• Abdulla, F. A., , and Lettenmaier D. P. , 1997a: Development of regional parameter estimation equations for land surface hydrologic model. J. Hydrol., 197 , 230257.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abdulla, F. A., , and Lettenmaier D. P. , 1997b: Application of regional parameter estimation to simulate the water balance of large continental river. J. Hydrol., 197 , 258285.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abdulla, F. A., , Lettenmaier D. P. , , Wood E. F. , , and Smith J. A. , 1996: Application of a macroscale hydrologic model to estimate the water balance of the Arkansas-Red River basin. J. Geophys. Res., 101 , D3. 74497459.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1996: A Land Surface Model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user’s guide. NCAR Tech. Note NCAR/TN-417+STR, National Center for Atmospheric Research, 150 pp.

  • Chen, Z. K., 1985: China’s water resources and its utilization. GeoJournal, 10 , 167171.

  • Cosby, B. J., , Hornberger G. M. , , Clapp R. B. , , and Ginn T. R. , 1984: A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res., 20 , 682690.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Coauthors, 2003: The Common Land Model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dickinson, R. R., , Henderson-Sellers A. , , Kennedy P. J. , , and Wilson M. F. , 1986: Biosphere-Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-275+STR, National Center for Atmospheric Research, 69 pp.

  • Dickinson, R. R., , Henderson-Sellers A. , , and Kennedy P. J. , 1993: Biosphere-Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-387+STR, National Center for Atmospheric Research, 72 pp.

  • Duan, Q., , Sorooshian S. , , and Gupta V. K. , 1992: Effective and efficient global optimization for conceptual rainfall-runoff models. Water Resour. Res., 28 , 10151031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Coauthors, 2006: Model Parameter Estimation Experiment (MOPEX): An overview of science strategy and major results from the second and third workshops. J. Hydrol., 320 , 317.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • FAO, 1998: Digital soil map of the world and derived soil properties. Land Water Digital Media Series, Vol. 1, Food and Agriculture Organization, CD-ROM.

  • Hansen, M. C., , DeFries R. S. , , Townshend J. R G. , , and Sohlberg R. , 2000: Global land cover classification at 1 km spatial resolution using a classification tree approach. Int. J. Remote Sens., 21 , 13311364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang, M., , and Liang X. , 2003: A transferability study of model parameters for the Variable Infiltration Capacity land surface scheme. J. Geophys. Res., 108 .8864, doi:10.1029/2003JD003676.

    • Search Google Scholar
    • Export Citation
  • Hubert, B., , Francois L. , , Warnant P. , , and Strivay D. , 1998: Stochastic generation of meteorological variables and effects on global models of water and carbon cycles in vegetation and soils. J. Hydrol., 212–213 , 318334.

    • Search Google Scholar
    • Export Citation
  • Liang, X., , and Xie Z. , 2001: A new surface runoff parameterization with subgrid-scale soil heterogeneity for land surface models. Adv. Water Resour., 24 , 11731193.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liang, X., , Lettenmaier D. P. , , Wood E. F. , , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res., 99 , D7. 1441514428.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liang, X., , Lettenmaier D. P. , , and Wood E. F. , 1996: One-dimensional statistical dynamic representation of subgrid variability of precipitation in the two-layer Variable Infiltration Capacity model. J. Geophys. Res., 101 , D16. 2140321422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Maurer, E. P., , O’Donnell G. M. , , Lettenmaier D. P. , , and Roads J. O. , 2000: Evaluation of NCEP/NCAR reanalysis water and energy budgets using macroscale hydrological simulations as a benchmark. Observations and Modeling of the Land Surface Hydrological Processes, V. Lakshmi, J. Albertson, and J. Schaake, Eds., Amer. Geophys. Union, 137–158.

    • Search Google Scholar
    • Export Citation
  • Nash, J. E., , and Sutcliffe J. V. , 1970: River flow forecasting through conceptual models. Part I: A discussion of principles. J. Hydrol., 10 , 282290.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nijssen, B., , O’Donnell G. M. , , and Lettenmaier D. P. , 2001: Predicting the discharge of global rivers. J. Climate, 14 , 33073323.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rawls, W., , Ahuja L. R. , , Brakensiek D. L. , , and Shirmohammadi A. , 1993: Infiltration and soil water movement. Handbook of Hydrology, D. R. Maidment, Ed., McGraw-Hill, 5.1–5.51.

    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., , Mintz Y. , , Sud Y. C. , , and Dalcher A. , 1986: A Simple Biosphere Model (SiB) for use within general circulation models. J. Atmos. Sci., 43 , 505531.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Todini, E., 1996: The ARNO rainfall-runoff model. J. Hydrol., 175 , 339382.

  • Xia, J., , Heung W. , , and Wai C. I. , 2003: Water problems and sustainability in North China. Water Resources Systems—Water Availability and Global Changes, S. Franks et al., Eds., IAHS Press, 12–22.

    • Search Google Scholar
    • Export Citation
  • Xie, Z., , Su F. , , Liang X. , , Zeng Q. , , Hao Z. , , and Guo Y. , 2003: Applications of a surface runoff model with Horton and Dunne runoff for VIC. Adv. Atmos. Sci., 20 , 165172.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , Sellers P. J. , , Kinter J. L. , , and Shukla J. , 1991: A simplified biosphere model for global climate studies. J. Climate, 4 , 345364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yangtse River Conservancy Commission, 2002: Flood and Drought Disaster for Yangtse River Basin, China Flood and Drought Disaster Series. (in Chinese). China Water Conservancy and Water Electricity Press, 326 pp.

    • Search Google Scholar
    • Export Citation
  • Zeng, X., , Shaikh M. , , Dai Y. , , Dickinson R. E. , , and Myneni R. , 2002: Coupling of the Common Land Model to the NCAR Community Climate Model. J. Climate, 15 , 18321854.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    River basins and climate zones in China according to Table 2.

  • View in gallery

    The locations of the selected basins in China for calibration and verifications.

  • View in gallery

    The locations of the 740 meteorological stations in China.

  • View in gallery

    Monthly hydrographs of the observed and simulated flows (control case and calibrated) for the primary basins.

  • View in gallery

    Mean monthly hydrographs of the observed and simulated flows (control case and calibrated) for the primary basins.

  • View in gallery

    Schematic representation of the parameter regionalization scheme.

  • View in gallery

    Monthly hydrographs of the observed and simulated flows (control case, parameter transfer, and recalibrated) for secondary basins. Monthly hydrographs of the observed and simulated flows (control case, parameter transfer, and recalibrated) for secondary basins.

  • View in gallery

    Mean monthly hydrographs of the observed and simulated flows (control case and calibrated) for secondary basins.

