Advanced Search
Article Contents

Elucidating Dominant Factors Affecting Land Surface Hydrological Simulations of the Community Land Model over China


doi: 10.1007/s00376-022-2091-5

  • In order to compare the impacts of the choice of land surface model (LSM) parameterization schemes, meteorological forcing, and land surface parameters on land surface hydrological simulations, and explore to what extent the quality can be improved, a series of experiments with different LSMs, forcing datasets, and parameter datasets concerning soil texture and land cover were conducted. Six simulations are run for the Chinese mainland on 0.1° × 0.1° grids from 1979 to 2008, and the simulated monthly soil moisture (SM), evapotranspiration (ET), and snow depth (SD) are then compared and assessed against observations. The results show that the meteorological forcing is the most important factor governing output. Beyond that, SM seems to be also very sensitive to soil texture information; SD is also very sensitive to snow parameterization scheme in the LSM. The Community Land Model version 4.5 (CLM4.5), driven by newly developed observation-based regional meteorological forcing and land surface parameters (referred to as CMFD_CLM4.5_NEW), significantly improved the simulations in most cases over the Chinese mainland and its eight basins. It increased the correlation coefficient values from 0.46 to 0.54 for the SM modeling and from 0.54 to 0.67 for the SD simulations, and it decreased the root-mean-square error (RMSE) from 0.093 to 0.085 for the SM simulation and reduced the normalized RMSE from 1.277 to 0.201 for the SD simulations. This study indicates that the offline LSM simulation using a refined LSM driven by newly developed observation-based regional meteorological forcing and land surface parameters can better model reginal land surface hydrological processes.
    摘要: 本文利用不同模式参数化方案、气象强迫和地表参数设计了六个模拟试验,并与观测进行对比分析,揭示了影响中国区域陆面水文要素(土壤湿度、蒸散发和雪深等)模拟的主要因子。结果表明,气象强迫是陆面水文过程模拟的主要影响因子,土壤湿度和雪深的模拟分别与土壤质地信息和陆面模式参数化方案紧密相关。利用新发展的融合观测信息的中国区域气象强迫和地表参数信息驱动陆面模式模拟,显著提高了中国大部分区域陆面水文过程模拟精度,并减少了模拟的不确定性。土壤湿度模拟与观测的相关系数从0.46提高到0.54,雪深模拟与观测的相关系数从0.54提高到0.67;土壤湿度模拟与观测的均方根误差从0.093降低到0.085,雪深模拟与观测的均方根误差从1.277降低到0.201。
  • 加载中
  • Figure 1.  Spatial distribution for the five-year (2004–08) averaged volumetric SM derived from the six simulations, the in situ site measurements in the 0–10-cm soil layer, and the location of the eight basins in the Chinese mainland: (a) Prin_CLM3.5; (b) CMFD_CLM3.5; (c) CMFD_CLM4.5; (d) CMFD_CLM4.5_NS; (e) CMFD_CLM4.5_MICL; (f) CMFD_CLM4.5_NEW; (g) in situ site observations; and (h) the location of the eight basins in China. I: Songhua River basin; II: Haihe River basin; III: Yellow River basin; IV: Huaihe River basin; V: Heihe River basin; VI:Tarim River basin; VII: Yangtze River basin; VIII: Zhujiang River basin.

    Figure 2.  Difference in the R of the modeled SM against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_ CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_ CLM4.5; and (f) CMFD_CLM4.5_ NEW minus Prin_CLM3.5.

    Figure 3.  Difference in the RMSE of the modeled SM against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_ CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_CLM4.5; and (f) CMFD_CLM4.5_NEW minus Prin_CLM3.5.

    Figure 4.  Time series of the monthly volumetric SM (mm3 mm–3) for the 0–10-cm soil layer from in situ observations, Prin_CLM3.5, CMFD_CLM3.5, CMFD_CLM4.5, CMFD_CLM4.5_NS, CMFD_CLM4.5_MICL, and CMFD_CLM4.5_NEW during the period of 2008–13 in the eight major basins of China.

    Figure 5.  Difference in the R of the modeled ET against the observation-based ET product Obs_MTE: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_CLM4.5; and (f) CMFD_CLM4.5_ NEW minus Prin_CLM3.5.

    Figure 6.  Difference in the R of the modeled SD against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_ CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_ CLM4.5; and (f) CMFD_CLM4.5_NEW minus Prin_CLM3.5.

    Figure 7.  Time series of (a–c) the SD (cm) from in situ observations, Prin_CLM3.5, CMFD_CLM3.5, CMFD_CLM4.5, CMFD_CLM4.5_NS, CMFD_CLM4.5_MICL, and CMFD_CLM4.5_NEW; (d–f) snowfall (mm d–1) from Princeton forcing data and CMFD during the period of 2004–08 in the three major snow areas of China.

    Table 1.  Comparison of the major features of the two sets of land-surface parameters information data.

    DataSoil textureLand coverReference
    CLM4.5
    Default
    1. Derived from the FAO/UNESCO world soil map (1:5 million) and about 60 soil profile data;
    2. Resolution: 5';
    3. The soil column including nine layers with variable thickness;
    4. The IGBP soil dataset of 4931 soil mapping units and their sand and clay content for each soil layer were used to create a mineral soil texture dataset
    1. Spatial resolution: 0.05°;
    2. PFTs were derived from the MODIS satellite product from November 2000 to October 2001;
    3. Glaciers were derived from IGBP DISCover data;
    4. Lakes and wetlands were derived from perennial freshwater lakes and swamps/marshes;
    5. Urban areas were derived from population density data, Landscan 2004
    Lawrence and Chase (2007); Loveland et al. (2000); Oleson et al. (2013); Bonan et al. (2002)
    New1. Derived from the Second National Soil Survey (1979–85) China soil map at a scale of 1:1 million and about 9000 soil profiles;
    2. Resolution: 0.5°;
    3. The soil column of 0–2.296 m including eight layers with variable thickness, for convenience of use, the first two layers are too thin and combined;
    4. The soil dataset of 925 soil mapping units in China and their soil type were used to create soil properties including a soil texture dataset
    1. Spatial resolution: 1 km;
    2. Merging Vegetation Atlas of China (1:1 000 000), Land use map of China (1:100 000) in 2000, Glacier distribution map of China (1:100000), Swamp-wetland map of China (1:1 000 000) and MODIS2001 product using IGBP classification system
    Shangguan et al. (2012, 2013); Hou (2001); Wu and Li (2004); Zhang (2002); Friedl et al. (2002)
    * Food and Agriculture Organization, United Nations Educational, Scientific and Cultural Organization (FAO/UNESCO); International Geosphere-Biosphere Programme (IGBP); Moderate resolution Imaging Spectroradiometer (MODIS); Plant functional Types (PFTs); International Geosphere-Biosphere Programme Data and Information System Global 1-km Land Cover Data Set (IGBP DISCover).
    DownLoad: CSV

    Table 2.  List of experiments.

    ExperimentCLM3.5CLM4.5Princeton ForcingCMFD
    Forcing
    Original
    Soil
    New
    Soil
    Original land coverMICL
    Prin_CLM3.5YYYY
    CMFD_CLM3.5YYYY
    CMFD_CLM4.5YYYY
    CMFD_CLM4.5_NSYYYY
    CMFD_CLM4.5_MICLYYYY
    CMFD_CLM4.5_NEWYYYY
    DownLoad: CSV

    Table 3.  Mean R and RMSE/NRMSE between the simulated and observation-based SM, SD, and ET from 2004 to 2008.

