基于经验模型和机理模型的杉木人工林生物量估测对比研究  

Comparative study on biomass estimation of Cunninghamia lanceolata plantations based on empirical model and mechanism model

作　　者：李佳怡 阮红玉 邱思玉 梁瑞婷 朱兆廷 文毅 王成雨 王轶夫[1] 孙玉军[1] LI Jiayi;RUAN Hongyu;QIU Siyu;LIANG Ruiting;ZHU Zhaoting;WEN Yi;WANG Chengyu;WANG Yifu;SUN Yujun(State Forestry and Grassland Administration Key Laboratory of Forest Resources and Environmental Management,Beijing Forestry University,Beijing 100083,China;Jishui County Shiyang Forest Farm,Ji′an 331600,China)

机构地区：[1]北京林业大学森林资源和环境管理国家林业和草原局重点实验室,北京100083 [2]吉水县石阳林场,吉安331600

出　　处：《生态学报》2024年第8期3502-3514,共13页Acta Ecologica Sinica

基　　金：江西省林业局科技创新专项([2021]33);国家自然科学基金项目(31800536)。

摘　　要：为了揭示间伐干扰下杉木人工林生物量的变化规律,研究利用江西省吉水县石阳林场的36块杉木人工林样地的实测数据和研究区气候数据,通过基于经验的引入地位指数(SI)的生物量生长方程组和基于机理的3⁃PG模型,模拟并预估林分生物量,分析在间伐和非间伐的情况下,不同立地的林分其生物量0—50a的变化。结果表明:(1)构建了生物量生长方程组,并在参数a、b、c中引入地位指数SI,发现改进后的模型对于基础模型拟合精度更高,且对数似然比检验表明,改进效果显著(P<0.05)。(2)通过对3⁃PG模型预测精度验证发现,预估值和实测值之间有很高的一致性,各因子的决定系数(R^(2))在0.65—0.96之间,其中,胸径和树高的R^(2)均高于0.92;各因子的平均相对误差(MRE)不超过26%。(3)通过比较经验模型和机理模型的生物量预测发现,经验模型的预测误差MRE为16.50%,机理模型为23.52%,经验模型预估精度更高。进一步对未来预测对比分析表明,机理模型预估值高于经验模型。(4)两个模型模拟的杉木人工林生物量规律一致,即随着林龄的增加,杉木人工林林分总生物量均表现出先快速增加,后逐渐平稳的趋势;并且间伐不会改变这种趋势,但间伐林分在间伐后的生物量生长速率高于无间伐林分。此外,由于SI对经验模型影响显著,改进模型拟合效果更好,更具有生态学意义。参数化后的3⁃PG模型模预估精度较高,能够为江西杉木人工林生长规律研究提供依据。虽然经验模型和机理模型在对研究区杉木人工林生物量的预估上均具有较好的表现,但各具特点和局限性。经验模型参数较易获得,且经验模型预测生物量、林分胸高断面积和林分平均树高的R^(2)、MRE均优于机理模型;但模型对于建模数据内的评价效果较好,对于建模数据外的应用具有局限性,即经验模型更适合模拟生长期间的某一阶段的林In order to reveal the variation patterns of biomass of Cunninghamia lanceolata plantations under thinning interference and provide theoretical supports for the biomass prediction of Cunninghamia lanceolata plantations after thinning,this study used the measured data of 36 Cunninghamia lanceolata plantations plots and the climate data in Shiyang Forest Farm,Jishui county,Jiangxi province.Through the biomass growth equations with the site index(SI)introduced and the 3⁃PG model,the biomass of the stand was simulated and estimated,and the forest stand biomass in different sites was analyzed under the condition of thinning and non⁃thinning from 0 to 50 years.The results showed that:(1)The biomass growth equations were constructed,and the site index(SI)was introduced into the parameters a,b and c.It was found that the improved model had higher fitting accuracy compared with the basic model,and the logarithmic likelihood ratio test showed that the improvement effect was significant(P<0.05).(2)Through the verification of the prediction accuracy of the 3⁃PG model,there was strong consistency between the predicted values and the measured values.The determination coefficient(R^(2))of the linear regression equations between the measured and estimated values of each factor ranged from 0.65 to 0.96.The R^(2)of diameter at breast height(DBH)and tree height was higher than 0.92.The mean relative error(MRE)of each component was less than 26%.(3)By comparing the biomass prediction of the empirical model and the mechanism model,it was found that the prediction error MRE of the empirical model was 16.50%,and the mechanism model was 23.52%.The prediction accuracy of the empirical model was higher.Further comparative analysis of future predictions showed that the estimated value of the mechanism model was higher than that of the empirical model.(4)The trend of biomass of Cunninghamia lanceolata plantations simulated by the two models was consistent,that is,with the increase of forest age,the forest biomass showed a trend of rap

关 键 词：生物量生长方程组 3⁃PG模型 杉木人工林 间伐 林分生长 

分 类 号：S718.5[农业科学—林学]