1 Surface and groundwater dynamic interactions in the Upper Great Chao Phraya Plain of Thailand: semi-coupling of SWAT and MODFLOW [1] Werapol BEJRANONDA, [1] Sucharit KOONTANAKULVONG, [2] Manfred KOCH [1] Department of Water Resources Engineering, Faculty of Engineering, Chulalongkorn University, Thailand, [email protected][2] Department of Geohydraulics and Engineering Hydrology, University of Kassel, Germany [email protected]Abstract Surface and groundwater dynamic interaction in the Upper Great Chaophraya Plain of Thailand is explored. The surface-soil water model SWAT and the groundwater model MODFLOW are semi-coupled to determine the various hydrological components and their relevant flow behaviour. The coupled models are executed by running the simulations individually using the monthly river-groundwater interactions and the groundwater recharge to generate the surface-groundwater dynamic interactions. The calibrated results from coupled simulations show that the coupling-method improves both the streamflow and the groundwater simulations. From the water balance analysis a quantitative picture of the local surface and subsurface water interaction is obtained. The semi-coupled models result in streamflow- and groundwater level estimates that are improved by 12% and 2.3%, respectively, especially in the dry season. The dynamic simulations reveal that the interaction is seasonally dependent The improvement of the water balance analysis suggest that coupled modelling is a key to clarify the relationship between surface- and subsurface water, as well as their interaction. Keywords: Surface-groundwater interactions, semi-coupled model, SWAT, MODFLOW, Thailand 1. Introduction In spite of the tremendous steps made in recent years towards becoming an industrialised country, Thailand economically still defines itself as an agricultural country, owing to the fact that the export of agricultural products is still bringing in a large portion of its national revenue. Boosting up rice production and the, often precarious, living conditions of the rice farmer has thus been an active policy of the Thai government in recent years and has led it to develop agricultural price-subsidized schemes and many irrigation projects in the Great Chao Phraya Plain to support the local farmers (Koontanakulvong et al., 2005). At the same time both surface water and groundwater resources have been developed to respond to the increased water consumption. Since it is not possible to provide sufficient surface water for irrigation, conjunctive water uses are to be explored as part of a major national effort to optimize the water resources for the various stakeholders involved. Providing irrigation water through irrigation canals sipping adjacent streams has been the classical approach method in so-called irrigation projects in Thailand to augment the agriculturally available water there during a dry season or drought periods, the latter becoming more frequent in recent years, due to climate change. As the canal irrigation water has to be increasingly complemented by pumped groundwater, complex surface-groundwater interactions occur which need to be understood for the setup of a comprehensive water balance analysis. This is the objective of the present conjunctive water use study wherein a new semi-coupled surface- groundwater model will be employed. Historically the coupling of surface- and subsurface flow models started with the setup of relationships between river stages and groundwater storage (Pinder and Sauer, 1971). One widely used computational model to solve for the complex flow between surface and subsurface has been the stream-aquifer interaction program MODBRANCH (Swain, 1996), which couples the groundwater flow model MODFLOW (McDonald and Harbaugh, 1998) and the river network program BRANCH. Orhan and Mustafa (2004) coupled a one-dimensional channel network flow model that uses the St. Venant equation with a 2D groundwater flow model. Smits and Hemker (2004) modeled the interaction of surface-water and groundwater flow by linking Duflow to Microflow. Ellingson and Schwartzman. (2004) integrated both an unsaturated zone flow model and a groundwater model into the regional HSPF Model. Coupling of the surface-water model SWAT (Arnold and Fohrer, 2005;
11
Embed
Surface and groundwater dynamic interactions in the Upper ......Study and design: The watershed model SWAT-2005 and the groundwater flow model MODFLOW-2003 are selected to simulate
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Surface and groundwater dynamic interactions in the Upper Great Chao
Phraya Plain of Thailand: semi-coupling of SWAT and MODFLOW
1. Introduction In spite of the tremendous steps made in recent years towards becoming an industrialised country,
Thailand economically still defines itself as an agricultural country, owing to the fact that the export
of agricultural products is still bringing in a large portion of its national revenue. Boosting up rice
production and the, often precarious, living conditions of the rice farmer has thus been an active policy
of the Thai government in recent years and has led it to develop agricultural price-subsidized schemes
and many irrigation projects in the Great Chao Phraya Plain to support the local farmers
(Koontanakulvong et al., 2005). At the same time both surface water and groundwater resources have
been developed to respond to the increased water consumption. Since it is not possible to provide
sufficient surface water for irrigation, conjunctive water uses are to be explored as part of a major
national effort to optimize the water resources for the various stakeholders involved.
