Advanced Search
Article Contents

Evaluating Common Land Model Energy Fluxes Using FLUXNET Data

doi: 10.1007/s00376-017-6251-y

  • Given the crucial role of land surface processes in global and regional climates, there is a pressing need to test and verify the performance of land surface models via comparisons to observations. In this study, the eddy covariance measurements from 20 FLUXNET sites spanning more than 100 site-years were utilized to evaluate the performance of the Common Land Model (CoLM) over different vegetation types in various climate zones. A decomposition method was employed to separate both the observed and simulated energy fluxes, i.e., the sensible heat flux, latent heat flux, net radiation, and ground heat flux, at three timescales ranging from stepwise (30 min) to monthly. A comparison between the simulations and observations indicated that CoLM produced satisfactory simulations of all four energy fluxes, although the different indexes did not exhibit consistent results among the different fluxes. A strong agreement between the simulations and observations was found for the seasonal cycles at the 20 sites, whereas CoLM underestimated the latent heat flux at the sites with distinct dry and wet seasons, which might be associated with its weakness in simulating soil water during the dry season. CoLM cannot explicitly simulate the midday depression of leaf gas exchange, which may explain why CoLM also has a maximum diurnal bias at noon in the summer. Of the eight selected vegetation types analyzed, CoLM performs best for evergreen broadleaf forests and worst for croplands and wetlands.
    摘要: 陆面模式描述陆地表面与大气之间物质, 能量和动量交换过程, 因其可以为大气环流模式提供准确的下垫面条件而对全球气候变化研究非常重要. 随着人们对天气预报和气候预测精度的需求越来越高, 用观测资料验证和改进陆面模式也逐渐成为提高模式模拟能力的一种重要手段. 本研究利用覆盖不同气候区不同植被类型的20个FLUXNET站点观测数据检验通用陆面模式(Common Land Model, CoLM)的能量平衡模拟情况. 在对比分析观测和模拟的感热, 潜热, 净辐射和地表热通量时, 采用分解方法将每个观测值或模拟值分解为月平均, 日平均残差和逐时残差三个部分. 综合相关系数(R), 均方根误差(RMSE)和标准化误差(Nbias)对不同时间尺度和不同植被类型的能量通量分析, 结果发现CoLM能够很好地模拟4个地表能量通量, 其中模拟净辐射最好, 模拟潜热通量好于感热通量; 模式能够很好地模拟能量通量的季节变化, 但在有明显干湿季的站点低估潜热通量, 这可能与模式对干季土壤水的模拟存在缺陷有关; CoLM在夏季中午产生较大模拟误差, 这很可能是因为CoLM不能准确模拟光合午睡现象; 就不同的植被类型而言, CoLM在所选的8种植被类型中的常绿针叶林表现最好.
  • 加载中
  • Aitken A. P., 1973: Assessing systematic errors in rainfall-runoff models.J. Hydrol.,20, 121-136, doi: 10.1016/0022-1694(73) 90035-8.
    Baldocchi, D., Coauthors, 2001: FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide,water vapor, and energy flux densities. Bull. Amer. Meteor. Soc., 82, 2415-2434, doi: 10.1175/1520-0477 (2001)082<2415:FANTTS>2.3.CO;2.
    Baldocchi, D., Coauthors, 2016: Role of flux networks in benchmarking land atmosphere models. 2016 International Land Model Benchmarking (ILAMB) Workshop, Washington, DC, ILAMB.
    Best, M., Coauthors, 2015: The plumbing of land surface models: Benchmarking model performance. Journal of Hydrometeorology,16, 1425-1442, doi: 10.1175/JHM-D-14-0158.1.
    Blyth E., J. Gash, A. Lloyd, M. Pryor, G. P. Weedon, and J. Shuttleworth, 2010: Evaluating the JULES land surface model energy fluxes using FLUXNET data. Journal of Hydrometeorology,11, 509-519, doi: 10.1175/2009JHM1183.1.
    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, 150 pp.'s_guide?ev=auth_pub
    Bonan G. B., 2015: Ecological Climatology: Concepts and Applications. 3rd ed., Cambridge University Press.
    Bonan G. B., K. W. Oleson, M. Vertenstein, S. Levis, X. B. Zeng, Y. J. Dai, R. E. Dickinson, and Z. L. Yang, 2002: The land surface climatology of the community land model coupled to the NCAR community climate model. J. Climate,15, 3123-3149, doi: 10.1175/1520-0442(2002)015<3123:TLSCOT>2. 0.CO;2.
    Burgess S. S. O., M. A. Adams, N. C. Turner, and C. K. Ong, 1998: The redistribution of soil water by tree root systems. Oecologia,115, 306-311, doi: 10.1007/s004420050521.
    Caldwell M. M., T. E. Dawson, and J. H. Richards, 1998: Hydraulic lift: Consequences of water efflux from the roots of plants. Oecologia,113, 151-161, doi: 10.1007/ s004420050363.
    Collatz G. J., J. T. Ball, C. Grivet, and J. A. Berry, 1991: Physiological and environmental regulation of stomatal conductance,photosynthesis and transpiration: A model that includes a laminar boundary layer. Agricultural and Forest Meteorology, 54, 107-136, doi: 10.1016/0168-1923(91)90002-8.
    Dai Y. J., Q. C. Zeng, 1997: A land surface model (IAP94) for climate studies part I: Formulation and validation in off-line experiments.Adv. Atmos. Sci.,14, 433-460, doi: 10.1007/ s00376-997-0063-4.
    Dai, Y. J., Coauthors, 2003: The common land model.Bull. Amer. Meteor. Soc.,84, 1013-1023, doi: 10.1175/BAMS-84-8-1013.
    Dai Y. J., R. E. Dickinson, and Y.-P. Wang, 2004: A two-big-leaf model for canopy temperature,photosynthesis, and stomatal conductance. J. Climate, 17, 2281-2299, doi: 10.1175/1520-0442(2004)0172281:ATMFCT>2.0.CO;2.
    Dickinson R. E., A. Henderson-Sellers, P. J. Kennedy, and M. F. Wilson, 1986: Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-275-+STR, doi: 10.5065/D6668B58.
    Groenendijk, M., Coauthors, 2011: Assessing parameter variability in a photosynthesis model within and between plant functional types using global FLUXNET eddy covariance data. Agricultural and Forest Meteorology,151, 22-38, doi: 10.1016/j.agrformet.2010.08.013.
    Henderson-Sellers A., Z. L. Yang, and R. E. Dickinson, 1993: The project for intercomparison of land-surface parameterization schemes.Bull. Amer. Meteor. Soc.,74, 1335-1349, doi: 10.1175/1520-0477(1993)074<1335:TPFIOL>2.0.CO;2
    Huang C. L., X. Li, and J. Gu, 2008: Estimation of regional soil moisture by assimilating multi-sensor passive microwave remote sensing observations based on ensemble Kalman filter. Proc. IEEE International Geoscience and Remote Sensing Symposium, Boston, MA, IEEE, III-1052-III-1055, doi: 10.1109/IGARSS.2008.4779534.
