Anda di halaman 1dari 9

Proceedings of the 19th IAHR-APD Congress 2014, Hanoi, Vietnam

ISBN 978604821338-1

APPLICATION OF THE DETERMINISTIC HYDROLOGICAL MODEL FOR LARGE CATCHMENT


CASE STUDY AT VU GIA-THU BON CATCHMENT - VIET NAM.

VO NGOC DUONG(1), PHILIPPE GOURBESVILLE(2)

(1)Nice Sophia Antipolis University, Polytech Nice Sophia, Innovative City lab URE 005,vo.ngoc.duong@etu.unice.fr

(2)Nice Sophia Antipolis University, Polytech Nice Sophia, Innovative City lab URE 005, gourbes@unice.fr

ABSTRACT
In order to create a tool to help hydrologists and authorities having good understandings about happenings in stream
flow regime and its variation in the future under the impact of climate change in Vu Gia Thu Bon catchment, a
deterministic hydrological model MIKE SHE from DHI- is constructed. The model covers the major processes in
hydrologic cycle and includes process models for evapo transpiration, overland flow, unsaturated flow, groundwater
flow, channel flow, and their interactions. Each of these processes can be represented at different levels of spatial
distribution and complexity, according to the goals of the modeling study, the availability of field data and the
modeler's choices. This model is calibrated and validated against the daily data in period 1991-2000 and 2001-2010,
respectively. The performance of model are showed by good statistical coefficients at seven gauging stations. The Nash
Sutcliffe and relation coefficients could reach to 0.78 to 0.82 and 0.89 to 0.92, respectively when comparing the discharge.
With the water level, because the limitation of resolution, these coefficients are lower than comparing with the
discharge. However, these are still high in upstream when The Nash Sutcliffe and relation coefficients could reach to
0.63 to 0.77 and 0.81 to 0.89, respectively. This study is confirmed again the performance of deterministic distributed
model in simulating the hydrological process and the capacity of this kind of model to projected climate change, at least
confirm the usefulness of this model with ungagged catchment or large catchment. This study also prove the role of
multi calibration in increasing the accuracy of hydrological model for a large catchment.

Keywords: Deterministic hydrological modeling, sensitivity, multi-calibrate, long-term simulations, Vu Gia Thu Bon catchment.

1. INTRODUCTION as conceptual and physical based, and according to the


spatial description of catchment processes as lumped and
The water is seen the premier constitutive factor of
distributed (Refsgaard, 1996). Each of them has
human being. Thus, it always plays an important role to
advantages and inconveniences for simulating the
the development of human society. However, water
hydrological process. Lumped models are one of simple
distribution is not equal due to time and space. These
hydrological model which assume that all characteristics
unbalances conduct many negative impacts to the
are constant across the catchment (Chow, 1972) such as
human. Annually, natural disaster related on water issue
HBV model, MIKE11/ NAM Model, Tank, Topmodel,
such as flood, drought, stormbring out lots of severe
Xinanjiang. Lumper parameter models are much simpler
damages. These are not only on the property but also on
in their treatment of spatial variation, which each model
the life. According to IPCC (2007) in recent years, under
parameter described by a value that is uniform for the
the impact of climate change, the consequences of flood,
whole catchment so these could not be determined
drought catastrophes are more serious. Getting the deep
directly from physical characteristics of the catchment
knowledge on the hydrological factors in a catchment is
under consideration. These parameters are generally
an essential objective of hydrologist community to be
determined via calibration (Chow, 1972, Madsen, 2003).
able to transform disadvantages to advantages or
In contrast, distributed model is constructed on dividing
mitigate the natural catastrophic damage to people.
catchment into sub units which each unit represents all
Nowadays, with the development of mathematics and
physical characteristics for a real area. These kinds of
computer science, the simulating the hydrological cycle
models maintain the physical details at a given grid size
for catchment becomes easier and more accurate. These
and consider the distributed nature of hydrological
progressions help hydrologists might have concrete and
properties such as soil type, slope and land use
truthful insights about what happens in hydrology to
(Refsgaard, 1997, Vansteenkiste et al, 2013). In principle,
make good decisions for reducing natural hazards on the
parameters of distributed model could be gotten from the
catchment (Cunge, 2003).
catchment data. For this reason, distributed model is
Up to date, many hydrological models have been evaluated to be able to translate more accurately and
developed with different theories to simulated concretely the hydrologic process in a catchment. One
catchments hydrological phenomenon. They may be more advantage of distributed model is that the outputs
classified according to the description of physical process such as water level, discharge, hydrological factors could