Providing irrigation water through irrigation canals sipping adjacent streams has been the classical
approach method in so-called irrigation projects in Thailand to augment the agriculturally available
water there during a dry season or drought periods, the latter becoming more frequent in recent years,
due to climate change. As the canal irrigation water has to be increasingly complemented by pumped
groundwater, complex surface-groundwater interactions occur which need to be understood for the
setup of a comprehensive water balance analysis. This is the objective of the present conjunctive water
use study wherein a new semi-coupled surface- groundwater model will be employed.
Historically the coupling of surface- and subsurface flow models started with the setup of relationships
between river stages and groundwater storage (Pinder and Sauer, 1971). One widely used
computational model to solve for the complex flow between surface and subsurface has been the
stream-aquifer interaction program MODBRANCH (Swain, 1996), which couples the groundwater
flow model MODFLOW (McDonald and Harbaugh, 1998) and the river network program BRANCH.
Orhan and Mustafa (2004) coupled a one-dimensional channel network flow model that uses the St.
Venant equation with a 2D groundwater flow model. Smits and Hemker (2004) modeled the
interaction of surface-water and groundwater flow by linking Duflow to Microflow. Ellingson and
Schwartzman. (2004) integrated both an unsaturated zone flow model and a groundwater model into
the regional HSPF Model. Coupling of the surface-water model SWAT (Arnold and Fohrer, 2005;
2
Arnold et al., 1998) with the groundwater flow model MODFLOW was endeavoured by Il-Moon
Chung (2006). However the full coupling of surface- and subsurface water model is still complicated
for practical applications. Semi-coupling appears to be a more viable approach for practical
simulation. Here the surface-water model SWAT and the groundwater flow model MODFLOW are
semi-coupled. The dynamic interaction is achieved by running the two models individually, though
with an intermediate interchange of information between the surface- and subsurface compartments
using hydraulic functions. This semi-coupled model is applied to the Upper Great Chao Phraya Basin.
2. Study area
The Upper Great Chao Phraya Plain of Thailand (Fig.1) covers an area of ~38,000 km2, extending
over 8 provinces, with a population of 4 million people. The main land-use is 63% agricultural, out of
which 21% is irrigated, and 24% forest. More than 90,000 groundwater wells exist in the region
(Faculty of Engineering, 2007). The main groundwater basin is dissected by five major rivers that flow
from north to south and which, geologically, form a depositional flood plain. The basin is surrounded
in the east and west by mountains of volcanic rocks. The average elevation of the basin is 40-60
m.MSL. The basin drains into the lower basin in the south, though the free discharge is partially
obstructed by crystalline rocks there. The 900 - 1,450 mm annual rainfall within the study region is
apportioned to 81 % in the wet (Apr.-Sep.) and to 19 % in the dry season (Oct.-Mar.).
3. Theoretical basis of the coupled surface- and groundwater flow model
3.1 Coupling-process development The development of the new coupled model can be divided into three major steps: 1) study and design,
2) coupling and testing and, 3) application & evaluation (see Fig. 2).
Study and design: The watershed model SWAT-2005 and the groundwater flow model MODFLOW-
2003 are selected to simulate the surface water- and groundwater balances. The original versions of
SWAT and MODFLOW are set up for a simple test case and used to explore the relationships between
surface- and subsurface water. The study of the coupling process is designed to cover three major
subjects, namely, coupling components, areal coupling, and temporal coupling,.
