双集合卡尔曼滤波LAI同化结合BEPS模型的竹林生态系统碳通量模拟  

Simulating of carbon fluxes in bamboo forest ecosystem using BEPS model based on the LAI assimilated with Dual Ensemble Kalman Filter

作　　者：李雪建[1,2] 毛方杰[1,2] 杜华强[1,2] 周国模[1,2] 徐小军[1,2] 李平衡[1,2] 刘玉莉[1,2] 崔璐[1,2] 

机构地区：[1]浙江省森林生态系统碳循环与固碳减排重点实验室,浙江临安311300 [2]浙江农林大学环境与资源学院,浙江临安311300

出　　处：《应用生态学报》2016年第12期3797-3806,共10页Chinese Journal of Applied Ecology

基　　金：浙江省杰出青年科学基金项目(LR14C160001);国家自然科学基金项目(31670644;31370637;31500520);浙江省自然基金项目(LQ15C160003);浙江农林大学农林碳汇与生态修复研究中心研究基金项目;浙江省林学一级重中之重学科学生创新计划项目(201511)资助~~

摘　　要：叶面积指数(LAI)是森林生态系统碳循环研究的重要观测数据,也是驱动森林生态系统模型模拟碳循环的重要参数.本文以毛竹林和雷竹林为研究对象,首先利用双集合卡尔曼滤波,同化两种竹林生态系统观测站点2014—2015年MODIS LAI时间序列数据,然后将同化的高质量毛竹LAI和雷竹LAI作为输入数据驱动BEPS模型,模拟两种竹林生态系统总初级生产力(GPP)、净生态系统碳交换量(NEE)和总生态系统呼吸(TER)等碳循环数据,并用通量站实际观测值评价模拟结果;另外,还对比不同质量LAI对碳循环模拟的影响.结果表明:双集合卡尔曼滤波同化得到的毛竹林和雷竹林LAI与实测LAI之间的相关关系极为显著,R^2分别为0.81和0.91,且均方根误差和绝对偏差均较小,极大地提高了MODIS LAI的产品精度;在同化得到的LAI驱动下,BEPS模型模拟的毛竹林GPP、NEE和TER与实际观测值之间的R^2分别为0.66、0.47和0.64,雷竹林分别为0.66、0.45和0.73,模拟结果均好于三次样条帽盖算法平滑LAI模拟得到的GPP、NEE和TER,其中,毛竹林、雷竹林NEE的模拟精度提高幅度最大,分别为11.2%和11.8%.LAI is one of the most important observation data in the research of carbon cycle of forest ecosystem, and it is also an important parameter to drive processbased ecosystem model. The Moso bamboo forest （MBF） and Lei bamboo forest （LBF） were selected as the study targets. Firstly, the MODIS LAI time series data during 2014-2015 was assimilated with Dual Ensemble Kalman Filter method. Secondly, the high quality assimilated MBF LAI and LBF LAI were used as input dataset to drive BEPS model for simulating the gross primary productivity （GPP）, net ecosystem exchange （NEE） and total ecosystem respiration （TER） of the two types of bamboo forest ecosystem, respectively. The modeled carbon fluxes were evaluated by the observed carbon fluxes data, and the effects of different quality LAI inputs on carbon cycle simulation were also studied. The LAI assimilated using Dual Ensemble Kalman Filter of MBF and LBF were significantly correlated with the observed LAI, with high R^2 of 0.81 and 0.91 respectively, and lower RMSE and absolute bias, which represented the great improvement of the accuracy of MODIS LAI products. With the driving of assimilated LAI, the modeled GPP, NEE, and TER were also highly correlated with the flux observation data, with the R^2 of 0.66, 0.47, and 0.64 for MBF, respectively, and 0.66, 0.45, and 0.73 for LBF, respectively. The accuracy of carbon fluxes modeled with assimilated LAI was higher than that acquired by the locally adjusted cubicspline capping method, in which, the accuracy of mo-deled NEE for MBF and LBF increased by 11.2% and 11.8% at the most degrees, respectively.

关 键 词：竹林 碳通量 LAI同化 双集合卡尔曼滤波 BEPS模型 

分 类 号：S795[农业科学—林木遗传育种]