    James L. D., S. J. Burges, 1982: Selection, calibration, and testing of hydrologic models. Hydrologic Modelling of Small Watersheds, C. T. Haan, Ed., ASAE Michigan, 435- 472.
    Ji, D., Coauthors, 2014: Description and basic evaluation of BNU-ESM version 1.Geosci. Model Dev. Discuss.,7, 1601-1647, doi: 10.5194/gmdd-7-1601-2014.
    Jung, M., Coauthors, 2010: Recent decline in the global land evapotranspiration trend due to limited moisture supply. Nature,467, 951-954, doi: 10.1038/nature09396.
    Kottek M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World map of the Köppen-Geiger climate classification updated.Meteor. Z.,15, 259-263, doi: 10.1127/0941-2948/ 2006/0130.
    Lee J.-E., R. S. Oliveira, T. E. Dawson, and I. Fung, 2005: Root functioning modifies seasonal climate. Proceedings of the National Academy of Sciences of the United States of America,102, 17 576-17 581, doi: 10.1073/pnas.0508785102.
    Li Y., X. Liu, and W. P. Li, 2012: Numerical simulation of land surface process at different underlying surfaces in Tibetan Plateau. Plateau Meteorology, 31, 581- 591. (in Chinese)
    Luo, Y. Q., Coauthors, 2012: A framework for benchmarking land models. Biogeosciences,9, 3857-3874, doi: 10.5194/bg-9-3857-2012.
    Ma L. L., L. L. Tang, and Z. L. Li, 2006: Simulation of land ETs of China with CoLM. Proc. SPIE 6200, Remote Sensing of the Environment: 15th National Symposium on Remote Sensing of China, Guiyan City, China, SPIE, 62000N, doi: 10.1117/ 12.681768.
    Meng C. L., Z. L. Li, X. Zhan, J. C. Shi, and C. Y. Liu, 2009: Land surface temperature data assimilation and its impact on evapotranspiration estimates from the Common Land Model. Water Resour. Res., 45,W02421, doi: 10.1029/2008WR006971.
    Neumann R. B., Z. G. Cardon, 2012: The magnitude of hydraulic redistribution by plant roots: A review and synthesis of empirical and modeling studies. New Phytologist,194, 337-352, doi: 10.1111/j.1469-8137.2012.04088.x.
    Oleson, K. W., Coauthors, 2008: Improvements to the Community Land Model and their impact on the hydrological cycle. J. Geophys. Res., 113(G1),G01021, doi: 10.1029/2007 JG000563.
    Oleson, K. W., Coauthors, 2013: Technical Description of Version 4.5 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-503+STR,420 pp, doi: 10.5065/ D6RR1W7M.
    Oliveira R. S., T. E. Dawson, S. S. O. Burgess, and D. C. Nepstad, 2005: Hydraulic redistribution in three Amazonian trees. Oecologia,145, 354-363, doi: 10.1007/s00442-005-0108-2.
    Pessarakli M., 2005: Handbook of Photosynthesis. 2nd ed., CRC Press.
    Prieto I., C. Armas, and F. I. Pugnaire, 2012: Water release through plant roots: New insights into its consequences at the plant and ecosystem level. New Phytologist,193, 830-841, doi: 10.1111/j.1469-8137.2011.04039.x.
    Quijano J. C., P. Kumar, 2015: Numerical simulations of hydraulic redistribution across climates: The role of the root hydraulic conductivities.Water Resour. Res.,51, 8529-8550, doi: 10.1002/2014WR016509.
    Roerink G. J., Z. Su, and M. Menenti, 2000: S-SEBI: A simple remote sensing algorithm to estimate the surface energy balance. Physics and Chemistry of the Earth,Part B: Hydrology, Oceans and Atmosphere, 25, 147-157, doi: 10.1016/S1464-1909(99)00128-8.
    Sellers, P. J., Coauthors, 1997: Modeling the exchanges of energy,water, and carbon between continents and the atmosphere. Science, 275, 502-509, doi: 10.1126/science.275. 5299.502.
    Shangguan W., Y. J. Dai, Q. Y. Duan, B. Y. Liu, and H. Yuan, 2014: A global soil data set for earth system modeling. Journal of Advances in Modeling Earth Systems,6, 249-263, doi: 10.1002/2013MS000293.
    Stöckli, R., Coauthors, 2008: Use of FLUXNET in the community land model development. J. Geophys. Res., 113,G01025, doi: 10.1029/2007JG000562.
    Taylor K. E., 2001: Summarizing multiple aspects of model performance in a single diagram.J. Geophys. Res.,106, 7183-7192, doi: 10.1029/2000JD900719.
    Twine, T. E., Coauthors, 2000: Correcting eddy-covariance flux underestimates over a grassland. Agricultural and Forest Meteorology,103, 279-300, doi: 10.1016/S0168-1923(00) 00123-4.
    Wang G. L., C. Alo, R. Mei, and S. S. Sun, 2011: Droughts,hydraulic redistribution, and their impact on vegetation composition in the Amazon forest. Plant Ecology, 212, 663-673, doi: 10.1007/s11258-010-9860-4.
    Wilson, K., Coauthors, 2002: Energy balance closure at FLUXNET sites. Agricultural and Forest Meteorology,113, 223-243, doi: 10.1016/S0168-1923(02)00109-0.
    Xin Y. F., L. G. Bian, and X. H. Zhang, 2006: The application of CoLM to arid region of Northwest China and Qinghai-Xizang Plateau. Plateau Meteorology,25, 567-574, doi: 10.3321/ j.issn:1000-0534.2006.04.002. (in Chinese)
    Xu T. R., S. L. Liang, and S. M. Liu, 2011: Estimating turbulent fluxes through assimilation of geostationary operational environmental satellites data using ensemble Kalman filter. J. Geophys. Res., 116,D09109, doi: 10.1029/2010JD015150.
    Yan B. Y., R. E. Dickinson, 2014: Modeling hydraulic redistribution and ecosystem response to droughts over the Amazon basin using Community Land Model 4.0 (CLM4).J. Geophys. Res.,119, 2130-2143, doi: 10.1002/2014JG002694.
    Yang K., Y. Y. Chen, and J. Qin, 2009: Some practical notes on the land surface modeling in the Tibetan Plateau. Hydrology and Earth System Sciences,13, 687-701, doi: 10.5194/hess-13-687-2009.
    Yuan H., Y. J. Dai, Z. Q. Xiao, D. Y. Ji, and W. Shangguan, 2011: Reprocessing the MODIS Leaf Area Index products for land surface and climate modelling. Remote Sensing of Environment,115, 1171-1187, doi: 10.1016/j.rse.2011.01.001.
    Zeiger E., G. D. Farquhar, and I. R. Cowan, 1987: Stomatal Function. Stanford University Press.
    Zeng X. B., M. Shaikh, Y. J. Dai, R. E. Dickinson, and R. Myneni, 2002: Coupling of the common land model to the NCAR community climate model. J. Climate,15, 1832-1854, doi: 10.1175/1520-0442(2002)015<1832:COTCLM>2.0.CO;2.