    Prin_CLM3.5CMFD_CLM3.5CMFD_CLM4.5CMFD_CLM4.5_NSCMFD_CLM4.5_
    MICL
    CMFD_CLM4.5_
    NEW
    RRMSE/
    NRMSE
    RRMSE/
    NRMSE
    RRMSE/
    NRMSE
    RRMSE/
    NRMSE
    RRMSE/
    NRMSE
    RRMSE/
    NRMSE
    SM at 0–10-cm
    depth for
    308 stations*
    China0.46(77)0.09300.51
    (84, +10.9)
    0.0907
    (−2.5)
    0.52
    (85,
    +2.0)
    0.0912
    (+0.6)
    0.54
    (86, +3.8)
    0.0820
    (−10.1)
    0.52
    (86, 0)
    0.0935
    (+2.5)
    0.54
    (87, +3.8,
    +17.4)
    0.0854
    (−6.4, −8.2)
    SD for 1390 stations*China0.55(75)1.2770.71
    (92,+29.1)
    0.422
    (−66.9)
    0.76
    (97,+7.0)
    0.198
    (−53.1)
    0.75
    (97,−1.3)
    0.202
    (+2.02)
    0.75
    (97,−1.3)
    0.199
    (+0.51)
    0.76
    (97, +7.0)
    0.198
    (−53.1)
    ET**Songhua
    River basin
    0.95111.870.963
    (+1.26)
    11.03
    (−7.08)
    0.965
    (+0.21)
    11.70
    (+6.07)
    0.967
    (+0.21)
    11.49
    (−1.8)
    0.965
    (0)
    11.55
    (−1.28)
    0.968
    (+0.31, +1.79)
    11.40
    (−2.56, −3.96)
    Haihe River basin0.95113.090.956
    (+0.53)
    14.46
    (+10.5)
    0.960
    (+0.42)
    15.86
    (9.68)
    0.962
    (+0.21)
    14.56
    (−8.2)
    0.963
    (+0.31)
    15.14
    (−4.54)
    0.965
    (0.52, +1.47)
    15.09
    (−4.86, +15.3)
    Heihe River basin0.6309.200.605
    (−4.00)
    8.54
    (−7.17)
    0.640
    (+5.79)
    9.46
    (+10.7)
    0.647
    (+1.09)
    9.31
    (−1.59)
    0.647
    (+1.09)
    9.29
    (−1.8)
    0.656
    (+2.5, +4.13)
    9.14
    (−3.38, −0.65)
    Tarim River basin0.6629.690.722
    (+9.06)
    8.49
    (−12.4)
    0.730
    (+1.11)
    9.44
    (+11.2)
    0.738
    (+1.10)
    9.40
    (−0.42)
    0.742
    (+1.64)
    9.21
    (−2.44)
    0.749
    (+2.6, +13.14)
    9.16
    (−2.97, −5.47)
    Yellow River basin0.9259.830.926
    (+0.11)
    9.68
    (−1.53)
    0.925
    (−0.11)
    11.41
    (+17.9)
    0.926
    (+0.11)
    11.10
    (−2.72)
    0.932
    (+0.76)
    11.16
    (−2.19)
    0.934
    (+0.97,+0.97)
    11.03
    (−0.33,+12.2)
    Huaihe River basin0.92314.780.912
    (−1.2)
    16.76
    (+13.4)
    0.924
    (+1.32)
    23.98
    (+43.1)
    0.921
    (−0.33)
    22.96
    (−4.25)
    0.924
    (0)
    23.64
    (−1.42)
    0.922
    (−0.22, −0.11)
    23.84
    (−0.58, +6.13)
    Yangtze River basin0.93215.140.939
    (+0.75)
    15.84
    (+4.62)
    0.942
    (+0.32)
    19.55
    (+23.4)
    0.939
    (−0.32)
    19.22
    (−1.69)
    0.940
    (−0.21)
    19.64
    (+0.46)
    0.937
    (−0.53, +0.54)
    19.60
    (+0.26, +29.4)
    Zhujiang River basin0.89419.030.917
    (+2.57)
    20.68
    (+8.67)
    0.923
    (0.65)
    26.01
    (+25.7)
    0.914
    (−0.98)
    25.35
    (−2.54)
    0.920
    (−0.33)
    26.92
    (+3.5)
    0.912
    (−1.19, +2.01)
    26.59
    (+2.23, +39.7)
    China0.86812.180.881
    (+1.5)
    11.96
    (−1.81)
    0.883
    (+0.23)
    16.26
    (+35.9)
    0.883
    (0)
    16.03
    (−1.42)
    0.886
    (+0.34)
    16.01
    (−1.54)
    0.886
    (+0.34, +2.07)
    15.97
    (−1.78, +31.1)
    * The first number in the parentheses is the percentage of stations with a correlation that is statistically significant at the p = 0.05 level. The second and third numbers in the parentheses are the percentage increases relative to the paired experiments (i.e., the same as shown in Fig. 2). ** The number in the parentheses is the percentage of variation relative to the paired experiments (i.e., the same as shown in Fig. 2).
    DownLoad: CSV
  • Albergel, C., P. de Rosnay, G. Balsamo, L. Isaksen, and J. Muñoz-Sabater, 2012: Soil moisture analyses at ECMWF: Evaluation using global ground-based in situ observations. Journal of Hydrometeorology, 13, 1442−1460, https://doi.org/10.1175/JHM-D-11-0107.1.
    Bonan, G. B., S. Levis, L. Kergoat, and K. W. Oleson, 2002: Landscapes as patches of plant functional types: An integrating concept for climate and ecosystem models. Global Biogeochemical Cycles, 16, 5−1-5-23, https://doi.org/10.1029/2000GB001360.
    Chen, J., B. Z. Chen, T. A. Black, J. L. Innes, G. Y. Wang, G. Kiely, T. Hirano, and G. Wohlfahrt, 2013: Comparison of terrestrial evapotranspiration estimates using the mass transfer and Penman-Monteith equations in land surface models. J. Geophys. Res.: Biogeosci., 118, 1715−1731, https://doi.org/10.1002/2013JG002446.
    Dirmeyer, P. A., X. Gao, M. Zhao, Z. C. Guo, T. Oki, and N. Hanasaki, 2006: GSWP-2: Multimodel analysis and implications for our perception of the land surface. Bull. Amer. Meteor. Soc., 87, 1381−1398, https://doi.org/10.1175/BAMS-87-10-1381.
    Friedl, M. A., and Coauthors, 2002: Global land cover mapping from MODIS: Algorithms and early results. Remote Sensing of Environment, 83(1−2), 287−302, https://doi.org/10.1016/S0034-4257(02)00078-0.
    Gao, H., and G. S. Jia, 2013: Assessing disagreement and tolerance of misclassification of satellite-derived land cover products used in WRF model applications. Adv. Atmos. Sci., 30(1), 125−141, https://doi.org/10.1007/s00376-012-2037-4.
    He, J., K. Yang, W. J. Tang, H. Lu, J. Qin, Y. Y. Chen, and X. Li, 2020: The first high-resolution meteorological forcing dataset for land process studies over China. Scientific Data, 7, 25, https://doi.org/10.1038/s41597-020-0369-y.
    Hoffman, F. M., and Coauthors, 2017: 2016 international land model benchmarking (ILAMB) workshop report. Technical Report DOE/SC-0186, 159 pp,
    Hou, X. Y., 2001: Vegetation Atlas of China (1:1,000,000). Sciences Press, 260 pp. Beijing. (in Chinese)
    Jia, B. H., J. G. Liu, and Z. H. Xie, 2015: Evaluation of a multi-satellite soil moisture product and the Community Land Model 4.5 simulation in China. Hydrology and Earth System Sciences Discussions, 12, 5151−5186, https://doi.org/10.5194/hessd-12-5151-2015.
    Jia, B. H., J. G. Liu, Z. H. Xie, and C. X. Shi, 2018: Interannual variations and trends in remotely sensed and modeled soil moisture in China. Journal of Hydrometeorology, 19(5), 831−847, https://doi.org/10.1175/JHM-D-18-0003.1.
    Jung, M., and Coauthors, 2010: Recent decline in the global land evapotranspiration trend due to limited moisture supply. Nature, 467, 951−954, https://doi.org/10.1038/nature09396.
    Jung, M., and Coauthors, 2011: Global patterns of land-atmosphere fluxes of carbon dioxide, latent heat, and sensible heat derived from eddy covariance, satellite, and meteorological observations. J. Geophys. Res.: Biogeosci., 116, G00J07, https://doi.org/10.1029/2010JG001566.
    Kato, H., and Coauthors, 2006: Sensitivity of Land Surface Simulations to Model Physics, Parameters, and Forcings, at Four CEOP Sites. J. Meteor. Soc. Japan, 85A, 187−204, https://doi.org/10.2151/jmsj.85A.187.
    Lawrence, D. M., and Coauthors, 2011: Parameterization improvements and functional and structural advances in version 4 of the Community Land Model. Journal of Advances in Modeling Earth Systems, 3, M03001, https://doi.org/10.1029/2011MS00045.
    Lawrence, P. J., and T. N. Chase, 2007: Representing a new MODIS consistent land surface in the Community Land Model (CLM 3.0). J. Geophys. Res.: Biogeosci., 112, G01023, https://doi.org/10.1029/2006JG000168.
    Li, W. K., W. D. Guo, B. Qiu, Y. K. Xue, P. C. Hsu, and J. F. Wei, 2018a: Influence of Tibetan Plateau snow cover on East Asian atmospheric circulation at medium-range time scales. Nature Communications, 9, 4243, https://doi.org/10.1038/s41467-018-06762-5.
    Li, J. D., F. Chen, G. Zhang, M. Barlage, Y. J. Gan, Y. F. Xin, and C. Wang, 2018b: Impacts of land cover and soil texture uncertainty on land model simulations over the central Tibetan Plateau. Journal of Advances in Modeling Earth Systems, 10, 2121−2146, https://doi.org/10.1029/2018MS001377.
    Liang, X., and Y. J. Dai, 2008: A sensitivity study of the Common Land Model on soil texture and soil brightness. Climatic and Environmental Research, 13(5), 585−597. (in Chinese with English abstract)
    Liu, J. G., and Z. H. Xie, 2013: Improving simulation of soil moisture in China using a multiple meteorological forcing ensemble approach. Hydrology and Earth System Sciences, 17, 3355−3369, https://doi.org/10.5194/hess-17-3355-2013.
    Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, Z. Zhu, L. Yang, and J. W. Merchant, 2000: Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sens., 21(6−7), 1303−1330, https://doi.org/10.1080/014311600210191.
    Lu, H., D. H. Zheng, K. Yang, and F. Yang, 2020: Last-decade progress in understanding and modeling the land surface processes on the Tibetan Plateau. Hydrology and Earth System Sciences, 24, 5745−5758, https://doi.org/10.5194/hess-24-5745-2020.
    Makarieva, A. M., V. G. Gorshkov, D. Sheil, A. D. Nobre, P. Bunyard, and B. L. Li, 2014: Why does air passage over forest yield more rain? Examining the coupling between rainfall, pressure, and atmospheric moisture content Journal of Hydrometeorology, 15, 411−426, https://doi.org/10.1175/JHM-D-12-0190.1.
    Niu, G. Y., Z. L. Yang, R. E. Dickinson, L. E. Gulden, and H. Su, 2007: Development of a simple groundwater model for use in climate models and evaluation with Gravity Recovery and Climate Experiment data. J. Geophys. Res.: Atmos., 112, D07103, https://doi.org/10.1029/2006JD007522.
    Nearing, G.S., D.M. Mocko, C.D. Peters-Lidard, S.V. Kumar, and Y. Xia, 2016: Benchmarking NLDAS-2 soil moisture and evapotranspiration to separate uncertainty contributions. J. Hydrometeor., 17(3), 745−759, https://doi.org/10.1175/JHM-D-15-0063.1.
    Oleson, K. W., and Coauthors, 2007: CLM 3.5 documentation, 34 pp. [available online from http://www.cgd.ucar.edu/tss/clm/distribution/clm3.5/CLM3_5documentation.pdf.]
    Oleson, K. W., and Coauthors, 2008: Improvements to the Community Land Model and their impact on the hydrological cycle. J. Geophys. Res.: Biogeosci., 113, G01021, https://doi.org/10.1029/2007JG000563.
