Daily runoff simulation in Poyang Lake Intervening Basin based on remote sensing data

Available online at www.sciencedirect.com

Procedia Environmental Sciences 10 (2011) 2740 – 2747

2011 3rd International Conference on Environmental Science and Information Application Technology (ESIAT 2011)

Daily runoff simulation in Poyang Lake Intervening Basin based on remote sensing data Jiali GUO, Shenglian GUO, Tianyuan LI 1

State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan 430072, China [email protected]

Abstract Due to the frequent water exchange and turbulence between Poyang Lake and Yangtze River, the measured runoff hydrograph at the Hukou outlet control station is unable to be directly used in hydrological model calibration. As a result, the rainfall runoff relationship and water budget analysis in Poyang Lake regions is an unsolved problem so far. In this study, the basin area analogue and water budget methods was used to modified runoff hydrograph at the Hukou, and multiple-input single-output system model was used to derive intervening basin runoff. The Variable Infiltration Capacity (VIC) distributed hydrological model based on different types of remote sensing data was used to simulate daily runoff in the ungauged Poyang Lake Intervening Basin (PLIB). The comparison of estimated and simulated daily runoff hydrographs shows that the VIC model is performed relatively well. It is proved that with the help of remote sensed data, important hydrological characteristics of the PLIB can be reproduced and predicted.

© 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Conference © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of [name organizer] ESIAT2011 Organization Committee. Keywords: Poyang Lake, Runoff Simulation, VIC Distributed Model, Remote Sensing

1.

Introduction

There has been considerable focus on researches of introducing remote sensing data into the application of hydrological models, especially at large or medium scale basins level. Remote sensing offers a potentially powerful alternative to the use of ground observations that historically have provided the sole forcing, and even in some ungauged basins or data-sparse regions, remote sensing would be the only option. In addition, the use of remotely sensed information for data assimilation in such models has begun. [1] Due to the frequent water exchange and turbulence between the Poyang Lake and Yangtze River (See Fig. 1) as well as many other uncertain reasons, the runoff hydrograph at the Hukou outlet control station is unable to be directly used in hydrological model calibration. In particular, the bottomlands around control stations are often submerged when flood occurs; in such case water moves through other paths instead of the control sections, leading too much water amount can not be measured. The so-called “water escape” combined with reasons like human activities, low measure techniques and so forth; results

1878-0296 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Conference ESIAT2011 Organization Committee. doi:10.1016/j.proenv.2011.09.425

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

in the water unbalance problem of the intervening basin which hinders the development of flood control, water resources assessment and a lot of subsequent works. There is an urgent need to establish a method for determining the water yielding amount of this region and managing rainfall-runoff over the whole basin. However, the rainfall runoff relationship and water balance analysis in Poyang Lake region is an unsolved problem so far. This study focus on the understanding of controls on the ungauged PLIB hydrologic response by relating that response to meteorological data and underlying surface attributes which are interpreted by remote sensing data. Relationships between them allow a simulation to be made of the hydrologic response in the PLIB, based on climatologic and hydrological data as well as topography, land cover and soil attributes of that basin. The VIC distributed hydrological model is selected to simulate daily runoff in the ungauged PLIB. 2.

Study area

Fig. 1 Geographical position of the Poyang Lake basin and control stations

Fig. 2 The sub-basins and observation stations in the ungauged Poyang Lake intervening basin

Located in the Jiangxi Province, Poyang Lake is the largest freshwater lake in China and constitutes a major hydrological subsystem of the middle Yangtze River basin. The lake is a tributary of the Yangtze River (as shown in Fig. 1) and directly exchanges and interacts with it [2]. The Poyang flood plain is subject to massive changes in water level. In rainy season the size of the lake grows up to 4000 km2 while in dry season the size of the lake shrinks to less than 1000 km2. The average depth is 8 m, and the maximum depth is 23 m. It extends in a flat depression at very low elevation, only a few ten meters above sea level, surrounded by mountains near the boundaries of Jiangxi province. The five major rivers (Xiushui, Ganjiang, Fuhe, Raohe and Xinjiang) flow into Poyang Lake. The Poyang Lake drainage area is only a narrow outlet into the Yangtze River which lies on the northern border of the province. In this study, the PLIB with an area of 25 915km2, refers to the range between Hukou outlet control station and seven inflow control stations (Qiujin, Wanjiabu, Waizhou, Lijiadu, Meigang, Dufengkeng and Hushan ) of these five rivers, as illustrated in Fig.1 and Fig. 2. The intervening basin regularly floods during the summer wet season and embankment on the low-lying land failures occurs along the lake boundaries and the