1
be perfectly extracted at anywhere in the catchment. evaluate the sensitivity of model parameters and (3)
These efficiencies help to overcome difficulties in lack of compare the calibrated efficiency of multi-site versus
observed data, which have a great significance for single-site (4) analyze the uncertainty of input data and
simulating the hydrological process at a large catchment model structure.
or in developing countries. However, which model is the
best to simulate hydrological process? Until now this
2. VU GIA THU BON CATCHMENT
question has not been clearly answered yet. There are
some arguments on the advantage and limitation of this The Vu Gia Thu Bon, which originates on the eastern side
both kind of model. Even if distributed model have of the Truong Son mountain range and drains to the
drawbacks on computation time, initial parameter ocean near the cities of Da Nang and Hoi An, is the
definition or lack of spatial data for set up model as well biggest river system of the coastal province in the central
as validation (Parkin, 1996, Bormann et al, 2009, Benven, region of Viet Nam. This system has two main rivers, the
2011), most of modelers estimate highly the capacity of Vu Gia and Thu Bon rivers, which flow through many
distributed model than lumped model. The lumped complex topographies: the relatively narrow
models, in many case, perform just as well as distributed mountainous area with a maximum elevation of 2600m at
models with regard to rainfall runoff simulation when Ngoc Linh mountain which, features a large number of
sufficient calibration data exist. (Michaud and steep tributaries, and the flat coastal zone at the
Sorooshian, 1994, Refsgaard and Knudsen, 1996). downstream, prone to annual flooding consisting of a
Conversely, the distributed model could not prove complex interconnected coastal river system. This system
completely its performance in simulating the is located at a tropical monsoon climate region where
hydrological factor. Because the spatial input data such as weather phenomena, such as rain and storm happen
topography, soil type, land use nowadays might be complicatedly. With the typical characteristic of the
available to build the model, but it is really not easy to region, the climate pattern in Vu Gia Thu Bon basin is
find the spatial data for calibration or validation (Beven, influenced by Truong Son mountain in western with a
1996). This leads to quality estimation of the model in the quite high rainfall. The average annual rainfall of this
whole catchment is mostly unable, the calibration and area is from 2000 mm to 4000 mm. However, it differs in
validation are generally carried out against the data at season, 65 % to 80 % of the annual rainfall drops during
several gauging stations. One more weak point of the period of September December. The precipitation
distributed models is that this model type commonly has shown increasing trends from north to south and
consists of more parameters than lumped models so the from low to high elevation area. Moreover, this region is
calibrated process is more complicated and difficult to annually attacked by 2-4 typhoons that bring huge
achieve the acceptable values (Reed, 2004). However, rainfall and whirlwind (RETA 6470, 2011, To, 2005). It
Refsgaard, 1997 noted that more detailed physically base makes inundation disaster happen frequently more
and spatially distributed models are assumed to give a seriously. On the contrary, drought frequently occurs in
detailed and potentially more correct description of the the remaining months. Despite these complicated climate
hydrological process in the catchment whereby might conditions, the hydrological infrastructure in the basin is
provide more accurate predictions. From these still underdeveloped. The density of measuring stations is
preeminences, many distributed models have been sparse, especially the rain and flow gauge stations in
developed and applied in recent years such as MIKE tributaries. Although this area covers around 10350 km 2,
SHE, SWAT, SWIM, LISFLOOD and WETSPA. it has only fifteen stations (Figure 1) that record rainfall
data. Lack of data and data quality bring certain
The Vu Gia Thu Bon is one of the large catchment located
difficulties in forecasting natural phenomenon in current
at central of Viet Nam where annually confronts with
time as well as in future.
severe damages due to natural disasters such as flood,
drought catastrophe. Furthermore, according to the
prediction of IPCC's scenario (IPCC, 2007), under the
impact of global warming, sea level increase, the change
in hydrological cycle, flood and drought disaster,
abnormal phenomena e.g phenomena El Nino and La
Nina in Vu Gia Thu Bon basin are forecasted to be going
to happen more frequently and more extremely. It makes
the consequences of natural disasters to people,
livelihood, social economic development are more severe.
Hence, in order to mitigate the impact of these
catastrophes to human being, it is necessary to have an
efficient tool to help hydrologists and authorities having
good understandings about happenings in stream flow
regime and its variation in the future under the impact of
climate change . For this purpose, a fully deterministic
hydrological model MIKE SHE model from DHI- is
chosen to simulate the hydrological process in Vu Gia
Thu Bon catchment. The objective of this paper is to (1)
confirm the performance of distributed model type to
simulate the hydrological process and capacity to Figure 1. Vu Gia Thu Bon catchment in central Viet Nam, and
hydro meteorological network.
translate climate scenario, at least the advantage of this
model towards large catchment and ungauged area (2)