    Oleson, K. W., and Coauthors, 2013: Technical description of version 4.5 of the Community Land Model (CLM). NCAR Technical Note NCAR/TN-503+STR, 422 pp.
    Ran, Y. H., X. Li, L. Lu, and Z. Y. Li, 2012: Large-scale land cover mapping with the integration of multi-source information based on the Dempster-Shafer theory. International Journal of Geographical Information Science, 26(1), 169−191, https://doi.org/10.1080/13658816.2011.577745.
    Rodell, M., and Coauthors, 2004: The global land data assimilation system. Bull. Amer. Meteor. Soc., 85, 381−394, https://doi.org/10.1175/BAMS-85-3-381.
    Shangguan, W., Y. J. Dai, B. Y. Liu, A. Z. Ye, and H. Yuan, 2012: A soil particle-size distribution dataset for regional land and climate modelling in China. Geoderma, 171−172, 85−91,
    Shangguan, W., and Coauthors, 2013: A China data set of soil properties for land surface modeling. Journal of Advances in Modeling Earth Systems, 5(2), 212−224, https://doi.org/10.1002/jame.20026.
    Sheffield, J., G. Goteti, and E. F. Wood, 2006: Development of a 50-year high-resolution global dataset of meteorological forcings for land surface modeling. J. Climate, 19, 3088−3111, https://doi.org/10.1175/JCLI3790.1.
    Sheffield, J., and E. F. Wood, 2007: Characteristics of global andregional drought, 1950-2000: Analysis of soil moisture data from off-line simulation of the terrestrial hydrologic cycle. J. Geophys. Res., 112, D17115, https://doi.org/10.1029/2006JD008288.
    Shi, X. Y., J. F. Mao, P. E. Thornton, and M. Y. Huang, 2013: Spatiotemporal patterns of evapotranspiration in response to multiple environmental factors simulated by the Community Land Model. Environmental Research Letters, 8, 024012, https://doi.org/10.1088/1748-9326/8/2/024012.
    Sun, S. B., B. Z. Chen, Q. Q. Shao, J. Chen, J. Y. Liu, X. J. Zhang, H. F. Zhang, and X. F. Lin, 2017: Modeling evapotranspiration over China’s Landmass from 1979 to 2012 using multiple land surface models: Evaluations and analyses. Journal of Hydrometeorology, 18, 1185−1203, https://doi.org/10.1175/JHM-D-16-0212.1.
    Swenson, S. C., and D. M. Lawrence, 2012: A new fractional snow-covered area parameterization for the Community Land Model and its effect on the surface energy balance. J. Geophys. Res.: Atmos., 117, D21107, https://doi.org/10.1029/2012JD018178.
    Tian, Y., R. E. Dickinson, L. Zhou, and M. Shaikh, 2004: Impact of new land boundary conditions from Moderate Resolution Imaging Spectroradiometer (MODIS) data on the climatology of land surface variables. J. Geophys. Res.: Atmos., 109, D20115, https://doi.org/10.1029/2003JD004499.
    Wang, A. H., and X. B. Zeng, 2011: Sensitivities of terrestrial water cycle simulations to the variations of precipitation and air temperature in China. J. Geophys. Res.: Atmos., 116, D02107, https://doi.org/10.1029/2010JD014659.
    Wang, A. H., X. B. Zeng, and D. L. Guo, 2016: Estimates of global surface hydrology and heat fluxes from the Community Land Model (CLM4.5) with four atmospheric forcing datasets. Journal of Hydrometeorology, 17, 2493−2510, https://doi.org/10.1175/JHM-D-16-0041.1.
    Wang, K. C., and R. E. Dickinson, 2012: A review of global terrestrial evapotranspiration: Observation, modeling, climatology, and climatic variability. Rev. Geophys., 50, RG2005, https://doi.org/10.1029/2011RG000373.
    Wu, L. Z., and X. Li, 2004: China Glacier Information System. China Ocean Press, 135 pp. (in Chinese)
    Xia, Y. L., J. Sheffield, M. B. Ek, J. R. Dong, N. Chaney, H. L. Wei, J. Meng, and E. F. Wood, 2014: Evaluation of multi-model simulated soil moisture in NLDAS-2,. Journal of Hydrology, 512, 107−125, https://doi.org/10.1016/j.jhydrol.2014.02.027.
    Xie, Z. P., Z. Y. Hu, Z. H. Xie, B. H. Jia, G. H. Sun, Y. Z. Du, and H. Q. Song, 2018: Impact of the snow cover scheme on snow distribution and energy budget modeling over the Tibetan Plateau. Theor. Appl. Climatol., 131(3−4), 951−965, https://doi.org/10.1007/s00704-016-2020-6.
    Yang, K., and J. He, 2019: China meteorological forcing dataset (1979−2018). National Tibetan Plateau Data Center,
    Yu, Y., Z. H. Xie, Y. Y. Wang, and F. Chen, 2014: Results of a CLM4 land surface simulation over China using a multisource integrated land cover dataset. Atmos. Ocean. Sci. Lett., 7, 279−285, https://doi.org/10.3878/j.issn.1674-2834.13.0083.
    Zhang, Q., and Coauthors, 2020: Causes and changes of drought in China: Research progress and prospects. J. Meteor. Res., 34(3), 460−481, https://doi.org/10.1007/s13351-020-9829-8.
    Zhang, S. Q., 2002: An introduction of wetland science database in China. Scientia Geographica Sinica, 22(2), 189. (in Chinese)
    Zhang, Y. F., and Z. L. Yang, 2016: Estimating uncertainties in the newly developed multi-source land snow data assimilation system. J. Geophys. Res.: Atmos., 121, 8254−8268, https://doi.org/10.1002/2015JD024248.
    Zheng, H., and Z. L. Yang, 2016: Effects of soil-type datasets on regional terrestrial water cycle simulations under different climatic regimes. J. Geophys. Res.: Atmos., 121, 14 387−14 402,
  • [1] LI Tao, ZHENG Xiaogu, DAI Yongjiu, YANG Chi, CHEN Zhuoqi, ZHANG Shupeng, WU Guocan, WANG Zhonglei, HUANG Chengcheng, SHEN Yan, LIAO Rongwei, 2014: Mapping Near-surface Air Temperature, Pressure, Relative Humidity and Wind Speed over Mainland China with High Spatiotemporal Resolution, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 1127-1135.  doi: 10.1007/s00376-014-3190-8
    [2] Zhang Qiang, Cao Xiaoyan, Wei Guoan, Huang Ronghui, 2002: Observation and Study of Land Surface Parameters over Gobi in Typical Arid Region, ADVANCES IN ATMOSPHERIC SCIENCES, 19, 121-135.  doi: 10.1007/s00376-002-0039-3
    [3] LI Fang, ZENG Xiaodong, SONG Xiang, TIAN Dongxiao, SHAO Pu, ZHANG Dongling, 2011: Impact of Spin-up Forcing on Vegetation States Simulated by a Dynamic Global Vegetation Model Coupled with a Land Surface Model, ADVANCES IN ATMOSPHERIC SCIENCES, 28, 775-788.  doi: 10.1007/s00376-010-0009-0
    [4] Jianguo LIU, Binghao JIA, Zhenghui XIE, Chunxiang SHI, 2016: Ensemble Simulation of Land Evapotranspiration in China Based on a Multi-Forcing and Multi-Model Approach, ADVANCES IN ATMOSPHERIC SCIENCES, 33, 673-684.  doi: 10.1007/s00376-016-5213-0
    [5] XIE Zhenghui, SU Fengge, LIANG Xu, ZENG Qingcun, HAO Zhenchun, GUO Yufu, 2003: Applications of a Surface Runoff Model with Horton and Dunne Runoff for VIC, ADVANCES IN ATMOSPHERIC SCIENCES, 20, 165-172.  doi: 10.1007/s00376-003-0001-z
    [6] LI Hongqi, GUO Weidong, SUN Guodong, ZHANG Yaocun, FU Congbin, 2011: A New Approach for Parameter Optimization in Land Surface Model, ADVANCES IN ATMOSPHERIC SCIENCES, 28, 1056-1066.  doi: 10.1007/s00376-010-0050-z
    [7] LIANG Miaoling, XIE Zhenghui, 2008: Improving the Vegetation Dynamic Simulation in a Land Surface Model by Using a Statistical-dynamic Canopy Interception Scheme, ADVANCES IN ATMOSPHERIC SCIENCES, 25, 610-618.  doi: 10.1007/s00376-008-0610-7
    [8] CHEN Feng, XIE Zhenghui, 2011: Effects of Crop Growth and Development on Land Surface Fluxes, ADVANCES IN ATMOSPHERIC SCIENCES, 28, 927-944.  doi: 10.1007/s00376-010-0105-1
    [9] Guo Weidong, Sun Shufen, Qian Yongfu, 2002: Case Analyses and Numerical Simulation of Soil Thermal Impacts on Land Surface Energy Budget Based on an Off-Line Land Surface Model, ADVANCES IN ATMOSPHERIC SCIENCES, 19, 500-512.  doi: 10.1007/s00376-002-0082-0
    [10] Dai Yongjiu, Zeng Qingcun, 1997: A Land Surface Model (IAP94) for Climate Studies Part I: Formulation and Validation in Off-line Experiments, ADVANCES IN ATMOSPHERIC SCIENCES, 14, 433-460.  doi: 10.1007/s00376-997-0063-4
    [11] Dai Yongjiu, Xue Feng, Zeng Qingcun, 1998: A Land Surface Model (IAP94) for Climate Studies Part II: Implementation and Preliminary Results of Coupled Model with IAP GCM, ADVANCES IN ATMOSPHERIC SCIENCES, 15, 47-62.  doi: 10.1007/s00376-998-0017-5
    [12] Fuqiang YANG, Li DAN, Jing PENG, Xiujing YANG, Yueyue LI, Dongdong GAO, 2019: Subdaily to Seasonal Change of Surface Energy and Water Flux of the Haihe River Basin in China: Noah and Noah-MP Assessment, ADVANCES IN ATMOSPHERIC SCIENCES, 36, 79-92.  doi: 10.1007/s00376-018-8035-4
    [13] SONG Xiang and ZENG Xiaodong*, , 2014: Investigation of Uncertainties of Establishment Schemes in Dynamic Global Vegetation Models, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 85-94.  doi: 10.1007/s00376-013-3031-1
    [14] Jinliang Liu, Han-Ru Cho, 2001: Effects of Topographic Slopes on Hydrological Proecsses and Climate, ADVANCES IN ATMOSPHERIC SCIENCES, 18, 733-741.
    [15] Chenxi WANG, Zhihua ZENG, Ming YING, 2020: Uncertainty in Tropical Cyclone Intensity Predictions due to Uncertainty in Initial Conditions, ADVANCES IN ATMOSPHERIC SCIENCES, 37, 278-290.  doi: 10.1007/s00376-019-9126-6
    [16] HU Shujuan, CHOU Jifan, 2004: Uncertainty of the Numerical Solution of a Nonlinear System's Long-term Behavior and Global Convergence of the Numerical Pattern, ADVANCES IN ATMOSPHERIC SCIENCES, 21, 767-774.  doi: 10.1007/BF02916373
    [17] Deliang CHEN, Christine ACHBERGER, Jouni R¨AIS¨ANEN, Cecilia HELLSTR¨OM, 2006: Using Statistical Downscaling to Quantify the GCM-Related Uncertainty in Regional Climate Change Scenarios: A Case Study of Swedish Precipitation, ADVANCES IN ATMOSPHERIC SCIENCES, 23, 54-60.  doi: 10.1007/s00376-006-0006-5
    [18] Chengjun XIE, Tongwen WU, Jie ZHANG, Kalli FURTADO, Yumeng ZHOU, Yanwu ZHANG, Fanghua WU, Weihua JIE, He ZHAO, Mengzhe ZHENG, 2023: Spatial Inhomogeneity of Atmospheric CO2 Concentration and Its Uncertainty in CMIP6 Earth System Models, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 2108-2126.  doi: 10.1007/s00376-023-2294-4
    [19] Yuejian ZHU, 2005: Ensemble Forecast: A New Approach to Uncertainty and Predictability, ADVANCES IN ATMOSPHERIC SCIENCES, 22, 781-788.  doi: 10.1007/BF02918678
    [20] Zhang Yu, Lu Shihua, 2002: Development and Validation of a Simple Frozen Soil Parameterization Scheme Used for Climate Model, ADVANCES IN ATMOSPHERIC SCIENCES, 19, 513-527.  doi: 10.1007/s00376-002-0083-z