2741

2742

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

lower sections at down-streams of tributary rivers, resulting in extensive agricultural losses, damage to several cities and many agricultural villages, and massive population relocation. The Poyang Lake water level is determined primarily by the water surface elevation of the Yangtze River, and to a lesser extent by the runoff from the Jiangxi’s rivers [3]. The rainy season in Jiangxi usually begins in April. Typically, the Jiangxi’s rivers runoff increases from April to June, raising the level of water in Poyang Lake, which drains into the Yangtze River. From July to September, the Jiangxi’s rivers runoff decreases. However, at the same time the Yangtze River water level increases because of the summer concentration of rainfall and snowmelt in the mountainous headwaters region in western China. As a result, usually in mid July, the direction of water flow from the lake into the Yangtze River reverses and water begins to flow from the Yangtze River into the Poyang Lake. Maximum runoff from the Yangtze River typically occurs during the mid to late summer months. The most severe floods in the Poyang Lake region occur when a high runoff from the Jiangxi’s rivers occur later than normal in summer while the level of the Yangtze River is also high. For many years, the Poyang Lake area has been suffered from flood disaster in rain season and suffered from drought disaster in dry season. In the early of twenty first century, the flood and drought disasters have becoming more and more frequent. In addition to abnormal weather, as pointed out by [4], artificial effects as extreme deforestation and reclamation of lakes, which reduces the buffer storage capacity, are the main causes of the flooding. 3.

Estimation of daily runoff hydrograph

To avoid any confusion, the daily runoff hydrograph calculated in this section is defined as “estimated”, while the “simulated” refer to the output of the VIC model in the next section. Modification of daily runoff hydrograph at Hukou. The basin area analogue and water budget methods were used to modify observed runoff hydrograph at the Hukou outlet. The intervening basin runoff hydrograph consists of two parts, water-surface runoff and land-surface runoff. It is supposed that all the rainfall fallen on water surface of lake contributes to water-surface runoff by subtracting evaporation at the same synchronic time step. The accumulating area of four control stations, Hushan, Dufengkeng, Wangjiabu and Qiujin in the Poyang Lake region was selected as the analogical basin to obtain the daily land-surface runoff hydrograph and then modified by annual water balance factors. The observed inflows of seven inlet control stations together with water surface flow and land surface flow of PLIB were routed to the Hukou by considering different lag times as

QtHukou

WS QtWaizhou  QtLijiadu  QtMeigang  QtQiujin  QtWanjiabu  QtDufengkeng  QtHushan  QtLS 2 3 3 1 1 2 2 1  Qt .

(1) where

Hukou t

Q

is to the runoff at the time t ; the first seven items at the right-side are inflows of seven

control stations, respectively, the last two items are land surface runoff and water surface runoff in the PLIB, respectively. The annual runoff volume at Hukou control station was calculated by following two methods, one is based on water budget equation (QHukou,1) and the other is the sum of routed daily runoff of Hukou(QHukou,2). Annual precipitation and evaporation were calculated based on the mean value of seventy-two rainfall gauged stations and eight evaporation stations around the PLIB as shown in Fig.2 and annual runoff depth equals annual precipitation minus evaporation by assuming the variation of soil water content unchanged. The results of them during 1999~2009 were listed in Table 1. The ratio of two types of annual runoff volume is used to modify the routed daily runoff hydrograph since the annual water budget estimator is much more reliable. Annually Modified routed runoff hydrograph was integrated into the observed Hukou runoff data to

2743

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

form the ultimately estimated runoff hydrograph at Hukou control station according to the following two steps. Firstly, the modified routed hydrograph was only used to directly replace the controversial segments (such as the reverse parts) of the observed Hukou runoff hydrograph. Then the inevitable zigzag parts were smoothed manually based on water budget method. Multiple-input and single-output model. Before being used to derive the runoff hydrograph of PLIB, the reliability of modified runoff hydrograph at Hukou must be verified by the multiple-input and single-output (MISO) system model [5, 6]. Table 1 List of annual hydrological characteristics and runoff volumes QPLIB