2
Due to the violence of climatological events, the fragile applied the MIKE SHE model to simulate the
economic condition and the underdeveloped hydrological process of Senegal River Basin. This model
infrastructure, the natural disasters related to river flow was developed on an area 375000 km 2, and included all
deeply affect the population in this region. In addition, of hydrologic components. The result was relatively
the farming habits formed agricultural production and preventative of the characteristic of this catchment with
cultural customs also influence negatively on the good obtained statistical coefficients. Thomson, et al,
prevention of inhabitants against natural catastrophes. (2004) used this model to simulate the hydrological
Furthermore, local authorities although have made system in lowland wet grassland in southeast England.
efforts to prepare for these disasters and to overcome the These authors used MIKE SHE coupling with MIKE 11 to
damage, these works seem still insufficient. They have present the hydrologic factors in Elmley Marshes
not yet had a complete strategy to help population avoid catchment during 36 months. This research gave
the catastrophic effects. Consequently, the population in remarkable results in simulating surface flooding,
central Vietnam, especially in Vu Gia Thu Bon basin, groundwater and flow in the channel. The application of
annually sustains considerable damages to people and the coupled MIKE SHE/MIKE 11 modelling system to
property. the Elmley Marshes has demonstrated its potential to
represent complex hydrological systems found within
3. METHODOLOGY
many wetland environments. By simulating the stream
flow process at catchment (<100 km 2) in China, and at in
3.1 Deterministic distributed hydrological model
Hawaii, USA, the works of Zhang et al, (2008) and
In many current hydrological models, deterministic Sahoo et al (2006) already proved the capacity of MIKE
distributed model is likely to have more advantages. The SHE to describe the flow in mountainous region. This
main interest of a deterministic distributed hydrological model is likely to be preeminent to simulate hydrology in
model is to be able to provide hydrological data at any semi-arid area with the studies of McMichael et all,
locations within the catchment. This possibility allows to (2006). These strong points of fully deterministic
investigate, in depth the hydrological dynamic. Several distributed hydrological model MIKE SHE model are
tools are today available and could be used for such expected to increase the accuracy when modelling the
analysis. The typical model is MIKE SHE developed and river flow in current and could translate better the
extended by DHI Water & Environment since the last climate scenarios in the future.
decades of the 20st century. MIKE SHE covers the major
3.2 Input data and model setup
processes in hydrologic cycle and includes process
models for evapo transpiration, overland flow, In order to assess the variation of run off at Vu Gia Thu
unsaturated flow, groundwater flow, channel flow, and Bon catchment happening in the end of this century, a
their interactions. Each of these processes can be deterministic hydrological model - MIKE SHE model -
represented at different levels of spatial distribution and has been built. All modules of MIKE SHE model are used
complexity, according to goals of the modeling study, the to be able to describe honestly the hydrological process in
availability of field data and the modeler's choices (Butts this region and help to reduce the uncertainties when
et al, 2004). The representation of catchment translating the climate scenario to hydrological model.
characteristics and input data is provided through the Vu Gia Thu Bon is a big river system which plays an
discretization of the catchment horizontally into an important role with the developing of Da Nang city and
orthogonal network of grid squares. In this way spatial Quang Nam province. Its run off variation affects
variability in parameters such as elevation, soil type (soil significantly the social economy of this area. However,
hydraulic parameters), land cover, precipitation and observed data of this river system are still not sufficient
potential evapo-transpiration can be represented. Within to give a good simulation. The lack of data is always a big
each grid square, the vertical variations in soil and hydro problem when modelling hydrology in this catchment.
geological characteristics are described in a number of Therefore in this study, the data is collected from
horizontal layers with variable depths. Lateral flow different sources and with the support from many
between grid squares occurs as either overland flow or agencies.
subsurface saturated zone flow. The one-dimensional
* Topography: The elevation data using in the model is
Richards' equation employed for the unsaturated zone
taken from STRM DEM with the resolution 90 m of
assumes that horizontal flow is negligible compared to
NASA (Figure 1) (http://www.cgiar-csi.org).
vertical flow (Thompson et al, 2004). With its
performance, MIKE SHE has been used in a broad range * Precipitation: The rainfall which is a key factor in
of applications. It is being used operationally in many hydrologic process. Its spatial variation affects heavily on
countries around the world by organizations ranging both runoff generation and hydrologic process in a
from universities and research centers to consulting catchment (Moon et al., 2004). The spatial variability in
engineers companies (Refsgaard and Storm, 1995). MIKE rainfall may introduce significant uncertainty in model
SHE has been used for the analysis, planning and parameter during the calibration process (Chaubey et al.,
management of a wide range of water resources and 1999). While the quality of spatial rainfall distribution
environmental and ecological problems related to surface usually depends on the characteristic of the study area
water and groundwater, such as: River basin and other factors, in particular, the rain gauge density.
management and planning, Water supply design, Therefore, the network of rain gauge station in Vu Gia
management and optimization, irrigation and drainage, Thu Bon is sparse, averagely one station observes for an
Soil and water management, groundwater management, area 700 km 2. Hence, redistributing spatial rainfall is
interactive between water surface et ground water, necessary. According to the study of Vo and Gourbesville
ecological evaluations flood plain studies, impact of land (2014), the Kriging is the most suitable method to
use and climate change. Graham and Butts, (2005) interpolate the rainfall in this area. This simulation uses