  • View in gallery

    Mean monthly hydrographs of the observed and simulated flows for secondary basins.

  • View in gallery

    Annual mean distribution of (a) simulated runoff in the control case; (b) simulated runoff in the transfer case; (c) difference of simulated runoff [(b)−(a)]; and (d) precipitation.

  • View in gallery

    Comparison of the simulated runoff in control case and transfer case for climate zones in China.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 152 152 28
PDF Downloads 108 108 10

Regional Parameter Estimation of the VIC Land Surface Model: Methodology and Application to River Basins in China

View More View Less
  • 1 Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
  • | 2 Lawrence Livermore National Laboratory, University of California, Livermore, California
  • | 3 Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
© Get Permissions
Full access

Abstract

This paper presents a methodology for regional parameter estimation of the three-layer Variable Infiltration Capacity (VIC-3L) land surface model with the goal of improving the streamflow simulation for river basins in China. This methodology is designed to obtain model parameter estimates from a limited number of calibrated basins and then regionalize them to uncalibrated basins based on climate characteristics and large river basin domains, and ultimately to continental China. Fourteen basins from different climatic zones and large river basins were chosen for model calibration. For each of these basins, seven runoff-related model parameters were calibrated using a systematic manual calibration approach. These calibrated parameters were then transferred within the climate and large river basin zones or climatic zones to the uncalibrated basins. To test the efficiency of the parameter regionalization method, a verification study was conducted on 19 independent river basins in China. Overall, the regionalized parameters, when evaluated against the a priori parameter estimates, were able to reduce the model bias by 0.4%–249.8% and relative root-mean-squared error by 0.2%–119.1% and increase the Nash–Sutcliffe efficiency of the streamflow simulation by 1.9%–31.7% for most of the tested basins. The transferred parameters were then used to perform a hydrological simulation over all of China so as to test the applicability of the regionalized parameters on a continental scale. The continental simulation results agree well with the observations at regional scales, indicating that the tested regionalization method is a promising scheme for parameter estimation for ungauged basins in China.

Corresponding author address: Zhenghui Xie, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China. Email: zxie@lasg.iap.ac.cn

Abstract

This paper presents a methodology for regional parameter estimation of the three-layer Variable Infiltration Capacity (VIC-3L) land surface model with the goal of improving the streamflow simulation for river basins in China. This methodology is designed to obtain model parameter estimates from a limited number of calibrated basins and then regionalize them to uncalibrated basins based on climate characteristics and large river basin domains, and ultimately to continental China. Fourteen basins from different climatic zones and large river basins were chosen for model calibration. For each of these basins, seven runoff-related model parameters were calibrated using a systematic manual calibration approach. These calibrated parameters were then transferred within the climate and large river basin zones or climatic zones to the uncalibrated basins. To test the efficiency of the parameter regionalization method, a verification study was conducted on 19 independent river basins in China. Overall, the regionalized parameters, when evaluated against the a priori parameter estimates, were able to reduce the model bias by 0.4%–249.8% and relative root-mean-squared error by 0.2%–119.1% and increase the Nash–Sutcliffe efficiency of the streamflow simulation by 1.9%–31.7% for most of the tested basins. The transferred parameters were then used to perform a hydrological simulation over all of China so as to test the applicability of the regionalized parameters on a continental scale. The continental simulation results agree well with the observations at regional scales, indicating that the tested regionalization method is a promising scheme for parameter estimation for ungauged basins in China.

Corresponding author address: Zhenghui Xie, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China. Email: zxie@lasg.iap.ac.cn

1. Introduction

The land component of climate models represents many processes that control exchanges of energy, momentum, and water between soil, vegetation, and the atmosphere, and has long been recognized as important for hydrological forecasting and climate change studies (Dickinson et al. 1993; Bonan 1996; Zeng et al. 2002; Dai et al. 2003). To apply a land surface model to a specific region, the model parameters for the region must be determined a priori (Duan et al. 1992, 1996). Studies have shown that land surface models can perform well if their model parameters are appropriately estimated on the basis of calibration with observations, but they perform poorly if their model parameters are not assigned properly (Huang and Liang 2003). Therefore, a practical parameter estimation method, especially for data-sparse areas, is critical to hydrological modeling. Most land surface models (LSMs) are a hybrid of physically based representations and conceptual components, such as Variable Infiltration Capacity (VIC; Liang et al. 1994, 1996; Liang and Xie 2001), the Biosphere–Atmosphere Transfer Scheme (Dickinson et al. 1986), LSM (Bonan 1996), Simple Biosphere model (Sellers et al. 1986; Xue et al. 1991), and the Common Land Model (Dai et al. 2003). The runoff generation mechanisms in most LSMs tend to be conceptual in nature, and the energy exchange at the atmosphere–land surface interface is usually based on physical principles. The parameters for the physically based part of the model are often determined based on land surface characteristics such as soil and vegetation classes. Most LSMs use a lookup table approach in which land surface attributes are kept constant within each land surface class, and vary only between different classes (Nijssen et al. 2001). This simplifies the application and greatly reduces the number of parameters that need to be specified. The parameters for the conceptual part of the model can be calibrated against observed streamflow data for basins with streamflow observations. For basins without streamflow observations, parameters have to be estimated from other sources of information, such as neighboring basins, or taken from tabulated values from the literature, or otherwise assumed based on expert judgment. The choice of basins from which information is to be transferred is usually based on some sort of similarity measure; that is, one tends to choose those basins that are most similar to the site of interest. One common similarity measure is spatial proximity. This is based on the premise that basins that are close to each other will have a similar runoff regime, as climate and basin conditions will only vary smoothly in space. An example of such an approach was given by Nijssen et al. (2001), in which a methodology of parameter transfer is developed to reduce the number of basins requiring direct calibration. Another alternative similarity measure is the use of basin attributes such as land use, soil type, and topographical characteristics. For instance, Abdulla et al. (1996) and Abdulla and Lettenmaier (1997a, b) used a family of regression equations that relate the two-layer VIC (VIC-2L) model parameters to local land surface and climatological characteristics to derive the spatial distribution of vegetation and hydrological parameters over the Arkansas–Red River basin and obtained good simulation results. In principle, one would assume that the model parameters are closely related to basin attributes, as the model parameters are designed to represent the functional behavior of basin response which, in turn, should be controlled by the physical characteristics of the basins such as land use and land cover.

