinzelbach W 2005. Coupled flow and salinity transport
modelling in semi-arid environments: the Shashe River Valley, Botswana. J Hydrol. 316(1–4):163–
183.

CGWB 1999. Groundwater Resources and Developmental Potentials of Balasore District, Orissa.
Technical Report, Series ‘D’, Central Ground Water Board (CGWB), Bhubaneswar, Orissa, India, p
50

Faunt, C.C., D’Agnese, F.A., O’Brien, G.M., 2004. Hydrology, Chapter D of Death Valley Regional
Groundwater Flow System, Nevada and California- Hydrogeological Framework and Transient
Groundwater Flow Model. U.S. Geological Survey. Scientific Investigation Report 2004-5205.

Hou, G.C., Zhang, M.S., 2008. Exploration and Study of Groundwater in the Ordos Basin, 563 pp.
Geological Publishing House, Beijing (in Chinese).

Nishikawa T, 1998. Water resources optimization model for Santa Barbara, California. J Water
Resource Plan Manage ASCE, 124(5):252–263

Refsgaard, J.C., van der Sluijs, J.P., Brown, J., van der Keur, P., 2006. A framework for dealing with
uncertainty due to model structure error. Advances in Water Resources, 29, 1586-1597

Reed M. Maxwell& Stefan J. Kollet, 2008.Interdependence of groundwater dynamics and land-energy

feedbacks under climate change. Nature Geoscience, 1, 665 - 669.

Rejani R, Jha MK, Panda SN, Mull R, 2003.Hydrologic and hydrogeologic analyses in a coastal
groundwater Basin, Orissa, India. ApplEngAgric, 19(2):177–186
Scibek J, Diana M. Allen, Alex J. Cannon, Paul H. Whitfield, 2007. Groundwater–surface water
interaction under scenarios of climate change using a high-resolution transient groundwater model,
Journal of Hydrology, Volume 333, Issues 2–4, Pages 165–181.

Shao, J.L., Zhao, Z.Z., Cui, Y.L., Wang, R., Li, C.Q., Yang, Q.Q., 2009. Application of groundwater
modeling system to the evaluation of groundwater resources in North China plain. Resources Science
313, 361e367 (in Chinese).

Sun, R.J., Johnson, R.H., 1994. Regional Aquifer System Analysis Program of the U.S. Geological
Survey, 1978e1992. U.S. Geological Survey. Circular 1099.

Sun, R.J., Weeks, J.B., Grubb, H.F., 1997. Bibliography of Regional Aquifer System Analysis
Program of the U.S. Geological Survey, 1978e1996. U.S. Geological Survey Water Resources,
Austin, Texas. Investigation Report 97-4074

Ting CS, Zhou Y, Vries JJ de., Simmers, 1998. Development of a preliminary groundwater flow
model for water resources management in the Pingtung Plain, Taiwan. Ground Water, 35(6):20–35.

Umar R, Izrar Ahmed, FakhreAlam, Mohammad Muqtada Khan, 2008. Hydrochemical characteristics
and seasonal variations in groundwater quality of an alluvial aquifer in parts of Central Ganga Plain,
Western Uttar Pradesh, India. Environmental Geology, Volume 58, Issue 6, pp 1295-1300

Wang, L.Y., Han, J.P., Liu, J.R., Ye, C., Zheng, Y.J., Wan, L.Q., Li, W.P., Zhou, Y.X., 2009.
Modeling of regional groundwater flow in Beijing plain. Journal of Hydrogeology and Engineering
Geology 36 (1), 11-19 (in Chinese with English abstract).

YangxiaoZhoua, Wenpeng Lib, 2011. A review of regional groundwater flow modelling. Geoscience
Frontiers,Volume 2, Issue 2, Pages 205–214.
Table 1 Initial Conductivity and PEST Calibrated Conductivity values in the x, y and z
directions.
Initial Conductivity PEST Calibrated Conductivity
Conductivity Kx[m/d] Ky[m/d] Kz[m/d] Kx[m/d] Ky[m/d] Kz[m/d]
Zone
1 15 15 0.15 15 15 0.71
2 16. 16 0.1 8.38 8.38 0.42
3 15 15 0.2 21.67 21.67 0.53
4 20 20 0.25 23.39 23.39 0.91
5 10 10 0.1 10 10 0.35
6 23 23 0.14 14.835 14.835 0.52
7 22 22 0.19 23.86 23.86 0.60
8 11 11 0.20 18.47 18.47 0.50
 Table 2 Statistics on growth of errors at various stages of model runtime

Length of RMSE Correlation Max Residual Min Residual Standard

model runtime (m) Error(m)
(days)
730 5.84 0.99 -14.5 0.199 1.5
1100 4.84 0.99 +12.5 0.25 1.2
1300 4.626 0.99 6.69 0.72 1.1
1560 6.50 0.98 12.5 1.002 1.5
Highlights

1. Regional scale groundwater model for Ganga river basin.

2. Helps in management of groundwater and surface water resources.

3. Provides a first step in understanding river -aquifer interaction.