3
the rainfall data which is redistributed spatially based on the measurements, remaining ones are extracted from the
daily rainfall data from 15 rain gauge stations with the DEM.
Kriging method (cell size 1000 m).
* Overland flow: The overland flow appears after the net
* Evapotranspiration: This data is inherited from the rainfall rate exceeds the infiltration capacity of the soil,
study of Vu et al (2008). water is ponded on ground surface. The main parameter
supplies MIKE SHE to calculate this flow is
* Land use and Soil map: The land use and soil data are
Stickler roughness coefficient M. For Vu Gia Thu Bon,
provided by researcher of project Land Use and Climate
this parameter is determined depending on Land use
Change Interaction in Central Viet Nam (LUCCI), and
map and in 2 to 90 m 1/3/s (Nguyen Thong, 2005; DHI
project Impacts of Climate Change in Mi-Central Viet
2012; Vieux, 2001)
Nam (P1-08 VIE). In order to simplify the calibrated
process, the original land use and soil map data is * Unsaturated zone: DHI suggests three methods for
preliminarily treated by merge from many kinds of soil describing the flow in this zone. However, the difference
and land use to typical types of Vu Gia Thu Bon about accuracy between them is not so high. In this
Catchment. simulation, the simple two-layer water balance method is
chosen to diminish the computation time. The physical
property of each kind of soil is presented via the water
content at saturation, water content at field capacity,
water content at wilting point, saturated hydraulic
conductivity. Initial values using in this model is
consulted on the work of Do and Nguyen (2000),
Domenico (1998).
* Saturated zone: The groundwater is supplied by Central
Viet Nam Division of Water Resources Planning and
Investigation (http://www.ceviwrpi.gov.vn). The
characteristic of aquifer is mainly presented by
Horizontal hydraulic and vertical hydraulic conductivity
that is taken from the study of Environmental Science
Division (EVS) a Division of Argonne National
Laboratory.

4. SENSITIVITY ANALYSIS, CALIBRATION


AND VALIDATION

In principle, the model parameters of hydrological


distributed model should be assessable from catchment
data (Refshaard et al, 1995). However, simulating for a
large catchment with the input data is not very accurate
as well as the components are simplified. Furthermore,
the input data resolution is still coarse, which affects
significantly on the representation of geology,
topography, soil, vegetation and climate factors into
model. As a result, the model could not represent
precisely the physical property of catchment due to both
these reasons. (Gurtz et al, 1999). These limitation make
reduction on the simulation performance of distributed
model. Consequently, the run off from uncalibrated
model may be different from the measurement. In order
to improve the weakness, the calibration is carried out to
Figure 2. Soil map (upper), land use (lower) at Vu Gia Thu Bon
find an optimal set of parameter values that simulate the
catchment for this study.
behavior of the watershed as accurate as possible.
* River and lakes: the Vu Gia Thu Bons river network is (Madsen, 2003, Wang et al, 2012). Nevertheless,
so complicated. It flows through different terrain Refsgaard and Storm (1995) note that the number of
morphologies, mountain at the western and delta coastal parameters subject to adjustment during calibration of
region at eastern. It is composed from two main rivers, MIKE SHE should be kept as small as possible. This issue
Vu Gia and Thu Bon rivers which include lots of is depended so much on the accuracy of initial set up
tributaries at upstream and linking branches at value of model. For that requirement, the initial value
downstream. In order to simulate better the river flow, have to define carefully base on the physical property of
MIKE SHE model is coupling with a hydrodynamic catchment.
MIKE 11 model. This model is developed on 44 big
branches with length varying from 20 km to 202 km. The 4.1 Sensitivity analysis
geometry of each river branch is specified via cross
section. The cross section applies in this model is from Calibrating a fully distributed model which consists of
two sources, few of them at downstream are taken from many components and mutually dependent parameters