To simulate streamflow for river basins in China, this paper presents a methodology for regional parameter estimation for the LSM called the three-layer VIC (VIC-3L) model. This methodology is based on combination of climatic zone and large river basins and designed to obtain estimates of runoff-related parameters from a limited number of calibrated basins and regionalize them to uncalibrated basins. Thirty-three basins from different climatic zones in China were selected for model calibration and verification, among which 14 basins from different climatic zones were used for calibration and 19 basins from different climatic zones were used for verification. Seven runoff-related parameters of the VIC-3L model were calibrated for each of the 14 basins and the calibrated parameters were then transferred (i.e., regionalized) to uncalibrated basins based on climate and large river basin characteristics of the land area in China. Hydrological simulations for the 19 independent basins mentioned above using both the regionalized parameters and the default a priori parameters were conducted and analyzed to test the efficiency of the parameter regionalization method. To verify the applicability of the regionalized parameters on a continental scale, the calibrated model parameters were transferred to the entire continental China based on climate and large river basin zones. Hydrologic simulations over the entire domain were then performed. The space grid and time step of VIC model applications in this study are at a 50 × 50 km2 grid and 24-h time step, respectively.

In comparison with previous works such as Abdulla et al. (1996), Abdulla and Lettenmaier (1997a, b), and Nijssen et al. (2001), this paper is the first study on hydrological parameter regionalization conducted over continental China with more accurate hydrologic data than those used in the global studies. Further, the regionalization scheme based on combined climatic characteristics and basin domains is very important, especially for river basins in China, for which hydrological data are generally sparse and often hard to obtain for most areas. In addition, the regionalization scheme is of potential global portability. This scheme could help general circulation models to reasonably represent land surface processes, as quite a few basins studied in given climatic zones in China have similarities with basins with very sparse site data in other parts of the world, which is in good need of parameter transfer methods.

This paper is organized as follows. Section 2 briefly describes the VIC-3L model structure and parameters. The study domain and datasets are then presented in section 3. Section 4 discusses model calibration methodology and calibration results. Section 5 explains the parameter transfer scheme and verifications results. Finally, section 6 provides a summary and conclusions of this study.

2. Model description

a. Model structure

The VIC-3L land surface model is a soil vegetation atmospheric transfer scheme that considers both energy and water balances. It includes a top thin soil layer to represent quick bare soil evaporation following small rainfall events, a middle soil layer to represent the dynamic response of the soil to rainfall events, and a lower layer to characterize the seasonal soil moisture behavior (Liang et al. 1994, 1996; Liang and Xie 2001). VIC-3L explicitly represents the effects of multiple vegetative covers on water and energy budgets. It uses physically based formulations for the calculation of the sensible and latent heat fluxes, and the conceptual ARNO formulation for base flow (Todini 1996) to simulate runoff generation from the deepest soil layer. It also uses the conceptual surface runoff model with the Philip infiltration formulation that dynamically represents both the saturation and infiltration excess runoff processes in a model grid cell with consideration of subgrid-scale soil heterogeneity (Liang and Xie 2001; Xie et al. 2003) to simulate runoff generation from the upper two soil layers. For a detailed description of the VIC-3L in this paper, readers are referred to Liang and Xie (2001) and Xie et al. (2003).

b. VIC-3L model parameters

The VIC-3L model assigns many of its parameters based on two primary types of land surface variables: vegetation type and soil texture. Table 1 provides a list of VIC-3L model parameters and their estimation methods. Vegetation-related parameters such as architectural resistance, minimum stomata resistance, leaf-area index, albedo, roughness length, zero-plane displacement, and fraction of root depth of each soil layer based on the University of Maryland’s (UMD) land cover classification (Hansen et al. 2000; http://www.geog.umd.edu/landcover/1km-map.html) were estimated according to the Land Data Assimilation Systems developed by the National Aeronautics and Space Administration (http://ldas.gsfc.nasa.gov/). Soil texture information was derived from the 5-min Food and Agriculture Organization (FAO) dataset (FAO 1998). The soil parameters fall into two general categories. The first category is estimated based on FAO soil texture maps. These are related to soil characteristics such as porosity θs, saturated soil potential ψs, saturated hydraulic conductivity Ks, and the exponent c of the unsaturated hydraulic conductivity curve, which is based on Cosby et al. (1984) and Rawls et al. (1993). The second category of soil parameters is subject to calibration based on the agreement between simulated and observed hydrographs. These include the infiltration parameter b, which controls the amount of water that can infiltrate into the soil; the three soil-layer thicknesses di (i = 1, 2, 3), which affect the maximum storage available for transpiration; the three parameters in the base flow scheme including the maximum velocity of base flow Dm, the fraction of maximum base flow Ds, and the fraction of maximum soil moisture content of the third layer Ws at which a nonlinear base flow response is initiated, which determines how quickly the water stored in the third layer is depleted. These parameters are calibrated here because it is difficult to determine average soil properties over a large area, and soil depths are generally not well known over large areas. The ARNO base flow model routing parameters only determine how quickly the water stored in the third layer is depleted. The transport of water through the second layer is largely determined by the parameters selected for calibration, because the VIC model calculates the hydraulic conductivity following Cosby et al. (1984). The soil moisture storage is dynamically determined by the model. A change in the thickness of the second layer affects not only the hydraulic conductivity, but also the maximum storage available in the second layer and consequently the water available for transpiration.

3. Description of the study domain and datasets

a. Study domain

We selected the entire continental China as the spatial domain for this study, with special focus on 33 basins selected from six distinct climatic zones, as defined based on the Köppen classification rules (see Table 2) (Hubert et al. 1998). The six climatic zones and nine major river basins in China are plotted in Fig. 1. Located in the east of the Asian continent and on the western shore of the Pacific Ocean, China has a land area of about 9.6 million km2. From north to south, the territory of China stretches from the center of the Heilong River north of the town of Mohe with a latitude of a little more than 55°N to the Zengmu Reef at the southernmost tip of the Nansha Islands with a latitude of about 4°N, covering a distance of 5500 km. From east to west, the nation extends from the confluence of the Heilong and Wusuli Rivers with a longitude of about 73°E to the Pamirs with a longitude of a little more than 135°E, covering a distance of 5200 km. China’s topography is formed around the emergence of the Qinghai–Tibetan Plateau, which rises continuously to become the “roof of the world,” averaging more than 4000 m above sea level. The terrain in China then gradually descends from west to east like a staircase. Most of China has a continental monsoon climate. From September to April, dry and cold winter air masses blow from Siberia and the Mongolian Plateau, resulting in cold and dry winters and great differences in the temperatures between north and south China. From April to September, warm and humid summer monsoons blow from the seas in the east and south, resulting in overall high temperatures and plentiful rainfall, and little difference in the temperatures from north to south. In terms of temperature, the nation can be sectored from south to north into equatorial, tropical, subtropical, warm–temperate, temperate, and cold–temperate zones. Precipitation gradually declines from the southeastern to the northwestern inland area, and the average annual precipitation varies greatly from place to place. In southeastern coastal areas, it reaches over 1500 mm, while in the northwestern areas, it drops to below 200 mm.