Coupling and testing: The designed coupling processes are applied to a basic case to test the coupling
techniques by comparing their output with results of a simple analytical CN rainfall-runoff equation.
Once the coupled simulations match the analytical results and the theoretical hydrological processes
involved, the coupled model will be applied to the real field data.
Application and evaluation: The original SWAT and MODFLOW models are applied to the Upper
Great Chao Phraya Plain where they are also calibrated and verified. The theoretical coupling-process
is also included in the modeling procedure. Through comparison with the observed data, the
uncoupled and coupled simulations are assessed for the ongoing coupling-process development. The
coupling-process is then evaluated by measuring the quantitative improvements in the model accuracy,
defined as the difference of the calibration errors between the uncoupled and coupled models.
3.2 Coupling methodology
3.2.1 General aspects The interactions between the surface- and groundwater-modeled groundwater recharge and the river-
aquifer interactions are used to couple the models. The percolation, i.e. the infiltration of soil water to
the shallow aquifers, as included in SWAT, and the river-groundwater interaction, i.e. the exchange of
water between the river body and the groundwater as included in MODFLOW, are selected for the use
in the coupling process and for the interaction of the components. While the model is operating in the
“coupling process”-mode, SWAT and MODFLOW are connected through some selected components
that exhibit substituted duplicated functions. For example, in the coupled model the groundwater
recharge is replaced by the percolation, as computed by SWAT, and the baseflow is substituted by the
river-groundwater interaction, as computed by MODFLOW (cf. Figs. 3 to 5). The coupled model is
executed by a VISUAL BASIC interface program which has been written to operate and transmit each
of the two model’s information to the coupling process.
3
Fig. 1. Study area showing river networks and alluvial aquifers.
Fig. 2. Study scheme of the coupled model.
3.2.2 Areal coupling The space dimensions of both the surface- and the subsurface water model are calculated in different
ways. SWAT, a distributed hydrological model, is set up with polygon-basins and a river-network,
whereas the FD-model MODFLOW is constructed using grid-cells and river-nodes. These differences
in the spatial distributions of the two models require the coupling process to be accounted for by
interactions of (1) the land- and (2) the river phase. The land phase describes the one-way! recharge
from SWAT polygons to MODFLOW grid-cells, and the river phase the two-way! interaction of the
SWAT stream polyline with the MODFLOW river nodes (see Fig.4). The river recharge/baseflow in
the various sub-basins is calculated with the river-groundwater interaction function as programmed in
MODFLOW’s RIVER-modul. In order to define the interaction for the entire study area, the surface-
water model boundary is extended sufficiently outwards so that the subsurface boundary is included.
GIS assists in the coupling process by joining SWAT sub-basin-polygons to MODFLOW grid-cells.
Southeast Asia
Thailand
Ping river
Yom river Nan river
Chao Phraya river
Sakae Krang river
study area
alluvial
aquifers
Y.14
N.12A
P.7A
C.2
4
SURFACE WATER SYSTEM
GROUNDWATER SYSTEM
SURFACE AQUIFER
PRECIPITATION
1283
DIRECT
RUNOFF
342
E/T
768
SOIL PROFILE
187
PERCOLATION1
74(28.9*)
RIVER-GW
INTERACTION
5.4
LEAKAGE
0.14
PUMPAGE
12.4
LATERAL FLOW
0.02
STREAMFLOW
(OUT)
444(#42)
GROUNDWATER SYSTEM
LOWER AQUIFER
STREAMFLOW (IN)
403
WATER USE
63
STREAM
-FLOW
LATERAL
FLOW
96
SOIL
STORAGE
16.6
AQUIFER STORAGE
13.0
PUMPAGE
2.4
AQUIFER STORAGE
11.7
RIVER-GW
INTERACTION
0.55
LATERAL FLOW
0.77
PERCOLATION2
(15.8*)
SURFACE
ABSTRACTION
512
**Coupled
component Unit : mm. – depth of water
Fig. 3. Surface water and groundwater hydrological components with water balance analysis results.