4
is really difficult, it makes modeler must attach special Saturated hydraulic conductivity 8.5 10-6 m/s
importance to many parameters, even several factors - Light clay
influence insignificantly hydrographs. Hence, the Water content at saturation 0.41
calibration is no need to apply for all parameters. Only Water content at field capacity 0.12
Water content at wilting point 0.04
several parameters, which have important and great
Saturated hydraulic conductivity 2.085 10-4 m/s
effects on the model are chosen. These parameters are
- Sand
decided relying on the results of sensitivity process which Water content at saturation 0.37
is done manually by varying individually the values of Water content at field capacity 0.1
parameters one by one. The sensitivity also quantifies the Water content at wilting point 0.03
dependent rate of runoff on the change of these Saturated hydraulic conductivity 2.89 10-4 m/s
parameters. As a result, these rates conduct the * Saturated zone
calibration to more easily obtain the acceptable value. It is Horizontal hydraulic conductivity 6.7 10-5 m/s
seen a prior step of calibration process. When doing this Vertical hydraulic conductivity 10-4 m/s
work for MIKE SHE model in Vu Gia Thu Bon
catchment, the sensitivity of each parameters is analyzed This big catchment has a complicated river system.
based on the response of discharge in Nong Son and Although the length of two main rivers reaches to 200
Thanh My station. Only few of model parameters are km, there is only one flow measured station in the middle
chosen to calibrate. These are show at the Table 1. of each main river, Nong Son and Thanh My stations.
This situation makes several difficulties in the
4.2 Calibration and validation comparison of the results between simulation and
observation. Especially this is an inconvenient point to
The MIKE SHE model is built in an 11 year period from predict flood risk for downstream region, and also is a
1990 to 2000 for calibration and 2000 to 2010 for factor producing the uncertainty when assessing the
validation. This model simulates on all of the available impact of climate change to river run off. The lack of
components of MIKE SHE model, e.g overland flow, observation data for comparing with the simulation
river and lake, unsaturated flow, evapotranspiration, decline the performance of distributed model. Many
saturated flow to make the model will be able to describe hydrologists have suggested to realize calibration on
the most accurate hydrological cycle in Vu Gia Thu Bon multi-site, with not only against the discharge but also
catchment, as well as to reduce the uncertainty when the water level. Accordingly, The water lever at Ai Nghia
simulating for future scenarios. In other to stabilize the station, Cam Le Station on Vu Gia branch and Hiep Duc,
model and establish proper initial condition for next step, Giao Thuy, Cau Lau stations on Thu Bon branch is as
the first year of each period is used for warming up the well compared in this Mike She model to increase the
model. Hence, in this study, only ten years of daily data confidence of the model and simultaneously reduce
is taken to calibrate and validate. The optimal values uncertainty in projected climate scenario, beside compare
reached from calibrated process are show at Table 1. the discharge at these two stations. The model assessment
Table 1: Calibrated parameter values of MIKE SHE is performed with statistical measures of the root mean
model. squared error (RMSE), the correlation coefficient (R), and
Nash-Sutcliffe coefficient (E).
Key parameter Optimal value
* River bed resistance - Strickler Coefficient
- Tributary and upstream of Vu Gia 18 m(1/3)/s (1)
- Tributary and upstream of Thu Bon 25 m(1/3)/s
- Linking branch 30 m(1/3)/s
- Downstream 40 m(1/3)/s
* Overland flow - Strickler Coefficient
- Planted forest 5 m(1/3)/s
- Rural settlement 8 m(1/3)/s
- Rice 16 m(1/3)/s (2)
- Annual crops 8 m(1/3)/s
- Perennial crops 8 m(1/3)/s
- Unsed land 5 m(1/3)/s
- Natural forest 2 m(1/3)/s
- Urban 90 m(1/3)/s (3)
- Water surface 33 m(1/3)/s
* Unsaturated flow - soil property where the Xobs is observed value and X model is modelled
- Clay 0.5 value at time/ place i.
Water content at saturation 0.32
* Calibration and validation with discharge:
Water content at field capacity 0.2
Water content at wilting point 1.2 10-8 m/s Hydrographs in Figure 3 demonstrate that the model
Saturated hydraulic conductivity simulates relatively accurately the runoff in Vu Gia Thu
- Suit loam
Bon catchment. Simulated base flows at the two stations
Water content at saturation 0.39
Nong Son and Thanh My are similar to the
Water content at field capacity 0.18
Water content at wilting point 0.1
measurements. However, it seems that the peak of sub-
Saturated hydraulic conductivity 2.45 10-6 m/s main flood is not presented well. This limitation may
- Loamy Sand causes by the quality of observation data. In dry season,
Water content at saturation 0.43 the data is only captured once or twice per day, so it
Water content at field capacity 0.32 could not present precisely the time of sub-main flood
Water content at wilting point 0.18 appearance. It is really difficult to overcome the problem