    Zhang S. W., D. Q. Li, and C. J. Qiu, 2011: A multimodel ensemble-based Kalman filter for the retrieval of soil moisture profiles.Adv. Atmos. Sci.,28, 195-206, doi: 10.1007/s00376-010-9200-6.
    Zheng J., Z. H. Xie, Y. J. Dai, X. Yuan, and X. Q. Bi, 2009: Coupling of the common land model (CoLM) with the regional climate model (RegCM3) and its preliminary validation. Chinese Journal of Atmospheric Sciences,33, 737-750, doi: 10.3878/j.issn.1006-9895.2009.04.08. (in Chinese)
  • [1] Daeun KIM, Yoon-Jin LIM, Minseok KANG, Minha CHOI, 2016: Land Response to Atmosphere at Different Resolutions in the Common Land Model over East Asia, ADVANCES IN ATMOSPHERIC SCIENCES, 33, 391-408.  doi: 10.1007/s00376-015-5059-x
    [2] Jing ZOU, Zhenghui XIE, Chesheng ZHAN, Feng CHEN, Peihua QIN, Tong HU, Jinbo XIE, 2019: Coupling of a Regional Climate Model with a Crop Development Model and Evaluation of the Coupled Model across China, ADVANCES IN ATMOSPHERIC SCIENCES, 36, 527-540.  doi: 10.1007/s00376-018-8160-0
    [3] Lei ZHANG, Xiquan DONG, Aaron KENNEDY, Baike XI, Zhanqing LI, 2017: Evaluation of NASA GISS Post-CMIP5 Single Column Model Simulated Clouds and Precipitation Using ARM Southern Great Plains Observations, ADVANCES IN ATMOSPHERIC SCIENCES, 34, 306-320.  doi: 10.1007/s00376-016-5254-4
    [4] Samuel Takele KENEA, Young-Suk OH, Jae-Sang RHEE, Tae-Young GOO, Young-Hwa BYUN, Shanlan LI, Lev D. LABZOVSKII, Haeyoung LEE, Robert F. BANKS, 2019: Evaluation of Simulated CO2 Concentrations from the CarbonTracker-Asia Model Using In-situ Observations over East Asia for 2009-2013, ADVANCES IN ATMOSPHERIC SCIENCES, 36, 603-613.  doi: 10.1007/s00376-019-8150-x
    [5] Jianfeng WANG, Ricardo M. FONSECA, Kendall RUTLEDGE, Javier MARTÍN-TORRES, Jun YU, 2020: A Hybrid Statistical-Dynamical Downscaling of Air Temperature over Scandinavia Using the WRF Model, ADVANCES IN ATMOSPHERIC SCIENCES, 37, 57-74.  doi: 10.1007/s00376-019-9091-0
    [6] Susannah M. BURROWS, Aritra DASGUPTA, Sarah REEHL, Lisa BRAMER, Po-Lun MA, Philip J. RASCH, Yun QIAN, 2018: Characterizing the Relative Importance Assigned to Physical Variables by Climate Scientists when Assessing Atmospheric Climate Model Fidelity, ADVANCES IN ATMOSPHERIC SCIENCES, 35, 1101-1113.  doi: 10.1007/s00376-018-7300-x
    [7] YAN Qing*, WANG Huijun, Ola M. JOHANNESSEN, and ZHANG Zhongshi, 2014: Greenland Ice Sheet Contribution to Future Global Sea Level Rise based on CMIP5 Models, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 8-16.  doi: 10.1007/s00376-013-3002-6
    [8] Zhou Tianjun, Zhang Xuehong, Yu Yongqiang, Yu Rucong, Wang Shaowu, 2000: The North Atlantic Oscillation Simulated by Versions 2 and 4 of IAP/ LASG GOALS Model, ADVANCES IN ATMOSPHERIC SCIENCES, 17, 601-616.  doi: 10.1007/s00376-000-0023-8
    [9] ZHOU Tianjun, WU Bo, WEN Xinyu, LI Lijuan, WANG Bin, 2008: A Fast Version of LASG/IAP Climate System Model and Its 1000-year Control Integration, ADVANCES IN ATMOSPHERIC SCIENCES, 25, 655-672.  doi: 10.1007/s00376-008-0655-7
    [10] CAO Ning, REN Baohua, ZHENG Jianqiu, 2015: Evaluation of CMIP5 Climate Models in Simulating 1979-2005 Oceanic Latent Heat Flux over the Pacific, ADVANCES IN ATMOSPHERIC SCIENCES, 32, 1603-1616.  doi: 10.1007/s00376-015-5016-8
    [11] Xiaoyong YU, Chengyan LIU, Xiaocun WANG, Jian CAO, Jihai DONG, Yu LIU, 2022: Evaluation of Arctic Sea Ice Drift and its Relationship with Near-surface Wind and Ocean Current in Nine CMIP6 Models from China, ADVANCES IN ATMOSPHERIC SCIENCES, 39, 903-926.  doi: 10.1007/s00376-021-1153-4
    [12] Masoud NOBAKHT, Bahram SAGHAFIAN, Saleh AMINYAVARI, 2021: Skill Assessment of Copernicus Climate Change Service Seasonal Ensemble Precipitation Forecasts over Iran, ADVANCES IN ATMOSPHERIC SCIENCES, 38, 504-521.  doi: 10.1007/s00376-020-0025-7
    [13] DONG Siyan, XU Ying, ZHOU Botao, SHI Ying, 2015: Assessment of Indices of Temperature Extremes Simulated by Multiple CMIP5 Models over China, ADVANCES IN ATMOSPHERIC SCIENCES, 32, 1077-1091.  doi: 10.1007/s00376-015-4152-5
    [14] Botao ZHOU, Ying SHI, Ying XU, 2016: CMIP5 Simulated Change in the Intensity of the Hadley and Walker Circulations from the Perspective of Velocity Potential, ADVANCES IN ATMOSPHERIC SCIENCES, 33, 808-818.  doi: 10.1007/s00376-016-5216-x
    [15] Fang ZHOU, Hong-Li REN, 2017: Dynamical Feedback between Synoptic Eddy and Low-Frequency Flow as Simulated by BCC_CSM1.1(m), ADVANCES IN ATMOSPHERIC SCIENCES, 34, 1316-1332.  doi: 10.1007/s00376-017-6318-9
    [16] Daosheng XU, Jeremy Cheuk-Hin LEUNG, Banglin ZHANG, 2023: A Time Neighborhood Method for the Verification of Landfalling Typhoon Track Forecast, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 273-284.  doi: 10.1007/s00376-022-1398-6
    [17] Xiaojing YU, Lixia ZHANG, Tianjun ZHOU, Jianghua ZHENG, 2024: Assessing the Performance of CMIP6 Models in Simulating Droughts across Global Drylands, ADVANCES IN ATMOSPHERIC SCIENCES, 41, 193-208.  doi: 10.1007/s00376-023-2278-4
    [18] Joshua TALIB, Omar V. MÜLLER, Emma J. BARTON, Christopher M. TAYLOR, Pier Luigi VIDALE, 2023: The Representation of Soil Moisture−Atmosphere Feedbacks across the Tibetan Plateau in CMIP6, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 2063-2081.  doi: 10.1007/s00376-023-2296-2
    [19] SHI Xiaoying, MAO Jiafu, WANG Yingping, DAI Yongjiu, TANG Xuli, 2011: Coupling a Terrestrial Biogeochemical Model to the Common Land Model, ADVANCES IN ATMOSPHERIC SCIENCES, 28, 1129-1142.  doi: 10.1007/s00376-010-0131-z
    [20] WANG Bo, and HUO Zhenhua, 2013: Extended application of the conditional nonlinear optimal parameter perturbation method in the Common Land Model, ADVANCES IN ATMOSPHERIC SCIENCES, 30, 1213-1223.  doi: 10.1007/s00376-012-2025-8