Fig. 4. Conceptual coupling of the surface- and the groundwater flow model.
6
13
16
4
19
3
20
9
1
2
8
10
12
5
14
18
15
7
21
22
11
17
6
13
16
4
19
3
20
9
1
2
8
10
12
5
14
18
15
7
21
22
11
17
SWAT MODFLOW
land phase
coupling
river phase
coupling
SWAT
Polygons
MODFLOW
Grid-cells
model overlay
5
3.2.3 Temporal coupling As groundwater movement is usually very slow when compared to river discharge, different time-
scales of the surface- and groundwater simulations are needed. Thus, whereas the time-step chosen for
the groundwater model MODFLOW is one month, that of the surface water model SWAT is one day,
though the output for the latter is checked usually once a week. Therefore, during the coupling process
model input/output data is exchanged at specific coupling intervals to link the models’ different time
dimensions, namely, one month in the present situation. Although other exchange time intervals could
be selected, in the case of which an interpolation is required if the MODFLOW time-step is not an
integer multiplier of the SWAT time step, the one-month step turned out to be optimal, as this is also
the sampling time of the pumping data.
4. Model application to the Upper Great Chao Phraya river basin First the traditional SWAT and MODFLOW models are set up and run individually. Subsequently,
they are coupled together. The conceptual build-up of the coupled model is endeavored using
information on the topography and the hydrogeology (aquifers’ layouts and parameters, recharge
behavior, etc.). The surface- and subsurface boundaries of the coupled-model are defined identical to
those of the individual models SWAT and MODFLOW. The river network and the watershed
boundaries are used to describe the coupling interaction at the interface of specific coupling zones.
During the coupling development, the surface- and subsurface compartments of the Upper Great Chao
Phraya Plain are connected to each other through 5 river course and 22 watersheds (cf. Fig. 5).
4.1 Surface water model The major hydrological processes conceptualized in the SWAT surface-water model are rainfall,
streamflow and soil-water. Scattered rain falling on areas of the earth’s surface of various land-use
and soil type causes a different behavior of the overall water movement, whereby some rainfall is
converted to runoff, flowing out of the basin, and some infiltrates into the ground to become soil-
water, before recharging the shallow aquifer and/or adjacent streams.
The watershed area is used to define the model’s boundary. The P.7A, Y.14, N.12A stream gauges are
defined as vertices of the flow-in boundary, whereas the stream gauge C.2 at the Chao Phraya river in
the south is assigned the major basin’s outlet. Using a DEM of the total basin, 22 additional sub-basins
are defined, covering the Ping, Yom, Nan Sakae Krang and Chao Phraya river (Fig.1). Infiltration
(areal recharge) is derived from rainfall, land-use characterization soil properties.
4.2 Groundwater model The MODFLOW groundwater conceptual model, namely the aquifers and their confining boundaries,
are defined using the concept of hydrostratigraphic units. The aquifer system in this study is set up as a
two-layer aquifer, whereby the thickness of the upper, semi-confined layer varies between 40 and 100
m and that of the lower, confined layer between 100 and 300 m. The 3-D block-centred grid model
representing the groundwater basin has a grid-size of 10 km x 10 km (Fig.4), resulting in 320 elements
in the upper and 346 elements in the lower layer. The western, eastern and northern borders of the
model are assumed to be an impermeable body of consolidated rock and are defined as specific inflow
boundaries, using information from the available head distribution there. The southern boundary
which is partially blocked by impermeable rocks and forms a narrow trough between the mountains in
the east and west is set up as an outflow boundary. The hydraulic properties of the aquifer, namely,
hydraulic conductivity, transmissivity and specific storage have been estimated by means of pumping
test data. The river-aquifer interaction of the five main rivers which are mainly responsible for the
monthly recharge are derived from the hydraulic properties of the rivers’ bed materials, cross-
sections, river stages, and the computed, seasonally varying groundwater table. The recharge, the
river stages, as well as the surface- and groundwater use, especially for rice farm irrigation, has been
adapted in response to the seasonal climatic conditions, namely, in terms of the available amount of
rainfall and reservoir storage. As for the possibility of return flow of irrigated water into the canals, it
is considered to be insignificant in the study area since, during the dry seasons, the drainage canals are
nearly drying out and the irrigation area covers only 13% of the effective recharge area of entire model
(Bejranonda et al., 2006).