For calibration and verification, we have chosen 33 basins, with drainage areas ranging from 1683 to 1 010 000 km2, from within the Yellow, Haihe, Yangtze, Heihe, Songhuajiang, and Pearl River basins. These basins are located in different climate zones, as shown in Fig. 2. They were partitioned into two groups. The primary 14 basins, shown in Table 3, were used for calibration as described in section 4, and the 19 secondary basins, shown in Table 4, are used for verification as described in section 5. Tables 3 and 4 also provide basin descriptive information and annual statistics of precipitation, runoff, and temperature for the two basin groups, respectively. As shown in Tables 3 and 4, the Qinan, Nanhechuan, Yanjiaping, and Heishiguan basins are located in the middle reaches of the Yellow River, where the annual mean temperature is about 8°–14°C, the annual mean precipitation is 466 mm, and the runoff depth is 73.0 mm, belonging to the arid and semiarid climate zones. The Xiahui, Xiabao, Xiangshuibao, and Luanxian basins are located in the Haihe River basin, where the annual mean precipitation is 400–500 mm, with 75%–85% of the rainfall occurring in the flooding season. The Haihe River basin is located in north China with a tremendous conflict in the water supply and demand (Chen 1985; Xia et al. 2003). Extensive anthropogenic water withdrawal from river channels and underground for agricultural and industrial production influences the natural hydrological cycle in the Haihe River basin greatly. The Wuhouzhen, Madao, Herong, Xindianpu, Hanzhong, Chadianzi, Yuxiakou, Guotan, Jian, Yichang, and Gongtan basins are located in the Yangtze River basin, which is the longest river in China with the most abundant water volume. The Yangtze River basin occupies 18.75% of China’s area, with bulk water volume at about 9.560 × 1013 m3 yr−1 (Yangtse River Conservancy Commission 2002). The Xixian, Bantai, and Luohe basins are located in the Huaihe River basin, where the annual mean temperature is about 11°–16°C and the annual mean precipitation is 700–1300 mm. Rainfall is variable in the Huaihe River basin, with the maximum annual precipitation being 3–4 times the minimum one. The Zhamashike and Yingluoxia basins are located in the upper reach of the Heihe River, which is the second longest innercontinental river in northeast China, where the annual mean precipitation is about 350 mm and the annual mean temperature is below 2°C. The Nianzishan, Changjiangtun, Lanxi, and Yixin basins are located in the Songhuajiang River basin in northeast China. The annual mean temperature of the Songhuajiang River basin is about 1°–5°C and the annual mean precipitation is about 300–900 mm with great spatial and temporal variability. The Heyuan, Shijiao, Hengshi, Nanning, and Boluo basins are located in the Pearl River basin, where the annual mean temperature is about 14°–22°C and the annual mean precipitation is about 1200–2200 mm.

b. Streamflow data

Streamflow data for the 33 basins mentioned above were obtained from the Chinese Ministry of Water Resources. These data for the 25 basins without the asterisk marker shown in Tables 3 and 4 satisfy the following requirements: 1) Observed monthly flows are available for at least part of the period 1980–2001; 2) the basins are only minimally affected by extractions, diversions, and dams; 3) the basins cover a variety of the main large river basins in China and the climatic zones. For the eight basins with the asterisk marker shown in Table 4, Hengshi, Boluo, Nanning, Jian, Yichang, Gongtan, Heishiguan, and Luanxian, only mean monthly streamflow data are available. The annual mean runoff depths for all basins were calculated from the observed streamflow, and these are shown in Tables 3 and 4. In general, runoff depth is lower for the arid region and higher for the humid. Runoff of the basins located in the Yellow and Haihe River basins is lower, ranging from merely 17.2 to 152.4 mm yr−1. However, the Heihe River and Songhuajiang River basins are arid, but the local temperature is low, so these contribute a small amount of evaporation, and the runoff of these river basins is a little higher than that of the Yellow and Haihe River basins. Most of the basins in the humid region are abundant in runoff production, ranging from about 189.7 to 1054.3 mm yr−1, especially for the basins in the Pearl River basin, which has the most ample rainfall in China.

c. Meteorological forcing data

The meteorological data consist of daily time series of precipitation (for the period 1980 through 2001), maximum temperature, and minimum temperature to drive the VIC-3L model with water balance. Solar radiation in the model is calculated according to latitude and Julian day (Rawls et al. 1993), and vapor pressure, atmospheric pressure, downward shortwave radiation, and longwave radiation are calculated according to the daily maximum and minimum temperatures. Precipitation and air temperature data were obtained by interpolating station values from 740 meteorological stations in China, which are shown in Fig. 3. We used the linear interpolation weighted by the inverse squared distances between the rain gauges and the grid cells. The interpolation scheme is given as follows: If there are n (≥1) meteorological stations located in a study grid cell, the meteorological variable Zx,y for the study grid cell is calculated according to the following formula:
i1525-7541-8-3-447-e1
where Zi and di (i = 1, . . . , n) represent the meteorological variable for the ith station located in the study grid cell and the distance between the location of the ith meteorological station and the central coordinates of the study grid cell, respectively, and n represents the total number of stations located in the study grid cell. If there is no station located in the study grid cell, the grid cell is extended by including neighboring grid cells until there are at least three stations included in the extended area. The meteorological variable Zx,y for the study grid cell is then calculated according to the Eq. (1). However, when there is a station close to the central point of the study grid cell, the meteorological variable Zx,y for the study grid cell is set according to the observed value at the station.

d. Vegetation dataset

Vegetation types and their cover fraction at the 50 × 50 km2 resolution in China were obtained from the UMD land cover data at a 1 × 1 km2 global resolution, which has a total of 14 different land cover classes labeled by integers ranging from 0 to 13 (Hansen et al. 2000). In this study, only 11 types (indexed from 1 to 11) are considered; water body (0), bare ground (12), and urban and built-up areas (13) are excluded here. For each 50 × 50 km2 grid cell, the fraction of each vegetation type is calculated according to how many image pixels of the vegetation type are located in the grid cell. For example, suppose that the UMD evergreen–needleleaf forest vegetation type makes up 20% of the image pixels (2500). In this case, the vegetation fraction would be 0.2. In this study, only those vegetation types whose proportions over the computational grid cell are greater than 10% are involved in computing the water and energy balances. For each UMD vegetation type, the four largest vegetation fractions of each vegetation type at the given 50 × 50 km2 grid cell are kept, while the others are excluded. These four fractions are scaled up so that they sum up to 1 and become “final” vegetation fractions. For example, assume that the vegetation fractions associated with a UMD vegetation type in a given 50 × 50 km2 grid cell were 0.40, 0.20, 0.15, 0.15, and 0.10. In this case, the value 0.10 would be excluded, and the other values would be scaled up such that they became 0.44, 0.22, 0.17, and 0.17. For each type of vegetation, the vegetation parameters such as architectural resistance, minimum stomata resistance, leaf-area index, albedo, roughness length, zero-plane displacement, and rooting depth and fraction for each soil layer are specified. The vegetation parameters used in VIC-3L for different vegetation classes are presented in Table 5.