Get Citation+


Share Article

Manuscript History

Manuscript received: 13 October 2016
Manuscript revised: 02 May 2017
Manuscript accepted: 08 May 2017
通讯作者: 陈斌,
  • 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

Evaluating Common Land Model Energy Fluxes Using FLUXNET Data

  • 1. College of Global Change and Earth System Science, Beijing Normal University, Beijing 100875, China
  • 2. School of Atmospheric Sciences, Sun Yat-sen University, Guangzhou 510275, China
  • 3. Department of Geological Sciences, the University of Texas at Austin, Austin, Texas 78712, USA

Abstract: Given the crucial role of land surface processes in global and regional climates, there is a pressing need to test and verify the performance of land surface models via comparisons to observations. In this study, the eddy covariance measurements from 20 FLUXNET sites spanning more than 100 site-years were utilized to evaluate the performance of the Common Land Model (CoLM) over different vegetation types in various climate zones. A decomposition method was employed to separate both the observed and simulated energy fluxes, i.e., the sensible heat flux, latent heat flux, net radiation, and ground heat flux, at three timescales ranging from stepwise (30 min) to monthly. A comparison between the simulations and observations indicated that CoLM produced satisfactory simulations of all four energy fluxes, although the different indexes did not exhibit consistent results among the different fluxes. A strong agreement between the simulations and observations was found for the seasonal cycles at the 20 sites, whereas CoLM underestimated the latent heat flux at the sites with distinct dry and wet seasons, which might be associated with its weakness in simulating soil water during the dry season. CoLM cannot explicitly simulate the midday depression of leaf gas exchange, which may explain why CoLM also has a maximum diurnal bias at noon in the summer. Of the eight selected vegetation types analyzed, CoLM performs best for evergreen broadleaf forests and worst for croplands and wetlands.

摘要: 陆面模式描述陆地表面与大气之间物质, 能量和动量交换过程, 因其可以为大气环流模式提供准确的下垫面条件而对全球气候变化研究非常重要. 随着人们对天气预报和气候预测精度的需求越来越高, 用观测资料验证和改进陆面模式也逐渐成为提高模式模拟能力的一种重要手段. 本研究利用覆盖不同气候区不同植被类型的20个FLUXNET站点观测数据检验通用陆面模式(Common Land Model, CoLM)的能量平衡模拟情况. 在对比分析观测和模拟的感热, 潜热, 净辐射和地表热通量时, 采用分解方法将每个观测值或模拟值分解为月平均, 日平均残差和逐时残差三个部分. 综合相关系数(R), 均方根误差(RMSE)和标准化误差(Nbias)对不同时间尺度和不同植被类型的能量通量分析, 结果发现CoLM能够很好地模拟4个地表能量通量, 其中模拟净辐射最好, 模拟潜热通量好于感热通量; 模式能够很好地模拟能量通量的季节变化, 但在有明显干湿季的站点低估潜热通量, 这可能与模式对干季土壤水的模拟存在缺陷有关; CoLM在夏季中午产生较大模拟误差, 这很可能是因为CoLM不能准确模拟光合午睡现象; 就不同的植被类型而言, CoLM在所选的8种植被类型中的常绿针叶林表现最好.

1. Introduction
  • Land surface processes, including the exchanges of energy, water vapor, momentum and carbon fluxes between the land surface and atmosphere, play important roles in determining changes in atmospheric circulations and the global climate. Surface energy fluxes, such as sensible heat flux (H S), latent heat flux (H L), net radiation (R net), and ground heat flux (H G), are fundamental components of the global energy equilibrium (Bonan, 2015); thus, their accurate estimation is critical for global climate change research. In recent years, much work has focused on the representation and evaluation of these energy fluxes (Roerink et al., 2000).

    Land surface models (LSMs) are common tools to describe the aforementioned processes and provide lower boundary conditions for atmospheric general circulation models (Sellers et al., 1997). The first generation of LSMs consisted of only simple energy balance schemes. With increased physical knowledge, more sophisticated processes have been included in the current models, such as descriptions of land-based biogeochemical, biogeophysical and landscape processes (Oleson et al., 2013). The Common Land Model (CoLM) is a typical example. It contains improved treatments of soil moisture, soil temperature, turbulence fluxes, and other variables (Dai et al., 2003, 2004) and is based on a combination of an LSM (Bonan, 1996), the Biosphere-Atmosphere Transfer Scheme (Dickinson et al., 1986), and the Chinese Academy of Sciences Institute of Atmospheric Physics LSM (Dai and Zeng, 1997). Given the advantages of CoLM, it has been used as the land component in many fully coupled earth system models, such as the Community Climate Model from the National Center for Atmospheric Research (NCAR) (Zeng et al., 2002) and the Beijing Normal University-Earth System Model (BNU-ESM) (Ji et al., 2014). CoLM has also been used for some data assimilation schemes (Huang et al., 2008; Meng et al., 2009; Xu et al., 2011; Zhang et al., 2011).

    Previous studies have demonstrated the excellent performance of CoLM. (Xin et al., 2006) performed a validation experiment with CoLM at three arid stations with irrigated agriculture and demonstrated that the simulated results were consistent with observations, especially in modeling the diurnal and annual changes of the vertical profiles of soil temperature. (Ma et al., 2006) compared the regional evapotranspiration simulated by CoLM with remote sensing observations in China and concluded that CoLM can successfully simulate the magnitudes and spatial patterns of the monthly average evapotranspiration in China for certain months of 1991. In addition, (Zheng et al., 2009) coupled CoLM with version 3 of the Regional Climate Model and found that the coupled model produced reasonable results for precipitation, near-surface air temperature, and general atmospheric circulation.