Get Citation+

Export:  

Share Article

Manuscript History

Manuscript received: 17 April 2022
Manuscript revised: 23 July 2022
Manuscript accepted: 17 August 2022
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Elucidating Dominant Factors Affecting Land Surface Hydrological Simulations of the Community Land Model over China

    Corresponding author: Binghao JIA, bhjia@mail.iap.ac.cn
  • 1. School of Mathematics and Computational Science, and Key Laboratory of Intelligent Control Technology for Wuling-Mountain Ecological Agriculture in Hunan Province, Huaihua University, Huaihua, Hunan 418008, China
  • 2. Department of Geological Sciences, The John A. and Katherine G. Jackson School of Geosciences, University of Texas at Austin, Austin, Texas 78712-1722, USA
  • 3. State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
  • 4. National Meteorological Information Center, China Meteorological Administration, Beijing 100081, China

Abstract: In order to compare the impacts of the choice of land surface model (LSM) parameterization schemes, meteorological forcing, and land surface parameters on land surface hydrological simulations, and explore to what extent the quality can be improved, a series of experiments with different LSMs, forcing datasets, and parameter datasets concerning soil texture and land cover were conducted. Six simulations are run for the Chinese mainland on 0.1° × 0.1° grids from 1979 to 2008, and the simulated monthly soil moisture (SM), evapotranspiration (ET), and snow depth (SD) are then compared and assessed against observations. The results show that the meteorological forcing is the most important factor governing output. Beyond that, SM seems to be also very sensitive to soil texture information; SD is also very sensitive to snow parameterization scheme in the LSM. The Community Land Model version 4.5 (CLM4.5), driven by newly developed observation-based regional meteorological forcing and land surface parameters (referred to as CMFD_CLM4.5_NEW), significantly improved the simulations in most cases over the Chinese mainland and its eight basins. It increased the correlation coefficient values from 0.46 to 0.54 for the SM modeling and from 0.54 to 0.67 for the SD simulations, and it decreased the root-mean-square error (RMSE) from 0.093 to 0.085 for the SM simulation and reduced the normalized RMSE from 1.277 to 0.201 for the SD simulations. This study indicates that the offline LSM simulation using a refined LSM driven by newly developed observation-based regional meteorological forcing and land surface parameters can better model reginal land surface hydrological processes.

摘要: 本文利用不同模式参数化方案、气象强迫和地表参数设计了六个模拟试验,并与观测进行对比分析,揭示了影响中国区域陆面水文要素(土壤湿度、蒸散发和雪深等)模拟的主要因子。结果表明,气象强迫是陆面水文过程模拟的主要影响因子,土壤湿度和雪深的模拟分别与土壤质地信息和陆面模式参数化方案紧密相关。利用新发展的融合观测信息的中国区域气象强迫和地表参数信息驱动陆面模式模拟,显著提高了中国大部分区域陆面水文过程模拟精度,并减少了模拟的不确定性。土壤湿度模拟与观测的相关系数从0.46提高到0.54,雪深模拟与观测的相关系数从0.54提高到0.67;土壤湿度模拟与观测的均方根误差从0.093降低到0.085,雪深模拟与观测的均方根误差从1.277降低到0.201。

    • The land surface plays a crucial role in connecting the water cycle and the other parts of the climate system (Oleson et al., 2008). For example, soil moisture (SM) controls the partitioning of net radiation into sensible and latent heat fluxes at the soil surface through soil–atmosphere interactions (Lawrence and Chase, 2007). Evapotranspiration (ET) supplies water vapor to the atmosphere and affects the local climate through precipitation (Makarieva et al., 2014). Snow cover affects the radiation balance because of its high albedo and snowmelt (Xie et al., 2018; Li et al., 2018a). Thus, accurately obtaining high-resolution spatiotemporal land surface hydrological information is not only important for weather forecasts and climate predictions but also for monitoring extreme events such as droughts and floods (Albergel et al., 2012; Wang and Dickinson, 2012; Wang et al., 2016; Jia et al., 2018).

      At present, land surface models (LSMs) have been widely used to provide estimates of land surface hydrological variables on regional and global scales (Oleson et al., 2008; Shi et al., 2013; Chen et al., 2013; Wang et al., 2016; Sun et al., 2017). Compared to in situ observations and satellite retrievals, LSM simulations are process-based and have the ability to describe dynamic variations in a consistent fashion both in time and space, so they can provide high spatiotemporal resolution and long-term estimates of land surface hydrological variables. However, hydrological modeling using offline LSMs still suffers from large errors. Hydrological modeling accuracy depends largely on the employed land surface parameterization scheme (Dirmeyer et al., 2006; Oleson et al., 2008, 2013; Chen et al., 2013; Zhang and Yang, 2016), the quality of the meteorological forcing (Wang and Zeng, 2011; Liu and Xie, 2013; Wang et al., 2016), and the land surface parameters. Numerous studies have found that the information on land surface parameters is very important in land surface and climate simulations (Tian et al., 2004; Lawrence and Chase, 2007; Gao and Jia, 2013). Land surface hydrological modeling using offline LSMs is sensitive to soil texture and land cover information (Liang and Dai, 2008; Yu et al., 2014; Zheng and Yang, 2016; Li et al., 2018b). Recently, there have been several efforts to understand the sources of model uncertainty and their impacts on the LSM simulations (Kato et al., 2006; Nearing et al., 2016). Moreover, there have been considerable efforts in improving LSMs (e.g. Niu et al., 2007; Lawrence et al., 2011), producing meteorological forcing datasets (e.g. Sheffield et al., 2006; He et al, 2020), and developing new land-surface parameter datasets such as soil texture datasets (Shangguan et al., 2012, 2013) and land cover datasets (Ran et al., 2012). However, it remains unknown how these efforts are reducing the simulation errors in a systematic fashion.

      In this study, a series of experiments are conducted to: (1) compare the impacts of the above three error sources on offline LSM simulations; and (2) determine to what extent newly developed regional meteorological forcing data, land surface parameter data, and LSM refinements improve hydrological simulations over China. These experiments include six simulations that use the Community Land Model versions 3.5 (CLM3.5, Oleson et al., 2007) and 4.5 (CLM4.5, Oleson et al., 2013), which are driven by different forcings using different land surface parameter datasets.

      This paper is organized as follows. Section 2 briefly describes the two different LSMs, the two meteorological forcings, and the four datasets of land surface parameters. In section 3, we briefly describe the experiment design and analysis methods. In section 4, we focus on validating and comparing our results, which is followed by discussions in section 5. Finally, we give a summary and our conclusions in section 6.

    2.   Models and data
    • The LSMs used in this study are CLM3.5 (Oleson et al., 2007) and CLM4.5 (Lawrence et al., 2011; Swenson and Lawrence, 2012; Oleson et al., 2013) released by the National Center for Atmospheric Research (NCAR). CLM4.5 is an extensive update of CLM 3.5 and CLM4, with the following major changes: (1) a 15-layer total soil column of 0–42.1032 m with variable thickness ranging from 0.0175 to 13.8512 m; (2) new treatments of soil column–groundwater interactions and soil evaporation; (3) revised hydraulic properties of frozen soils, which increases the consistency between soil water state and water table position and allows for a perched water table above icy permafrost ground; and (4) a revised snow model (Lawrence et al., 2011; Swenson and Lawrence, 2012; Oleson et al., 2013). For further details, the reader is referred to Oleson et al. (2013).

