Using approximate Bayesian computation to infer photosynthesis model parameters
ZENG Ji-Ye1, TAN Zheng-Hong2, *, , SAIGUSA Nobuko1
1National Institute for Environmental Studies, Tsukuba 305-8506, Japanand 2Department of Environmental Science, Hainan University, Haikou 570228, China
We developed a method, namely Adaptive Population Monte Carlo Approximate Bayesian Computation (APMC), to estimate the parameters of Farquhar photosynthesis model. Treating the canopy as a big leaf, we applied this method to derive the parameters at canopy scale. Validations against observational data showed that parameters estimated based on the APMC optimization are un-biased for predicting the photosynthesis rate. We conclude that APMC has greater advantages in estimating the model parameters than those of the conventional nonlinear regression models.
Keywords:Monte Carlo
;
big-leaf model
;
Farquhar photosynthesis model
;
net ecosystem exchange
ZENGJi-Ye, TANZheng-Hong, SAIGUSANobuko. Using approximate Bayesian computation to infer photosynthesis model parameters. Chinese Journal of Plant Ecology, 2017, 41(3): 378-385 https://doi.org/10.17521/cjpe.2016.0067
基于将冠层视为一片大叶的思维抽象, 叶片尺度的光合模型便得以直接应用于冠层之上, 简称大叶模型。本文采用了Farquhar模型来模拟冠层尺度的光合作用, 同时采用Ball-Berry模型来模拟气孔导度关系(Ball et al., 1987)。具体使用到的关系式见表1; 各缩写和符号代表的含义和参数初值见表2。模型细节请参考Medlyn等(2002a), von Caemmerer等(2009)和Damour等(2010)的文献。
图1 温度修正模型比较。实线: von Caemmerer等(2009)的曲线(S = 710, H = 220 000), 点线: 增加S约10%的曲线(S = 790, H = 220 000), 虚线: 本文用的曲线(Cm = 0.3, Tm = 37)。
Fig. 1 Comparison of the temperature correction models. Solid line is the curve of von Caemmerer et al. (2009) when S = 710 and H = 220 000; dotted line is the curve when S is increased by 10% (S=790, H=220 000); dashed line is the curve calculated according to our new response curve (Cm = 0.3, Tm = 37).
对每一个APMC粒子 For all particles for k = 1 to Nobs do(对所有的观测数据) For all observations 用APMC粒子选定的V25, J25,Rd25, g1, Ea, Cm, Tm Use parameter values selected by AMPC 计算Vcmax, Jmax,Rd, gb Estimate target model parameters 设t = 1 (Ci的迭代计算次数) At initial time 设\(C_i^0=0.7C_a\) Set the intercellular CO2 equal to 70% of air CO2 设\(\Delta C_i >1\) Set the intercellular CO2 not in equilibrant While \(\Delta C_i >1 \) do 计算 Ac, Aj Compute the two rate. 计算\(A=\frac{A_c+A_j-\sqrt{(A_c+A_j)^2-4\times 0.98A_cA_j}}{2\times 0.98} \\ Compute the joint rate 计算\\ C_i^t=C_a-\frac{A}{g_b}-\frac{g_bC_aA-A^2}{g_bg_1h_sA-g_0A+g_0Ag_bg_oC_a}\) Compute the intercellular CO2 \( \Delta C_i=|C_i^t-C_i^{t-1}|\) Check equilibrant state 设t = t + 1 Advance time end while end for 将|A-NEE|的平均值作为APMC粒子的r Calculate the difference between modeled photosynthesis rate and the observed rate
BallJT, WoodrowJT, BerryJA (1987). A model predicting stomatal conductance and its contribution to the control of photosynthesis under different environmental conditions. In: Biggins J ed. Progress in Photosynthesis Research, Vol. 4. Proceedings of the 7th International Congress on Photosynthesis. Matins Nijhoff, Dordrecht, the Netherlands. 221-224.