5
concerning to missing data. Generally, peak floods are such as rainfall. Therefore, this component of runoff is
almost the same to observation data. Occasionally, some more stable through long time and closer with the
peaks are higher than reality but the difference is not very measurement if the model is set well. On the contrary,
high, it is reasonably acceptable. In theory, this problem peak flow is influenced by rainstorm event. The time step
could be completely controlled. With simulation for small of rainfall data input decides hugely to concentration
area and in short time, this issue seems uncomplicated to time, concretely to peak discharge. For this reason, the
reach a better situation by modifying parameters related rainfall data input using large step might be the main
on concentration time and infiltration. However, for ten cause of the differences of peak discharge in this MIKE
years simulation in a large area, this problem becomes SHE which use daily rainfall input. Nevertheless, the
more complex. It needs to take more time to calibrate. higher discharge in wet season when simulating the
Selecting between an acceptable results and higher impact of climate change might result in better
performed model must be considered carefully. protections. The quality of this simulation also prove
Furthermore, the more accuracy of MIKE SHE model in through validated period.
describing the runoff in dry season than flood season
might be understandable under the data aspect. The base
flow is mainly composed from groundwater, so this
component is not dependent so much on data factor,

Figure 3. Calibrated and validated hydrographs of discharge

6
Figure 4. Calibrated and validated hydrographs of water level.

The flow in ten years is regenerated approximately with probably adds a certain value to show the correlative
the observation. The simulated flow in these ten years level of model result with real data.
also indicated that the tendency of simulated peak
discharge is higher than measurement. Table 2: Statistical indices of MIKE SHE model in
Vu Gia Thu Bon catchment.
The efficiency of MIKE SHE model is also shown through
the statistical coefficients in Table 2. Daily and monthly Calibaration Validation
discharges are compared between simulation and Station (1991-2000) (2001-2010)
RMSE R E RMSE R E
observation. These numbers prove the accuracy of this
H Thanh My 0.77 0.86 0.67 0.68 0.83 0.61
model in describing the hydrological process in Vu Gia
H Ai Nghia 0.7 0.81 0.63 0.66 0.78 0.56
Thu Bon catchment. The R and E coefficients at Nong Son H Cam le 0.26 0.83 0.12 0.28 0.8 -0.46
and Thanh My in the calibration period reach 0.92, 0.89 H HiepDuc 0.77 0.89 0.77 0.91 0.83 0.59
and 0.82, 0.78, respectively. In the validation period, these H Nong Son 0.89 0.88 0.76 0.84 0.86 0.72
factors reduce, but not very low, R and E coefficients at H Giao Thuy 0.85 0.85 0.61 0.73 0.82 0.6
Nong Son station is 0.92 and 0.82 and at Thanh My is 0.90 H Cau Lau 0.44 0.84 0.56 0.47 0.83 0.16
and 0.69. The RMSE coefficient at Nong Son and Thanh Q Thanh My 132.3 0.89 0.78 123.2 0.9 0.69
My in both period are relatively small. In ten years, the Q Nong Son 288.7 0.92 0.82 250.5 0.91 0.82
RMSE of simulation are only 132.3 m 3/s at Thanh My,
and 288.7 m 3/s at Nong Son while the RMSE of One more important, if we refer on the rate between
validation at Thanh My is 123.2 m 3/s and at Nong Son is statistical coefficients of water level on statistical
250.5 m3/s. coefficients of discharge at Nong Son and Thanh My
stations where have both of these kinds of observed data
* Calibration and validation with water level: and hypothesize that this relation is similar to other
The similar shape between model and observation in stations, it is completely possible to believe that the flow
Figure 4 is again to illustrate the simulated capacity of at these stations could be evidently reproduced well. It is
MIKE SHE. Nevertheless, precision of simulated water a strong point to make more confident on the estimating
level is not perfect like discharge. These differences the variation tendency of stream flow in future, in
conduct to the statistical coefficients of water level particular in location do not have data.
comparison between simulation and measurement are
not as high as these one of discharge case (Table 2). The Through these hydrographs and coefficients, the flow in
relation coefficient in all stations is around 0.8 to 0.9 but Vu Gia Thu Bon catchment is reproduced rather
Nash-Sutcliffe coefficients are low, it varies from 0.6 to impressively, not only via discharge but also via water
0.7 at upstream station and smaller in station under the level. These results demonstrate the performance of Mike
effect of tidal level. The source of this inaccuracy might She model when simulating the hydrological process and
be from the coarse resolution of topography data and the this model is able to estimate the variability of stream
lack of measured cross section, as well as the big distance flow under the impact of climate change.
of two computed note (dx) in MIKE 11. Related on the
resolution of data and distance dx, there are up till now 5. CONCLUSIONS
many studies on these problems. The contrasted relation
between cell size, quality of simulation and computation With the aim of providing a tool to simulating the
time is an insolvable issue, in particular towards big area hydrological process and estimating the impacts of
and longtime period (Xevi, 1997, Vzquez, 2002). Thus, climate change on runoff of Vu Gia Thu Bon river
satisfying all of these requirements are not easy. Despite system, a deterministic hydrological model MIKE SHE
the fact that statistical coefficients of water level are still model- is built. This model accounted mostly the runoff
low, calibrating the model relying on this factors factors, from surface flow to groundwater flow, and
evapotranspiration, so this model is expected to