    • In this study, we used two sets of forcing data developed by different institutions to drive CLM3.5 or CLM4.5 over China. The first forcing is the global dataset developed by Princeton University (Sheffield et al., 2006, hereafter Princeton). It covers the period of 1948–2008 with a spatial resolution of 1° × 1° and a temporal resolution of 3 hr. More details on the Princeton meteorological forcing data can be found in Sheffield et al. (2006). Several researchers have assessed or tested this dataset in offline LSM simulations and suggested that it is a reliable atmospheric forcing dataset for LSMs (Sheffield and Wood, 2007; Wang and Zeng, 2011). The other regional forcing dataset is the China Meteorological Forcing Dataset (CMFD) established and updated by the Institute of Tibetan Plateau Research, Chinese Academy of Sciences (Yang and He, 2019; He et al., 2020). The CMFD includes 3-h 0.1° meteorological variables for China from 1979 to 2018 and combines the Global Land Data Assimilation System (GLDAS) dataset (Rodell et al., 2004) and a suite of observation-based products from China. Several previous studies have found that the CMFD is a reliable atmospheric forcing dataset for LSM simulation in China (Liu and Xie, 2013; Sun et al., 2017; Lu et al., 2020).

    • In this study, two sets of soil texture datasets and land cover datasets developed by different institutions are employed in CLM4.5. For the two soil texture datasets, one is the default global soil texture dataset in CLM4.5, which is derived from the Food and Agriculture Organization/United Nations Educational, Scientific and Cultural Organization (FAO/UNESCO) soil map of the world at the scale of 1:5 million. The other is an alternative soil texture dataset of the Chinese mainland developed by Shangguan et al. (2012, 2013) at Beijing Normal University in China that uses the Second National Soil Survey (1979–85) China soil map at a scale of 1:1 million and about 9000 soil profiles, which we coded as new soil (NS) texture. For a more detailed data description, see Shangguan et al. (2012, 2013). The first land cover datasets is the default global land cover data in CLM4.5, which contains distributions of plant functional type, glaciers, lakes and wetlands, and urban areas (Table 1). The other land cover dataset is the multisource integrated China land cover dataset (MICL) developed by Ran et al. (2012) at the Cold and Arid Regions Environmental and Engineering Institute (CAREEI), CAS that includes merged multisource land cover/use data based on Dempster–Shafer theory. For more details on the MICL data, the reader is referred to Ran et al. (2012).

      DataSoil textureLand coverReference
      CLM4.5
      Default
      1. Derived from the FAO/UNESCO world soil map (1:5 million) and about 60 soil profile data;
      2. Resolution: 5';
      3. The soil column including nine layers with variable thickness;
      4. The IGBP soil dataset of 4931 soil mapping units and their sand and clay content for each soil layer were used to create a mineral soil texture dataset
      1. Spatial resolution: 0.05°;
      2. PFTs were derived from the MODIS satellite product from November 2000 to October 2001;
      3. Glaciers were derived from IGBP DISCover data;
      4. Lakes and wetlands were derived from perennial freshwater lakes and swamps/marshes;
      5. Urban areas were derived from population density data, Landscan 2004
      Lawrence and Chase (2007); Loveland et al. (2000); Oleson et al. (2013); Bonan et al. (2002)
      New1. Derived from the Second National Soil Survey (1979–85) China soil map at a scale of 1:1 million and about 9000 soil profiles;
      2. Resolution: 0.5°;
      3. The soil column of 0–2.296 m including eight layers with variable thickness, for convenience of use, the first two layers are too thin and combined;
      4. The soil dataset of 925 soil mapping units in China and their soil type were used to create soil properties including a soil texture dataset
      1. Spatial resolution: 1 km;
      2. Merging Vegetation Atlas of China (1:1 000 000), Land use map of China (1:100 000) in 2000, Glacier distribution map of China (1:100000), Swamp-wetland map of China (1:1 000 000) and MODIS2001 product using IGBP classification system
      Shangguan et al. (2012, 2013); Hou (2001); Wu and Li (2004); Zhang (2002); Friedl et al. (2002)
      * Food and Agriculture Organization, United Nations Educational, Scientific and Cultural Organization (FAO/UNESCO); International Geosphere-Biosphere Programme (IGBP); Moderate resolution Imaging Spectroradiometer (MODIS); Plant functional Types (PFTs); International Geosphere-Biosphere Programme Data and Information System Global 1-km Land Cover Data Set (IGBP DISCover).

      Table 1.  Comparison of the major features of the two sets of land-surface parameters information data.

      Table 1 summarizes the primary features, compositions, and the differences between the two types of land-surface parameters data, including a soil texture dataset and a land cover dataset.

    • In this study, several comprehensive observation datasets, which include observation-based gridded ET data, in situ SM, and in situ snow depth (SD), were employed to evaluate and compare the six simulations.

    • The in situ SM observations from 778 agricultural meteorological stations covering the growing season (March to October) from 1992 to 2013 with three SM values per month (i.e., on the 8th, 18th, and 28th of each month) were obtained from the China Meteorological Administration (CMA) National Meteorological Information Center (NMIC, http://data.cma.cn). The in situ SM observations at 0–10-cm depths from 308 stations from 2004 to 2008 were used in this study. We processed the in situ site SM measurements into monthly volumetric water content data by the quality-control method described in Jia et al. (2015). Since the simulated results represent the average value of each grid cell, we selected the grid cell closest to the observation station for matching. In addition, the 308 SM sites were grouped based on the eight major basins over the Chinese mainland. Then, eight sub-regions were obtained (Fig. 1g).

      Figure 1.  Spatial distribution for the five-year (2004–08) averaged volumetric SM derived from the six simulations, the in situ site measurements in the 0–10-cm soil layer, and the location of the eight basins in the Chinese mainland: (a) Prin_CLM3.5; (b) CMFD_CLM3.5; (c) CMFD_CLM4.5; (d) CMFD_CLM4.5_NS; (e) CMFD_CLM4.5_MICL; (f) CMFD_CLM4.5_NEW; (g) in situ site observations; and (h) the location of the eight basins in China. I: Songhua River basin; II: Haihe River basin; III: Yellow River basin; IV: Huaihe River basin; V: Heihe River basin; VI:Tarim River basin; VII: Yangtze River basin; VIII: Zhujiang River basin.

    • The global monthly ET product used in this study is based on observations and covers from 1982 to 2011 with 0.5° spatial resolution (Jung et al., 2010, 2011; https://www.bgc-jena.mpg.de/geodb/projects/Data.php). We code this observation-based ET data as Obs_MTE. This product has been used to compare and evaluate with other ET products at regional and global scales in many previous studies (Jung et al., 2010; Shi et al., 2013; Sun et al., 2017).

    • The in situ SD data were obtained from NMIC at CMA (http://data.cma.cn). The original data were collected daily from 2417 operational meteorological stations (including national general reference stations and national base stations) from 1951 and are constantly updated.

      We performed a simple quality control procedure on the original data in terms of observation frequency and threshold detection (plausible value check), and the percentage of valid measurements during the snow period (November to April) from 2004 to 2008 in China was over 80%. In addition, we removed all stations where their mean SD measurement records during the snow period (November to April) from 2004 to 2008 were less than 1 mm. Finally, monthly mean daily SD values at 1390 stations during the snow period from 2004 to 2008 in China were used to evaluate the six simulations of SD. The matching method for the site measurements and simulated values was the same as that for SM. Here, we considered three main snow regions in China: Northeast China (NE: 35°–54°N, 115°–136°E), Xinjiang area (XJ: 34°–49°N, 73°–96°E), and the Tibetan Plateau (TP: 26°–41°N, 73°–105°E; see Fig. 6a). To avoid partial representation of SD site measurements in the simulated grid, the same method as that for SM was used for the three areas of China, and the mean SD values on each area were compared.

    3.   Experimental design
    • As described in section 1, the performances of the hydrological simulations using LSMs are subject to uncertainties arising from meteorological forcing data, LSM parameterization schemes, and land surface parameters. In this study, a series of experiments were designed to compare the uncertainty influences of these factors and explore to what extent new meteorological forcing data, land surface parameter information, and the updated CLM4.5 could improve hydrological modeling for China. Table 2 gives the list of experiments.

      ExperimentCLM3.5CLM4.5Princeton ForcingCMFD
      Forcing
      Original
      Soil
      New
      Soil
      Original land coverMICL
      Prin_CLM3.5YYYY
      CMFD_CLM3.5YYYY
      CMFD_CLM4.5YYYY
      CMFD_CLM4.5_NSYYYY
      CMFD_CLM4.5_MICLYYYY
      CMFD_CLM4.5_NEWYYYY

      Table 2.  List of experiments.

      First, the Princeton forcing data and CMFD forcing data were used to drive the same CLM3.5 model that consisted of CLM default land surface parameter information; these experiments were coded as Prin_CLM3.5 and CMFD_CLM3.5. We compared these two cases with each other to highlight the influence of forcing on offline LSM modeling. Second, the same CMFD forcing data were used to drive CLM3.5 and CLM4.5 using the same CLM default land surface parameter information; these experiments were coded as CMFD_CLM3.5 and CMFD_CLM4.5. These two cases were compared with each other to evaluate the effects of the considered LSM parameterization scheme on the offline LSM simulations. Finally, the same CMFD forcing data were used to drive CLM4.5 using four different land surface parameter information datasets, such as soil texture and land cover (i.e., default, replaced with NS only, replaced with MICL only, and replaced both NS and MICL); these experiments were coded as CMFD_CLM4.5, CMFD_CLM4.5_NS, CMFD_CLM4.5_MICL, and CMFD_CLM4.5_NEW, respectively. We compared these four cases with each other to evaluate the impact of the surface parameter information on the offline LSM modeling.