e. Soil dataset

The soil texture information is derived from the 5-min Food and Agriculture Organization dataset (FAO 1998). The classification for the upper 0–30-cm soil depth represents the whole soil layer in one grid cell. Soil parameters such as porosity θs, saturated soil potential ψs, saturated hydraulic conductivity Ks, and the exponent c of the unsaturated hydraulic conductivity curve were determined from FAO data. Table 6 shows the soil-related parameters in VIC-3L. The FAO soil data have 16 classification types indexed by integers ranging from 1 to 16. In this study, only 12 types (indexed from 1 to 12) are used, and organic materials (13), water (14), bedrock (15) and other materials (16) are not taken into account. For each 50 × 50 km2 grid cell, the fraction of each soil type is calculated according to how many image pixels of the soil type are located in the grid cell, as it was done above for vegetation in section 3d. In this study, we use the parameter values of the soil type with the highest proportion over the study grid cell as the parameters for the whole computational grid.

4. Model calibration

a. Parameter calibration scheme

Fourteen basins were selected as the primary basins to implement model calibration, which is achieved by matching the total annual flow volume and the shape of the mean monthly hydrograph to corresponding observations. In the calibration study, all primary basins were grouped by climate and large river basin zones and the parameters of all basins within the same climate and large river basin zone were assumed to have the same values.

During the calibration process, the infiltration parameter b and the depths of the three soil layers (d1, d2, and d3), which were treated as the primary calibration parameters, were changed to a uniform set of values in a given climate and large river basin zone. Calibrations were performed according to the following procedure: (a) set the estimated values for the depths of the three soil layers (d1, d2, and d3), commonly with deeper depths for arid and semiarid regions and lower depths for humid regions; (b) calibrate the ARNO model parameters (Dm, Ds, and Ws) so as to fit the low flow; (c) adjust the infiltration parameter b to match the observed flow peaks, with a higher value to increase the peak and a lower value to decrease the peak; (d) make a fine adjustment on these parameters to get the best simulation results. Consequently, after calibration the texture-based soil hydraulic parameters (ks and θs) varied spatially, while b, di (i = 1, 2, 3), Dm, Ds, and Ws were constant within each region. Generally, the thickness of the second soil layer was increased to allow for more storage. The calibrated infiltration parameter b tended to be smallest in the arid climates, in an effort to reduce runoff production.

Two criteria were selected for model calibration as follows:

  • 1) The relative error (Bias; %) between simulated and observed mean annual runoff, which reflects the error of the total annual flow volume:
    i1525-7541-8-3-447-e2
    where Qc and Qo are the simulated and observed mean annual runoffs (mm), respectively.
  • 2) The Nash–Sutcliffe coefficient (CE; Nash and Sutcliffe 1970), which describes the matching extent of the hydrograph between the simulated and observed values:
    i1525-7541-8-3-447-e3
    where Qi,o is the observed streamflow (m3 s−1), Qi,c is the simulated streamflow (m3 s−1), and Qo is the mean observed streamflow (m3 s−1).

The calibrated parameters, their typical ranges, and the effect of each parameter on simulated hydrograph are given in Table 7 and are briefly described below: 1) Dm, typically ranging from 0 to 30 mm day−1, is the maximum base flow from the lowest soil layer; 2) Ds, typically ranging from 0 to 1, is the fraction of Dm where nonlinear (rapidly increasing) base flow begins. With a higher value of Ds, the base flow will be higher at lower water content in the lowest soil layer; 3) Ws, typically ranging from 0 to 1, with DsWs, is a fraction of the maximum soil moisture (of the lowest soil layer) where nonlinear base flow occurs. This is analogous to Ds. A higher value of Ws will raise the water content required for rapidly increasing, nonlinear base flow, which will tend to delay runoff peaks; 4) b, typically ranging from 0 to 8, defines the shape of the variable infiltration capacity curve. It describes the amount of available infiltration capacity as a function of the relative saturated grid cell area and controls the quantity of runoff generation directly and the water balance. A higher value of b gives lower infiltration and yields higher surface runoff; 5) d1, d2, and d3, ranging typically from 0.1 to 2.0 m. Soil depths d1 and d2 have a great effect on many model variables. In general, for runoff generation, thicker soil depths slow down seasonal peak flows and increase the loss due to evapotranspiration.

b. Calibration results

We have calibrated seven VIC-3L model parameters for the 14 primary basins using the procedure described in section 4a. Table 7 shows the calibrated parameter values as well as the default parameter values. The hydrologic simulations using the calibrated parameters were compared to those using the default parameter set for VIC (the control case).

Figure 4 shows the observed and simulated monthly streamflow time series for the 10 primary basins for the control case (without calibration) and for the calibrated case. Figure 5 shows the observed and simulated mean monthly hydrographs for the 14 primary basins using those two parameter sets. The model performance was considerably better when using the calibrated parameters than those using default parameters. The model in the control case overestimated the streamflows for the Qinan, Nanhechuan, Xiahui, Xiabao, and Xindianpu basins and underestimated the streamflows for the Wuhouzhen, Madao, Herong, Zhamashike, and Changjiangtun basins, but the simulated streamflows using the calibrated parameters matched the observed values well.

Table 8 lists the statistical results for the 14 primary basins using control case parameters and calibrated parameters. In terms of CE and the relative root-mean-squared error (RRMSE), the model in the control case and in the calibration case provided good simulation results for the Wuhouzhen, Madao, Xixian and Bantai basins, but the simulated streamflows in the calibration case were much closer to the observed ones as compared with the simulated results in the control case. In general, the calibration improved the results for all instances, although in some basins the final calibration was still unsatisfactory, especially for arid basins such as the Haihe River, which flows through a region with strong human activity. Calibration reduced the mean bias from to 71.0% to 7.3% and the mean RRMSEs from 30.5% to 7.9%. After calibration for all the basins, the CEs were all higher than 60% except for the Xiabao station (21.8%) in the Haihe River basin. In addition, the model simulations in dry season in some of basins are unsatisfactory. Generally speaking, these results indicate that the VIC-3L model with properly calibrated parameters provides better streamflow simulations for the primary basins than with default parameters. In section 5, we will demonstrate that these calibrated parameters can be transferred to the secondary basins and ultimately to the entire continental China with reasonably good results.

