摘要
定量描述植被总初级生产力(GPP)对于全球碳循环和全球气候变化研究具有重要意义。针对MODIS MOD_17 GPP(MOD_17)产品在通量站点低估的现象,通过3个实验依次改进了模型输入参数(气象数据和吸收的光合有效辐射吸收比例(f PAR))和模型本身的参数(最大光能利用率),分析了各个参数对模拟结果的不确定性影响,结果表明各参数对模拟结果都有不同程度的影响。在阿柔草地站,最大光能利用率的重新标定对结果影响最大,GPP估算结果的提高最为明显;在关滩森林站利用广义神经网络算法得到的GLASS f PAR代替原始MODIS f PAR产品,比其他参数的改进效果更明显,GPP的值更接近涡动通量观测值。利用改进的MOD_17模型重新估算了黑河上游2001–2012年间植被GPP,通过趋势分析得出该研究时段内GPP以9.58 g C·m^(–2)·a^(–1)的平均速率呈上升趋势。同时计算了气候因子(温度、降水和饱和水汽压差(VPD))与时间序列GPP的偏相关性,分析了植被GPP对气候变化的响应情况,2001–2012年平均温度和VPD与年GPP大部分区域呈正相关,体现了温度和VPD对植被生长的促进作用;2001–2012年的降水量与年GPP无明显相关,且大部分区域呈负相关。
Aims Quantifying the gross primary productivity（GPP） of vegetation is of primary interest in studies of global carbon cycle. This study aims to optimize the MODIS GPP model for specific environments of a fragile waterhead ecosystem, by performing simulations of long-term（from 2001 to 2012） GPP with optimized MOD17 model, and to analyze the response of GPP to the local climatic variations. Methods The original MODIS GPP products that underestimate GPP were validated against two years（2010^–2011） of eddy covariance（EC） data at two sites（i.e. an alpine pasture site and a forest site, respectively） in the upstream of Heihe River Basin. Three comparative experiments were then conducted to analyze the effects of input parameters derived from three sources（i.e. meteorological, biome-specific, and fraction of absorbed photosynthetically active radiation（f PAR） parameters） on the model behavior. After refining the model-driven parameters, long-term GPPs of the study area were estimated using the optimized MOD17 model, and the Least Absolute Deviation method was applied to analyze the partial correlations between interannual GPPs and climatic variables（temperature, precipitation and vapor pressure deficit（VPD））. Important findings The uncertainties in the original MODIS GPP products are attributable to biome-specific parameters, input data（e.g. meteorological and radiometry data） and vegetation maps. At the pasture site, the light use efficiency had the strongest impact on the GPP simulations. The refined f PAR calculated from the leaf area index（LAI） products of Global Land Surface Satellite（GLASS） greatly improved the GPP estimates, especially at the forest site. The GPPs from the optimized MOD17 model well matched the EC data（R2 = 0.90, root mean squared error（RMSE） = 1.114 g C·m^–2·d^–1 at the alpine pasture site; R2 = 0.91, RMSE = 0.649 g C·m^–2·d^–1 at theforest site）. The time series of GPPs displayed an up trend at an average rate of 9.58 g C·m^
出处
《植物生态学报》
CAS
CSCD
北大核心
2016年第1期1-12,共12页
Chinese Journal of Plant Ecology
基金
国家重点基础研究发展计划"973计划"(2013CB733404)
中央级公益性科研院所基金(IFRIT201302)