      All six simulations were run at a spatial resolution of 0.1° × 0.1° from 1979 to 2008 in China. To achieve an equilibrium state in CLM3.5 and CLM4.5, CLM3.5 was run forced by Princeton forcing data from 1948 to 2008 for 61 years, and the first file on 1 January 2009 was saved and used to initialize all six simulations.

    4.   Results
    • In this subsection, six simulations of SM at 0–10-cm depths were evaluated using in situ site observations from 308 stations in China. Here, the multiple soil layers in the CLM were adjusted to the observed 0–10-cm depths using the weighted averages of the soil layer thicknesses. Because of the strong spatial heterogeneity of SM, assessing the representativeness of site measurements is a challenging task. Previous studies have used the regional average of the simulated SM values when comparing with station observations (e.g., Xia et al., 2014; Wang et al., 2016). Here, we employ three methods to obtain relatively robust evaluation results: (1) individual station; (2) statistics from 308 stations over the Chinese mainland; (3) the average of these stations over eight major basins of China.

      Figure 1 shows the spatial distribution of the five-year (2004–08) averaged volumetric SMs derived from the six simulations, the in situ site measurements in the 0–10-cm soil layer for the growing season, and the locations of the eight basins in the Chinese mainland. Compared with the observed SM values (Fig.1g), the six simulated SMs (Figs.1af) all generally captured the spatial pattern and dry/wet centers of the SM in most cases, but they displayed different spatial variations.

      The correlation coefficient (R) and root-mean-square error (RMSE) were used to examine the performance of the six SM simulations. Figure 2 shows the Differences in the R (RD) of the six modeled SM results against the in situ site measurements. Compared to Prin_CLM3.5, CMFD_ CLM3.5 improved the simulation of SM over most of China (RD>0, for 216 of the 304 stations), except for most of the Xinjiang province in northwestern China. The most pronounced increase (RD>0.2) of R was seen over most of northeastern China and the north China plain (the Huang-Huai-Hai plain; 45 stations) (Fig. 2a). Compared to CMFD_CLM3.5, CMFD_CLM4.5 improved the simulation of SM in many areas (161 of the 304 stations) (Fig. 2b). Compared to CMFD_CLM4.5, CMFD_CLM4.5_NS and CMFD_ CLM4.5_NEW both improved the simulation of SM in many areas (188 of the 308 stations and 187 of the 308 stations, respectively). Compared to CMFD_CLM4.5_NS, CMFD_CLM4.5_NEW did not improve significantly. This indicates that the new soil texture data have a large influence on SM simulation and can significantly improve SM simulation over most of China. Moreover, meteorological forcing and soil texture information are the two major sources of uncertainty in SM modeling by LSMs (Figs. 2ce). Compared to Prin_CLM3.5, CMFD_CLM4.5_NEW, which employed the LSM refinements driven by a forcing that merges more observation information using new soil texture data and land cover data, significantly improved the simulation of SM over most of China (RD>0, for 213 of the 308 stations; RD>0.2, for 75 of the 308 stations).

      Figure 2.  Difference in the R of the modeled SM against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_ CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_ CLM4.5; and (f) CMFD_CLM4.5_ NEW minus Prin_CLM3.5.

      Figure 3 shows the differences in the RMSE (RMSED) of the six SM simulations against the in situ site measurements. From Fig. 3f, we can see that CMFD_CLM4.5_NEW had reduced RMSE in general over most of China (183 of the 308 stations), and the most pronounced reduction (> 0.05) of the RMSE can be seen over most of northeastern China and the north China plains (37 stations). A reduction of RMSE can be seen easily in Fig. 3a (182 of 308 stations), Fig. 3c (190 of 308 stations), and Fig. 3e (168 of 308 stations). This indicates that meteorological forcing and soil texture data were the two most important sources of biases in the SM simulations when using a LSM.

      Figure 3.  Difference in the RMSE of the modeled SM against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_ CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_CLM4.5; and (f) CMFD_CLM4.5_NEW minus Prin_CLM3.5.

      Table 3 shows the mean R and RMSE of the SM between the simulations and the site observations at 0–10-cm depths for the growing season over all 308 stations of the Chinese mainland during 2004–08. The mean R varied from 0.46 in Prin_CLM3.5 to 0.54 in CMFD_CLM4.5_NEW, and CMFD_CLM4.5_NEW displayed a significant improvement in the SM simulation (around 17.4%). The mean RMSE values varied from 0.093 in Prin_CLM3.5 to 0.085 in CMFD_CLM4.5_NEW, which showed an obvious reduction of around 8.2% in the latter model. Another pronounced reduction of the RMSE was found between CMFD_CLM4.5 and CMFD_CLM4.5_NS (about 10.1%), which indicated the soil texture data are an important uncertainty source in SM modeling. The largest contribution to performance for SM among these uncertainty sources was the meteorological forcing data, where the mean R value increased from 0.46 in Prin_CLM3.5 to 0.51 in CMFD_CLM3.5 (an improvement of around 10.9%), and the soil texture data came in second (around 3.8%).

      Prin_CLM3.5CMFD_CLM3.5CMFD_CLM4.5CMFD_CLM4.5_NSCMFD_CLM4.5_
      MICL
      CMFD_CLM4.5_
      NEW
      RRMSE/
      NRMSE
      RRMSE/
      NRMSE
      RRMSE/
      NRMSE
      RRMSE/
      NRMSE
      RRMSE/
      NRMSE
      RRMSE/
      NRMSE
      SM at 0–10-cm
      depth for
      308 stations*
      China0.46(77)0.09300.51
      (84, +10.9)
      0.0907
      (−2.5)
      0.52
      (85,
      +2.0)
      0.0912
      (+0.6)
      0.54
      (86, +3.8)
      0.0820
      (−10.1)
      0.52
      (86, 0)
      0.0935
      (+2.5)
      0.54
      (87, +3.8,
      +17.4)
      0.0854
      (−6.4, −8.2)
      SD for 1390 stations*China0.55(75)1.2770.71
      (92,+29.1)
      0.422
      (−66.9)
      0.76
      (97,+7.0)
      0.198
      (−53.1)
      0.75
      (97,−1.3)
      0.202
      (+2.02)
      0.75
      (97,−1.3)
      0.199
      (+0.51)
      0.76
      (97, +7.0)
      0.198
      (−53.1)
      ET**Songhua
      River basin
      0.95111.870.963
      (+1.26)
      11.03
      (−7.08)
      0.965
      (+0.21)
      11.70
      (+6.07)
      0.967
      (+0.21)
      11.49
      (−1.8)
      0.965
      (0)
      11.55
      (−1.28)
      0.968
      (+0.31, +1.79)
      11.40
      (−2.56, −3.96)
      Haihe River basin0.95113.090.956
      (+0.53)
      14.46
      (+10.5)
      0.960
      (+0.42)
      15.86
      (9.68)
      0.962
      (+0.21)
      14.56
      (−8.2)
      0.963
      (+0.31)
      15.14
      (−4.54)
      0.965
      (0.52, +1.47)
      15.09
      (−4.86, +15.3)
      Heihe River basin0.6309.200.605
      (−4.00)
      8.54
      (−7.17)
      0.640
      (+5.79)
      9.46
      (+10.7)
      0.647
      (+1.09)
      9.31
      (−1.59)
      0.647
      (+1.09)
      9.29
      (−1.8)
      0.656
      (+2.5, +4.13)
      9.14
      (−3.38, −0.65)
      Tarim River basin0.6629.690.722
      (+9.06)
      8.49
      (−12.4)
      0.730
      (+1.11)
      9.44
      (+11.2)
      0.738
      (+1.10)
      9.40
      (−0.42)
      0.742
      (+1.64)
      9.21
      (−2.44)
      0.749
      (+2.6, +13.14)
      9.16
      (−2.97, −5.47)
      Yellow River basin0.9259.830.926
      (+0.11)
      9.68
      (−1.53)
      0.925
      (−0.11)
      11.41
      (+17.9)
      0.926
      (+0.11)
      11.10
      (−2.72)
      0.932
      (+0.76)
      11.16
      (−2.19)
      0.934
      (+0.97,+0.97)
      11.03
      (−0.33,+12.2)
      Huaihe River basin0.92314.780.912
      (−1.2)
      16.76
      (+13.4)
      0.924
      (+1.32)
      23.98
      (+43.1)
      0.921
      (−0.33)
      22.96
      (−4.25)
      0.924
      (0)
      23.64
      (−1.42)
      0.922
      (−0.22, −0.11)
      23.84
      (−0.58, +6.13)
      Yangtze River basin0.93215.140.939
      (+0.75)
      15.84
      (+4.62)
      0.942
      (+0.32)
      19.55
      (+23.4)
      0.939
      (−0.32)
      19.22
      (−1.69)
      0.940
      (−0.21)
      19.64
      (+0.46)
      0.937
      (−0.53, +0.54)
      19.60
      (+0.26, +29.4)
      Zhujiang River basin0.89419.030.917
      (+2.57)
      20.68
      (+8.67)
      0.923
      (0.65)
      26.01
      (+25.7)
      0.914
      (−0.98)
      25.35
      (−2.54)
      0.920
      (−0.33)
      26.92
      (+3.5)
      0.912
      (−1.19, +2.01)
      26.59
      (+2.23, +39.7)
      China0.86812.180.881
      (+1.5)
      11.96
      (−1.81)
      0.883
      (+0.23)
      16.26
      (+35.9)
      0.883
      (0)
      16.03
      (−1.42)
      0.886
      (+0.34)
      16.01
      (−1.54)
      0.886
      (+0.34, +2.07)
      15.97
      (−1.78, +31.1)
      * The first number in the parentheses is the percentage of stations with a correlation that is statistically significant at the p = 0.05 level. The second and third numbers in the parentheses are the percentage increases relative to the paired experiments (i.e., the same as shown in Fig. 2). ** The number in the parentheses is the percentage of variation relative to the paired experiments (i.e., the same as shown in Fig. 2).