Determination of the gas exchange phenology in an evergreen coniferous forest from 7 years of eddy covariance flux data using an extended big-leaf analysis.
b). Temperature response of parameters of a biochemically based model of photosynthesis. I. Seasonal changes in mature maritime pine (Pinus pinaster Ait.). Plant,
von CaemmererS, FarquharG, BerryJ (2009). Biochemical model of C3 photosynthesis. In: Laisk A, Nedbal L, Govindjee eds. Photosynthesis in Silico: Understanding Complexity from Molecules to Ecosystems. Springer, Dordrecht, the Netherlands. 209-230.
Net ecosystem CO2 exchange over a larch forest in Hokkaido, Japan.
Atmospheric Environment, 38, 7021-7032.
[32]
WangYP, LeuningR (1998).
A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I: Model description and comparison with a multi- layered model.
Seasonal trends in photosynthetic parameters and stomatal conductance of blue oak (Quercus douglasii) under prolonged summer drought and high temperature.
A simple new equation for the reversible temperature dependence of photosynthetic electron transport: A study on soybean leaf.
2004
Determination of the gas exchange phenology in an evergreen coniferous forest from 7 years of eddy covariance flux data using an extended big-leaf analysis.
1
2013
... 涡度相关技术(eddy covariance technique)作为目前直接测定地表-大气间CO2和水热通量的标准方法, 为原位、无破坏监测生态系统的光合作用(实质上植被大气之间的CO2交换)提供了新的选择(de Pury & Farquhar, 1997; Wang & Leuning, 1998; Dai et al., 2004; Groenendijk et al., 2011; Kosugi et al., 2013).在过去的30余年里, 该方法有了快速的发展, 得到了广泛的应用, 目前已成为国际通量观测网络(FLUXNET)的主要技术手段.据不完全统计, 目前全球安装涡度相关监测系统的研究站已经超过5 000个, 仅中国就超过了300个.考虑到涡度相关法可以提供冠层光合作用的实测值, 而叶片尺度的光合模型也很明确, 笔者认为可以在叶片光合模型的基础上提出一种算法, 直接反演出Vcmax和Jmax这两个重要的生理参数在冠层上的数值.如果这一想法能够实现, 我们就有可能从生理学角度切入, 对大量冠层尺度的监测数据进行更深入的解读和挖掘.作为一个初步的尝试, 为了简化模型的数据和计算, 本文通过将冠层抽象为一片大叶, 提出了一种相对简洁却行之有效的方法来反演冠层的生理参数.下文中, 笔者将对该方法的技术细节进行分解, 希望该方法能引起读者的兴趣, 在读者的批评中获得进一步提高和完善. ...
Adaptive approximate Bayesian computation for complex models.
2013
Sequential importance sampling for nonparametric Bayes models: The next generation.
1
1999
... 其中, p(y|θ)为似然函数(likelihood function), p(y)为边缘似然函数(marginal likelihood function).当似然函数和边缘似然函数易得, 或可以通过某种近似法, 诸如马尔可夫链蒙特卡洛法(Markov chain Monte Carlo, Andrieu et al., 2010)、重点取样法(importance sampling, MacEachern et al., 1999)以及序贯蒙特卡洛法(sequential Monte Carlo, Gao & Zhang, 2012)等算出时, 这些方法便可用来求解待定参数的后验分布.然而对复杂的非线性多参数模型而言, 似然函数往往不可知, 或者计算边缘似然函数有困难.在此背景下, 近似贝叶斯计算法(Approximate Bayesian Computation)应运而生(Toni et al., 2009; Beaumont, 2010; Marin et al., 2012), 并被广泛应用(Csilléry et al., 2010; Vrugt & Sadegh, 2013; Hartig et al., 2014).在此, 我们介绍如何应用Lenormand等(2013) 提出的自调近似贝叶斯计算法(APMC)来估计植物光合模型的几个重要参数. ...
Approximate Bayesian computational methods.