Q7obs

QHukou,1

QHukou,2

(108m3)

(108m3)

(108m3)

(108m3)

0.705

370

1465

1835

1991

1.085

824

0.558

207

1123

1330

1439

1.083

178

1211

1389

1478

1.065 0.978

P

E

R

(mm)

(mm)

(mm)

1999

2093

618

1475

2000

1476

652

Year

Rc

Ratio

2001

1368

660

709

0.518

2002

1870

630

1240

0.663

311

1615

1926

1884

2003

1599

752

847

0.530

212

1088

1300

1474

1.134

2004

1375

699

676

0.491

169

733

902

985

1.092

2005

1543

686

857

0.555

215

1178

1393

1577

1.132

2006

1357

668

689

0.508

173

1307

1480

1564

1.056

2007

1137

673

464

0.408

116

815

931

1075

1.155

2008

1550

687

863

0.557

216

1016

1232

1395

1.132

542

0.425

136

833

969

1112

1.148

2009

1274

732

The MISO model presumes a simple linear relationship between various inputs x

(1)

, x ( 2) ,",

i.e.

the upstream tributary inflows and/or the average rainfall over the intervening basin, and output y , the runoff recorded at the gauging station. In discrete form, this model is expressed by the Eq. 2 m (1)

yi

¦x j 1

(l )

(1) (1) i  j 1 j

h

m ( 2)



( 2) ( 2) i  j 1 j

¦x j 1

h

m(l )

" ¦ x j 1

(l ) (l ) i  j 1 j

h

m( J )

"

¦x

(J ) (J ) i  j 1 j

h

 ei .

(2)

j 1

where h j is the j th ordinate of the (l ) th pulse response function relating input (l ) to a component in the output, J is the number of inputs, m(l ) is the memory length of the system corresponding to the (l ) th input series, and ei is the i th error term. The Nash-Sutcliffe efficiency ( R 2 ) and the relative error of the volumetric fit ( RE ) criteria were used to justify the performance of MISO model. Eleven-year of daily runoff data was divided into calibration period (1999~2005) and verification period (2006~2009). Eight inflowsˈincluding seven inflow control stations and average rainfall of PLIB, were taken as inputs; the modified runoff of the Hukou station as output, and the MISO model was solved by least square method, in which the memory lengths of inputs were determined by trail and error method. The results of calibration and verification of MISO model were listed in Table 2. It is shown that the model efficiencies are 99.42% and 98.40%, the relative errors are 0.41% and 1.77% for calibration and verification period, respectively. Fig. 3 shows that MISO can fit modified runoff hydrographs very well during verification period.

2744

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747 Table 2 The results of calibration and verification at the Hukou and PLIB

model MISO VIC

period

Num. of years

R 2 (%)

RE (%)

Calibration(1999-2005)

7

99.42

0.41

verification (2006-2009)

4

98.40

1.77

Calibration(1999-2005)

7

68.59

2.69

verification (2006-2009)

4

73.56

-6.53

18000 Estimated

16000

MISO simulated

3 -1

Discharge(m ·s )

14000 12000 10000 8000 6000 4000 2000 0 1-Jan-06 1-Jul-06

1-Jan-07 1-Jul-07

1-Jan-08 1-Jul-08 1-Jan-09 1-Jul-09

Time(Day-Month-Year) Fig. 3 The comparison of estimated and simulated runoff hydrograph at Hukou station during verification period 2006~2009

Estimating PLIB daily runoff hydrograph. Based on the Eq. 2, the output y is calculated by inputs of seven inlet control stations multiply with pulse response functions. The difference between modified runoff at the Hukou outlet station and output y is the runoff hydrograph of PLIB, whereas, some ordinates of hydrograph would be negative and need be modified. Two runoff hydrographs of PLIB have been estimated by basin area analogue and water budget methods and MISO model as discussed above. The former is better in hydrological process, while the later is more accurate in water budget. In order to eliminate the negative ordinates of the second hydrograph, the verified hydrograph equals to the mean of two runoff hydrograph with the corresponding ordinates. 4.