      Table 3.  Mean R and RMSE/NRMSE between the simulated and observation-based SM, SD, and ET from 2004 to 2008.

      Figure 4 shows time series comparisons between the observed and six modeled monthly average SM values over the eight major basins of the Chinese mainland. All six simulations captured the seasonal cycle of the observed SM reasonably well. For the temporal evolution, CMFD_CLM4.5_NS significantly increases R values over most basins of the eastern monsoon area such as the Songhua River Basin (from 0.59 to 0.73), the Yellow River Basin (from 0.83 to 0.90), the Yangtze River Basin (from 0.69 to 0.82), and the Zhujiang River Basin (from 0.50 to 0.56). And compared to Prin_CLM3.5, CMFD_CM3.5 increases R values from 0.59 to 0.68 in the Songhua River Basin, from 0.83 to 0.90 in the Yellow River Basin, and from 0.69 to 0.74 in the Yangtze River Basin. For the amplitude, all six simulations overestimated the SM values over all six basins of the eastern monsoon area and underestimated the SM values over the Heihe River Basin and the Tarim River Basin of northwestern China. The lowest SM occurs in 2007, and it reflects the severe drought over most of China in 2007 (Zhang et al., 2020).

      Figure 4.  Time series of the monthly volumetric SM (mm3 mm–3) for the 0–10-cm soil layer from in situ observations, Prin_CLM3.5, CMFD_CLM3.5, CMFD_CLM4.5, CMFD_CLM4.5_NS, CMFD_CLM4.5_MICL, and CMFD_CLM4.5_NEW during the period of 2008–13 in the eight major basins of China.

      Figures 14 demonstrate that meteorological forcing is the largest contributor to performance for SM over the Chinese mainland among the above three uncertainty sources, especially in the humid and semi-humid region of the eastern monsoon area. Former studies have shown that using accurate precipitation would effectively improve SM simulation (Wang and Zeng, 2011; Liu and Xie, 2013). To further understand the impact on precipitation of the meteorological forcing used in SM modeling and the time evolution of SM, we computed the mean monthly precipitation from the Princeton, CMFD, and CN05.1 precipitation datasets based on CMA gauge observations over eight major basins of the Chinese mainland (figure not shown). Precipitation from the Princeton and CMFD datasets display similar patterns and agree with CN05.1 precipitation very well, and CMFD generally agrees more closely with CN05.1 over most of the river basins, especially for the peak values. The magnitudes and patterns of SM shown in Fig. 4 are generally consistent with those of precipitation over most of the river basins. The SM variation extremes seen in Fig. 4 lag about 1–2 months behind precipitation extremes. For example, Huang-Huai-Hai Plain (i.e., the Yellow River Basin, the Huaihe River Basin, and the Haihe River Basin) experienced severe SM drought (the lowest SM value) in May 2007, while the smallest precipitation occurred in April 2007 over these basins.

      After meteorological forcing data, SM simulations are most sensitive to soil texture. CMFD_CLM4.5_NS increases R values over most arid and semi-arid regions of northern China, especially in the Haihe River Basin (From 0.68 to 0.75) and the Tarim River Basin (from 0.39 to 0.51). This is because the new regional observation-based soil texture dataset shows less sand than the default soil texture dataset in these basins, and it displays more consistency with the in situ measurements (Shangguan et al., 2012, Zheng and Yang, 2016). Overall, the results show that meteorological forcing is the most important factor governing output. Beyond that, SM seems to be also very sensitive to soil texture information.

    • We firstly compared the six simulations of the mean annual ET against the Obs_MTE product from 2004 to 2008 in China (figure not shown). The six ET simulations all capture the spatial pattern very well, but they underestimate the mean annual ET over most of the Chinese mainland. The CMFD_CLM4.5_NEW experiment displayed greater spatial variability and produced a closer simulation to the mean annual ET obtained using the Obs_MTE product.

      Figure 5 displays the RD of the six simulated ETs against the Obs_MTE product. Figure 5 shows that the soil texture and land cover information had a relatively small impact (–0.03<RD<0.03) on the ET simulation (Figs. 5ce). Replacing the default land cover information with regional land cover information can increase the R value in the ET simulation (Fig. 5d, RD>0 in 58% of the Chinese mainland). Figure 5a shows pronounced differences (RD>0.06 or RD <–0.06) over northwestern and southeastern China (around 22% of the Chinese mainland), which indicates that the meteorological forcing data was an important uncertainty source in the ET modeling. Compared to Prin_CLM3.5, CMFD_CLM4.5_NEW can increase the R value in ET simulation over most of China (RD>0, around 22% of the Chinese mainland; Fig. 5f).

      Figure 5.  Difference in the R of the modeled ET against the observation-based ET product Obs_MTE: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_CLM4.5; and (f) CMFD_CLM4.5_ NEW minus Prin_CLM3.5.

      We further compared the RMSED of the six simulated ET against observation-based product Obs_MTE (figure not shown); there is a pronounced reduction (RMSED<–2) in the RMSE for most of northeastern China and northwestern China (around 25% of the Chinese mainland); a pronounced increase (RMSED>2) can be seen in southeastern China and most of the TP (around 55% of the Chinese mainland). Compared with CLM3.5, CLM4.5 produced larger RMSE in the ET simulation over most of the Chinese mainland. Results of CMFD_CL4.5_NEW minus Prin_CLM3.5 display the combined influence of meteorological forcing and the LSM parameterization scheme; it has a similar pattern as CMFD_CLM4.5 minus CMFD_CLM3.5 and further indicates that the LSM parameterization scheme is the major source of bias in ET simulation.

      To further quantify the relative contribution of each uncertainty source to the performance of the ET simulation, the R and RMSE values of the modeled ETs against Obs_MTE product in the Chinese mainland and its eight major basins are summarized in Table 3. For all of the Chinese mainland, the mean R values varied from 0.868 in Prin_CLM3.5 to 0.886 in CMFD_CLM4.5_NEW, which indicates a slight improvement in ET simulation (around 2.07%) in CMFD_CLM4.5_NEW. The slight increases of R were located over arid and semi-arid areas of northern China such as the Tarim River basin (from 0.662 to 0.749), the Heihe River basin (from 0.63 to 0.656), the Haihe River basin (from 0.951 to 0.968), and the Yellow River basin (from 0.925 to 0.934). Meteorological forcing had a larger impact on the R values of the modeled ET relative to the Obs_MTE product. Compared to Prin_CLM3.4, CMFD_CLM3.5 increased the R value from 0.662 to 0.772 in the Tarim River basin, from 0.951 to 0.963 in the Songhua River basin, and from 0.951 to 0.956 in the Haihe River basin. In terms of the mean RMSE, a pronounced increase of RMSED>5 was seen in CMFD_CLM4.5_NEW relative to Prin_CLM3.5 over most of the southern humid region of China such as most of the TP, the Yangtze River basin (from 15.14 to 19.6), the Huaihe River basin (from 14.78 to 23.84), and the Zhujiang River basin (from 19.03 to 26.59). This represented the major influence of LSM parameterization scheme. A slight decrease of RMSE (0<RMSE<1) was found for most of northwestern China, which demonstrates the major influence of meteorological forcing data. The influence of land cover data is seen over most arid and semi-arid areas such as the Heihe River basin, the Tarim River basin, the Haihe River, and the Yellow River basin; CMFD_CLM4.5_MICL had a reduced RMSE compared to CMFD_CLM4.5. The new land cover dataset exhibits less bare soil than the default land cover dataset over these areas and displays more consistency with measurements (Ran et al., 2012, Yu et al., 2014).

      This indicates that the new parameterization scheme for ET in CLM4.5 slightly increases R values over most of the Chinese mainland, but it underestimates ET values (the corresponding figure is not shown in this paper) and produces lager RMSEs over most humid areas such as the Tibetan Plateau and southeastern China. The systematic biases of ET simulation in terms of RMSE when using CLM4.5 are mainly caused by the new parameterization scheme for ET. Another partial reason could be errors in the Obs_MTE products, which lead to overestimates in the ET value compared to the nine representative flux tower measurements of ET (Sun et al., 2017).

    • SD is an important hydrological variable on the land surface. We compare and evaluate the six modeled SD results during the snow period against the in situ SD measurements in this subsection.

      Figure 6 displays the RD of six simulated SD results against in situ site measurements. From Fig. 6a, we can see that CMFD_CLM3.5 significantly improved the simulation of SD over most of China (RD>0.2, for 779 of 1384 stations). Figure 6b also displays some obvious differences (RD>0.2, for 446 of 1384 stations). Only slight differences are seen in Figs. 6ce, where most have RD ranges from –0.1 to 0.1(1268, 1280, and 1216 of 1384 stations, respectively). This indicates that the choice of soil texture data and land cover data had only a small impact on the SD simulations compared to the choice of meteorological forcing data and snow parameterization.

      Figure 6.  Difference in the R of the modeled SD against the in situ site observations: (a) CMFD_CLM3.5 minus Prin_CLM3.5; (b) CMFD_CLM4.5 minus CMFD_CLM3.5; (c) CMFD_CLM4.5_NS minus CMFD_CLM4.5; (d) CMFD_CLM4.5_MICL minus CMFD_ CLM4.5; (e) CMFD_CLM4.5_NEW minus CMFD_ CLM4.5; and (f) CMFD_CLM4.5_NEW minus Prin_CLM3.5.