7
reproduce the happening of hydrological cycle in this Domenico, P. A., & Schwartz, F. W. (1998). Physical and
catchment. Relying on these considerable progress, this chemical hydrogeology (Vol. 44). New York: Wiley.
model is hoped to reduce a part of uncertainty when Do, D.S, Nguyen, N.B (2000), Assessment of potential
assessing the impact of climate change for the region. On productivity of forest lands in Viet Nam, Statistical
the other hand, another advance of fully deterministic publishing house.
distributed model is to help to overcome the weakness of Dong NC., Araki H., Yamanishi H., and Koga K. (2005).
lack of the data. The difference between the actual and Simulation of groundwater flow and environmental
future runoff regime could be compared in anywhere effects resulting from pumping. Environmental
over the catchment. Hence, an overview of change in run Geology, 47 (3), 361-374.
off regime in the whole catchment could be generated Graham, D. N., & Butts, M. B. (2005). Flexible, integrated
without difficultly. These great points confirm the watershed modelling with MIKE SHE. Watershed
capacity of distributed model in modeling the models, 849336090, 245-272.
hydrological process and climate change. This model is Gurtz, J., Baltensweiler, A., & Lang, H. (1999). Spatially
calibrated and validated against daily data and monthly distributed hydrotope-based modelling of
data in the period of 1990-2000 and 2001-2010, evapotranspiration and runo in mountainous basins.
respectively. The performance of model is showed via the Hydrological Processes, 13, 2751-2768.
shapes of hydrograph, and via statistical indices Managing water in Asias river basins: Charting progress
comparing with seven gauging stations in this catchment. and facilitating investment. RETA 6470 (2011), The Vu
These comparisons are carried out not only the discharge Gia Thu Bon Basin.
best also on the water lever in different locations of the Madsen, H. (2003). Parameter estimation in distributed
Vu Gia Thu Bon catchment whereby prove the necessary hydrological catchment modelling using automatic
to compare the measurement and simulation at multi- calibration with multiple objectives. Advances in water
site. The multi-site calibration help to increase the resources, 26(2), 205-216.
accuracy in translating what happen in the nature to McMichael, C. E., Hope, A. S., &Loaiciga, H. A. (2006).
model as well as minimize the uncertainty of model. This Distributed hydrological modelling in California
model still contains many uncertainties. These semi-arid shrublands: MIKE SHE model calibration
uncertainties are from many sources such as, model and uncertainty estimation. Journal of
structure, cell size, and input data. Eliminating these Hydrology, 317(3), 307-324.
source is likely unable. Hence, it is necessary to realize on Michaud, J., & Sorooshian, S. (1994). Comparison of
many model with different cell sizes and data accurate simple versus complex distributed runoff models on a
definition to minimize highly the uncertainty. midsized semiarid watershed. Water Resources
Research, 30(3), 593-605.
ACKNOWLEDGMENTS Moon, J., Srinivasan, R., & Jacobs, J. H. (2004). Stream flow
estimation using spatially distributed rainfall in the
Thanks to the Hydro meteorological Center in mid
Trinity River basin, Texas. Transactions of the
central Viet Nam, the Central Viet Nam Division of Water
ASAE, 47(5), 1445-1451.
Resources Planning and Investigation, LUCCI project, P1-
Nguyen Thong, (2005) supply water and drainage.
08 VIE project where provided the data for this study.
Construction publishing house.
Pachauri, R. K., &Reisinger, A. (2007). IPCC fourth
REFERENCES assessment report. IPCC, Geneva.
Beven, K. J. (1996). A discussion of distributed Parkin, G., O'donnell, G., Ewen, J., Bathurst, J. C.,
hydrological modelling (pp. 255-278). Springer O'Connell, P. E., & Lavabre, J. (1996). Validation of
Netherlands. catchment models for predicting land-use and climate
Beven, K. J. (2011). Rainfall-runoff modelling: the primer. change impacts. 2. Case study for a Mediterranean
John Wiley & Sons. catchment. Journal of Hydrology, 175(1), 595-613.
Refsgaard, J. C. (1997). Parameterisation, calibration and
Bormann, H., Breuer, L., Giertz, S., Huisman, J. A., & validation of distributed hydrological models. Journal
Viney, N. R. (2009). Spatially explicit versus lumped of Hydrology, 198(1-4), 69-97.
models in catchment hydrologyexperiences from two Refsgaard, J. C. (1996). Terminology, modelling protocol
case studies. In Uncertainties in environmental and classification of hydrological model codes.
modelling and consequences for policy making (pp. 3- In Distributed hydrological modelling (pp. 17-39).
26). Springer Netherlands. Springer Netherlands.
Butts, M. B., Payne, J. T., Kristensen, M., & Madsen, H. Refsgaard, J. C., & Knudsen, J. (1996). Operational
(2004). An evaluation of the impact of model structure validation and intercomparison of different types of
on hydrological modelling uncertainty for streamflow hydrological models. Water Resources Research, 32(7),
simulation. Journal of Hydrology, 298(1), 242-266. 2189-2202.
Chaubey, I., Haan, C. T., Grunwald, S., & Salisbury, J. M. Reed, S., Koren, V., Smith, M., Zhang, Z., Moreda, F., Seo,
(1999). Uncertainty in the model parameters due to D. J., & DMIP, P. (2004). Overall distributed model
spatial variability of rainfall. Journal of intercomparison project results. Journal of
Hydrology, 220(1), 48-61. Hydrology, 298(1), 27-60.
Chow, V.T (1972). Hydrologic modelling. Journal of the Refshaard, J. C., Storm, B., & Singh, V. P. (1995). MIKE
Boston society of Civil Engineering, 60, 1-27. SHE. Computer models of watershed hydrology., 809-
Cunge, J. (2003). Of data and models. Journal of 846.
Hydroinformatics, 5, 75-98. Sahoo, G. B., Ray, C., & De Carlo, E. H. (2006). Calibration
DHI (2012), MIKE SHE User's Manual. Mike by DHI. and validation of a physically distributed hydrological
DHI (2012), MIKE ZERO User's Manual. Mike by DHI. model, MIKE SHE, to predict streamflow at high