5. Parameter transfer

a. Parameter transfer scheme

As shown in Fig. 6, for a large study area, climatic zones are first classified based on climatic forcing data, and also are divided into large river basin areas. A limited number of basins from those climate and river basin zones are selected for calibration as described in section 4. The calibrated parameters are used to represent the hydrologic characteristics for the climate and large river basin zones and then regionalized to uncalibrated basins based on climate characteristics and large river basin subareas.

In this work, the parameter regionalization scheme mentioned above is implemented for river basins in China. Continental China was grouped into six distinct climate zones as shown in Fig. 1, which include tropical climate, dry and cold climate, rainy and midlatitude climate, continental climate with hot summer, continental climate with cool summer, and continental climate with short cool summer. It can be seen from Fig. 1 that the Köppen climate zones, to a great extent, reflect the actual climatic characteristics in China with southern and eastern China being humid and northern and western China being semiarid or arid. The calibrated infiltration parameter (b) and the depths of the three soil layers (d1, d2, and d3), and the ARNO model parameters (Dm, Ds, and Ws) were transferred to the uncalibrated river basins in China according to climate and large river basin zones.

The parameter transfer process is described in detail as follows:

The transferred parameters for the climate and large river basin zones in China are given in Table 9.

b. Parameter transfer verification results

To test the parameter transfer scheme, we conducted a verification study on the 19 independent basins in China mentioned in section 3a. Eleven of the secondary basins were further calibrated before transferring the parameters from all of the calibrated basins to the remaining land grid cells.

Figure 7 shows the observed and the simulated monthly streamflow time series for the eight secondary basins in the control case, parameter transfer, and recalibration. Figure 8 shows the observed and the simulated mean monthly hydrographs for the 11 secondary basins in the three cases. Figure 9 shows the observed and the simulated mean monthly hydrographs for the eight secondary basins in the parameter transfer case. The model performance was better when using the transferred parameters than those using default parameters. The simulated streamflows using the transferred parameters matched the observed values well.

Table 10 shows the calibration and parameter transfer statistics for the 11 secondary basins. Compared with the no calibration case, the parameter transfer process improved the simulated flow volume in eight basins with the mean value of absolute bias being reduced from 88.4% to 18.8%, and worsened the simulated volume in three basins (Luohe, Yinghuoxia, and Lanxi) with the absolute value of the bias being increased from 14.8% to 18.0%, 10.9% to 73.6%, and 39.5% to 42.1%, respectively. The transferred parameters reduced the RRMSEs for all the basins except the Luohe, Yingluoxia, and Lanxi basins, and increased all of the CEs other than those of the Luohe and Lanxi basins. The runoff simulation for the Yanjiaping basin without calibration was unsuccessful, with the CE being −649.9% and the RRMSE being 47.8%. However, in the case using transferred parameters, the modeling performance for the Yangjiaping basin was largely improved; the CE and the RRMSE became 84.3% and 6.9%, with an improvement of 731.2% and 40.9%, respectively. Compared with the simulations without calibration, the CEs of the seven basins (Hanzhong, Chadianzi, Yuxiakou, Guotan, Yingluoxia, Yixin, and Shijiao) using transferred parameters were improved by 23.3%, 20.6%, 9.5%, 2.2%, 1.9%, 12.0%, and 4.8%, respectively, and the RRMSEs of the six basins (Yangjiaping, Hanzhong, Chadianzi, Yuxiakou, Guotan, Yixin, and Shijiao) using the transferred parameters were decreased by 40.9%, 8.9%, 9.3%, 1.5%, 0.2%, 1.2%, and 0.7%, respectively. As a whole, the parameter transfer scheme improved the streamflow simulations for most of the secondary basins.

Subsequent recalibration of all basins further enhanced the modeling performance. Although the Xiangshuibao basin in the Haihe River basin is involved in intense human activity where runoff simulation is a difficult task, its simulation was still improved in the recalibration case. For the 11 secondary basins, the recalibrated model reduced the averaged RRMSE from 11.5% to 8.1% and the average absolute value of bias from 25.8% to 7.8%, and increased the mean CE from 80.1% to 84.7%, where the CE of the Xiangshuibao basin is not in the statistic. Compared with the case using transferred parameters, the improvements of the results using the recalibrated parameters were not remarkable: only 3.4% in RRMSE and 4.6% in CE. This phenomenon implies that the parameter transfer scheme is very successful when applied in the study regions, and this method can be potentially used in hydrological modeling for the ungauged basins in China.

c. Testing of the regionalized VIC parameters on continental China

As described above, the calibrated model parameters were transferred to the whole of continental China based on climate and large river basin zones. The regionalized model parameters were applied to runoff simulation over continental China.

Figures 10a–d show the annual mean simulated runoff distribution for the river basins in China without parameter calibration, with parameter transfer, their differences, and the annual mean distribution of precipitation, respectively. It can be seen from Figs. 10a and 10b that both cases (that without calibration and that with parameter transfer) indicate a similar annual runoff distribution over the whole area of China. In both cases, the annual mean runoff over continental China tends to increase from the north to the south and from the west to the east. The least amount of annual runoff (less than 50 mm) occurs in the northwestern China, such as Sinkiang and the west of Inner Mongolia, where precipitation is scarce and most of land is covered by desert. Because of plentiful rainfall, as show in Fig. 10d, southeastern China is abundant in runoff, generally more than 600 mm, and the most ample runoff appears in the Hainan Island and the southern Guangdong Province, where annual mean runoff is more than 1000 mm. The runoff depth of the central and northern China falls between that of southeastern China and that of northeastern China, being around 100 to 400 mm yr−1. In a word, the spatial distribution of precipitation determines the spatial distribution of runoff depth to some extent, where more rainfall produces more runoff and less rainfall generates less runoff. Figure 10c shows that there exist differences between the calculated annual mean runoff distribution without calibration and that with transferred parameters. Compared with the case with no calibration, the VIC simulation with transferred parameter has the tendency of overestimating annual runoff in the region of southern and eastern China and underestimating annual runoff in northern and western China.