      The regional-averaged R values from the simulations relative to the mean site measurements over the Chinese mainland were compared to quantify the influences of each uncertainty source (Table 3). For all 1390 stations of the Chinese mainland, the mean R value varied from 0.55 in Prin_CLM3.5 to 0.76 in CMFD_CLM4.5_NEW, indicating a significant improvement in SD simulation (around 38.2%) by using CLM4.5, which includes newly developed regional meteorological forcing data and land surface parameter data. The percentage of stations with a correlation that was statistically significant at the p=0.05 level increased from 75% in Prin_CLM3.5 to 97% in CMFD_CLM4.5_NEW, which further indicates better performance of CMFD_CLM4.5_NEW in the six simulations. The largest contributor to the performance of the SD simulations among these error sources was the meteorological forcing data, where the mean R increased from 0.55 in Prin_CLM3.5 to 0.71 in CMFD_CLM4.5 (an improvement of around 29.1%).

      Another statistical index, normalized RMSE (NRMSE, i.e., RMSE/maximum of the sample), was used to evaluate the performances of the six simulations of SD. We computed the regional-averaged NRMSE values from the six simulations against the mean site measurements for the three major snow areas and the Chinese mainland (figure not shown). The NRMSE was reduced significantly from 1.277 for Prin_CLM3.5 to 0.201 for CMFD_CLM4.5_NEW (around 84.3%). NRMSE and Table 3 support similar conclusions as the above R index; the choices of atmospheric forcing data and the snow parameterization scheme in the LSM have the most important influences on SD simulation that used an LSM; CMFD_CLM4.5_NEW shows significant improvements in SD simulation by using an LSM.

      To further examine the performance of the six SD simulations, we compared the observed and six modeled monthly average SD time series during the snow period over the three major snow areas of the Chinese mainland China (Figs. 7ac). For the temporal evolution, compared to Prin_CLM3.5, CMFD_CLM4.5_NEW significantly increases R values over all three snow areas. For example, it increases R values from 0.89 to 0.97 over the northeast area, from 0.91 to 0.98 over the Xingjiang area, and from 0.75 to 0.94 over the TP area. For the amplitude, the six simulations showed large differences with each other. The four simulations using CLM4.5 overestimated the SD values, and the other simulations using CLM3.5 underestimated the SD values over all three areas. Figures 7ac indicate that the choice of snow parameterization scheme is the most important factor for the SD simulations over the three major snow areas of China. In CLM3.5, SD is computed from the snow water equivalent and snow density prior to the snow cover fraction computation. The snow parameterization in CLM3.5 produced lower SD than the observation. Compared to CLM3.5, the new snow parameterization in CLM4.5, where the SD is computed as a function of the change of snow water equivalent and the snow cover fraction (Swenson and Lawrence, 2012), simulated greater SD and higher extremes (SDV>1), and significantly increased the R values.

      Figure 7.  Time series of (a–c) the SD (cm) from in situ observations, Prin_CLM3.5, CMFD_CLM3.5, CMFD_CLM4.5, CMFD_CLM4.5_NS, CMFD_CLM4.5_MICL, and CMFD_CLM4.5_NEW; (d–f) snowfall (mm d–1) from Princeton forcing data and CMFD during the period of 2004–08 in the three major snow areas of China.

      The source of meteorological forcing data is another very important factor influencing the SD simulations, and we compared the mean monthly snowfall averaged in the three snow areas during the snow period using the Princeton and CMFD sources (Figs. 7df). The magnitudes and patterns of SD shown in Figs. 7ac are generally consistent with those of observed snowfall in these areas. The snowfall from the Princeton and CMFD forcing data displays similar patterns and time evolutions over all three snow areas. There is more snowfall from CMFD than from Princeton over the TP area, and CMFD_CLM3.5 simulated greater SD than Pin_CLM3.5 in this area. However, there is less snowfall from CMFD than from Princeton over the XJ area, and CMFD_CLM3.5 simulated lower SD than Pin_CLM3.5 in this area. This indicates that snowfall plays an important role in the SD simulations.

    5.   Discussion
    • In this study, several observation experiments were conducted to investigate the influence of the uncertainties arising from atmospheric forcing data, LSM parameterization scheme, and land surface parameters, and we explored to what extent the newly developed regional meteorological forcing data, land-surface parameter information data, and LSM refinements improved the land surface hydrological simulations over the Chinese mainland. We quantified the relative impacts of choice of atmospheric forcing data source, LSM, and land surface parameters in offline LSM simulations through comparing R and RMSE (or NRMSE) of simulated SM, ET, and SD values against the observations in the paired experiments.

      CMFD_CLM4.5_NEW generally increased the R values of analyzed SM, ET, and SD, and reduced the RMSE (or NRMSE) values of analyzed SM and SD over the Chinese mainland. However, it did not perform best for all variables over all areas; it showed a reduction of R and an increase in RMSE (or NRMSE) for some areas. For example, in the ET simulation, CMFD_CLM4.5_NEW had an increased R value over most of the Chinese mainland but also an increased RMSE over some areas such as southeastern China and the Tibetan Plateau. An ensemble approach was found to be an effective strategy for improving LSM modeling. For example, a multiple LSM ensemble [e.g., GLDAS, (Rodell et al., 2004); Global Soil Wetness Project (GSWP) (Dirmeyer et al., 2006); International Land Model Benchmarking (ILAMB) Project (Hoffman et al., 2017)], a multiple forcing ensemble (Wang et al., 2016), and an advance ensemble method such as Bayesian model averaging (BMA) have been found to further improve LSM simulations (Liu and Xie, 2013; Liu et al., 2018).

      In this study, there are some limitations to be noted. Some uncertainties in land surface hydrological simulations are beyond the scope of this current study, such as the error in observations, the error from the resolution difference between two datasets, and the error from the scale mismatch between simulations and observations. These error sources may be additional reasons for the differences between simulations and observations. In this study, to partially avoid the uncertainty from mismatch, we chose the LSM grid that most closely matched the distance between local site measurements. We first gave the statistics from all stations in China, and then analyzed the spatial average of site measurements over specific regions. Another error source results from the resolution difference between the two chosen datasets of forcing and land surface parameter information. So, we adjusted the resolution from the different datasets to the simulation resolution using the interpolation program from CLM. We did not consider the uncertainty resulting from this resolution difference in the present study. The third scale difference between the SM simulations and the observations is in the thicknesses of the observed soil layers, which are different from those required in the CLM models. We adjusted the multiple soil layer thicknesses in the LSM to the observed soil layers using a weighted-average approach based on the soil layer thicknesses in the LSM.

    6.   Summary and conclusions
    • LSMs have been widely used to produce long-term and high-resolution land surface hydrological variables. However, the quality of offline LSM modeling depends on the chosen meteorological forcing data, LSM parameterization scheme, and land surface parameter data. In this study, a series of paired experiments that employed different LSMs (CLM3.5 and CLM4.5), were driven by different meteorological forcing data (Princeton and CMFD), and considered different land surface parameters concerning soil texture and land cover [default, NS only, MICL only, and NEW (both NS and MICL)], were conducted. The aim was to examine the influences of the uncertainties arising from multiple sources in LSM modeling, and explore how the new forcing data, parameter data, and LSM refinements improve the land surface hydrological simulations over the Chinese mainland and its eight major basins. Our main conclusions are stated as follows.

      (1) For SM simulation at 0–10-cm depths, CMFD_CLM4.5_NEW significantly increased the R value from 0.46 to 0.54 (around 17.4%), especially in the arid and semi-arid areas [e.g., the Heihe River basin (around 20.8%), the Tarim River basin (around 25%), and the Songhua River basin (around 35.1%)], and decreased the RMSE value from 0.093 to 0.085 (around 8.2%), especially in the Huaihe River basin (around 33.1%) and the Songhua River basin (around 12.8%). The meteorological forcing is the major source of biases in SM simulation, especially in the humid and semi-humid region of the eastern monsoon area, and the soil texture is another important source of biases in SM simulation, especially over most arid and semi-arid regions of northern China.

      (2) For ET simulations, CMFD_CLM4.5_NEW had an increase in the R value from 0.868 to 0.886, especially over arid areas [e.g., the Tarim River basin (from 0.662 to 0.749)], where the meteorological forcing data was the main contributor to the observed improvement. However, it also increased the RMSE value from 12.18 to 15.97, especially over humid areas (e.g., the Zhujiang basin), where the different LSMs showed large differences and the LSM parameterization scheme is the largest source of bias in most of the southern humid region of China. The land cover information imparted a slight impact on the ET simulation over most arid and semi-arid areas.

      (3) For the SD simulation, CMFD_CLM4.5_NEW significantly increased the R value from 0.54 to 0.67 (around 37%), especially on the TP (from 0.42 to 0.76), and it pronouncedly reduced the NRMSE value from 1.277 to 0.201 (around 84.1%). The choice of meteorological forcing data is the most important factor impacting the SD simulation, while the new snow parameterization in the LSM came in second.

      The meteorological forcing is the most important factor governing output. Beyond that, SM simulation seems to be also very sensitive to soil texture information; SD simulation is also very sensitive to snow parameterization scheme in the LSM. CMFD_CLM4.5_NEW significantly improved the simulations in most cases over the Chinese mainland and its eight basins. This study indicates that the offline LSM simulation using a refined LSM driven by newly developed observation-based regional meteorological forcing and land surface parameters can better model regional land surface hydrological processes. It can provide regional high-quality and high spatiotemporal-resolution data of land surface hydrological variables for further investigation of the regional terrestrial water cycle and monitoring of local floods and droughts.

      Acknowledgements. This research was supported by the Natural Science Foundation of Hunan Province (Grant No. 2020JJ4074), the Second Tibetan Plateau Scientific Expedition and Research Program (STEP) (Grant No. 2019QZKK0206), the Youth Innovation Promotion Association CAS (2021073), the National Key Scientific and Technological Infrastructure project “Earth System Science Numerical Simulator Facility” (EarthLab), and the Huaihua University Double First-Class Initiative Applied Characteristic Discipline of Control Science and Engineering.

Reference

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return