8
frequency in a flashy mountainous Hawaii
stream. Journal of Hydrology,327(1), 94-109.
Thompson, J. R., SRENSON, H. R., Gavin, H., &
Refsgaard, A. (2004). Application of the coupled
MIKE SHE/MIKE 11 modelling system to a lowland
wet grassland in southeast England. Journal of
Hydrology, 293(1), 151-179.
To, T. N (2005), Strategic plan for integrated water
resources management of the Vu Gia Thu Bon Basin
Organization. Economic and Social commission for Asia
and the Pacific - United Nation.
Vansteenkiste, T., Tavakoli, M., Ntegeka, V., Willems, P.,
De Smedt, F., &Batelaan, O. (2013). Climate change
impact on river flows and catchment hydrology: a
comparison of two spatially distributed
models. Hydrological Processes, 27(25), 3649-3662.
Vzquez, R. F., Feyen, L., Feyen, J., & Refsgaard, J. C.
(2002). Effect of grid size on effective parameters and

model performance of the MIKESHE

code.Hydrological Processes, 16(2), 355-372.


Vieux, B. E. (2001). Distributed hydrologic modeling using
GIS (pp. 1-17). Springer Netherlands.
Vo, N.D. Gourbesville, P. (2014). Rainfall uncertainty in
distributed hydrological modelling in large
catchments: an operational approach applied to the vu
gia-thu bon catchment - Viet Nam. 3rd IAHR Europe
Congress, Book of Proceedings, 2014, Porto Portugal.
Vu, V. N., Do, D. D., & Dang, T. L. (2008). Potential
evapotranspiration estimation and its effect on
hydrological model response at the Nong Son Basin.
Wang, S., Zhang, Z., Sun, G., Strauss, P., Guo, J., Tang, Y.,
& Yao, A. (2012). Multi-site calibration, validation,
and sensitivity analysis of the MIKE SHE Model for a
large watershed in northern China. Hydrology and
Earth System Sciences,16(12), 4621-4632.
Xevi, E., Christiaens, K., Espino, A., Sewnandan, W.,
Mallants, D., Srensen, H., & Feyen, J. (1997).
Calibration, validation and sensitivity analysis of the
MIKE-SHE model using the Neuenkirchen catchment
as case study. Water Resources Management, 11(3), 219-
242.
Zhang, Z., Wang, S., Sun, G., McNulty, S. G., Zhang, H.,
Li, J., ...& Strauss, P. (2008). Evaluation of the MIKE
SHE Model for Application in the Loess Plateau,
China1.

Anda mungkin juga menyukai