植物生态学报 ›› 2018, Vol. 42 ›› Issue (12): 1131-1144.DOI: 10.17521/cjpe.2018.0231

• 研究论文 •    下一篇

Biome-BGC模型模拟阔叶红松林碳水通量的参数敏感性检验和不确定性分析

李旭华,孙建新()   

  1. 北京林业大学林学院, 北京 100083
  • 收稿日期:2018-09-18 修回日期:2018-12-06 出版日期:2018-12-20 发布日期:2019-04-04
  • 通讯作者: 孙建新
  • 基金资助:
    国家林业公益性行业科研专项(201404201)

Testing parameter sensitivities and uncertainty analysis of Biome-BGC model in simulating carbon and water fluxes in broadleaved-Korean pine forests

LI Xu-Hua,SUN Osbert Jianxin()   

  1. College of Forestry, Beijing Forestry University, Beijing 100083, China
  • Received:2018-09-18 Revised:2018-12-06 Online:2018-12-20 Published:2019-04-04
  • Contact: Osbert Jianxin SUN
  • Supported by:
    Supported by the Forestry Research for the Public Benefits of Ministry of Finance of China(201404201)

摘要:

生态过程模型的发展为研究者在长时间序列和区域尺度的研究提供了便利, 但模型模拟的准确性受到模型自身结构、模型参数估计合理性的影响。敏感性分析能够定量或定性筛选出对模型模拟结果影响较大的敏感参数, 是模型参数校准过程中的重要工具, 也是建模和应用的先决条件。该文以阔叶红松林为研究对象, 采用全局敏感性分析方法——傅里叶幅度灵敏度检验扩展法(EFAST)对Biome-BGC模型的生理生态参数进行了敏感性分析, 分别分析了红松(Pinus koraiensis)和阔叶树的净初级生产力(NPP)、蒸散(ET)对参数变化的敏感性。结果表明: (1)模拟红松NPP的不确定性高于阔叶树, 但二者的模拟ET的不确定性均较小。阔叶树的NPPET对生理生态参数的敏感性总体上都小于红松。(2)无论是红松、阔叶或其他植被类型, 模拟NPP均表现出对叶片碳氮比、细根碳氮比、比叶面积(SLA)和冠层截留系数的敏感性, 这4个参数的高敏感性主要是由模型自身结构所决定的, 与植被类型和研究地区的关系较小。对模拟ET而言, 细根与叶片碳分配比、新茎与新叶碳分配比和SLA均是影响红松和阔叶树ET的敏感参数, 但红松ET主要受参数与参数间的二阶或多阶交互作用的间接影响, 而阔叶树ET则主要是受到敏感参数直接效应的影响。(3)除了上述影响红松和阔叶树碳水通量的共性参数外, 诸如核酮糖-1,5-二磷酸羧化酶中叶氮含量、叶片与细根周转率、所有叶面积与投影叶面积之比等也是对模拟结果有影响的重要参数, 但是其敏感程度随物种不同和研究区不同而不同, 所以这类参数可以根据具体情况进行参数本地化, 对于其他不敏感参数则可以采用模型缺省值。

关键词: 敏感性分析, 傅里叶幅度灵敏度检验扩展法, 净初级生产力, 蒸散, Biome-BGC模型

Abstract:

Aims The emergence and application of ecosystem process models have provided useful tools for studying carbon and water balances of terrestrial ecosystems at large spatiotemporal scales, but the accuracy of model simulations is affected by the parameterization of key variables among many factors. Sensitivity analysis is commonly used to screen the critical parameters that have predominant influences on model simulations. The objective of this study was to identify the critical ecophysiological parameters in Biome-BGC model in simulating annual net primary productivity (NPP) and evapotranspiration (ET) of broadleaved-Korean pine forests in Northeast China.

Methods We simulated carbon and water fluxes of broadleaved-Korean pine forests with the Biome-BGC (version 4.2) at a daily time step based on site- and species-specific parameters. Daily meteorological data for the period 1958-2015 was obtained from the China Meteorological Administration. Initialization parameters such as geographical position, soil depth, and soil texture of the site were obtained from field measurements. Among the 43 ecophysiological parameters represented in the model, 30 were derived either from field measurements or from published data for the study sites in literature, and the default values were used for 13 of the parameters. The modeled forest NPP was compared with the tree-ring width index to test the model’s ability to simulate the inter-annual variations in forest productivity. The modeled NPP and ET were also compared with existing remote sensing products for the period 2000-2014 for validation purpose. Sensitivity analysis was conducted using a variance-based sensitivity analysis method—Extended Fourier Amplitude Sensitivity Test (EFAST) to acquire the first order and total order sensitivity index of the parameters.

Important findings Our locally parameterized Biome-BGC model well simulated the carbon and water fluxes of the broadleaved-Korean pine forests. The uncertainty of simulated NPP is higher for Korean pine trees than for broad-leaved trees, while that of ET was small for both tree types. Both NPP and ET of broad-leaved trees were generally less sensitive to ecophysiological parameters than Korean pine. Leaf carbon to nitrogen ratio, fine root carbon to nitrogen ratio, specific leaf area (SLA), and water interception coef?cient were among the highly sensitive parameters affecting the modeled NPP; while fine root carbon to new leaf carbon allocation, new stem carbon to new leaf carbon allocation and SLA were the highly sensitive parameters influencing ET. In addition, fraction of leaf N in Rubisco, leaf and fine root turnover, ratio of all sided to projected leaf area are also critical parameters affecting the output of Biome-BGC simulations. The degree of sensitivity of the critical parameters varied with species and sites, highlighting the need to adopt local parametrization of Biome-BGC model in simulating regional forest carbon and water fluxes. For other non-sensitive parameters, model default value can be readily used.

Key words: sensitivity analysis, extended fourier amplitude sensitivity test method, net primary production, evapotranspiration, Biome-BGC model