基于模型数据融合的千烟洲亚热带人工林碳水通量模拟

Modeling of carbon and water fluxes of Qianyanzhou subtropical coniferous plantation using model-data fusion approach

人工林生态系统是我国森林生态系统的重要组成部分,在全球碳平衡中的作用越来越受到重视。利用千烟洲亚热带人工针叶林通量观测站的碳水通量和气象观测数据,通过模型数据融合方法对碳水循环过程模型——SIPNET模型关键参数进行反演,模拟了2004—2009年千烟洲人工林生态系统的碳水通量。结果表明:仅用碳通量观测数据优化模型参数时,净生态系统碳交换量(NEE)模拟效果较好(R2=0.934),而生态系统蒸散(ET)模拟效果较差(R2=0.188);同时用碳水通量观测数据优化时,NEE模拟效果稍差(R2=0.929),但ET模拟效果显著提升(R2=0.824),说明利用碳水通量观测数据同时优化,SIPNET模型才能较好地模拟试验站点碳水通量。在此基础上,开展了人工林生态系统碳通量对降水变化响应的敏感性分析,发现降水量减少对光合作用的影响比对呼吸作用的影响更为强烈,且碳水通量同时参与优化时模型才能较好地模拟碳通量随降水减少而快速降低的趋势,表明如果不能同时利用碳水通量进行参数优化,模型无法正确揭示生态系统碳循环对降水变异的响应。

As one of the most widespread forest types, China＇s plantation plays an important role in global carbon balance. It is crucial to reduce the uncertainties in the estimation of carbon and water fluxes of plantation ecosystems, and model-data fusion technique provides an effective way. The purpose of this research is to improve the modeling accuracy of SIPNET model, the simplified Photosynthesis and Evapo-Transpiration （ET） model through two experiments, namely NEE alone and NEE ＆ ET multi-constraints. The model-data fusion method used here is a combination of Metropolis-Hastings algorithm and Simulated Annealing algorithm. Based on eddy fluxes and meteorological observation data of Qianyanzhou subtropical eoniferous plantation during 2004--2009 in ChinaFLUX （ Chinese Terrestrial Ecosystem Flux Research Network ）, we estimated the key parameters of SIPNET model and simulated the corresponding carbon and water fluxes. Comparisonsbetween the measured and modeled net ecosystem exchange of carbon dioxide （NEE） showed that the SIPNET model had approximately equivalent fits to the observed NEE under two optimization procedures （R2 decreased from 0. 934 to 0. 929, and RMSE increased from 0. 736 g C/m2 to 0. 763 g C/m2 ）. In the case of ET, the NEE and ET parameterization produced a markedly better fit to the observed ET than the NEE parameterization （R2 increased from 0. 188 to 0. 824, and RMSE decreased from 0. 152 cm to 0. 053 cm）. As for transpiration, when optimized by observed NEE alone, SIPNET largely underestimated annual accumulated transpiration in 2004 compared with the measurements of sap flow technique. In comparison, while optimization based on NEE and ET, SIPNET led to a better fit of annual cumulative estimation of transpiration in 2004 to the sap flow measurement. These results indicated that the SIPNET model parameterized using NEE and ET observed fluxes could well reproduce the characteristics of carbon and water fluxes. In other words, more information can be extracted from simultaneous optimization, since there is additional process information in water flux observation data. Furthermore, we conducted a sensitivity test of precipitation on carbon fluxes through reduction of precipitation. We found that photosynthesis was more sensitive to precipitation reduction than respiration, and the model optimized using NEE and ET reproduced the response of NEE to precipitation reduction better than that optimized using NEE alone. In addition, we detected that the difference of NEE response to precipitation reduction in two optimizations of the SIPNET model was caused by gross ecosystem production rather than ecosystem respiration. Therefore, parameter estimation using NEE and ET altogether improved the performance of SIPNET model. And without optimization using both NEE and ET, the response of ecosystem carbon cycle to precipitation variation may be misrepresented.