Simulating PLIB daily runoff by VIC distributed model

The Variable Infiltration Capacity (VIC) distributed model is a macro-scale hydrological model based on Soil-Vegetation-Atmosphere Transfer Scheme (SVATS), which was originally developed to describe the land surface in numerical weather prediction and climate, as well as to take account of the variation and transfer of water and energy. The model is designed for application at horizontal resolutions ranging from several to hundreds of miles, and has widely been tested and applied for runoff simulation study with good performance to various basins of different scales by many authors [7~11]. One of the advantages of this regular squared model is that it is well suited for using remote sensing data. In this paper, RS data was used to build the vegetation and soil input files and to be an auxiliary

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

means to delineate the boundary and to construct the routing network of PLIB. Establishing VIC model. The VIC model was used to simulate daily runoff of the ungauged PLIB. DEM, soil and land cover data etc are required for VIC model calibration. DEM data of 0.00083 degree (around 100m×100m cell size) spatial resolution for the intervening basin was derived and used to delineate the basin boundary and stream network, and the whole intervening basin was divided into a 1374-cell 5km×5km grid. However, it is difficult to automatically execute such operations by GIS soft wares in lake regions since the embedded D8 method alone fails in flat area or depressions. To avoid these problems, the digital river and lake network (DRLN) which use TM remote sensing data as input in addition to the DEM has been adopted to develop the river network of the PLIB [12]. The required soil texture classification was based on the information of global 5 min data of FAO (Food and Agriculture Organization). In the case of the FAO soil data including two layers, the top-layer of 0~30cm depth, as shown in Fig. 4(a), and the sub-layer of 30~100cm depth, as shown in Fig. 4(b), we defined that these two layers represent the thin layer, upper and lower layers of VIC model, respectively. The individual soil parameters, such as porosity T s , saturated soil potential\ s , saturated hydraulic conductivity Ks and exponent b, were derived based on the works of [13] and [14].

(a) top-layer

(b) sub-layer

Fig. 4 The soil data used in the intervening basin for VIC model

The land cover of the study area has been extracted from University of Maryland’s 1km Global Land Cover product. This product was sorted into fourteen kinds of vegetation patterns. As the Fig. 5 illustrates that the vegetation types within the PLIB are closed shrub-land, cropland, deciduous broadleaf forest, deciduous broadleaf forest, grassland, mixed cover, open shrub-land, urban and built-up, water body, wooded grass land and wood land. Vegetation parameters were determined by Land Data Assimilation System (LDAS) [15].

Fig. 5 The vegetation data used in the intervening

Fig. 6 The parameter distribution of the

2745

2746

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

basin for VIC model intervening basin for VIC model Since the VIC model was employed in water balance mode, the required forcing input data were daily minimum and maximum air temperature, and daily precipitation. The forcing data were obtained based on seventy-two rainfall stations and twelve national weather stations around the intervening basin. The data were mapped to the resolution of 5km×5km grids through Inverse Distance Weight (IDW) method. Parameterization procedures. The VIC model has six parameters that need to be calibrated for each cell. The final parameters type distribution was given in Fig.6. The main parameterization procedures for each cell were described as follows: (1) Identify the grid cells that cover the ungauged PLIB by integer numbers from 1 to 1374. Classify all the identified cells into four categories, i.e. A, B, C and D. A-cell is covered by the lake water surface, in which the runoff equals rainfall minus evaporation at the synchronic time step. (2) Two gauged sub-basins (Xi River and Boyang River as shown in Fig. 2) within the PLIB were selected to calibrate VIC model’s parameters. Assuming B-cell and C-cell underlying conditions were similar to that of the Xi River and Boyang River, therefore their VIC parameters could be represented by the calibrated parameter values, respectively. (3) The remaining cells are denoted as D-cell, and parameters of them should be optimized according to the objective function. Two sets of initial parameter values of D-cell were given, one was the average of values of A-cell and B-cell, and the other was the parameters similar to the lower Hanjiang River basin. The final parameter values were determined by the trail and error method.