6
Fig. 5. The schematics of the coupled surface-water and groundwater model in the study area.
4.3 Coupled model The sensitivity study of the groundwater model (Bejranonda et al., 2006) indicated that the most
significant parameters affecting groundwater levels are the areal groundwater recharge and the degree
of river-aquifer interaction. Therefore, in the present study, the percolation (areal recharge) computed
by the SWAT surface-water model is applied on the top layer and on the outcropping sections of the
groundwater model, and the river-aquifer interaction (channel recharge) module of the MODFLOW
groundwater model is linked to the streamflow network of the SWAT model. The conceptual set-up of
the coupled model is based on the original SWAT and MODFLOW models in the study area, with
particular consideration of the connection of the rivers, aquifers and the soil water.
During the operation of the coupled model, the major rivers contribute water to the aquifers and vice-
versa – as calculated by the river-aquifer interaction function - and 22 sub-basins provide recharge to
the groundwater - regulated by the infiltration function - (Fig.5). Both surface- and groundwater
models are bonded through the groundwater recharge and the river-aquifer interaction, using a
VISUAL BASIC interface program (Bejranonda, 2007) to exchange input and output parameter
between the two models. The coupled parameters are transferred back and forth two times, 1) after the
original SWAT and MODFLOW have been run individually, the time-series output of the SWAT
modeled percolation is sent as input to the groundwater model, 2) then the groundwater model is run
again with these new input parameters and the model-given river-groundwater interaction terms to
recalculate the streamflow discharge computed earlier with the surface water model. The final output
of the coupled model represents then the results of the individual hydrological components as
incorporated in the original SWAT and MODFLOW model as well as the dynamic interaction at their
interface zones (Fig. 5).
4.4 Calibration and verification The calibration and verification of the SWAT surface-water model is performed using monthly
streamflow discharge obtained from the stream gauge at the basin outlet. The SCS CN-numbers of
three different land-uses, forest, rice-growing and agriculture, as well as the available water capacity
(AWC) of 5 different soil-types are calibrated in the time-interval 1993-1998. The subsequent
verification of the surface-water model has then been performed with 1998-2003 historical river
discharge data and results in relative seasonally and monthly errors of 53%/36% and 77%/59%,
respectively, in the wet/dry season. As these results represent the ability of the uncoupled surface
water model to simulate the basin discharge, they will subsequently be compared with those of the
coupled model.
The MODFLOW groundwater-model calibration and verification is performed in steady state as well
as in transient state. Following the seasonal crop pattern, the seasonal stress periods are used in the
calibration of the historical groundwater levels, recorded in two-week intervals. Since during 2001-
2003, due to an invariant situation for the surface water, the groundwater use was almost stable, the
precipitation
Surface water model
Groundwater model
surface water use
streamflow
(out)
lateral flow
(out)
Ping, Yom,
Nan,Sakae Krang,
Chao Phraya river
22 sub-basins River-aquifer
interaction
infiltration
streamflow (in)
lateral flow
(in)
groundwater
use
7
average water levels during the dry season of 2003 are selected to be the representative steady-state
water levels in the calibration. 13 zonal groups of the hydraulic conductivity are adjusted during the
steady-state calibration process. The transient-state calibration is carried out, using the 1993-2003
historical water levels, whereby in addition groups of the specific storage parameter are calibrated.
The transient simulation is initialised from an average wet-season water level and the pumping rates
weights are fine-tuned during the process, as these are often prone to errors.
