中文摘要：

准确估计和预测陆地生态系统碳循环时空变化是预测气候变化的基础,也是目前全球变化研究中最为重要的前沿领域之一.最大光能利用率（εmax）是遥感估算陆地生态系统初级生产力的关键参数之一,本研究基于CASA（Carnegie-Ames-Stanford Approach）模型,采用马尔科夫链-蒙特卡罗（Markov Chain Monte Carlo,简称MCMC）方法,利用中国陆地生态系统通量观测研究网络（ChinaFLUX）8个野外台站的涡度相关通量观测数据对εmax进行反演,得到εmax的最优估计值及其不确定性,并利用优化后的εmax对2003～2008年各生态系统总初级生产力（Gross Primary Productivity,简称GPP）及其不确定性进行了模拟.结果表明：8个生态系统εmax后验估计结果均呈近似正态分布,森林、农田和草地生态系统εmax分别为0.737±0.026 ～0.850±0.035g C/MJ·PAR,1.056±0.090g C/MJ· PAR和0.199±0.068～0.469±0.043g C/MJ·PAR;εmax估计的不确定性将导致内蒙、当雄和海北草地生态系统GPP年总量的模拟值产生9.17％ ～ 14.20％的误差,长白山、鼎湖山、千烟洲和西双版纳4个森林生态系统GPP年总量的误差为3.52％～7.79％,禹城农田生态系统GPP年总量的误差为8.52％.对εmax进行优化后,GPP年总量模拟值的相对误差显著降低,有效改善了原模型对内蒙草地生态系统GPP年总量的高估和除当雄草地生态系统外其他6个生态系统GPP年总量的低估.

英文摘要：

Accurate estimation and forecasting of terrestrial ecosystem dynamic carbon cycles and its spatial-temporal pattern are crucial for climate prediction in the context of global change.The maximum light use efficiency （εmax） is one of key parameters of remote sensing models to estimate primary production in terrestrial ecosystems.Based on the CASA model and eddy covariance flux observations at 8 ChinaFLUX sites （Changbaishan temperate mixed forest,Qianyanzhou subtropics evergreen needle leaf forest,Dinghushan subtropics evergreen broadleaf and needle leaf mixed forest,Xishuangbanna tropical evergreen broadleaf forest,Inner Mongolia typical temperate grassland,Haibei alpine meadow,Dangxiong alpine steppe-meadow and Yucheng warmer temperate dry farming cropland）,we used the Markov Chain Monte Carlo （MCMC） method to inverse εmax and calculated the optimum values and uncertainties from the posterior probability distributions.Using optimized values of εmax,we modeled the gross primary production （GPP） and uncertainties for each ecosystem in 2003～ 2008.The results show that estimated posterior probability distributions of εmax at eight sites followed approximately normal distributions.Estimated values of εmax were 0.737±0.026 ～ 0.850±0.035g C/MJ PAR at forest sites,1.056±0.090g C/MJ PAR at cropland site,and 0.199±0.068～0.469±0.043g C/MJ PAR at grassland sites.The uncertainties of modeled annual GPP caused by the error of estimating εmax ranged from 9.17％ to 14.20％ at three grassland sites,3.52％ to 7.79％ at four forest sites,and 8.52％ at cropland site.The relative error of annual GPP was reduced after using optimized εmax values.The optimization of εmax improved the original model that overestimated annual GPP in Inner Mongolia grassland ecosystem and underestimated in the other six ecosystems except Dangxiong grassland ecosystem.