Discharge(m3·s -1)

9000 8000

VIC Simulated

7000

Estimated

6000 5000 4000 3000 2000 1000 0 1-Apr

1-May

31-May

30-Jun

30-Jul

29-Aug

28-Sep

28-Oct

Time(Day-Month)

Fig. 7 Comparison of estimated and simulated runoff hydrographs of PLIB during 2000 flood season

Result discussion. The Nash-Sutcliffe efficiency ( R 2 ) and the relative error of the volumetric fit ( RE ) criteria were used to justify the performance of the VIC model. Daily runoff data of eleven-year was divided into calibration period (1999~2005) and verification period (2006~2009). Table 2 lists the results of VIC model in PLIB. The values of R 2 and RE are 68.59% and 73.56%, 2.69% and -6.53% in the calibration and verification periods, respectively. Fig. 7 shows the comparison between estimated and simulated runoff hydrographs of PLIB during 2000 flood season. It is shown that the VIC distributed model is able to simulate and predict daily runoff hydrograph of the intervening basin.

Jiali GUO et al. / Procedia Environmental Sciences 10 (2011) 2740 – 2747

5.

Conclusions

An approach combining basin area analog method, water budget method and multi-input, single-output system model was proposed to modify runoff hydrographs at the Hukou control station and estimate runoff in the PLIB. The VIC distributed hydrological model based on remote sensing data was established and used to simulate daily runoff of PLIB. Application results show that the proposed methods are able to simulate and predict runoff hydrograph reasonably well. It is proved that with the help of remote sensed data, important hydrological characteristics of the PLIB can be reproduced and predicted. 6.

Acknowledgement

This study is financially supported by the Ministry of Science and Technology (2009BAC56B02) and Ministry of Water Resources (200901001) of China. The authors would also like to thank the editor and anonymous reviewers for their reviews and valuable comments related to this manuscript. References [1]

R. Dubayah, D. Lettenmaier, E. F. Wood and J. Rhoads in: Remote Sensing and Hydrology 2000, edited by M. Owe, K.

Brubaker, J. Richchie and A. Rango chapter 3, IAHS Publication no.267 (2001). [2]

D. Shankman, B. D. Keim, and J. Song: Int. J. Climatol. Vol. 26(2006), p.1255

[3]

R. Andreoli, Y. Y sou, J. Li, Y-L. Desnos, H. Shifeng and De Fraipont P.: Annals of GIS Vol. 13 (2007), p.24

[4]

H. F. Yin, and C. G. Li: Geomorphology Vol.41 (2001), p.105

[5]

G. C. Liang, R. K. Kachroo, W. Kang and X. Z. Yu: J. Hydrol. Vol. 133 (1992), p.99

[6]

X. Li, S.L. Guo, P. Liu, and G.Y. Chen: J. Hydrol. Vol. 391 (2010), p.126

[7]

X. Liang, D. P. Lettenmaier, E. F. Wood and S. J. Burges: Journal of Geophysical Research Vol.99 (1994), p.415

[8]

Z. H. Xie, F. G. Su, X. Liang, Q. C. Zeng, Z. C. Hao and Y. F. Guo: Adv. Atmos. Sci. Vol.20 (2003), p.165

[9]

F. Yuan, Z. H. Xie, Q. Liu, H. W. Yang, F. G. Su, X. Liang and L. L. Ren: Canadian Journal of Remote Sensing Vol.

30 (2004), p. 680 [10]

J. Z. Guo, X. Liang, L. R. Leung: J. Hydrol. Vol.298 (2004), p.311

[11]

S. L. Guo, J. Guo, J. Zhang and H. Chen: Sci. China Ser. E Vol. 52 (2009), p. 3234

[12]

R. Turcotte, J. P. Fortin, A. N. Rousseau, S. Massicotte and J. P. Villeneuve: J. Hydrol. Vol.240 (2001), p.225

[13]

B.J. Cosby, G. M. Hornberger, R. B. Clapp and T. R. Ginn: Water Resour. Res. Vol. 20 (1984), p.682

[14]

W. J. Rawls, L. R. Ahuja, D. L. Brakensiek, A. Shirmohammadi, in: Infiltration and Soil Water Movement edited by

D. R. Maidment McGraw-Hill Inc. (1992) [15]

Information on http://www.geog.umd.edu/landcover/1km-map.html

2747