As a final result of the calibration process the root mean square errors (RMS) of the hydraulics heads
are found to be 3.70 m in steady-state and 5.67 and 4.79 m for the wet and dry season, respectively, in
transient mode. An a posteriori transient-state verification/forecast, using two years of groundwater
level data (2004-2005) and water level data from 50 extra observation wells collected during the study
period (2005) has additionally been carried out, resulting in an RMS of 5.95 m. Again, these results of
the uncoupled groundwater model are the reference for the test of the following coupled model.
Using the new coupled model composed of the calibrated SWAT and MODFLOW models, the entire
historical data of both surface- and groundwater used above is re-simulated. The coupled model
results in model-fit errors for the streamflow discharge of 42%/27% and 70%/45% for the seasons and
the months, respectively, for the wet/dry seasons, whereas the RMS of the groundwater heads are
5.62 and 4.69 m for the wet and dry season, respectively. Comparing these results with those obtained
for the individual stand-alone models above, clearly indicates a simulation improvement of the
coupled model, especially as far as streamflow is concerned
4.5 Results of the coupled model
4.5.1 Surface- and groundwater flow results The simulation of both surface water and groundwater in the study region is performed with both the
uncoupled and coupled models, thus allowing to evaluate the beneficial effects of the latter. The
calibrated results of the coupled simulations for the 1993-2003 period provide monthly water flow
parameters within both the surface and the sub-surface compartment of the hydrologic cycle,
including its interaction. The simulation results are analyzed relative to the average yearly water
situation, the latter being categorized, respectively, as drought, dry, normal and wet.
Compared with the uncoupled surface-water model, the coupled surface water simulation shows the
calculated streamflow discharge at the basin outlet to be more accurate, especially, during low flow of
the dry season. The overall simulated hydrograph pattern is more accordant to the observed
streamflow, particularly in normal and wet years. The streamflow calculation of the uncoupled and the
coupled models is significantly modified. The shadowed area in Fig.6 shows the differences in the
calculated discharge of two model approaches which sum up to a total average of 240 million
m3/month. The streamflow hydrograph (Fig.6) is composed of low-flow and high-flow periods. The
model coupling increases the lower baseflow into the river during the low-flow period, but reduces
excessive baseflow during flood-events in within high-flow periods. According to the calculated
streamflow hydrograph, the coupling process reduces the discharge errors enormously in low-flow
and, somewhat less, in high-flow times.
The results of the coupled model shows that the seasonally groundwater levels are, on average, about 4
m below ground surface (GS) in the wet season, but drop to 6-9 m below GS in the dry season (Fig.7).
Significant head drops of 2.5-7 m are observed between the wet and the dry season in one year, mainly
in the dry season of a drought year. Benefitting from the coupling process, the accuracy of the
groundwater level simulations are improved considerably in the dry season, though somewhat less in
the wet season. Compared with the uncoupled simulations, there is a only a little bit of difference in
the groundwater levels of the order of 0.03-0.12 m. The coupled model produces a reduction of the
groundwater recharge, resulting in a groundwater level decrease of 0.11 and 0.12 m for the wet and
dry season, respectively. Because the groundwater recharge is dominated by surface water recharge
(Fig.8), the river-aquifer interaction has less influence on the groundwater levels. Nevertheless, for
the groundwater level close to the main rivers, where usually high hydraulic conductivities are
encountered, Fig.9. illustrates that the changes are rather insignificant.
8
4.5.2 Quantitative evaluation of the benefits of coupling Both uncoupled and coupled simulations are performed for the purpose to prove and to evaluate the
advancements achieved by using the coupling methodology. The results of the coupled simulations
provide monthly streamflow discharge and two-week’s groundwater levels. The residual errors in the
simulations represent the uncoupled and coupled models’ ability in better mimicking the coupled
surface- groundwater flow. A more quantitative analysis is provided by means of Eqs. (1) and (2)
which evaluate the differences of the errors for the two model approaches for, namely the streamflow
discharge and the average absolute groundwater levels, respectively. The results are listed in Table 1.
One notes the considerable improvement of the coupled model over the compartmental ones,
particularly, as far as streamflow is concerned, and more so for the dry than for the wet season.