Figure 11 shows the annual precipitation, runoff for the control case, that for the parameter transfer case, and observed runoff for each climate zone in China. For dry, cold; midlatitude, rainy; continental, short cool summer; and tropical climate zones, the simulated runoff in the parameter transfer case increases by about 13, 115, 101, and 184 mm, respectively, when compared with the control case. While for continental hot summer and continental short summer climate zones, the simulated runoff using the transferred parameters is reduced by 36 and 41 mm, respectively, when compared with the control case. It was shown that the model performance was better when using the transferred parameters than those using default parameters for those climate zones except the tropical zone. The differences of the simulated runoff for the two cases can be attributed to the following two reasons. The first reason is that the infiltration parameter b describes the amount of available infiltration capacity as a function of the relative saturated grid cell area and controls the quantity of runoff generation directly and the water balance. A higher value for b gives lower infiltration and yields higher surface runoff. In the case where no parameter calibration is performed, the default value of b, equal to 0.3, is set for all the grids covering the whole area of China, while in the case of parameter transfer, the value of b is set according to the climate and large river basin or climate zones. Most areas of the western and northern China belong to the continental climate zone with hot summer and the continental climate zone with cool summer, where b is assigned as 0.048 and 0.13, respectively. The smaller b in the parameter transfer case can lead to lower runoff over western and northern China compared with the no calibration case. Southern and eastern China is characterized by the rainy, midlatitude climate and the tropical climate, where b is set as 0.4 to 2.175 for the former climate and 0.5 for the latter—a little larger than the default value. Therefore, when using the transferred parameters, the bigger b can result in more runoff in the southern and eastern China than that simulated using default parameters.

The second reason is that the depths of the first two soil layers (d1 and d2) also have a great effect on runoff generation. Thicker soil depths slow down seasonal peak flows and increase the loss due to evapotranspiration. In general, the depth of the unsaturated zone tends to be deeper in the arid regions of China such as western China and northern China, but tends to be lower in the humid areas of China such as eastern and southern China. Thus the parameter transfer scheme sets deeper depths for the continental climate zone with hot summer (d1 = 0.5 m and d2 = 1.75 m) and the continental climate zone with cool summer (d1 = 0.3 m and d2 = 1.25 m), and assigns lower depths for the tropical climate zone (d1 = 0.1 m and d2 = 0.5 m) and the rainy, midlatitude climate zone (d1 = 0.075–0.1 m and d2 = 0.45–0.5 m). However, in the control case, d1and d2 are set to be 0.1 and 0.5 m, respectively, for all the grids in China. As a result, the deeper soil depths in western and northern China (belonging to the continental climate zone with hot summer and the continental climate zone with cool summer) produce less runoff, and the lower soil depths in the southern and eastern China (belonging to tropical climate zone and the rainy, midlatitude climate zone) generate more runoff compared with the control case. Because the parameters are set to the same values as those in the Pearl River basin for the tropical zone as mentioned in section 5a, which belongs to the rainy and midlatitude climate zone, and the simulated runoff with transferred parameters is not as the control case, more work needs to be investigated for the tropical zone in China.

6. Conclusions

Streamflow is important not only for water resource studies, but it is also the most commonly available component of the surface water balance, and thus it can be used in a variety of diagnostic studies (Maurer et al. 2000). In most macroscale hydrological models, the calibration of model parameters is critical to the hydrological modeling. In the VIC-3L model, most of the vegetation and soil parameter values may be estimated using lookup tables from the literature, but model parameters such as the infiltration parameter (b), the depths of the three soil layers (d1, d2, and d3), and the ARNO model parameters (Dm, Ds, and Ws) cannot be determined well based on the available soil information. These parameters have to be calibrated by matching the simulated streamflow with the observed streamflow data. However, this can be achieved only if discharge data are available. For data-sparse areas, parameter transfer is needed, which transfers the model parameters from data-rich areas to data-sparse areas. In this paper, a parameter estimation scheme is given to simulate streamflow for river basins in China, which is represented by 4355 cells in a 50 × 50 km2 grid. The land area in China is grouped by climate and large river basin zones, and a limited number of basins from those climate and river basin zones are for calibration. The calibrated parameters are then regionalized to uncalibrated basins based on climate characteristics and large river basin subareas. The transferred parameters were then used to simulate the water balance in the river basins in China. The simulated daily runoffs of VIC-3L with transferred parameters and uncalibrated parameters were routed to the outlets of the river basins, and compared to the monthly observed streamflows at the related basins. The results show that the model for the transferred parameters can simulate the observations well and the parameter transfer framework is promising in estimating the VIC-3L model parameters for data-sparse areas in China. This suggests that this parameter transfer method may be viable for hydrological simulations of large continental rivers basins, which will eventually be essential for climate and numerical weather prediction models that incorporate a fully interactive land surface hydrology representation. While encouraging, it should be emphasized that the data used in this study are not sufficient enough for hydrological modeling over the large river basins in China, especially for the Qinghai–Tibetan Plateau and the Tarim basin, and the method of climate and large river basin zone grouping is a little too coarse to characterize the climate and river basins in China in detail, which can generate quite a few uncertainties in the hydrological modeling. Therefore, more data sources are needed to calibrate model parameters so as to improve the modeling performance, and more work needs to be done to investigate a more reasonable grouping method.

Acknowledgments

This work was supported by the National Natural Science Foundation of China under Grant 90411007, the National Basic Research Program under Grant 2005CB321704, and CAS International Partnership Creative Group “The Climate System Model Development and Application Studies.” The authors thank Dr. Fayez Abdulla and the two anonymous reviewers for valuable comments and suggestion on this manuscript. The third author gratefully acknowledges the support of K. C. Wong Education Foundation, Hong Kong.

REFERENCES

  • Abdulla, F. A., , and Lettenmaier D. P. , 1997a: Development of regional parameter estimation equations for land surface hydrologic model. J. Hydrol., 197 , 230257.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abdulla, F. A., , and Lettenmaier D. P. , 1997b: Application of regional parameter estimation to simulate the water balance of large continental river. J. Hydrol., 197 , 258285.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abdulla, F. A., , Lettenmaier D. P. , , Wood E. F. , , and Smith J. A. , 1996: Application of a macroscale hydrologic model to estimate the water balance of the Arkansas-Red River basin. J. Geophys. Res., 101 , D3. 74497459.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1996: A Land Surface Model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user’s guide. NCAR Tech. Note NCAR/TN-417+STR, National Center for Atmospheric Research, 150 pp.

  • Chen, Z. K., 1985: China’s water resources and its utilization. GeoJournal, 10 , 167171.

  • Cosby, B. J., , Hornberger G. M. , , Clapp R. B. , , and Ginn T. R. , 1984: A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res., 20 , 682690.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Coauthors, 2003: The Common Land Model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dickinson, R. R., , Henderson-Sellers A. , , Kennedy P. J. , , and Wilson M. F. , 1986: Biosphere-Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-275+STR, National Center for Atmospheric Research, 69 pp.