    Given the critical role of LSMs in global climate research, an increasing number of researchers have evaluated these models (Stöckli et al., 2008; Blyth et al., 2010). From the early "Project for Inter-comparison of Land Surface Parameterization Schemes" (Henderson-Sellers et al., 1993) to the recent benchmark studies of LSMs (Luo et al., 2012; Best et al., 2015), the accuracy of LSMs has been greatly improved, especially after FLUXNET (Baldocchi et al., 2001) was established. For example, (Stöckli et al., 2008) used FLUXNET to develop and validate the Community Land Model (CLM). (Blyth et al., 2010) also evaluated Joint UK Land Environment Simulator energy fluxes using FLUXNET data.

    A good model should depict physical processes on different spatial and temporal scales based on different climate backgrounds. Despite the above studies, CoLM has not yet been sufficiently studied to verify its performance for different vegetation types in various climate zones. FLUXNET provides a useful dataset for the development of LSMs (Baldocchi et al., 2001), and this product can be used to evaluate the performance of CoLM for different biomes.

    In this study, we focused on the evaluation of land surface energy fluxes, such as H S, H L, R net and H G, as simulated by CoLM. This is the initial phase of a broader research effort by our group, as the accurate simulation of energy fluxes is a central premise of a good LSM. This work was conducted for ten different climate zones and eight different land-cover types. Simulations with different timescales were analyzed first, and then the results for several specific land-cover types from different climate zones were examined.

2. Methods
  • The initial version of CoLM was developed as a state-of-the-art LSM for climate studies. Two versions of this model have been developed based on the first version. One is CLM, developed by NCAR, which was first coupled with version 3 of the Community Climate Model, then with version 2 of the Community Atmosphere Model, and is currently coupled with the Community Earth System Model (Bonan et al., 2002; Zeng et al., 2002; Oleson et al., 2008). The other is the current version of CoLM, developed at BNU (Dai et al., 2003, 2004), which has been coupled with BNU-ESM (Ji et al., 2014). The two coupled models participated in phase 5 of the Coupled Model Intercomparison Project and provided projections of future climate for the IPCC's Fifth Assessment Report.

    CoLM utilizes an improved two-stream approximation model for calculating canopy radiation and a two-big-leaf (Dai et al., 2004) model, which considers shaded leaves and sunlit leaves separately. Its performance in simulating stomatal conductance, photosynthesis, and leaf temperature has been improved relative to the earlier version of CoLM. In CoLM, energy fluxes are computed by solving the energy balance equation at the land-atmosphere interface (Dai et al., 2003), which is expressed by Eq. (1), given the following three assumptions:

    (1) Convective heat transfer is assumed to be negligible within the canopy, soil and snow layers;

    (2) Vaporization and sensible heat transfer are assumed to be negligible within the snow and soil layers;

    (3) Heat conductance is assumed to be negligible within the canopy. \begin{equation} c\Delta z\dfrac{{\rm d}T}{{\rm d}t}=R_{\rm net}-H_{\rm S}-H_{\rm L} \ \ (1)\end{equation} In Eq. (1), c is volumetric heat capacity, ∆ z is the depth of the soil layer, T is the soil temperature, and the net radiation R net is the sum of the net shortwave radiation (R net,s) and net longwave radiation (R net,l): \begin{eqnarray} R_{\rm net}&=&R_{\rm net,s}+R_{\rm net,l} ,\ \ (2)\\ R_{\rm net,s}&=&S\downarrow(1-\alpha) ,\ \ (3)\\ R_{\rm net,l}&=&aL\downarrow+b\sigma T_{\rm c}^4+d\sigma T_{\rm g}^4 . \ \ (4)\end{eqnarray}

    In Eqs. (3) and (4), S↓ denotes downward solar radiation; α is albedo; L↓ denotes downward longwave radiation; T c denotes canopy temperature; T g denotes ground temperature; and a, b and d are parameters related to the fraction of the vegetated area and emissivity of the ground.

    The turbulent heat fluxes H S and H L are calculated as the heat resistance multiplied by the temperature/water gradient, and are described by Eqs. (5) and (6), respectively: \begin{eqnarray} H_{\rm S}&=&r(T_{\rm c}-\theta_{\rm a}) ,\ \ (5)\\ H_{\rm L}&=&r(q_{\rm c}-q_{\rm a})+E_w+E_{\rm tr}. \ \ (6)\end{eqnarray}

    In Eqs. (5) and (6), r denotes heat resistance; Ew denotes evaporation from the ground; E tr denotes transpiration; θ a stands for air temperature; and q c and q a are the specific humidity of the canopy and atmosphere, respectively.

  • FLUXNET is a network of globally distributed flux towers that measure the energy, water vapor, and carbon dioxide exchanges between the land surface and atmosphere, using eddy covariance methods. Currently, FLUXNET consists of more than 800 sites, covering a variety of biome types and climate zones (Baldocchi et al., 2001, 2016). FLUXNET provides the most comprehensive dataset of the terrestrial ecosystem, and its measurement uncertainties can be accurately quantified at different levels. The released datasets cover eleven land cover types, based on information from the International Geosphere-Biosphere Programme, and seventeen subclasses of the five main classes of climate zones (Kottek et al., 2006).

    To evaluate the performance of CoLM for different vegetation types in different climate zones, we selected sites that represented different vegetation types, allowing cases of as many climate zones as possible to be considered. For example, the grassland (GRA) sites cover three of the main climate zones [B (arid), C (warm-temperate), and D (snow)] and five subclasses [warm temperate climate with hot summer (Cfa), temperate climate with warm summer (Cfb), snow climate with warm summer (Dfa), steppe climate (BSk), and warm temperate climate with dry summer (Csa)]. In this study, eight vegetation types were included: cropland (CRO), deciduous broadleaf forest (DBF), evergreen broadleaf forest (EBF), evergreen needleleaf forest (ENF), wetland (WET), GRA, mixed forest (MF) and savanna (SAV).

    The selected sites also meet the requirements of having data that cover at least two continuous years, and containing few gaps. Ultimately, twenty FLUXNET sites covering eight plant function types and ten climate zone subclasses were chosen, and their locations, data coverage, and climate information are listed in Table 1.

  • CoLM requires eight atmospheric forcing variables for offline simulations: downward solar radiation at the surface (W m-2); downward longwave radiation (W m-2); precipitation (mm s-1); temperature at the reference height (K); wind speed in both the eastward and northward directions (m s-1); atmospheric pressure at the surface (Pa); and specific humidity at the reference height (kg kg-1). Five times the length of the observational record was used for model spin-up to provide stable initial conditions. All forcing data used in this research were in-situ measurements from FLUXNET.

    We utilized the improved MODIS Leaf Area Index (LAI) products (Yuan et al., 2011) for each site with consistent land-cover type settings in the model. The other land surface characteristic datasets and soil and vegetation properties were set to their default values; the parameters for the different sites were not optimized. All analyses are based on the original output.