1
2012
... 其中, p(y|θ)为似然函数(likelihood function), p(y)为边缘似然函数(marginal likelihood function).当似然函数和边缘似然函数易得, 或可以通过某种近似法, 诸如马尔可夫链蒙特卡洛法(Markov chain Monte Carlo, Andrieu et al., 2010)、重点取样法(importance sampling, MacEachern et al., 1999)以及序贯蒙特卡洛法(sequential Monte Carlo, Gao & Zhang, 2012)等算出时, 这些方法便可用来求解待定参数的后验分布.然而对复杂的非线性多参数模型而言, 似然函数往往不可知, 或者计算边缘似然函数有困难.在此背景下, 近似贝叶斯计算法(Approximate Bayesian Computation)应运而生(Toni et al., 2009; Beaumont, 2010; Marin et al., 2012), 并被广泛应用(Csilléry et al., 2010; Vrugt & Sadegh, 2013; Hartig et al., 2014).在此, 我们介绍如何应用Lenormand等(2013) 提出的自调近似贝叶斯计算法(APMC)来估计植物光合模型的几个重要参数. ...
a). Temperature response of parameters of a biochemically based model of photosynthesis. II. A review of experimental data. Plant,
b). Temperature response of parameters of a biochemically based model of photosynthesis. I. Seasonal changes in mature maritime pine (Pinus pinaster Ait.). Plant,
... 温度修正模型比较.实线: von Caemmerer等(2009)的曲线(S = 710, H = 220 000), 点线: 增加S约10%的曲线(S = 790, H = 220 000), 虚线: 本文用的曲线(Cm = 0.3, Tm = 37). ...
... Variables and parameters used in the photosynthesis model and their reference values mainly from Caemmerer et al. (2009) ...
Toward diagnostic model calibration and evaluation: Approximate Bayesian computation.
1
2013
... 其中, p(y|θ)为似然函数(likelihood function), p(y)为边缘似然函数(marginal likelihood function).当似然函数和边缘似然函数易得, 或可以通过某种近似法, 诸如马尔可夫链蒙特卡洛法(Markov chain Monte Carlo, Andrieu et al., 2010)、重点取样法(importance sampling, MacEachern et al., 1999)以及序贯蒙特卡洛法(sequential Monte Carlo, Gao & Zhang, 2012)等算出时, 这些方法便可用来求解待定参数的后验分布.然而对复杂的非线性多参数模型而言, 似然函数往往不可知, 或者计算边缘似然函数有困难.在此背景下, 近似贝叶斯计算法(Approximate Bayesian Computation)应运而生(Toni et al., 2009; Beaumont, 2010; Marin et al., 2012), 并被广泛应用(Csilléry et al., 2010; Vrugt & Sadegh, 2013; Hartig et al., 2014).在此, 我们介绍如何应用Lenormand等(2013) 提出的自调近似贝叶斯计算法(APMC)来估计植物光合模型的几个重要参数. ...
Net ecosystem CO2 exchange over a larch forest in Hokkaido, Japan.
2004
A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I: Model description and comparison with a multi- layered model.
1
1998
... 涡度相关技术(eddy covariance technique)作为目前直接测定地表-大气间CO2和水热通量的标准方法, 为原位、无破坏监测生态系统的光合作用(实质上植被大气之间的CO2交换)提供了新的选择(de Pury & Farquhar, 1997; Wang & Leuning, 1998; Dai et al., 2004; Groenendijk et al., 2011; Kosugi et al., 2013).在过去的30余年里, 该方法有了快速的发展, 得到了广泛的应用, 目前已成为国际通量观测网络(FLUXNET)的主要技术手段.据不完全统计, 目前全球安装涡度相关监测系统的研究站已经超过5 000个, 仅中国就超过了300个.考虑到涡度相关法可以提供冠层光合作用的实测值, 而叶片尺度的光合模型也很明确, 笔者认为可以在叶片光合模型的基础上提出一种算法, 直接反演出Vcmax和Jmax这两个重要的生理参数在冠层上的数值.如果这一想法能够实现, 我们就有可能从生理学角度切入, 对大量冠层尺度的监测数据进行更深入的解读和挖掘.作为一个初步的尝试, 为了简化模型的数据和计算, 本文通过将冠层抽象为一片大叶, 提出了一种相对简洁却行之有效的方法来反演冠层的生理参数.下文中, 笔者将对该方法的技术细节进行分解, 希望该方法能引起读者的兴趣, 在读者的批评中获得进一步提高和完善. ...
Seasonal trends in photosynthetic parameters and stomatal conductance of blue oak (Quercus douglasii) under prolonged summer drought and high temperature.