  • Dickinson, R. R., , Henderson-Sellers A. , , and Kennedy P. J. , 1993: Biosphere-Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-387+STR, National Center for Atmospheric Research, 72 pp.

  • Duan, Q., , Sorooshian S. , , and Gupta V. K. , 1992: Effective and efficient global optimization for conceptual rainfall-runoff models. Water Resour. Res., 28 , 10151031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Coauthors, 2006: Model Parameter Estimation Experiment (MOPEX): An overview of science strategy and major results from the second and third workshops. J. Hydrol., 320 , 317.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • FAO, 1998: Digital soil map of the world and derived soil properties. Land Water Digital Media Series, Vol. 1, Food and Agriculture Organization, CD-ROM.

  • Hansen, M. C., , DeFries R. S. , , Townshend J. R G. , , and Sohlberg R. , 2000: Global land cover classification at 1 km spatial resolution using a classification tree approach. Int. J. Remote Sens., 21 , 13311364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang, M., , and Liang X. , 2003: A transferability study of model parameters for the Variable Infiltration Capacity land surface scheme. J. Geophys. Res., 108 .8864, doi:10.1029/2003JD003676.

    • Search Google Scholar
    • Export Citation
  • Hubert, B., , Francois L. , , Warnant P. , , and Strivay D. , 1998: Stochastic generation of meteorological variables and effects on global models of water and carbon cycles in vegetation and soils. J. Hydrol., 212–213 , 318334.

    • Search Google Scholar
    • Export Citation
  • Liang, X., , and Xie Z. , 2001: A new surface runoff parameterization with subgrid-scale soil heterogeneity for land surface models. Adv. Water Resour., 24 , 11731193.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liang, X., , Lettenmaier D. P. , , Wood E. F. , , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res., 99 , D7. 1441514428.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liang, X., , Lettenmaier D. P. , , and Wood E. F. , 1996: One-dimensional statistical dynamic representation of subgrid variability of precipitation in the two-layer Variable Infiltration Capacity model. J. Geophys. Res., 101 , D16. 2140321422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Maurer, E. P., , O’Donnell G. M. , , Lettenmaier D. P. , , and Roads J. O. , 2000: Evaluation of NCEP/NCAR reanalysis water and energy budgets using macroscale hydrological simulations as a benchmark. Observations and Modeling of the Land Surface Hydrological Processes, V. Lakshmi, J. Albertson, and J. Schaake, Eds., Amer. Geophys. Union, 137–158.

    • Search Google Scholar
    • Export Citation
  • Nash, J. E., , and Sutcliffe J. V. , 1970: River flow forecasting through conceptual models. Part I: A discussion of principles. J. Hydrol., 10 , 282290.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nijssen, B., , O’Donnell G. M. , , and Lettenmaier D. P. , 2001: Predicting the discharge of global rivers. J. Climate, 14 , 33073323.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rawls, W., , Ahuja L. R. , , Brakensiek D. L. , , and Shirmohammadi A. , 1993: Infiltration and soil water movement. Handbook of Hydrology, D. R. Maidment, Ed., McGraw-Hill, 5.1–5.51.

    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., , Mintz Y. , , Sud Y. C. , , and Dalcher A. , 1986: A Simple Biosphere Model (SiB) for use within general circulation models. J. Atmos. Sci., 43 , 505531.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Todini, E., 1996: The ARNO rainfall-runoff model. J. Hydrol., 175 , 339382.

  • Xia, J., , Heung W. , , and Wai C. I. , 2003: Water problems and sustainability in North China. Water Resources Systems—Water Availability and Global Changes, S. Franks et al., Eds., IAHS Press, 12–22.

    • Search Google Scholar
    • Export Citation
  • Xie, Z., , Su F. , , Liang X. , , Zeng Q. , , Hao Z. , , and Guo Y. , 2003: Applications of a surface runoff model with Horton and Dunne runoff for VIC. Adv. Atmos. Sci., 20 , 165172.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , Sellers P. J. , , Kinter J. L. , , and Shukla J. , 1991: A simplified biosphere model for global climate studies. J. Climate, 4 , 345364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yangtse River Conservancy Commission, 2002: Flood and Drought Disaster for Yangtse River Basin, China Flood and Drought Disaster Series. (in Chinese). China Water Conservancy and Water Electricity Press, 326 pp.

    • Search Google Scholar
    • Export Citation
  • Zeng, X., , Shaikh M. , , Dai Y. , , Dickinson R. E. , , and Myneni R. , 2002: Coupling of the Common Land Model to the NCAR Community Climate Model. J. Climate, 15 , 18321854.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

River basins and climate zones in China according to Table 2.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 2.
Fig. 2.

The locations of the selected basins in China for calibration and verifications.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 3.
Fig. 3.

The locations of the 740 meteorological stations in China.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 4.
Fig. 4.

Monthly hydrographs of the observed and simulated flows (control case and calibrated) for the primary basins.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 5.
Fig. 5.

Mean monthly hydrographs of the observed and simulated flows (control case and calibrated) for the primary basins.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 6.
Fig. 6.

Schematic representation of the parameter regionalization scheme.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 7.
Fig. 7.

Monthly hydrographs of the observed and simulated flows (control case, parameter transfer, and recalibrated) for secondary basins. Monthly hydrographs of the observed and simulated flows (control case, parameter transfer, and recalibrated) for secondary basins.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 8.
Fig. 8.

Mean monthly hydrographs of the observed and simulated flows (control case and calibrated) for secondary basins.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 9.
Fig. 9.

Mean monthly hydrographs of the observed and simulated flows for secondary basins.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 10.
Fig. 10.

Annual mean distribution of (a) simulated runoff in the control case; (b) simulated runoff in the transfer case; (c) difference of simulated runoff [(b)−(a)]; and (d) precipitation.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Fig. 11.
Fig. 11.

Comparison of the simulated runoff in control case and transfer case for climate zones in China.

Citation: Journal of Hydrometeorology 8, 3; 10.1175/JHM568.1

Table 1.

VIC-3L parameters and their estimating methods.

Table 1.
Table 2.

Grouping of Köppen climate zones.

Table 2.
Table 3.

Selected basins for calibration (primary basins).

Table 3.
Table 4.

Selected basins for verification (secondary basins).

Table 4.
Table 5.

Vegetation-related parameters in VIC-3L.

Table 5.
Table 6.

Soil-related parameters in VIC-3L.

Table 6.
Table 7.

Calibrated parameters for the primary basins and default parameter values.

Table 7.
Table 8.

Calibration statistics for the primary basins.

Table 8.
Table 9.

Transferred parameters for the climate and large river basin zones.

Table 9.
Table 10.

Calibration and parameter transfer statistics for the 11 secondary basins.

Table 10.
Save