  • To resolve the imbalance issue of the surface energy budgets (Wilson et al., 2002), which refers to the inconsistencies of the independent measurements of the major energy balance flux components, we adjusted H S and H L and assumed that the measured Bowen ratio was correct. The tuning equations are as follows: \begin{eqnarray} H_{\rm S,cor}&=&AH_{\rm S,u}=\dfrac{R_{\rm net}-H_{\rm G}}{H_{\rm S,u}+H_{\rm L,u}}H_{\rm S,u}\ \ (7)\\ H_{\rm L,cor}&=&AH_{\rm L,u}=\dfrac{R_{\rm net}-H_{\rm G}}{H_{\rm S,u}+H_{\rm L,u}}H_{\rm L,u} . \ \ (8)\end{eqnarray}

    In Eqs. (7) and (8), the variables with the subscript cor denote the corrected variables, and those with u indicate the uncorrected variables. The tuning factor A will make up the under-measurement of H S and H L. When H G is missing, the alternative factor A* is used and is expressed as A*=R net/(H S,u+H L,u) (Twine et al., 2000). The corrected surface energy fluxes for the FLUXNET observations were used to validate the model in this study.

    In this research, the alternative factor A* was applied only to sites US-Ha1 and FI-Hyy of the 20 selected sites. The vegetation type of US-Ha1 is deciduous broadleaf forest and that of FI-Hyy is evergreen needleleaf forest, both of which are prone to small H G because of dense vegetation. In general, the method may have its limitations, but the lack of a surface energy balance has troubled the research community for a long time. (Twine et al., 2000) reviewed the surface fluxes and energy-balance closure, discussed improved energy-balance methods, and then suggested that the Bowen-ratio closure method may be the most appropriate. This Bowen-ratio closure method has been widely used; (Jung et al., 2010) even used it to process 253 globally distributed flux towers when producing a global evapotranspiration dataset. As the Bowen-ratio method may not always be ideal at a specific site, its rationale should be re-evaluated when further validation at specific sites is performed.

  • In this study, both the simulated and observed results were decomposed into three parts: annual cycle, daily residuals, and diurnal bias (stepwise perturbation), as described by Eq. (9): \begin{equation} X_{\rm i}=X_{\rm m}+X_{\rm d}+X_{\rm s} .\ \ (9) \end{equation}

    In Eq. (9), X i stands for the simulation or observation at each time step, X m stands for the monthly average of the month in which X i is observed, X d is calculated as the average over the entire corresponding day, and X s is the residual when the monthly mean and daily mean are removed from X i. The data used for the calculation of X d are equal to the averages of the original values minus their corresponding monthly means. Taking one day in May as an example, there were 48 values within this single day, i.e., X i=X1,X2,X3,…,X48. X m is the monthly average for May, such that X d=1/48∑148(X i-X m). The annual cycle, which is composed of the monthly averages, can capture seasonal issues, such as the deficiencies in modeled vegetation phenology or soil drying after wet seasons, by identifying the times that the model may produce biases. Similarly, the diurnal cycle, which consists of hourly averages, enables the identification of the particular time in a given day that a model underperforms. The diurnal cycles for different seasons were analyzed in this research to evaluate the performance of CoLM under different vegetation types during different stages of vegetation growth and water availability.

  • Three statistical indexes——the correlation coefficient (R), the variance ratio, and the root-mean-square error (RMSE)——were summarized within a Taylor diagram (Taylor, 2001) to evaluate the model performance. The Taylor diagram was used to check the performance of CoLM in simulating the surface energy fluxes from stepwise, to daily, to monthly averages, and thus indicated how the model performance differed at different timescales.

    Quantitative measurements of the differences between the simulations and observations are recommended to evaluate the performance of hydrological models (Aitken, 1973; James and Burges, 1982). RMSE, R, and the normalized bias (N bias) were used as the statistical performance criteria in the evaluation of this work. N bias is the bias between the model simulation and observational data normalized by the downward solar radiation at the same time step, as shown in Eq. (10). \begin{equation} N_{\rm bias}=\dfrac{X_{\rm model}-X_{\rm obs}}{R_{\rm solar}} .\ \ (10) \end{equation}

    Because the solar radiation should not be zero (since it is in the denominator), all N bias-based analyses are restricted to the daytime.

3. Results
  • As mentioned earlier, this study seeks to evaluate the performance of CoLM in simulating the surface energy fluxes for various vegetation types in multiple climate zones. In this section, the general model performance for simulating H S, H L, and R net on different timescales is assessed using all three indexes (R, RMSE, and N bias). The performance of CoLM for four specific land-cover types is then discussed separately to observe how the model performs for different climate zones and a given vegetation type.

  • The comparisons between the observed and simulated H S, H L, and R net values shown in Fig. 1 and Table 2 reflect the performance of the model across different timescales——namely, stepwise, daily, and monthly. The Taylor diagrams show that the R values of H S, H L, and R net mostly vary from 0.6 to 0.95, 0.7 to 0.95, and 0.95 to 1.0, respectively, and the normalized standard deviations of the three fluxes mostly vary from 0.9 to 1.5, 0.7 to 1.2, and 0.9 to 1.2, respectively. Obviously, the overall variability (normalized standard deviation) of H S is higher than that of H L, and the R range of H L is smaller than that of H S. Overall, CoLM simulates H S, H L, and R net better at monthly and stepwise levels than it does at a daily scale, and produces a better simulation of H L than H S at all timescales.

    Figure 1.  Performance of CoLM simulations for eight land cover types (1, CRO; 2, DBF; 3, EBF; 4, ENF; 5, WET; 6, GRA; 7, MF; 8, SAV) and three variables: sensible heat flux (H S); latent heat flux (H L); and net radiation (R net). Statistics are based on stepwise (X s), daily (X d), and monthly (X m) simulated and observed H S, H L, and R net results.

    The bold italic numbers in Table 2 show the best performances among the three timescales (the largest R or the smallest RMSE values of the four energy fluxes). Obviously, CoLM simulates the stepwise H S better than its daily average and annual cycle in terms of the associated R values, and provides better simulations of H L and H G at the annual scale. An excellent performance in simulating R net is observed at all three temporal scales, likely because the downward solar and longwave radiations are prescribed in the model.

    Figure 2.  Annual cycles of observed (Obs, black circles) and simulated (CoLM, black line) (a) H S, (b) H L, (c) H G and (d) R net at the 20 FLUXNET sites, with their names and plant functional types denoted in panel (c).

    Figure 2 compares the monthly averages of the energy fluxes between the model simulations and observations. Almost all modeled H S values at the selected 20 sites follow the seasonal cycles of the observations, with bias varying by the different vegetation types. For example, the modeled H S is lower than that of the observations for DBF and WET but higher for GRA, whereas the modeled H L is higher than that of the observations for DBF and WET. The modeled H L at all 20 sites, except for site US-Blo, follows the observed seasonal cycle and exhibits different biases for different vegetation types. The biases for both H L and H S are quite large in the wetlands (sites CA-Mer and FI-Kaa). H S is overestimated during the dry periods at sites with typical dry and wet seasons, such as at US-Bkg, US-Var, and BW-Ma1.

    Similar to the results of (Stöckli et al., 2008), in our study, H L is greater in magnitude than H S at most sites. The simulated H L is more consistent with the observed values than H S in terms of both their magnitudes and temporal patterns. However, the fact that H L is underestimated during dry periods whereas H S is overestimated at sites with typical distinct dry and wet seasons, may be associated with the weakness of CoLM when simulating soil water in dry seasons. The deficiencies of the model in handling dry soil moistures may be caused by the lack of a full description of plant physiology——for example, the process of hydraulic redistribution (HR) (Burgess et al., 1998; Yan and Dickinson, 2014).

    HR is defined as the passive water movement from wet soil to dry soil through plant roots, driven by water potential gradients. In the HR process, water can move passively through the roots, upwards or downwards, whenever a gradient in the soil water potential between soil layers exists and is stronger than the overall gradient between the soil and atmosphere (especially during the night, when transpiration is weak) (Oliveira et al., 2005). In dry seasons, HR allows water to move during the night or on cloudy days——specifically, from deep moist soil to dry shallow soil, where plant roots are more abundant. In wet seasons, water can also move down from the shallow layers to the deep layers via roots to reduce the loss through surface runoff or evaporation. HR has been recognized as a process that is widespread, occurs across several plant species and climates (Caldwell et al., 1998; Neumann and Cardon, 2012; Prieto et al., 2012; Quijano and Kumar, 2015), and may have an important influence on local and global climate change (Lee et al., 2005; Wang et al., 2011).

    Figure 3 shows the solar-normalized diurnal bias of H S in different seasons. CoLM generates different magnitudes of biases in different seasons and exhibits different bias patterns for different vegetation types. It overestimates H S for almost all eight vegetation types at noon during the summer (here, we took the summer season as June-July-August in the Northern Hemisphere, although three SAV sites and one EBF site in this study are located in the Southern Hemisphere, but these sites are located in either tropical or subtropical regions), except those in wetlands, and the U-shaped pattern of the normalized bias also exhibits relatively higher values when the downward solar radiation is higher. The biases of H S show different patterns during different seasons, with magnitudes mostly within 10% of the solar radiation. Compared to the normalized diurnal bias of H S, that of H L exhibits a totally opposite pattern (Fig. 4), although the values fall within the range of 20% of the downward solar radiation.

    A diurnal maximum bias occurs at noon, especially in the growing season, likely because CoLM could not explicitly capture the midday depression of the leaf gas exchange (Zeiger et al., 1987). A midday depression is one pattern of photosynthetic diurnal variation, with two marked peaks that occur in the late morning and late afternoon, separately (Pessarakli, 2005); this pattern occurs in many plants and will have a significant impact on crop yields. The physiological factors responsible for the midday depression consist of stomatal closure, the enhancement of respiration and photorespiration, an increase in the mesophyll resistance of CO2, and a decrease in the leaf water potential. However, the stomatal resistance in CoLM is coupled with leaf photosynthesis in a manner similar to that described by (Collatz et al., 1991) and does not consider the mesophyll diffusion of CO2. All of these deficiencies in simulating the leaf gas exchange may cause a bias in simulating H L because one important part of H L is transpiration [see Eq. (6)], and the calculation of transpiration relies on stomatal resistance and leaf boundary layer resistance.

    In CoLM, the H G at the land surface is computed as the residual of the net radiation, based on the energy balance equation: H G=R net-H S-H L (Dai et al., 2003). The discrepancy between the simulated and observed H G (Fig. 2c) indicates the embedded errors of estimating H S and H L using the simulated H G. CoLM produces a positive H G bias during daytime and a negative H G bias at night. The bias does not exhibit any seasonal dependence (not shown).

    A satisfactory agreement is found between the observed and simulated net radiation (Fig. 2d and Table 2), with R shown in the Taylor diagram. The high-quality simulation of the net radiation may be the result of the prescribed downward solar and longwave radiation and the accurate estimates of the surface albedo in the model.

  • The 20 selected sites in this study cover eight vegetation types to provide useful comparisons. Table 3 lists the means of the simulated biases of the four energy fluxes over different vegetation types. CoLM underestimates H S for more than half of these eight vegetation types and exhibits a negative H L bias for CRO, GRA, and SAV, and a positive bias for the other five types. Similarly, CoLM exhibits a negative R net bias for EBF, DBF, and SAV, and a positive bias for the other land-cover types. Although H G is small in the energy balance equilibrium, its simulation error may not be negligible compared to the other three fluxes.

    Figure 3.  Normalized diurnal bias of H S over eight different vegetation types in different seasons: red for spring (March-April-May, MAM); blue for summer (June-July-August, JJA); green for autumn (September-October-November, SON); black for winter (December-January-February, DJF).

    Figure 4.  As in Fig. 3 but for H L.

    3.2.1. Deciduous Broadleaf Forest

    DBF covers most of the eastern United States, central and northern Europe, eastern China, and Japan. It has a long warm growing season. In this study, the three DBF sites are in temperate (FR-Hes), subtropical-Mediterranean (IT-Ro1), and temperate-continental with warm summers (US-Ha1) climate zones. Interestingly, despite the climate discrepancies (site IT-Ro1 is hotter and drier in the summer than FR-Hes and US-Ha1, and US-Ha1 has a colder winter) and because of the abundant moisture in DBF, CoLM underestimates H S and overestimates H L at all three sites, based on the monthly N bias.

    According to (Stöckli et al., 2008), CLM3.0 overestimates H L in the winter but underestimates it in the summer. CoLM yields analogous seasonal performances. From the seasonality evident in Fig. 2, one can identify that the simulated H L exhibits a high consistency with the observations for the temperate-continental with warm summers (US-Ha1) climate zone. CoLM generates negative H L errors for the temperate (FR-Hes) zone and an H L phase error for the subtropical-Mediterranean (IT-Ro1) zone.

    At the stepwise scale, the R values of H S at the three DBF sites (0.90, 0.92, and 0.90 for FR-Hes, IT-Ro1, and US-Ha1, respectively) indicate a better performance of CoLM in simulating H L (0.81, 0.81, and 0.95, respectively). However, the simulated H S is not always better than the simulated H L at the monthly scale (R of H S is 0.78, 0.84, and 0.56, while that of H L is 0.88, 0.50, and 0.96). Therefore, CoLM can achieve a good performance in terms of modeling H S and H L at different timescales, even for the same vegetation type. Given the seasonal precipitation patterns of the three sites, it is obvious that the CoLM conducts better simulations at sites with even monthly rainfall, such as that of the temperate-continental climate zone.

    3.2.2. Evergreen Needleleaf Forest

    Woody plants cover more than 60% of ENF zones, and their canopies are normally higher than 2 m. Almost all trees remain green year-round in ENFs. The selected ENF sites in this study are located in temperate (DE-Tha), boreal (FI-Hyy), and subtropical-Mediterranean (US-Blo) climate zones. Obviously, the site US-Blo is the warmest and has the most abundant precipitation, while DE-Tha is warmer than FI-Hyy. However, site US-Blo has distinct dry and wet seasons.

    CoLM performs better when simulating H S than when simulating H L at the three ENF sites, based on a stepwise R analysis, and all of the R values of H S at the three sites (0.96, 0.94, and 0.94 for DE-Tha, FI-Hyy, US-Blo) are higher than those of H L (0.95, 0.90, and 0.72). Although CoLM overestimates H L, with a mean normalized stepwise bias of 0.05, 0.02, and 0.03, it underestimates H S at DE-Tha (-0.06) and FI-Hyy (-0.03), and overestimates H S at US-Blo (0.05).

    CoLM produces the best H L seasonal cycles at DE-Tha and FI-Hyy (see Fig. 2b), and produces a phase error at site US-Blo. While the model captures the general pattern of the H S seasonal cycle, it underestimates the annual range of H S at sites DE-Tha and FI-Hyy, and generates a greater range of simulated H S values than those observed.

    (Yang et al., 2009) noted that the soil surface resistance controls the surface evaporation of alpine deserts, but is not reasonably implemented in CoLM. After adjusting the soil porosity, hydraulic conductivity of saturated soil, and vertical soil layering scheme, (Li et al., 2012) found that CoLM could simulate the soil moisture distribution at three stations on the Tibetan Plateau. The lack of these adjustments may be the reason why CoLM could not yield satisfactory performances at US-Blo. Based on the literature and our experiments, we infer that high spatiotemporal resolution land surface data (Yuan et al., 2011; Shangguan et al., 2014) will improve the performance of CoLM.

    3.2.3. Grassland

    GRA is the land cover type that is dominated by herbaceous plants. The temperature varies more strongly from summer to winter, and the annual precipitation is lower in the temperate grasslands than in the savannas. Sites from subtropical-Mediterranean (IT-Amp and US-Var), temperate (NL-Ca1), temperate-continental with hot summers (US-Bkg), and dry (US-FPe) zones are used for the representations of GRA.

    From the results of the R and RMSE calculations, CoLM performs consistent well at IT-Amp and NL-Ca1 at the monthly scale, with larger R and smaller RMSE values for H S than those for H L, and at US-FPe, it exhibits the opposite pattern. At the daily timescale, CoLM exhibits consistently better performance in predicting H S at NL-Ca1 and US-Bkg. Interestingly, the hourly R and RMSE values indicate consistently better simulations of H S at NL-Ca1 alone. Based on the N bias results, CoLM consistently underestimates H L at all three timescales for IT-Amp, and consistently overestimates H L for US-Var. The different performances at the different timescales between these five sites and the different performances for the different sites from the different climate zones at the same timescales suggest that, for the same land-cover types, different climate conditions can affect the performance of the model.

    3.2.4. Savanna

    SAV describes grasslands with scattered individual trees and porous soil, favoring the rapid drainage of water. SAV has both dry and rainy seasons. Sites representing the SAV land-cover type were from tropical (AU-How), dry (BW-Ma1), and subtropical-Mediterranean (ES-LMa and ZA-Kru) zones.

    Similar to those of GRA, the R and N bias exhibit consistent patterns for SAV in the different climate zones for the three timescales. For example, the monthly R at site ZA-Kru exhibits the same improved performance for the simulation of H S as its performance for the simulation of H L, which is also exhibited in the hourly simulations at sites AU-How and ES-LMa and the daily simulations at sites BW-Ma1, ES-LMa, and ZA-Kru. The hourly and monthly N bias values are consistent for the simulated H S, which is opposite to the behavior seen in the daily simulations.

    Given the different characteristics of these land-cover types, i.e., GRA and SAV differ in their surface roughness, zero-plane displacement, and soil parameters (e.g., root depth and coefficient of root profile), CoLM achieves different performances even for these two more-similar vegetation types. (Groenendijk et al., 2011) noted that the quality of the fluxes decrease when more general parameters are used because the choice of a parameter classification has a considerable impact on the quality of the simulated photosynthesis and transpiration fluxes.

4. Discussion and conclusions
  • This study evaluated the performance of CoLM in simulating energy fluxes across a global range of ecosystems, over eight vegetation types within ten subclasses of climate zones. The utilization of the decomposition method provided detailed information that could be used to assess the performance of the model on different timescales, and the correction method effectively served as the energy closure of the eddy covariance measurement. The simulated energy fluxes were first evaluated over different timescales for eight land-cover types, and then for certain land-cover types in different climate zones. Overall, the results demonstrate that CoLM generally exhibits a satisfactory modeling performance.

    Of the four energy fluxes, CoLM generally exhibits better performance for H L than for H S, based on the Taylor diagram, and performs well in simulating the seasonality at the 20 sites selected in this study. A strong agreement between the simulations and observational data is found in the seasonal cycle of R net. H L is underestimated during dry periods at sites with distinct dry and wet seasons, and H S is overestimated, which might be related to the weakness of the model when simulating soil water during dry seasons.

    At the diurnal scale, CoLM produces a maximum bias close to noon in the summer, and the bias varies between different vegetation types and climate zones. This provides modelers with a clue for how to improve the ability of the model to simulate the midday depression of leaf gas exchange. Given the physical factors that lead to this midday depression, the model performance may be improved by explicitly implementing the stomatal closure mechanism or incorporating mesophyll diffusion in the sub-model of photosynthesis.

    Regarding the different vegetation types, CoLM performs best for EBF among the eight vegetation types and worst for CRO and WET. Although different climate conditions will lead to different model behavior, even for the same land-cover type, CoLM produces better simulations for sites with even monthly rainfall.

    All three indexes utilized in this study——R, RMSE, and N bias——do not show a consistent model performance between different energy fluxes, different timescales of certain sites, or different climate zones of one land-cover type. Therefore, it is difficult to assess the model's merits and limitations using a particular single index. We should combine the different skills of these indexes when evaluating the model's performance.

    Based on earlier studies and our evaluation, the implementation of site-specific parameters, including soil porosity, hydraulic conductivity of saturated soil, and a vertical soil layering scheme, will significantly improve the performance of CoLM, whereas a more general classification of the implemented parameters will reduce the predicted fluxes. For global-scale simulations, high-resolution datasets of land-surface characteristics are needed to achieve high-accuracy results.

    Overall, CoLM simulates energy fluxes better at monthly and stepwise scales than at daily scales, and it underestimates H S and overestimates H L for most of the 20 sites selected in this study. CoLM performs best for the EBF vegetation type, and worst for CRO and WET. The reason why CoLM underestimates H L at some sites may be associated with its weakness in simulating soil water in dry seasons. In addition, the model's deficiencies in simulating the midday depression of leaf gas exchange may account for the maximum bias observed at noon in summer, which is a potential subject for future research.




    DownLoad:  Full-Size Img  PowerPoint